EMF Risk Assessment:
Exposure Systems for Large-Scale Laboratory and
Experimental Provocation Studies
Diss. ETH No. 18636
EMF Risk Assessment:
Exposure Systems for
Large-Scale Laboratory and
Experimental Provocation
Studies
A dissertation submitted to
ETH ZURICH
for the degree of
Doctor of Sciences
presented by
SVEN EBERT
Dipl. Phys., TU Berlin, Germany
B.Sc. Physics, UMIST, Great Britain
born December 25th, 1970
citizen of Germany
accepted on the recommendation of
Prof. Dr. W. Fichtner, examiner
Prof. Dr. N. Kuster, Prof. Dr. P. Niederer, co-examiner
2009
Contents
Summary
v
Zusammenfassung
ix
Acknowledgements
I
xiii
Introduction
1
1 Background and Motivation
1.1 Worldwide Mobile Communication .
1.2 Relevant Electromagnetic Spectrum
1.3 EMF Health Risk Assessment . . . .
1.4 Laboratory Studies . . . . . . . . . .
1.5 Specific Absorption Rate . . . . . . .
1.6 RF Safety Standard . . . . . . . . .
1.7 Motivation . . . . . . . . . . . . . .
1.8 Objectives and Chapter Overview . .
II
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Exposure Assessment
2 Thermal Regulatory and
olds
2.1 Abstract . . . . . . . .
2.2 Introduction . . . . . .
2.3 Material and Methods
3
3
5
6
8
10
10
12
13
17
Thermal Breakdown Thresh. . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . .
i
19
19
20
22
ii
CONTENTS
2.4
2.5
2.6
2.3.1 Animals . . . . . . . . . . . . . .
2.3.2 Exposure Protocol . . . . . . . .
2.3.3 Temperature Measurement . . .
2.3.4 Evaluation of Thermal Response
2.3.5 Radiofrequency Exposure . . . .
Results . . . . . . . . . . . . . . . . . . .
Discussion . . . . . . . . . . . . . . . . .
Conclusion . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
22
22
24
24
25
32
33
37
3 Assessment of ELF MF of Mobile Phones and Proposal
for Worst-Case Signal for Bio-Experiments
3.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Material and Methods . . . . . . . . . . . . . . . . . .
3.3.1 B-Field Probes . . . . . . . . . . . . . . . . . .
3.3.2 Measurement Procedure . . . . . . . . . . . . .
3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5 Generic Exposure Test Signal . . . . . . . . . . . . . .
3.6 Discussion and Conclusion . . . . . . . . . . . . . . . .
39
39
40
41
41
42
44
48
48
4 Add-on ELF MF Exposure Setup Proposal for
with Human Volunteers
4.1 Introduction . . . . . . . . . . . . . . . . . . . .
4.2 Setup Design . . . . . . . . . . . . . . . . . . .
4.3 B-Field Strength and Distribution . . . . . . .
4.4 Current Requirements . . . . . . . . . . . . . .
4.5 Summary ELF Add-on Setup . . . . . . . . . .
51
51
52
53
54
55
III
Studies
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Exposure Systems and Studies
5 Flexible Exposure Setup for Experimental Provocation
Studies at 884 MHz
5.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . .
5.3 Requirements . . . . . . . . . . . . . . . . . . . . . . .
5.4 Exposure System Design . . . . . . . . . . . . . . . . .
5.4.1 Low Weight Stacked Micropatch Antenna . . .
57
59
59
60
61
65
65
CONTENTS
5.5
5.6
5.7
iii
5.4.2 Heat Load . . . . . . . . . . . . . . . . . . . . .
5.4.3 Headset . . . . . . . . . . . . . . . . . . . . . .
5.4.4 Exposure Room, Monitoring and Control System
5.4.5 Exposure Signal . . . . . . . . . . . . . . . . .
Dosimetry and Validation . . . . . . . . . . . . . . . .
Conclusion . . . . . . . . . . . . . . . . . . . . . . . .
Study Abstract: PERFORM C . . . . . . . . . . . . .
71
72
77
82
84
92
93
6 Exposure Systems for Large-Scale In Vivo Laboratory
GSM/DCS Risk Assessment Studies
95
6.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . 96
6.3 Objectives and Requirements . . . . . . . . . . . . . . 97
6.3.1 Study Outline and Objectives . . . . . . . . . . 97
6.3.2 Exposure and Environmental Requirements . . 99
6.4 Setup Design . . . . . . . . . . . . . . . . . . . . . . . 100
6.4.1 Mechanical and Electrical Design . . . . . . . . 104
6.4.2 Signal Generation, Monitoring and Control of
Exposure . . . . . . . . . . . . . . . . . . . . . 109
6.4.3 Exposure Signal . . . . . . . . . . . . . . . . . 112
6.5 Dosimetry and Validation . . . . . . . . . . . . . . . . 116
6.5.1 Dosimetric Method, Tools and Models . . . . . 116
6.5.2 Initial Setup Challenges . . . . . . . . . . . . . 124
6.5.3 Results of Dosimetry, Uncertainty and Variations
126
6.6 Setup Performance during Study . . . . . . . . . . . . 134
6.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . 138
6.8 Study Abstracts: PERFORM A . . . . . . . . . . . . . 140
6.8.1 Fraunhofer ITEM (Germany) . . . . . . . . . . 141
6.8.2 RBM (Italy) . . . . . . . . . . . . . . . . . . . 143
7 SAR Uniformity in Ferris Wheel Setups
7.1 Abstract . . . . . . . . . . . . . . . . . .
7.2 Introduction . . . . . . . . . . . . . . . .
7.3 Objectives . . . . . . . . . . . . . . . . .
7.4 Examined Ferris Wheels . . . . . . . . .
7.5 Ferris Wheel Setup Design . . . . . . . .
7.6 Material and Methods . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
145
145
146
146
147
147
150
iv
CONTENTS
7.6.1
7.6.2
7.7
7.8
7.9
Experimental Investigations . . . . . . . . . . .
Temperature Method for SAR Uniformity Assessment . . . . . . . . . . . . . . . . . . . . .
7.6.3 Measurement Procedure . . . . . . . . . . . . .
7.6.4 Dummy Characteristics . . . . . . . . . . . . .
7.6.5 Numerical Evaluation . . . . . . . . . . . . . .
Results . . . . . . . . . . . . . . . . . . . . . . . . . . .
Discussion . . . . . . . . . . . . . . . . . . . . . . . . .
Conclusions . . . . . . . . . . . . . . . . . . . . . . . .
150
152
154
155
156
157
161
162
Epilogue
163
List of Acronyms
169
Publications
171
Bibliography
176
Summary
Various biological and health effects have been suggested as resulting from exposure to radiofrequency electromagnetic fields (RF-EMF)
as emitted by mobile phones and other wireless communication devices. With the success of mobile communication technologies since
the 1990’s, the numbers of people exposed to RF-EMF has increased
to such an extent that even small health effects become relevant. Particular concerns also focus on long-term effects such as the promotion of cancer, which might only become evident after a considerable
delay. In addition to epidemiological examinations, health risk assessments require laboratory studies. These studies can be divided
in in vitro, in vivo and human studies. For the assessment of health
effects, in vivo studies are most relevant.
So far, laboratory studies have not provided consistent
results. In addition to the probably large inherent variability
in biological responses, undefined exposure conditions, variations
in exposure and artifacts resulting from the exposure equipment are discussed as possible reasons for inconsistent study outcomes. To foster significant progress in the health risk assessment of low-level RF-EMF exposure from mobile phones, the
5th Framework program of the European Union started a large international toxicological/carcinogenic project (named PERFORM) with
a special focus on well-defined, well-controlled and artifact-free RF
exposures. The objectives of this dissertation were the development,
characterization and successful operation of optimized exposure systems for the in vivo and human studies of the European PERFORM
project. Significant contributions were achieved for all major requirements of such systems, to realize well-defined, well-controlled,
v
vi
SUMMARY
well-characterized, and artifact-free exposures.
For the large-scale, life-time in vivo laboratory study
PERFORM A, 1500 mice had to be exposed at two frequencies
and four groups (high dose, medium dose, low dose and sham). To
be most relevant for risk assessment, the high-dose level needs to be
chosen as high as possible but below the thermal threshold of inducing
health effects. An experimental study was conducted to assess the
thermal regulatory and breakdown thresholds for tube-restrained,
whole-body exposed mice. The regulation of temperature started
at specific absorption rate (SAR) levels between 2 and 5 W/kg; the
level of SAR induced heating that cannot be compensated by the
animal (SAR threshold for a thermal breakdown) was assessed to
start as low as 6 W/kg under these laboratory conditions. In the
PERFORM A study the exposure level of the high dose exposure
group was then set at 4 W/kg.
For the mouse studies of PERFORM A, two types of efficient
exposure setups for 900 and 1800 MHz signals were developed.
The systems allow double-blind study protocols and are fully selfcontrolled (monitoring exposure and further parameters as oxygen,
temperature and humidity) with an auto-detection of malfunctions.
The applied exposure signal has a complexity covering different exposure scenarios of mobile phone uses. Different exposure phases included extremely low frequency (ELF) spectral components resulting
from the DTX/non-DTX, power saving and environmental control
signals. The experimental and numerical dosimetry was conducted
with details that have not been provided in any previous study. Assessed were the whole-body averaged and organ averaged SAR for
the different exposure phases, including their uncertainties and variations occurring in weight, anatomy, dielectric parameters, posture
and positioning. Exposure verification was enabled through a new,
experimental, dosimetric temperature method which verifies the exposure at each mouse position in the exposure resonator. Within this
project, Dr. Veronica Berdiñas and Dr. Walter Oesch conducted the
detailed numerical dosimetry and system software, which are not part
of this thesis.
A study has also been performed to compare different mouse exposure systems based on the Ferris wheel concept (the setup used in the
widely discussed Adelaide study that has shown a two-fold increase
vii
in lymphoma cancer in transgenic mice and the new developed setups
for the exposure of mice in PERFORM A and B). Thereby, a major
drawback has been identified for Ferris-wheel-like resonant structures:
their sensitivity to asymmetrical load conditions, resonator size, quality factor, asymmetries of mechanical constructions, etc., which can
result in higher mode excitation and highly varying exposure conditions. It was shown that variations in whole-body SAR as large as
7.4 dB can occur for mice in the same setup. Several concepts to improve the uniformity and inter-animal variations have been identified
within this thesis (small units, use of high permittivity bricks, homogeneous load). In this way, the variations of the PERFORM A
setup were limited to less than 2.5 dB. By using a rotational scheme,
variations over the life-time study are even as low as 1.2 dB.
Furthermore an optimized exposure system was developed for the
human provocation study PERFORM C (a study considered to be
one of the largest studies with human volunteers within the EMF
risk assessment context). The study design foresees exposures of subjects for a duration of three hours or even longer. This requires an
exposure system enabling head movements during exposure and was
realized with a new, low-weight antenna mounted on a flexible and
comfortable headset. The setup also incorporated a heated plate at
the ear lobe of subjects on the exposed side to simulate the thermal
load from, e.g., contact pressure or battery heat of a mobile phone
during usage. The dosimetry and uncertainty assessments were conducted using measurements and simulations. Artifacts, such as field
and SAR distortion due to EEG electrodes, were considered.
The final project covered by this thesis was an accurate assessment of the B-fields resulting from the amplifier supply currents of
mobile phones. For the first time, B-field distributions and B-field
waveforms were accurately measured for several commercial phones.
The study showed that B-fields up to 75 µT and rise times of 1.4 T/s
are present. Although the maximum field strengths from several harmonics of 217 Hz exceeded the ICNIRP guideline levels, it still is not
expected that the relevant current density in the body is also above
the threshold, since exposure peaks are very local. This work resulted
in a proposal for a worst-case test signal to be used for combined
RF/ELF studies.
For the conclusion of the summary the biological results are pre-
viii
SUMMARY
sented here. The PERFORM A mouse studies covered two combined chronic toxicity and carcinogenicity studies performed at Fraunhofer Institute of Toxicology and Experimental Medicine (ITEM) in
Germany and one co-carcinogenicity study performed at Istituto di
Ricerche Biomediche Antoine Marxer (RBM) in Italy. The outcomes
of the three studies showed no toxic/carcinogenic effects, even for the
high-dose group with an exposure in the range where the thermal
regulation mechanism within the animals starts being triggered. The
study at RBM was also set up as a replication study of the Adelaide
study; however, their findings were not confirmed. The Department of
Health Science at the Karolinska Institute in Sweden examined in the
human study PERFORM C effects of RF fields on self-reported symptoms, as well as the detection of fields after a three hour exposure time.
A well-defined study group was chosen including subjects reporting
symptoms attributed to mobile phone use. The results revealed that
neither group were able to detect RF exposure better than by chance.
The results also showed that headaches were more commonly reported
after RF exposure than in the sham-exposed control group, mainly due
to an increase in the subject group not reporting symptoms attributed
to mobile phone use. The higher prevalence of headaches justifies and
requires the further investigation of possible physiological correlations.
Zusammenfassung
Elektromagnetische Felder von Mobiltelefonen und anderen drahtlosen Geräten sind im Verdacht verschiedene biologische und gesundheitliche Effekte auszulösen. Seit den 1990er Jahren verbreiteten sich
die mobilen Kommunikationstechnologien sehr erfolgreich; inzwischen
sind eine so grosse Anzahl von Menschen durch die elektromagnetischen Hochfrequenzfelder (ELF-RF) von Mobiltelefonen exponiert,
dass auch geringste gesundheitsgefährdende Auswirkungen sehr relevant wären. Sorge bereiten insbesondere auch Auswirkungen, die
erst mit einer zeitlichen Verzögerung auftreten, wie z. B. Krebs. Die
Einschätzung des Gesundheitsrisikos erfordert epidemiologische Studien und Laborstudien (in vitro, in vivo und Studien mit Freiwilligen),
wobei in vivo Studien als besonders relevant gelten.
Bisher lieferten Laborstudien inkonsistente Ergebnisse. Das wird
verschiedenen Ursachen zugeschrieben: u. a. einer grossen Bandbreite
inhärenter (biologischer) Reaktionen, ungenau definierten Expositionsbedingungen, Variationen in der Exposition oder auch durch Artefakte bei Expositionssystemen. Für einen bedeutenden Fortschritt in
der Gesundheitsrisikoeinschätzung von ELF-RF Expositionen durch
Mobiltelefone wurde im 5. Rahmenprogramm der Europäischen Union ein grosses, internationales, toxikologisch/karzinogenes Forschungsprojekt (mit dem Namen PERFORM) gestartet, u. a. mit dem Ziel
wohl-definierter, wohl-kontrollierter und artefaktfreier Expositionen.
Zielsetzungen dieser Dissertation waren die Entwicklung, Charakterisierung und erfolgreiche Durchführung optimierter Expositionsanlagen für die in vivo Studien und die Freiwilligenstudie des Europäischen
Projekts PERFORM. Dabei wurde ein bedeutender Beitrag für
alle Hauptanforderungen solcher Systeme erzielt, um wohl-definierte,
ix
x
ZUSAMMENFASSUNG
wohl-kontrollierte, wohl-charakterisierte und artefaktfreie Expositionen zu ermöglichen.
In der grossen in vivo Langzeit-Laborstudie PERFORM A
wurden 1500 Mäuse bei zwei Frequenzen exponiert, unterteilt in
Hochdosis-, Mitteldosis-, Niedrigdosis- und Sham-Gruppen. Um die
Risikoabschätzung zu optimieren, muss die höchste applizierte Dosis
so hoch wie möglich gewählt werden, aber unterhalb des Schwellenwerts liegen, ab dem gesundheitliche Auswirkungen auftreten. Dafür
wurden in einer experimentellen Studie die beiden Schwellenwerte Einsetzen und Zusammenbruch der Körpertemperatur-Regelung durch
Ganzkörper EMF-RF Exposition für in Röhren fixierte Mäuse bestimmt. Die Temperaturregulation setzte bei einer spezifischen Absorptionsrate (SAR) zwischen 2 und 5 W/kg ein und kann bereits ab
6 W/kg unter den applizierten Laborbedingungen zusammenbrechen.
In der PERFORM A Studie wurde daraufhin die Exposition der Hochdosisgruppe auf 4 W/kg eingestellt.
Für die PERFORM A Mausstudien wurden zwei effiziente
Expositionsanlagen entwickelt – für 900 MHz und für 1800 MHz
Signale. Die Anlagen ermöglichen Doppelblindstudien, sind selbstkontrollierend (Sensoren überwachen die Exposition und weitere Parameter wie Sauerstoff, Temperatur, Luftfeuchtigkeit, etc.) und erkennen
Störungen automatisch. Das applizierte Expositionssignal ist komplex
und stellte verschiedene Expositionsszenarien dar, die beim Gebrauch
von Mobiltelefonen auftreten, wie niederfrequente Spektralbestandteile resultierend aus den DTX/non-DTX, Energieeinsparungs- und
Umgebungskontrollsignalen. Es wurde eine umfangreiche experimentelle und numerische Dosimetrie und Unsicherheitsanalyse mit mehr
Details als in bisherigen Studien durchgeführt. Für die verschiedenen
Expositionsphasen wurden die durchschnittlichen Ganzkörper-SAR
Werte sowie die durchschnittlichen SAR Werte für einzelne Organe ermittelt, einschliesslich ihrer Unsicherheiten und Variationen in Bezug
auf Masse, Anatomie, dielektrischen Parametern, Lage und Position.
Dabei ermöglichte eine neue, experimentelle, dosimetrische Methode
mittels Temperaturmessungen die Untersuchung der Expositionen an
jeder Position im Resonator. Innerhalb des Projekts wurden die numerischen Simulationen von Dr. Veronika Berdiñas und die Systemsoftware von Dr. Walter Oesch erstellt, sie sind somit nicht Teil dieser
Dissertation.
xi
Eine weitere Studie vergleicht drei verschiedene MausExpositionssysteme, die auf dem Ferris-Wheel Konzept basieren.
Untersucht wurden das Ferris-Wheel Expositionssystem aus der viel
diskutierten Adelaide-Studie, die eine Verdoppelung der LymphomKrebs-Rate in RF exponierten, transgenen Mäusen festgestellt hat,
sowie die beiden neu entwickelten Maus-Expositionssysteme von
PERFORM A und B. Die Studie hat wesentliche Nachteile von
Ferris-Wheel Expositionssystemen aufgedeckt: ihre Empfindlichkeit
bezüglich einer asymmetrischen Lastverteilung, zu grosser Resonator
Dimensionen, des Qualitätsfaktors und mechanischer Resonator
Asymmetrien. Dadurch können höhere Moden angeregt werden und
stark variierende Expositionen auftreten. Die Studie zeigt, dass
Schwankungen der Ganzkörper-SAR bis zu 7.4 dB zwischen Mäusen
im gleichen Resonator auftreten können. Es werden verschiedene
Konzepte vorgeschlagen, die die Gleichförmigkeit der EinzeltierExposition sowie die Tier-zu-Tier-Variationen verbessern (kleiner
Resonator Durchmesser, Vergrösserung der Distanzen zwischen
Nachbarn, homogene Lastverteilung). Auf diese Weise konnten die
Variationen im PERFORM A Resonator auf unter 2.5 dB reduziert
werden. Durch den Einsatz eines Rotationsschemas wurden die
auftretenden Variationen während der Langzeit-Studie sogar auf
1.2 dB reduziert.
Desweiteren wurde ein optimiertes Expositionssystem für
PERFORM C entwickelt (die Studie gilt als eine der grössten Freiwilligenstudien im Kontext einer EMF-RF Risikobewertung). In dieser
Provokationsstudie wurden Probanden für mindestens drei Stunden
mit einem Mobilfunksignal exponiert. Das erforderte ein Expositionssystem, welches Kopfbewegungen bei gleichzeitig konstanter Exposition erlaubt. Das wurde mit einer neuen, leichten Antenne realisiert, befestigt an einem flexiblen und bequemen Kopfhalter. Zudem löste auf
der exponierten Seite eine beheizte Keramikplatte am Ohrläppchen
einen Temperaturreiz aus, der Kontaktdruck- bzw. Batteriewärme des
Mobiltelefons während der Benutzung simulierte. Die Dosimetrie und
Unsicherheiten des Expositionssystems wurden durch Messungen und
Simulationen bestimmt; Artefakte, z. B. Einfluss von EEG Kabeln,
wurden berücksichtigt.
Die letzte Studie dieser Dissertation untersucht B-Felder, die
durch die Verstärker-Versorgungsströme in Mobiltelefonen verur-
xii
ZUSAMMENFASSUNG
sacht werden. Zum ersten Mal wurden die Magnetfeldverteilung und
-wellenform für mehrere kommerzielle Mobiltelefone vermessen. Die
Untersuchung zeigt, dass B-Felder bis zu 75 µT und Anstiegszeiten
von 1.4 T/s auftreten. Obgleich die maximalen Feldstärken einiger
Harmonischen von 217 Hz die ICNIRP-Grenzwerte übersteigen, ist
nicht zu erwarten, dass die relevante Stromdichte im Körper ebenfalls über dem ICNIRP-Grenzwert liegt, da diese Expositionsspitzen
sehr lokal auftreten. Als Ergebnis dieser Studie wird ein Testsignal für
kombinierte RF/ELF Studien vorgeschlagen.
Zum Abschluss der Zusammenfassung werden die biologischen Studienresultate vorgestellt. Die PERFORM A Mausstudien beinhalten
die Durchführung von zwei kombinierten Giftigkeits- und KarzinogenStudien am Fraunhofer Institut für Toxikologie und Experimentelle
Medizin in Deutschland, sowie eine Co-Karzinogen-Studie am Istituto di Ricerche Biomediche Antoine Marxer (RBM) in Italien. Die Ergebnisse der drei Studien zeigen keine negativen Auswirkungen, auch
nicht in der Gruppe mit der höchsten Dosis, dessen Expositionsstärke
den Temperatur-Regulationsmechanismus in Mäusen auslösen kann.
Die Studie bei RBM wurde als Replikationsstudie der Adelaide-Studie
konzipiert, jedoch konnten die Resultate nicht bestätigt werden. Das
Department of Health Science am Karolinska Institute in Schweden
untersuchte in der Freiwilligenstudie PERFORM C die Wirkung von
RF-Feldern auf Probanden bezüglich selbst-berichteter Symptome
und inwieweit sie RF Felder nach einer dreistündigen Expositionsdauer wahrnehmen können. Die Untersuchung wurde mit einer definierten Expositionsgruppe durchgeführt, die auch Personen einschloss, die
über Symptome bei Mobilfunkbenutzung berichteten. Die Ergebnisse
zeigen, dass keine Probandengruppe RF Exposition häufiger wahrnehmen konnte als es durch Zufallstreffer möglich wäre. Ferner zeigen die
Ergebnisse, dass häufiger von Kopfschmerzen nach der RF-Exposition
berichtet wurde als nach der Sham-Exposition, sogar hauptsächlich
von Probanden, die sonst von keinen Symptomen bei Mobilfunkbenutzung berichteten. Das häufigere Auftreten von Kopfschmerzen rechtfertigt und erfordert weitere Untersuchungen auf mögliche physiologische Wechselbeziehungen.
Acknowledgements
The Foundation for Research on Information Technologies in
Society (IT’IS), the industrial partner Schmid & Partner Engineering AG (SPEAG) and the Bioelectromagnetics/EMC group of the
Integrated Systems Laboratory (IIS) at the Swiss Federal Institute
of Technology (ETH) provided an excellent environment for the enthralling tasks accomplished during this thesis.
Prof. Niels Kuster provided me with the opportunity to work
on the scientific relevant, international projects PERFORM A and
PERFORM C of the European Union as well as on the national
ELF project of the Federal Office of Public Health (FOPH) and
enabled participations worldwide on conferences and project meetings (e.g., Austria, Canada, China, Germany, Ireland, Italy, Sweden,
Switzerland, USA). Additionally his support and enormous experience
in bioelectromagnetics provided in tricky situations always a helping
hand and guidance. Thank you, Niels.
I would like to thank Prof. Wolfgang Fichtner for giving me the
opportunity to work within his excellent group and for supervising
this thesis. He was there in good and also in difficult times and I am
very grateful for this. Very special thanks also go to Dr. Dölf Aemmer,
Dr. Norbert Felber and Christine Haller, for their great support and
open door policy.
I also want to express my gratitude to Prof. Peter Niederer, for
being the second co-examiner of my thesis.
During my five years at IT’IS/IIS my work was formed and influenced by many colleagues and external partners, to whom I would
like to express my deepest gratitude.
xiii
xiv
ACKNOWLEDGEMENTS
Very special thanks go to Dr. Jürgen Schuderer, with whom I
shared many thoughts on scientific and private topics. Special thanks
also to Dr. Andreas Christ, Neviana Nikoloski, Dr. Jürg Fröhlich and
Dr. Fin Bomholt for the introduction to the simulation and measurement tools and the various fruitful discussions in the world of Bioelectromagnetics. Thank you, Dr. Veronika Berdiñas, Dr. Walter Oesch
and Peter Sepan for our common work on the PERFORM projects,
and thanks to my former student Markus Tuor for your diligence in
the FOPH project. Thank you, Albert Lenherr, for taking care of the
mechanical setup challenges and for your stamina during intensive
times. The accomplishments are also your success.
I would like to address my special thanks to the IT’IS and SPEAG
team. All types of installations, assemblies and mechanical challenges
were solved through Rainer Mertens, Urs Frauenknecht, Dr. David
Perels, Nico Vetterli, Marc Salin, Roy Hunziger, Peter Fleischmann,
Thomas Conner, Bruno Reumer, Manuel Lindemann, and Dominik
Schmid. Also, I would like to thank Dr. Jean-Claude Gröbli for our
common time. Further thanks to Denis Spät, Anja Klingenböck,
Dr. Nik Chavannes, Dr. Hans-Ueli Gerber, Emilio Cherubini, Peter
Futter, Dr. Sven Kühn, Sang-Jin Eom, Valentin Keller, Dr. Marc
Wegmüller, Dr. Katja Poković, Dr. Axel Kramer, Dr. Michael Oberle
and many more, for the discussions and the fun during coffee breaks.
I would like to thank Prof. Theodoros Samaras for our great time at
Dr. J’s wedding and also Prof. Quirino Balzano - it was a pleasure to
discuss with you.
The scientific results were only possible to accomplish with the excellent administrative support from Jacqueline Pieper, Monica Lewis,
Sibyl Akrong and Martin Daellenbach – thank you. My special thanks
also go to Michelle von Issendorff-Stubbs for proofreading all the presentations, reports and this thesis. Thanks also to Christof Wicki,
Manuel Oettiker, Tobi Oettiker, Dr. Fritz Zaucker and Jonathan
Gubler for our common IT’IS computer support challenge.
It was a great pleasure working with so many international partners. I am grateful for the experiences and memories – thank you.
In PERFORM A I would like to especially thank: Prof. Clemens
Dasenbrock, Dr. Thomas Tillmann, Dr. Bernd Goerlitz, Dr. Georg
Neubauer, Dr. Paul Smith, Dr. Ullrich Decker, Dr. Santi Tofani,
Dr. Germano Oberto, Dr. Ya Da, Prof. Xu Zhenping, Dr. Michael
xv
Frauscher, Richard Überbacher and several others; in PERFORM C
Dr. Clairy Wiholm, Prof. Bengt Arnetz, Dr. Arne Lowden, Dr. Thomas
Persson, Lars Stahre, Dr. Lena Hillert and for the FOPH project
Dr. Mirjam Moser. Thanks also to the financial sponsors, the European Union and the Swiss Federal Office of Public Health. Their
funding enabled the scientific relevant, important research in the context of EMF health risk assessment.
I also want to give my gratitude to all my friends - thank you for
your support and friendship. My deepest thanks go to my family, to
Ute Ebert and Jill Ebert.
Part I
Introduction
1
Chapter 1
Background and
Motivation
1.1
Worldwide Mobile Communication
The rapid advances in wireless communication technology are having
a major impact on our lifestyles. It is expected that even more sophisticated technologies and services will become available to more people
in more places - creating new opportunities for personal freedom, mobility and economic development. The way we communicate changed
and it is difficult imagining modern life without mobile communication.
Mobile communication continues to increase worldwide
(Figure 1.1). In several European countries the number of mobile telephones already exceeds the number of inhabitants and has
surpassed even the number of landline connections. Also in developing and thinly populated countries or rural areas, where building
landline communication was not cost effective, mobile communication
became the most important communication form. According to the
GSM Association1 and 3GAmericas2 close to four billion people
worldwide use mobile phones at the end of 2008. Today most of
1 GSM
2 3G
Association (www.gsmworld.com)
Americas (www.3gamericas.org)
3
4
CHAPTER 1. BACKGROUND AND MOTIVATION
Figure 1.1: Left: Number of mobile subscribers per 100% population
and the mobile growth rate between 2006 and 2007. Right: Development of worldwide mobile subscribers and the mobile penetration
level. Source: WCIS.
them use the GSM system (GSM: 81%, UMTS: 8%, CDMA: 10%,
AMP/NMT/PDC/etc: 1%, source: WCIS3 Nov 2008).
The usage of UMTS is progressing fast, which enables new forms of
communication due to its flexible data transfer technology. Additional
new communication forms will develop, e.g., on- and in-body communication devices, hearing aids which communicate wirelessly between
both ears to give depth of perception to the hearing-impaired, devices
for medical supervision which automatically send data to emergency
doctors and implants which measure and document medical parameters. Communication is therefore not only verbal communication but
also increasingly data exchange.
Accompanying the rapid increase of mobile communication services and devices since the 1990’s, radiofrequency electromagnetic
fields (RF-EMF) represent one of the most common and fastest growing environmental influences in our lives. The amount of people exposed to RF-EMF has increased to numbers that make even small
health effects become relevant. Various health effects, e.g., promotion
of cancer, have been suggested as resulting from exposure to EMF.
The potential health effects from exposure to static and time varying electric and magnetic fields have caused significant public and
3 World
Cellular Information Service (www.wcisdata.com)
1.2. RELEVANT ELECTROMAGNETIC SPECTRUM
5
occupational health concerns and require scientific clarification. It is
therefore a major task for scientist and standard setting bodies to examine and define conditions for a safe handling and usage of modern
communication equipment: the potential health risks must be assessed
and analyzed.
Health risks due to exposure to electromagnetic fields have been
studied for several decades. With the rapid growth and introduction
of the mobile communication technology to the general public in the
early 1990’s, the research received a new focus and was intensified,
especially since the technological distribution and progress were faster
than the health risk research could provide answers about possible
adverse health effects [1][2].
1.2
Relevant Electromagnetic Spectrum
All over the world people are exposed to electromagnetic fields to
varying degrees, and the levels of exposure might increase as technology advances further. The spectrum of electromagnetic radiation
comprises a range of frequencies from very-low-frequency energy (such
as electrical power) through visible light, to extremely high frequency
radiation (such as X-rays and gamma rays).
Figure 1.2 shows the electromagnetic spectrum including ionizing
and non-ionizing radiation as marked in the lower part of the figure.
Ionizing radiation has energy levels high enough to strip electrons from
atoms and molecules. Exposure to ionizing radiation can cause serious
biological damage, including the production of cancer. Non-ionizing
radiation lacks the energy needed to cause ionization but can produce
other types of biological effects, e.g., the exposure to high levels of
RF energy can rapidly heat biological tissue (see also Chapter 2).
This thermal effect can cause harm by increasing body temperature,
disrupting behavior, and damaging biological tissue.
Interesting for today’s research and risk assessment is the exposure
to RF energy levels below the occurrence of thermal effects in the
non-ioning frequency range. Mobile phones are designed to operate
at frequency levels in the GSM/DCS band (around 800, 900, 1800,
1900 MHz) and at the UMTS band (2100 MHz) with power levels well
below the threshold for known thermal effects. Consequently, the
6
CHAPTER 1. BACKGROUND AND MOTIVATION
Figure 1.2: In the electromagnetic spectrum, the frequencies of mobile
phones fall in between those of TV transmitters and microwave ovens
(see also [3]).
mobile phone health issue has generally focused on whether there are
any adverse effects from long-term or frequent exposure to low-power
RF signals that do not cause biologically significant heating (nonthermal effects). There is particular interest whether using mobile
phones could promote brain cancer, since a user’s head absorbs parts
of the RF energy when the phone is held to the ear during a call.
1.3
EMF Health Risk Assessment
The classic risk assessment paradigm includes hazard identification,
establishing a dose-response, exposure metrics definition, and risk
characterization. Limitations of risk assessment relate to uncertainties which cover lack of relevant exposure and biological data, lack
of knowledge about mechanisms, imprecisions in quantitative risk assessment models/methods and statistical variability/inter-individual
1.3. EMF HEALTH RISK ASSESSMENT
7
variability. Sound scientific assessment methodology and uncertainty
analysis are key prerequisites for risk assessment studies [4] [5].
The EMF health risk assessment distinguishes between concepts
of interaction, biological effects, perception, and health effects [6].
Interaction takes place because Coulomb forces act on charges and
molecules of living tissue. A biological effect is a detectable change in
function or structure above the molecular level. EMF exposures which
lead to measurable biological effects within the normal range of physiological compensation of living systems are not considered hazardous.
EMF exposures which lead to biological effects exceeding the normal
compensatory range of living tissue might be an actual or potential
health hazard. EMF leading to biological effects which influence an
individual’s ability to function properly without the possibility of recovery, must be considered health hazards, i.e., thermal effects that
cause tissue damage.
Thermal effects due to RF exposures are generally considered
to be well understood. The public is concerned about possible adverse biological effects far below the threshold values of established
effects. These effects, often categorized athermal and non-thermal,
might greatly depend on signal characteristics, especially on the ELF
components [7] [8]. Besides exposure dose, special care must also be
given to the exposure signal.
In order to maximize the outcome of risk assessment studies, especially as long as no interaction mechanism is established, the objective
should be to provide the maximized significance of negative findings.
If positive effects are found, a subsequent analysis of all parameters
shall be used to assess the actual health risk. The exposure in risk
studies should represent the worst case with respect to exposed tissues, exposure strength and signal characteristics [9] [10].
A major shortcoming of many previous studies regarding possible
health effects from electromagnetic fields is an inadequate definition
and description of the exposure setups as well as in poorly characterized dosimetry [11] [12]. In order to compare the results of investigations carried out in different laboratories and to ensure the quality
of the results obtained, it is very important that the conditions of
exposure are well defined and exactly controlled. Requirements for
exposure setup design can be found in [9] [13] [10], and dosimetry
should be performed according to [14] [15].
8
CHAPTER 1. BACKGROUND AND MOTIVATION
These above-mentioned scientific quality conditions should be followed for exposure setup and study design. EMF risk assessment
studies need - in addition to the right biological method and model
- exposure systems with a sound dosimetric assessment, uncertainty
and variability analysis. Furthermore the exposure systems should
provide well-defined, well-controlled, well-characterized, and artifactfree exposures.
Many studies have been performed and contribute to the health
risk assessment through the exposure to EMF. Some reviews on EMF
and human health can be found in [16] [17] [18] [19] [20]; however,
further research has been deemed necessary [21] [22] [23].
Various repositories of publications and reports on the biological effects of EMF are available, e.g., at the WHO International
EMF Research Database4 , the EMF-Portal5 of Femu, RWTH, Germany and database ELMAR6 of the University Basel and BUWAL,
Switzerland. There are also various EMF publication services providing up to date information, i.e., EMF Database Information Ventures Inc, USA, RF Gateway Resource Strategies Inc, USA, and
RFsciencefaqs Helpdesk, UK.
1.4
Laboratory Studies
Laboratory studies are used to assess EMF health risks. They are
performed under controlled conditions with respect to environmental
and exposure parameters and enable - with a blinded study design a direct sham/exposure comparison. Laboratory studies are classified
as follows:
• Human experimental studies examine the effects of low dose
levels of a potential toxic agent on humans. Though these studies are of the highest relevance for risk assessment of the agent,
the dose levels must be low enough to ensure safe exposures; in
addition the choice of end-points is limited (e.g., no cancer or
threshold studies) and usually also the number of tested persons is comparably small. However, due to the high relevance,
4 http://www.who.int/peh-emf/research/database/en/
5 http://www.emf-portal.de
6 http://www.elmar.unibas.ch
1.4. LABORATORY STUDIES
9
these studies are important in the context of human health risk
assessment.
• In vivo studies examine the interaction between an agent and
complete, living biological systems, i.e., laboratory animals. In
general the experimental animals must be an appropriate surrogate for human response to the tested agent. In RF exposure studies mice and rats of various breeds are commonly used.
In vivo studies are performed to conduct dose range-finding
studies for the assessment of thresholds for radiation exposure
limits. Small effects can be detected through large-scale studies. However, before the study results are interpreted as human
health risks (especially when knowledge of the dose-response relationships for the toxic effects is of concern), they need to be
replicated and confirmed with more than a single animal model,
since one animal species might be extra sensitive or insensitive
to a particular end-point. Study results are relevant for human
health risk assessment and are the standard method.
• In vitro studies examine isolated biological components, e.g.,
cell cultures, DNA and macromolecules. They are preferred
for basic research, are fast to conduct and usually cost effective.
They enable possible interaction mechanism studies which might
be masked in the complexity of the whole-animal level. In vitro
studies also allow high exposure levels which would not be possible using in vivo models. In vitro studies provide end-points and
exposure conditions for further in vivo examinations. Among
their drawbacks is that their results are not directly transferable
to in vivo models and that end-points always require subsequent
in vivo tests to include the regulating mechanisms of biological
systems. Study results therefore have only indirect relevance for
human health risk assessment.
Epidemiological studies are a further variant, though not a laboratory study. They present an observational technique for examining
patterns of illness in human populations and statistical associations
between an illness and a suspected causative agent. Drawbacks include the limitedly defined cause and dose of the examined agent.
The primary objective is therefore a statistical significance, but they
10
CHAPTER 1. BACKGROUND AND MOTIVATION
cannot be used for dose, threshold or mechanism finding studies. Most
epidemiological studies have used some form of cancer as an end-point.
Epidemiological studies are of high relevance for human health risk assessment but have a very high uncertainty of dose levels and must be
valued within their limitations.
1.5
Specific Absorption Rate
Radiofrequency EMF penetrates deeply into tissue and can cause an
increase in temperature by increasing the kinetic energy of random
molecular motion. However, whole-body or local temperature increases depend not only upon the amount of energy absorbed, but
also on passive heat dissipation and on the underlying complex thermoregulatory processies in the body.
Therefore the initial temperature increase after the onset of an
exposure cannot be used to define the local energy dose. The initial
rate of temperature increase or the power delivered to a local volume
are independent of the thermoregulatory process of a living system
and can be used to measure the exposure. This dosing parameter,
a dose rate, is called the specific absorption rate (SAR) [24] and is
the key to determining the levels at which a biological response might
occur.
dP
σ
dT
= E2 = c
(1.1)
SAR =
dm
ρ
dt
The specific absorption rate is defined by Equation 1.1 as the incremental electromagnetic power dP absorbed by an incremental mass
dm. SAR is also directly proportional to the conductivity σ of the
tissue at the location of interest with the mass density ρ and the magnitude of the squared electric field strength E 2 . Furthermore SAR is
proportional to the initial rate of temperature rise dT
dt in an exposed
tissue with the specific heat capacity c.
1.6
RF Safety Standard
The recommended RF exposure limits in the frequency range 10 MHz
to 300 GHz are provided by the International Commission on Non-
1.6. RF SAFETY STANDARD
11
Ionizing Radiation Protection (ICNIRP)[25]. National legislation regulations refer to the ICNIRP guideline (e.g., Switzerland). The guidelines recommend basic restrictions on SAR to ensure that harmful
temperature increases do not occur in the human body. Tables 1.1
and 1.2 show the reference levels. Further information can also be
found in [26].
The ICNIRP guidelines have different limits for the exposure of
the general public and occupational workers (exposure limits for the
general public are usually reduced by a factor of up to five). Shortterm time-averaging transient exposures may be averaged over specified periods before comparison with the restrictions. The restrictions
on localized SAR vary over different regions of the body and apply
to continuous tissues within a specified region. At higher frequencies,
energy absorption becomes increasingly confined to the surface layers
of the skin and a heating effect is directly related to the power density
of the incident radiation. The ICNIRP guidelines take into account
the different efficiency of the coupling of the electromagnetic fields
into the human body; so the reference levels vary according to the
frequency of the incident RF.
Exposure quantity
Occupational
General public
SAR averaged over the body
and over any 6 min period
0.4 W/kg
0.08 W/kg
SAR averaged over any 10g
in the head and trunk and over
any 6 min period
10 W/kg
2 W/kg
SAR averaged over any 10g
in the limbs and over any
6 min period
20 W/kg
4 W/kg
Table 1.1: ICNIRP basic restrictions on the exposure to electric and
magnetic fields in the frequency range 10 MHz to 10 GHz for occupational and general public exposure (source: [25]).
12
CHAPTER 1. BACKGROUND AND MOTIVATION
Exposure
group
Frequency
range
Electric
field
strength
(V/m)
Magnetic
field
strength
(A/m)
Power
density
(W/m2)
Occupational
10-400 MHz
400-2000 MHz
2-300 GHz
61
√
3 f
137
0.16 √
0.008 f
0.36
10
f/40
50
General
public
10-400 MHz
400-2000 MHz
2-300 GHz
28 √
1.375 f
61
0.073 √
0.0037 f
0.16
2
f/200
10
Table 1.2: ICNIRP reference levels for occupational and general public
exposure to electromagnetic fields in the frequency range 10 MHz to
300 GHz (f : frequency, source: [25]).
1.7
Motivation
To foster significant progress in the health risk assessment of low-level
radiofrequency electromagnetic exposure from mobile phones, the 5th
Framework Program of the European Union (EU) started a large international toxicological/carcinogenic project named PERFORM at
the start of the dissertation. The PERFORM project was coordinated by the University of Helsinki, which functioned as a firewall
ensuring that the financial sponsors had no influence on the study results. In addition to the funding through the 5th Framework Program
of the EU, sponsorship came also from the Mobile Manufacturers Forum, the GSM Association, Elettra2000 and the national authorities
of Switzerland and Austria.
The agenda included long-term in vivo bioassay studies
(PERFORM A), animal behavioral studies and in vitro cell culture
studies (PERFORM B) as well as an experimental provocation study
(PERFORM C). Within PERFORM A four laboratories in Europe
conducted six in vivo studies investigating whether mobile phone
signals at GSM-900 MHz or DCS-1800 MHz are carcinogenic or cocarcinogenic in rats and mice. Not within PERFORM A but with the
1.8. OBJECTIVES AND CHAPTER OVERVIEW
13
same exposure setup, a seventh in vivo cancer study with rats was conducted in China. The PERFORM B program included several in vivo
and in vitro replication studies. The PERFORM C human study investigated effects of GSM-900 MHz wireless communication signals on
subjective symptoms, physiological reactions, alertness, performance
and sleep. In the beginning of 2004 the PERFORM C project was
renamed to Mobile Phone and Direct Health Effects; however, in this
dissertation it is still referred to as PERFORM C.
The biological studies were conducted between 2001 and 2006.
The results combined with current and past studies on the effects
of RF radiation, have been included in the database for health risk
evaluation by the World Health Organization (WHO) [27] [28].
The Foundation for Research on Information Technologies in
Society (IT’IS) was the technical partner in the PERFORM project
and was responsible for providing suitable exposure setups, including
their dosimetry. The challenging tasks were the development, characterization, construction, testing and installation of the exposure systems as well as their maintenance during the studies, with special
focus on the provision of well-defined, well-controlled and artifact-free
exposures.
1.8
Objectives and Chapter Overview
The objectives of this dissertation were the development, characterization and successful operation of optimized exposure systems for the
PERFORM A mouse studies and the PERFORM C human study. For
the successful accomplishment of these large international studies, various tasks at scientific, engineering and also organizational levels were
conducted; significant, innovative contributions were introduced to
achieve all major requirements of such systems, to realize well-defined,
well-controlled, well-characterized, and artifact-free exposures.
The dissertation is organized as follows: chapters 2, 3 and 4 present
exposure assessment studies for key design parameters (exposure dose,
exposure signal); chapters 5, 6 and 7 present the exposure systems developed for the PERFORM A and PERFORM C studies and a comparison study of exposure systems based on the Ferris wheel concept.
The detailed content of the chapters is as follows:
14
CHAPTER 1. BACKGROUND AND MOTIVATION
• Chapter 2:
Thresholds
Thermal Regulatory and Thermal Breakdown
To optimize the relevance of in vivo health risk exposure studies, the exposure dose should be as high as possible but below the threshold of known adverse biological effects. This
study determined the thermal regulatory and thermal breakdown thresholds for tube-restrained B6C3F1 and NMRI mice
exposed to radiofrequency electromagnetic fields. For the first
time in vivo temperature measurements in a well-defined exposure setup enabled the precise determination of these biological
thermal thresholds which were relevant for setting the exposure
level of the high dose group in the PERFORM A study.
• Chapter 3: Assessment of ELF MF of Mobile Phones and
proposal for worst-case signal for Bio-Experiments
Little was known about the extremely low frequency magnetic
fields (ELF MF) of mobile phones caused through the relatively
high amplifier currents which generate the RF signal. To close
this knowledge gap, the B-field strengths and waveforms of five
commercial phones were accurately assessed with an extrapolation method. As a study result, the worst-case test signal for
use in combined ELF/RF exposure studies is proposed.
• Chapter 4: Add-On ELF MF Exposure Setup Proposal for
Studies with Human Volunteers
This chapter proposes a setup for human exposure studies with
the worst-case extremely low frequency magnetic field signal defined in Chapter 3. This exposure setup was proposed as an option for the PERFORM C study. However, the steering committee decided not to include ELF MF signals in the study design.
• Chapter 5: Flexible Exposure Setup for Experimental Provocation Studies at 884 MHz
Study designs which envision the exposure of subjects for a duration of three hours or even longer require exposure systems
which enable both, head movements and a constant exposure.
To realize such an exposure system, a novel, low-weight antenna
mounted on a flexible and comfortable headset is presented. The
1.8. OBJECTIVES AND CHAPTER OVERVIEW
15
setup also incorporated a heated plate at the ear lobe of subjects on the exposed side to simulate the thermal load from, e.g.,
contact pressure or the battery heat of a mobile phone during
usage. The dosimetry and uncertainty of the setup were assessed
using simulations and measurements.
With this exposure system the Department of Health Science
at the Karolinska Institute in Sweden examined in the human
study PERFORM C the effects of RF fields on self-reported
symptoms, as well as the detection of fields after a three-hour
exposure time.
• Chapter 6:
Exposure Setups for Large-Scale In Vivo
GSM/DCS Risk Assessment Studies
For the mouse studies of PERFORM A, two types of efficient
exposure setups for 900 and 1800 MHz signals were developed
which are presented in this chapter. The systems allow doubleblind study protocols and are fully self-controlled (monitoring
exposure and further parameters as oxygen, temperature and
humidity) with an auto-detection of malfunctions. The applied
complex exposure signal covered in different exposure phases
the extremely low frequency spectral components resulting from
the DTX/non-DTX, power saving and environmental control
signals. The experimental and numerical dosimetry was conducted with details that have not been provided in any previous
study. The whole-body averaged and organ averaged SAR for
the different exposure phases were assessed, including their uncertainties and variations occurring in weight, anatomy, dielectric parameters, posture and positioning. Exposure verification
was enabled through a new, experimental, dosimetric temperature method which verifies the exposure at each mouse position
in the exposure resonator.
With these setups two combined chronic toxicity and carcinogenicity studies were performed at Fraunhofer Institute of Toxicology and Experimental Medicine (ITEM) in Germany and one
co-carcinogenicity study was performed at Istituto di Ricerche
Biomediche Antoine Marxer (RBM) in Italy.
16
CHAPTER 1. BACKGROUND AND MOTIVATION
• Chapter 7: SAR Uniformity in Ferris Wheel Setups
This study compares three different mouse exposure systems
based on the Ferris wheel concept (the setup used in the widely
discussed Adelaide study that has shown a two-fold increase in
lymphoma cancer in transgenic mice and the newly developed
setups for the exposure of mice in PERFORM A and B). Several
concepts to improve the uniformity and the inter-animal variations of exposure have been identified as well as their effects on
the instant, weekly and life-averaged dosimetry. These improvements limited the variations of the PERFORM A setup to less
than 2.5 dB. By applying a rotational scheme, variations over
the life-time study remained even below 1.2 dB.
Part II
Exposure Assessment
17
Chapter 2
Thermal Regulatory
and Thermal
Breakdown Thresholds
2.1
Abstract
The objective of this study was the determination of the thermal regulatory and the thermal breakdown thresholds for in tubes restrained
B6C3F1 and NMRI mice exposed to radiofrequency electromagnetic
fields at 905 MHz.
Different levels of the whole-body averaged specific absorption rate
(SAR = 0, 2, 5, 7.2, 10, 12.6 and 20 W/kg) have been applied to the
mice inside the “Ferris wheel” exposure setup at 22± 2 ◦ C and 30-70%
humidity. The thermal responses were assessed by measurement of
the rectal temperature prior, during and after the two hour exposure
session. For B6C3F1 mice, the thermal response was examined for
three different weight groups (20 g, 24 g, 29 g), both genders and for
pregnant mice. Additionally, NMRI mice with a weight of 36 g were
investigated for an interstrain comparison.
The thermal regulatory threshold of in tubes restrained mice was
found at SAR levels between 2 W/kg and 5 W/kg, whereas the break19
20
CHAPTER 2. THERMAL RESPONSE THRESHOLDS
down of regulation was determined at 10.1 ± 4.0 W/kg (K=2) for
B6C3F1 mice and 7.7 ± 1.6 W/kg (K=2) for NMRI mice. Based on
a simplified power balance equation, the thresholds show a clear dependence upon the metabolic rate and weight. NMRI mice were more
sensitive to thermal stress and respond at lower SAR values with regulation and breakdown. The presented data suggest that the thermal
breakdown for in tubes restrained mice, whole-body exposed to radio frequency fields, may occur at SAR levels of 6 W/kg (K=2) at
laboratory conditions.
2.2
Introduction
Risk assessment studies examining the potential hazards of radiofrequency (RF) electromagnetic fields (EMF) focus on the question of
whether the postulated EMF interactions below the thermal threshold
may pose a health risk. These studies are faced with the methodological challenges that the thermal effects will occur shortly above local
exposure levels, as they occur during everyday usage of e.g., a mobile phone, so that the sensitivity towards EMF cannot be enhanced
by applying a substantially increased dose. To maximize potentially
risk relevant effects, studies are preferably performed just below the
thermal threshold levels. This poses high requirements on the experimental exposure conditions [9]; the exposure setup must deliver a well
defined dose with minimal variations in the applied EMF dose, and
this also requires that the threshold levels must be carefully assessed.
The following power balance equation, in analogy to the heat balance equation formulated by Adair et al. [29], is useful to discuss
thermal response:
M + PRF − C = S
(2.1)
Hereby M is the rate at which thermal energy is produced through
metabolic processes, PRF is the absorbed RF power, C is the cooling
rate determined by the heat exchange via convection, radiation and
evaporation and S is the rate of heat storage in the body. All units
are expressed in Watts.
When mice are exposed to a heat load, e.g., from an increase in
RF exposure, their initial main response is to linearly lower their
2.2. INTRODUCTION
21
metabolic rate in favor of stabilizing their body temperature [30] [31].
When they reach their basal metabolic rate, other cooling mechanisms
become dominant, such as an increase in evaporation [32], increase of
skin temperature through vasolidation [33] and behavioral changes
[34], such as spreading saliva or urine on their fur, or, if possible,
choosing a new environment [35].
Three regions of thermal response to RF are distinguished in this
paper: (1) a non-thermal-regulatory region, where no measurable temperature response occurs, (2) a thermal regulatory region, where the
regulatory system is able to compensate the absorbed energy and
regulates to achieve a stable body temperature and (3) a thermal
breakdown region, where the organism is not able to compensate the
temperature increase. The thermal response regions are separated
by two thresholds: the thermal regulatory threshold and the thermal breakdown threshold. Strictly defined, a thermal response may
also occur without leading to measurable body temperature changes,
e.g., as soon as a single thermal receptor detects a subtle temperature
change and leads to cellular responses. However, such effects cannot
be investigated by simple body temperature measurement and are
therefore not the subject of this study.
Most of the published research on thermophysiological responses
due to whole-body exposure to RF fields has been conducted with
rodents (e.g., mice, rats and hamsters) [36]. However, no clear separation between thermal regulation and breakdown was targeted. A
measurable thermal response in rodents was reported for SAR levels
at various levels between 2 W/kg up to 40 W/kg [37], [38], [39], [36],
[37], [29]. High discrepancies are reported even for the same strain of
mice.
A major shortcoming of many experiments undertaken is the
lack of well defined exposure conditions, missing online temperature
recording and poorly performed dosimetry [40], which might be an
explanation for the controversial results.
The objective of this study was to assess the thermal regulatory and breakdown threshold for multiple mouse models exposed
to electromagnetic fields at 905 MHz inside the setup configuration
“Ferris wheel” which has been applied in several long- and short term
studies [41] [42].
22
2.3
2.3.1
CHAPTER 2. THERMAL RESPONSE THRESHOLDS
Material and Methods
Animals
The thermal responses of the mice were examined for different body
weight groups (20g, 24g, 29g, 36g), both genders (f, m), pregnant
mice and two different mouse strains (B6C3F1 and NMRI). Charles
River Laboratories (Sulzbach, Germany) supplied the B6C3F1 male
and female mice at the ages of 4 and 12 weeks and the male NMRI
mice at an age of 12 weeks. Four groups of B6C3F1 mice and one
group of NMRI mice were examined. Each group consisted of at least
12 animals: eight animals used for exposure and four as sham exposed
animals. Sham exposed animals were treated identically as exposed
animals, but were not exposed to RF. The animal groups had the
following parameters: Group A (female B6C3F1, 20.4±0.9 g, 4 weeks
old), Group B (female B6C3F1, 23.6±0.8 g, 12 weeks old), Group C
(male B6C3F1, 28.8±0.9 g, 12 weeks old), Group D (pregnant female
B6C3F1, 31±2.4 g to 36±3.7 g, 12 weeks old) and Group E (male
NMRI, 36.4±3.0 g, 12 weeks old) (see also Table 2.4).
The animals were kept in standard cages (Makrolon, type II
(350 cm2 )) in pairs or individually (only the 12 week old males); in the
cages absorbent softwood was used as bedding material. The body
weight of the mice was monitored every morning. The laboratory
maintained an environmental temperature of 22±2 ◦ C and a relative
humidity of 30-70 %. An automatic timing device provided an 12-hour
light/dark cycle.
2.3.2
Exposure Protocol
Prior to the experimental period all mice were acclimatized and
trained to the restrainer tube housing required for the RF exposure
(Figure 2.1). The training also included a dummy temperature probe.
The training phase lasted for about three weeks, so that the animals
had an age of 7 or 15 weeks at the start of the study. The animals
were trained by increasing the amount of time spent in the tube every day for a duration of up to four hours. The experiment began
after the training phase. Although trained identically, B6C3F1 and
NMRI mice behaved differently to the tube restraint during exposure:
2.3. MATERIAL AND METHODS
23
B6C3F1 mice behaved very calmly, even partly falling asleep during
the exposure. In contrast, the NMRI mice were more active and never
reached the calmness of the B6C3F1 mice. The NMRI mice pawed
during the entire exposure time and in rare occasions even started to
hyperventilate. Due to their higher activity, a higher metabolic rate
is present compared to the B6C3F1 mice.
RF exposure was performed according to a standardized exposure protocol: (1) preparation of mice with temperature probes,
(2) fixation of mice in the restrainer tubes and in the exposure
setup, (3) 10-20 minutes recording of the temperature base line until a stable body temperature (∆T<0.1 ◦ C) was reached for all mice,
(4) application of the RF exposure for a duration of 120 minutes,
(5) 15-30 minutes recording of body temperature development after
the RF exposure and (6) replacement of mice to the cages.
During the experiments an animal care taker was always present
and a veterinarian close by. The exposure level and body temperature
of the mice was monitored online, so that the safety of the animals was
ensured at all times. If during the exposure the body temperature of
an animal increased to more than 41 ◦ C, the animal was replaced for
safety and ethical reasons. Additionally, the exposure was interrupted
if two or more animals reached 41 ◦ C and the body temperature of the
majority of the mice increased linearly to more than 40 ◦ C.
The mouse groups were exposed to arbitrarily chosen increasing
SAR levels. Groups A, C and E were exposed to sham, 2, 5, 7.2, 10,
12.6, 20 W/kg levels, group B to sham, 5, 7.2, 10, 12.6, 20 W/kg and
group D to sham, 5, 10, 12.6, 20 W/kg (see also Table 2.4). To document and monitor any environmental influences during the exposures,
four additional animals of the same breed, gender and weight group
were always sham exposed in a second setup.
It should be noted from the literature that an average SAR of
9.2 W/kg for a 25 g mouse corresponds to approximately the same
power as the metabolic rate for a non-stressed resting female mus
musculus mouse [43]. Therefore it is expected that the applied RF
exposures with SAR values between 2 and 20 W/kg induce a heat load
between 0.22 and 2.2 times the metabolic rate.
One exposure per mouse group per day was undertaken. Each
mouse group was exposed in a series to increasing SAR levels on consecutive days at the same time slot of each day. After the series was
24
CHAPTER 2. THERMAL RESPONSE THRESHOLDS
completed a few exposures were repeated at different times of the
day to exclude any influence of the exposure time, e.g., due to the
circadian temperature rhythm of the mice.
2.3.3
Temperature Measurement
The body temperature of the mice was continuously recorded, starting
at the placement of the temperature probes until the mice were taken
out of the exposure setup. The temperature was recorded by rectal measurements. Eight RF-immune thermistor probes (T1V3LA,
SPEAG, Switzerland) with a noise level of 0.005 ◦ C and an absolute accuracy of 0.1 ◦ C were continuously read in 1 s intervals by two
recording systems (EASY4, SPEAG, Switzerland). The temperatures
of the four sham exposed mice in the second setup were measured
with a 4-channel temperature device (FoTemp4, OPTOCON, Germany), based on an optical measurement process (accuracy: approx.
1 ◦ C, sample rate: 8 s).
The tip diameter of the two temperature measurement systems
is 1 mm (EASY4) and 1.3 mm (FoTemp4) and allowed rectal insertion at a depth of 15 mm, as required for core body temperature measurements [44]. Insertion accuracy during the exposure
was ensured by taping the probes to the tails of the animals.
During rectal measurement, the stability of the temperature was
increased considerably due to mice falling asleep and/or calming
down, and decreased due to mouse movement, body gases and natural body temperature drifts. For an evaluation period of 10 min.,
the noise level for the sham exposed mice varied between 0.09 ◦ C
and 0.17 ◦ C (group A: 0.12 ◦ C, group B: 0.09 ◦ C, group C: 0.11 ◦ C,
group D: 0.17 ◦ C, group E: 0.13 ◦ C).
2.3.4
Evaluation of Thermal Response
Three characteristic thermal responses of whole-body exposed mice
can be distinguished (as shown in Figure 2.3). (A) The body temperature does not change upon RF exposure, within the measurement
noise limits. (B) The regulatory system of the mouse compensates
the RF induced heat load; the characteristic response was not necessarily an increase in body temperature during exposure but a decrease
2.3. MATERIAL AND METHODS
25
in body temperature after the exposure. (C) The mouse was not able
to compensate the body temperature increase due to the RF; body
temperature increased linearly.
Thermal regulation and breakdown were determined in the following way:
• Regulation: Whenever the average body temperature at about
10 min after the RF exposure (evaluation period 5 min) is significantly lower (= three times the noise level) in comparison to
the body temperature in the last 10 min of exposure, the mouse
was regarded as being in a thermal regulatory state. The thermal regulatory threshold was set between the lowest examined
SAR at which no animal showed any regulation and the lowest
SAR for which at least one mouse responded.
• Breakdown: Whenever a mouse showed a linear body temperature increase and reached temperature levels about 41◦ C, it was
regarded as being in the thermal breakdown region. The thermal breakdown threshold was then determined by measuring
the slope of the temperature increase over time dT /dt dependent upon the applied SAR value. Extrapolating the resulting
linear regression curve to dT /dt = 0 reveals an SAR value which
was used as the breakdown threshold.
2.3.5
Radiofrequency Exposure
The mice were exposed inside a Mini Wheel Setup, which is based on
the Ferris wheel concept [45]. The Mini Wheel Setup has a monomode cavity and therefore well defined exposure conditions (cavity
diameter 332 mm, plate separation 120 mm). The setup enables the
exposure of up to eight mice simultaneously and is designed as a resonant waveguide structure of two parallel, circularly-shaped metallic
plates shorted with metallic bars along the outer edge and with an
isotropic antenna at the center. The mice were circularly positioned
in restrainer tubes at a fixed radial distance of 111.5 mm from the
antenna with the animal body axis orientated parallel to the E-field.
The system was matched for resonance (return loss ≤ -18 dB) using a
triple stub tuner (Type 1878B, Maury Microwave, USA). An air ventilation system forced a constant airflow of about 1 l/min (measured in
26
CHAPTER 2. THERMAL RESPONSE THRESHOLDS
Figure 2.1: Mini Wheel Setup with temperature probes enabling the
exposure of up to eight mice simultaneously with a 905 MHz signal.
SPEAG thermistor probes enabled online high resolution body temperature measurements during the RF exposure.
empty tubes) through the tubes. This ensured sufficient fresh air for
the mice while restrained in the one-size tubes and lead to about the
same airflow through the tubes for all mice, regardless of their size.
RF signals were generated by a signal generator (SMIQ 02B, Rohde &
Schwarz, Germany) and an amplifier (model 2449R2, LS Elektronik,
Sweden). For exposure control the forward and reflected powers as
well as the electric field inside the cavity were measured. Power measurements were performed with a dual directional coupler (Type 778D,
Hewlett Packard, USA) and a dual head power meter (E4419B EPM,
Agilent, USA). The incident E-field measurement inside the wheel was
performed by two monopole sensors (length = 10.5 mm; at a radius of
20 mm) connected to diode detectors (ACSP2663NZC15, Advanced
Control Components, USA).
A detailed numerical and experimental dosimetry and a comprehensive uncertainty assessment of the Mini Wheel Setup was performed together with data from S.-J. Eom [46]. The uncertainty budget was assessed according to [47], taking all the occurring variabili-
2.3. MATERIAL AND METHODS
27
ties and variations into account. The numerical dosimetry is based on
FDTD simulations (using SEMCAD 1.8, SPEAG, Switzerland) with
high resolution mouse models for different animal sizes (scaled models
corresponding to mouse weights of 20 g, 25 g, 30 g, 35 g) and postures
(restrained and non-restrained mouse in tube). The mouse models
were derived from Microtome slices of 1 mm resolution. The numerical examination included simulations of the full exposure setup filled
with eight mice as well as simulations in a one and three mouse sector model. The experimental dosimetry used mouse dummies (conical
bottles filled with tissue simulating liquid) to assess the dependence
of the exposure on the eight positions within the wheel and to verify
the numerical FDTD model by precise E-field using a near-field scanner (DASY4, SPEAG, Switzerland) and temperature measurements.
Examinations were performed on the influence of various parameters
on SAR as listed in Table 2.2 and Table 2.3. In addition to the wholebody SAR assessment, detailed organ/tissue specific SAR values were
determined as shown in Table 2.1.
An uncertainty assessment for the whole-body and organ averaged
SAR was performed according to [48]. The uncertainty budget includes the absolute uncertainties of the SAR assessment and the variabilities of SAR between different mice in the setup and due to mice
movement. The uncertainty of the SAR assessment of the whole-body
averaged SAR was estimated to be ± 1.2 dB (± 33%, K=1) (see also
Table 2.2). Variations in SAR during the experiment including movements, position within the wheel, posture, etc. were estimated to be
± 1.1 dB (± 27%, K=1) for the whole-body averaged SAR (Table 2.3).
With respect to thermosensitivity and thermoregulartory responses, organs with thermal receptors are of interest, such as the
hypothalamus in the brain and the skin [25]. The averaged tissue
specific SAR of the brain is -1.1 dB (-23%) less in relation to the
whole-body averaged SAR; the exposure of the skin in relation to the
whole-body averaged SAR was about 2.6 dB (80%) higher (Table 2.1).
However, a higher local exposure in SAR does not necessarily lead to
a higher local temperature; blood flow etc. can distribute the heat
rapidly. Figure 2.2 shows the SAR distribution in a center sagittal
section of the mouse model in a tube.
28
CHAPTER 2. THERMAL RESPONSE THRESHOLDS
Organ/Tissue
Type
Whole-Body
Avg. Brain
Avg. Muscle
Bladder
Blood
Bone (Cortical)
Bone Marrow (Infiltrated)
Cartilage
Cerebrospinal Fluid
Eye Tissues (Sclera)
Fat (Averaged Infiltrated)
Heart
Kidney
Liver
Lung (Deflated)
Nerve
Skin (Wet)
Small Intestine
Spleen
Stomach
Testis
Tongue
Trachea
Vitreous Humor
Organ/Tissue
Deviation1
[dB]
Uniformity2
Isotropy3
[dB]
[dB]
0.00
-1.14
0.88
0.97
1.90
-2.46
-7.21
-1.46
2.80
0.54
-5.16
-0.59
-1.76
-1.43
-1.08
-1.04
2.56
1.99
-4.59
-2.11
1.78
0.57
1.35
1.72
5.29
-4.88
5.23
-4.35
-3.12
4.79
-2.04
-0.96
0.29
-4.65
4.80
-4.23
-3.33
-2.90
-4.81
-2.15
7.59
-0.56
-3.85
-3.10
-0.11
-5.93
-5.38
-4.72
0.19
0.40
0.20
0.81
0.26
0.30
0.34
1.23
1.12
0.62
0.48
0.77
1.00
0.21
0.83
0.26
0.67
0.53
1.76
1.11
0.84
0.28
0.46
0.25
1
organ/tissue deviation: ratio in dB of organ/tissue averaged SAR to whole-body
2
uniformity: ratio in dB of the standard deviation of the organ/tissue averaged
SAR to the organ/tissue averaged SAR
isotropy: standard deviation in dB of the organ/tissue averaged SAR for the
different positions in the wheel
averaged SAR
3
Table 2.1: Simulated SAR for specific organs/tissues (source: [46])
2.3. MATERIAL AND METHODS
29
Figure 2.2: SAR distribution in the center sagittal cut through a 25 g
high-resolution mouse model (0 dB equals 0.07 W/kg, source: [41]).
Normal
Normal
Rectangular
Rectangular
Rectangular
Normal
Rectangular
Rectangular
Rectangular
Rectangular
0.42
0.61
0.61
0.10
0.00
0.29
0.61
0.15
0.24
Probability
Distribution
0.94
Tolerance
[dB]
4
2
Table 2.2: Uncertainty assessment of the whole-body averaged SAR.
real parameters. This value was assessed by varying the tissue parameters.
deviations of SAR between the applied FDTD voxel size of (0.5 mm)3 and the reference model with
a (0.2 mm)3 resolution (determined for a plane wave exposure)
uncertainty in SAR for the deviation of the used tissue conductivity and permittivity compared to
1.22
2.44
0.17
0.35
0.08
0.14
3
1.0
1.0
1.0
1.0
√
√3
√3
√3
3
0.42
0.35
0.35
0.06
0.00
0.94
Uncertainty
[dB]
deviation between measurements with dummies and simulations
deviation due to segmentation of the mouse model and for different strains
1.0
1.0
1.0
1.0
1.0
1.0
ci
1.0
√
√3
√3
3
1.0
1.0
Div
1
Combined Standard Uncertainty
Combined Extended Uncertainty (K=2)
Setup Model1
Transfer Calibration
- calibration of reference probe
- position of reference probe (±1 mm)
- load dependence of field transformation
- probe linearity
- accuracy of data logger
Mouse Model
- anatomy 2
- discretization (reference (0.5 mm)3 ) 3
- conductivity (∆σ = ± 10 %) 4
- permittivity (∆ = ± 10 %) 4
Error Description
30
CHAPTER 2. THERMAL RESPONSE THRESHOLDS
8
7
6
5
Rectangular
Rectangular
Rectangular
Rectangular
Rectangular
Rectangular
Normal
Rectangular
Rectangular
Rectangular
0.13
0.11
0.24
prob.
Distribution
0.15
0.20
0.79
1.30
0.04
0.94
0.01
Tolerance
[dB]
ci
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
Div
√
√3
√3
√3
√3
√3
3
1.0
√
√3
√3
3
Table 2.3: Estimation of the whole-body SAR variations.
isotropy in SAR within the setup loaded with reference dummies
deviation of SAR for a 20% lighter dummy between averaged weight reference dummies
deviation in SAR for stuffed vs. non-stuffed mouse model
averaged deviation for fit function SAR(m) with m as mouse weight while using scaled mice models
Combined Standard Uncertainty
Combined Extended Uncertainty (K=2)
power drift during exposure
isotropy within the setup 5
deviations due to unbalanced loading 6
variations in anatomy beside size
skin parameters (dry vs wet)
varying mouse postures 7
fit function for different mouse weights 8
mouse movements
- shift in x-direction (±2 mm)
- shift in y-direction (±2 mm)
- shift along body axis (z-direction) (±5 mm)
Error Description
1.06
2.12
0.08
0.06
0.14
0.09
0.12
0.46
0.75
0.02
0.54
0.01
Uncertainty
[dB]
2.3. MATERIAL AND METHODS
31
32
2.4
CHAPTER 2. THERMAL RESPONSE THRESHOLDS
Results
Table 2.4 summarizes the results for the different mouse groups and
exposure strengths. All mice were evaluated individually. Listed are
the amounts of particular mice out of the total number of mice which
reached either the thermal regulatory or the breakdown region at the
various exposure levels. No thermal regulation in any mouse was
observed at an exposure level of 2 W/kg. Thermal response was measured starting at 5 W/kg.
Figure 2.4 shows the detailed evaluation of the breakdown threshold. Plotted is the temperature increase per time of all mice in the
thermal breakdown range as a function of the applied SAR. The linear
regression for each data point was used to extrapolate the lowest SAR
value for thermal breakdown (dT/dt = 0). A value of 10.1 ± 4.0 W/kg
(K=2) for B6C3F1 mice and 7.7 ± 1.6 W/kg (K=2) for NMRI mice
was found (the variations correspond to the standard deviation of the
fit multiplied with the factor K=2 according to [47]).
No recognizable differences between the additional four sham exposed animals on any exposure day were observed. No influence from
the exposure time slot was observed when exposures were repeated at
different times of the day.
When comparing the individual mouse groups, further observations regarding the different behaviour with respect to weight classes,
gender, pregnancy and strains can be made: (1) Weight classes: by
comparing the number of animals with thermal responses across the
groups A to D, representing increasing body weights, the thermal response in mice increases with increasing body weight. (2) Gender:
when comparing the different genders of groups B and C, male mice
seem more sensitive to the RF induced heat load: a higher number
of animals responded in the thermal regulatory as well as breakdown
regions. Also when compared to the heavier females of group D, more
male animals of group C responded to the RF in the thermal regulating region, although fewer male animals responded in the breakdown
region. However, gender specific responses might be due to the effect of a higher body mass and/or different metabolic rates between
the groups. (3) Influence on pregnancy: the pregnant mice of group
D responded to the RF exposure with a similar thermal regulatory
response as the non-pregnant mice (group B). Slight differences were
2.5. DISCUSSION
33
observed for the breakdown threshold, which is by about 1 W/kg below the non-pregnant animals. This may also be an effect from the
higher body mass and/or different metabolic rates of the pregnant
mice. (4) Strain: the NMRI mice show a different thermal response
when compared to B6C3F1: more animals responded with thermal
regulation at the same SAR level, and the breakdown threshold is
reached significantly earlier, already at 7.7 W/kg vs. 9.7 - 10.8 W/kg
for B6C3F1. However, the differences may also be due to the different
body masses and metabolic rates.
2.5
Discussion
The results indicate that the metabolic rate is an important factor for
the thermal response: The literature describes the basal metabolic
rate of a mouse (approx. 0.3 W for a 30 g mouse) as being dependent upon the body mass according to the relationship mx with
0.7 < x < 0.8 [49], [50], [51]. Right below the thermal breakdown
threshold in the temperature steady state, the cooling capacity is
working at its limit. The threshold SAR is then directly connected
to the metabolic rate. The power from metabolic processes and RF
absorption must equalize with the cooling rate M +PRF = C (derived
from Equation 2.1). For example, assuming a 30 g mouse with a basal
metabolic rate of 0.3 W, the cooling rate becomes C = 0.6 W when
exposed to 10 W/kg, which is twice as high as the basal metabolic
rate.
Since the maximal cooling capacity is constant, higher metabolic
rates lead to increased temperature responses and therefore a lower
thermal breakdown threshold. If the heat balance equation also takes
into account additional heat sources like muscle movements or stress,
the thermal breakdown threshold is further reduced. This is manifested by the observed lowered threshold with increasing mass or
increased stress level of the NMRI mice. Furthermore, the results indicate that animals with excessive weight seem more sensitive for the
same SAR level.
The threshold for the thermal regulatory response is more difficult
to assess. However, the current data suggest a threshold between
2 to 5 W/kg, which is very close to the thermal break down level,
34
CHAPTER 2. THERMAL RESPONSE THRESHOLDS
Figure 2.3: Characteristic thermal response curves for below, in and
above the regulatory region. The RF on phase is marked as a gray
background. Figure (A) shows the thermal response of a mouse exposed to an SAR below the thermal threshold (no increase in temperature). Figure (B) shows the two cases in the regulatory region:
(1) an increase in body temperature occurs, but stabilizes at a higher
level and (2) no temperature increase is measurable during exposure,
but after the exposure was switched off. Figure (C) shows two thermal responses above the breakdown threshold for different SAR levels:
(1) at SAR 20 W/kg and (2) at SAR 12.6 W/kg.
2.5. DISCUSSION
35
Exposure Groups
Group
Gender
Type
Mass
[g]
Age at Delivery
[weeks]
A
B
C
D
E
female
female
male
pregnant, f.
male
B6C3F1
B6C3F1
B6C3F1
B6C3F1
NMRI
20.4±0.9
23.6±0.8
28.8±0.9
31±2.4 to 36±3.7
36.4±3.0
4
12
12
12
12
Thermal Regulatory Threshold
Group
2 W/kg
[# Animals]
5 W/kg
[# Animals]
7.2 W/kg
[# Animals]
10 W/kg
[# Animals]
A
B
C
D
E
0/8
-/0/8
-/0/8
1/8
4/8
7/8
4/8
8/8
5/8
6/8
8/8
-/8/8
6/8
8/8
8/8
7/8
8/8
Thermal Breakdown Threshold
Group
7.2 W/kg
[# Animals]
10 W/kg
[# Animals]
12.6 W/kg
[# Animals]
20 W/kg
[# Animals]
A
B
C
D
E
0/8
0/8
0/8
-/0/8
0/8
0/8
1/8
5/8
2/8
4/8
4/8
7/8
6/8
8/8
7/8
8/8
8/8
8/8
8/8
Table 2.4: The upper part of the table lists the five examined mouse
groups. The middle part shows the amount of animals out of the
total number of animals exposed which reached the thermal regulatory
threshold. The lower part shows the amount of animals which reached
the thermal breakdown threshold.
36
CHAPTER 2. THERMAL RESPONSE THRESHOLDS
Figure 2.4: Linear body heat storage increase/decrease rate for different SAR levels in the breakdown region. Each marker represents
the slope of the linear body temperature change in one mouse at the
exposed SAR level. The linear regression fits were evaluated for each
exposure group. The thermal breakdown threshold corresponds to the
SAR level with no heat storage change (group A: 10.5 ± 1.0 W/kg,
group B: 10.8 ± 0.6 W/kg, group C: 9.5 ± 0.4 W/kg, group D:
9.7 ± 1.6 W/kg and group E: 7.7 ± 0.8 W/kg). The averaged thermal breakdown threshold is 7.7 ± 0.8 W/kg for NMRI mice and
10.1 ± 2.0 W/kg for B6C3F1 mice. Mice are able to cool with a rate
of approx. 0.3◦ C/min after RF is switched off.
2.6. CONCLUSION
37
with a lower limit of 6 W/kg (K=2). It should be noted that the
presented results were obtained for whole-body exposure of restrained
mice in tubes (as applied in several risk assessment studies), where
the heat transfer is different when compared to free running animals
(see also Equation 2.1). However, the determined threshold values
are close to the ones on which the current safety standards [25] are
based, which rely on a thermal threshold value of 4 W/kg for adverse
health effects. Further examination and precise determination of the
thresholds, e.g., also for different species, seem necessary.
In general, the data of the study are also in line with the results of
Lu et. al [37], who derived a thermal regulatory threshold of 2 W/kg in
different strains of rats (exposed to a 120 Hz modulated 2.45 GHz signal for 2 h). Lu further determined a breakdown threshold of 9.2 to
11.5 W/kg. Additionally, disruption of food motivated learning behavior was found to occur between 3 to 9 W/kg across a number of
animal species and frequencies [26]. However, the uncertainties of the
dosimetry for these studies are not provided and difficult to estimate.
2.6
Conclusion
Three characteristic thermal responses were identified: (1) no thermal regulation, (2) response of the thermal regulatory system capable
of compensating the RF induced temperature load and (3) thermal
breakdown, i.e., uncompensated body temperature increase. No thermal regulation was measured for exposure levels of 2 W/kg. Thermal
regulation was observed at 5 W/kg. The thermal breakdown threshold
was determined for B6C3F1 mice to be 10.1 ± 4.0 W/kg (K=2) and
for NMRI mice to be 7.7 ± 1.6 W/kg (K=2). The thermal response
was found to be dependent upon the metabolic rate of the animal,
which is a function of body mass and also depends on the activity
level. The presented data suggest that the thermal breakdown for
in tubes restrained mice, whole-body exposed to RF, may occur at
SAR levels of 6 W/kg (K=2) at laboratory conditions. Considering
the uncertainty of the SAR assessment, the minimal level might even
be below this threshold. Furthermore it is expected that the threshold
is higher for free-running animals at the same conditions and lower
for restrained animals at higher environment temperatures.
Chapter 3
Assessment of ELF MF
of Mobile Phones and
Proposal for Worst-Case
Signal for
Bio-Experiments
3.1
Abstract
In addition to the radio frequency (RF) transmission of mobile phones,
extremely low frequency magnetic fields (ELF MF) are also present
in the closest vicinity of the handset as a result of the relatively large
supply currents of the RF amplifiers. Some scientists and health agencies debate the need for combined ELF RF exposure experiments in
the context of health risk assessments. The objectives of this study
were the evaluation of the ELF and RF exposures of typical GSM
mobile phones and to propose representative worst-case exposure conditions for bioexperiments. The exposures of five commercial mobile
phones were assessed (ELF & RF spectral field distributions). The
39
40
CHAPTER 3. ELF MF OF MOBILE PHONES
ELF MF values at the surface were between 35 - 75 µT on the back
and 8 - 20 µT on the front side, the peak SAR (10 g) ranged between
0.44 and 1.02 mW/g. As expected there was no spatial correlation between SAR and the ELF magnetic fields. Based on the measurements,
a worst-case signal was proposed for bioexperiments that includes the
maximum values for the pulse height, pulse slopes and pulse half width
of the measured waveform shapes as well as peak SAR values.
3.2
Introduction
The research being conducted to investigate possible adverse health
effects posed by mobile phones focuses on the radiofrequency electromagnetic field exposures. In addition to the RF, mobile phones
also generate static and extremely low frequency magnetic fields due
to the relatively large supply currents. Since the front-end amplifier
is the component of highest power consumption, the peak fields are
the strongest for systems providing high duty cycles, like time division multiple access (TDMA) systems. On the other hand, the fields
decay rapidly with the distance from the phone. However, some scientists and health agencies argue that the strongest RF exposures and
ELF MF fields occur at the same loci and have identified needs for
bioexperiments with combined exposures.
The objectives of the this study were to assess the signal strength
and characteristics of the ELF and RF exposures of GSM phones and
to develop a worst-case combined exposure scenario. The basic frame
of the GSM TDMA system has a length of 4.6 ms containing eight
bursts of 0.577 ms durations. In addition to the 217 Hz component a
8 Hz spectral component is present since every 26th basic frame of the
multi-frame structure is idle. The comfort noise in the Discontinuous
Transmission Mode (DTX) adds an additional 2 Hz component [52]
[53]. This study was performed together with the student Markus
Tuor and was sponsored through the Swiss Federal Office of Public
Health.
The ELF MF of mobile phones had been studied by several authors, e.g., [54] [55] [53] [56]. However, none of them provide the information required to define a worst-case exposure scenario for bioexperiments. Some did not measure the magnetic fields, just the battery
3.3. MATERIAL AND METHODS
41
currents [56]; others used measurement devices capable of only a few
hundred Hertz, so that the actual spectral power of higher harmonics was not sufficiently identified (e.g., [54]). The presented studies
generally show no spatial resolution. None of the studies presented
the magnetic field distribution, only peak values (e.g., [53]) and waveforms [56]. Systematic peak spatial-averaged RF exposures have also
been reported since the mid 1990’s, e.g., [57]. None of the studies
provide a complete picture of the exposure of the cells closest to the
handset which are the subject of this investigation.
3.3
3.3.1
Material and Methods
B-Field Probes
The assessment of these ELF fields requires advanced features of the
measurement probe. The determination of the pulse characteristics
requires a bandwidth of larger than 20 kHz. The evaluation of the
field distribution demands a spatial resolution of much better than
10 mm, and the ability to measure as close to the surface of a cell
phone as possible. The field strength decreases rapidly with distance
from the phone surface, so that the magnetic field probe needs to
be sensitive enough to measure low fields ( 1 µT). Since there was
no probe available with all these requirements, two probes were employed, i.e., one for the high resolution scanning and one for assessing
the spectral content of the signal.
A 3-axis Hall meter (Gaussmeter 7030, F.W. Bell, USA) with a
3-axis probe (type ZOA73-3208-05-T, F.W. Bell, USA) was used to
measure the spatial B-field distribution. The advantage of the Hall
sensor is its high local resolution due to the small probe dimensions.
The sensors are arranged orthogonally within a 4.8 x 4.8 x 5.8 mm3
cube with a distance between the tip of the probe and the average sensor location of just 5 mm. The effective frequency response was taken
into account and measured prior the experiments using a Helmholtz
coil setup.
The disadvantages include the low frequency range of the Hall sensor from DC to 400 Hz and a relatively high broadband noise level in
the time domain of 1 µT. However, when narrowing the measurement
42
CHAPTER 3. ELF MF OF MOBILE PHONES
bands to the 217 Hz component of the B-field signal, levels down to
60 nT were detectable.
A 3-axis coil probe (ELT-400, NARDA, USA) was used to measure the waveform of the B-field exposure in the time domain. The
advantages of the probe include its frequency range of up to 400 kHz
and sensitivity to resolve B-fields between 80 nT and 80 mT. The noise
level of 80 nT can be reduced down to 10 nT by applying time averaging over 100 ms. The frequency response was determined by means
of a long wire setup and was taken into account. The disadvantage
is the large coil diameter of 112 mm, which precludes measurements
close to the phone surface and provides limited local field information.
In order to achieve high immunity in the strong RF environment
near mobile phones, both magnetic field sensors and cables were thoroughly shielded with metallic mesh sleeves. The effectiveness of the
shielding was verified close to a dipole operating at 900 MHz with a
peak input power of 100 W and variable pulse frequencies between
0.1 Hz - 400 kHz.
3.3.2
Measurement Procedure
The mobile phones operated at maximum output power at GSM
900 MHz (2 W, non-DTX), controlled by a digital radio tester (Rohde&Schwartz, CTS 55, Germany) and were positioned in plastic
holders for measurement. The magnetic field probes were mounted
on an industrial robot (RX90L, Staeubli SA, France) of a computer
controlled measurement system (DASY4, SPEAG, Switzerland) which
positions the probes with a precision of better than 0.1 mm (see also
Figure 3.1). The magnetic field values from the sensors were acquired
using a data acquisition interface (PCI-MIO-16E-1 with Labview 7.0,
National Instruments, USA). The data were processed with Matlab 7
(Mathworks Inc, USA).
The B-field distribution and the waveform were measured with the
Hall sensor and Coil probe, respectively. Additionally, DC field and
SAR distributions with 1g and 10g average values were determined.
All five mobile phones were measured using the same procedure:
(1) Measurement of the B-field waveform with the coil probe
(1.6 MHz sampling rate) at the position where a maximal signal to
3.3. MATERIAL AND METHODS
43
Figure 3.1: Measurement system consisting of a magnetic field probe
mounted on a DASY4 robot and computer controlled field data processing unit.
noise ratio was detected. The full spectrum of the waveform was determined by a fast Fourier transformation (FFT).
(2) Measurement of the B-field distribution on the front and back
side surfaces of the mobile phone with the Hall meter detecting the
217 Hz component in a 10 mm grid with the probe touching the phone
case. The absolute field strength was determined by matching the
217 Hz component of the Hall meter with that of the coil probe.
(3) Measurement of the B-field decay perpendicular to the surface
of the phone with the Hall meter. The measurements were carried out
in 1 mm steps at the field maximum of the surface scan.
(4) Detection of the maximum DC field on the front and back
sides of the phone with the Hall sensor using surface and decay scans.
To suppress the DC background level during the measurements (e.g.,
from the magnetic field of the earth) the Hall sensor output was set
to zero when the phone was not yet in place. Then the maximal DC
level at the surface of the phone was determined. At the location
of the field maximum, the field decay was determined with a z-scan
perpendicular to the surface of the phone (1 mm steps).
(5) To obtain realistic DC and ELF MF values at the surface’s
44
CHAPTER 3. ELF MF OF MOBILE PHONES
of the phones, the fields were extrapolated by employing the simple
circular loop model [58].
(6) Determination of the SAR distributions of the phones including
the corresponding 1 g and 10 g peak spatial peak SAR values in accordance with IEEE P1528/D1.2 using the DASY4 near-field scanning
system (SPEAG, Switzerland) [59]. The extrapolated field values were
assessed using the polynomial extrapolation routines implemented in
the evaluation software of DASY4.
3.4
Results
Figure 3.2 and Figure 3.3 provide for all five mobile phones the B-field
distribution on front and back sides, the B-field pulse in the time domain and the B-field spectrum. The distribution, measured with the
Hall probe, shows the 217 Hz magnetic
field component as the sum
q
2
vector of all three axes (Bs = Bx + By2 + Bz2 ). The surface scans
were performed at an average distance of 5 mm between the phone
surface and the sensor position. The position of the phone (white
contour) and the battery (yellow contour) are sketched. The waveform, measured with the Coil probe, was extrapolated to the B-field
maximum on the phone surface. The B-field spectrum was derived
from the waveform and is shown together with the corresponding ICNIRP reference levels [25].
A summary of all measured and evaluated parameters is provided
in Table 3.1, showing B-field parameters of the front and back sides as
well as general parameters for each phone, such as waveform aspects,
DC fields and SAR values.
The upper half of Table 3.1 shows the maximum B-field values
of the front and back sides and the corresponding slope. Listed are
the pulse height as maximum measured B-field strength during the
surface scan (5 mm distance to surface) and the pulse height according
to the extrapolated maximum B-field strength on the phone surface
(0 mm distance to surface). The maximum slope of the up/down flank
was determined from the maximum pulse extrapolated to the phone
surface. Furthermore listed are the fit coefficients (current, radius and
position) used for the B-field extrapolation model to determine the
maximum B-field strength on the surface (0 mm distance to surface).
3.4. RESULTS
45
The lower half of Table 3.1 shows the pulse half width (full pulse
width at half rise of the pulse), the half rise time (the rise time until
half of the pulse height is reached), the measured maximal DC field
(at 5 mm to surface and extrapolated to the surface) and the spatial
peak SAR values (1g and 10g peak spatial average). The ratio of the
measured frequency components compared to ICNIRP reference levels
for the 217 Hz and its harmonic components is shown at the bottom
of the Table 3.1.
Figure 3.2: B-field distribution on front and back sides showing the
217 Hz component (sum vector of all three axes) at a distance of 5 mm
from the phone surface. The position of the phone (white contour) as
well as the battery (yellow contour) are marked.
46
CHAPTER 3. ELF MF OF MOBILE PHONES
A
B
C
D
E
Figure 3.3: Waveforms with corresponding spectral components in the
field maximum extrapolated to the mobile phone surfaces.
3.4. RESULTS
47
Unit
Phone
A
Phone
B
Phone
C
Phone
D
Phone
E
µT
µT
T/s
T/s
4.70
8.3
0.23
-0.21
7.52
12.4
0.10
-0.11
14.63
19.3
0.26
-0.28
6.09
8.3
0.13
-0.13
4.94
11.7
0.10
-0.10
µT
µT
T/s
T/s
29.46
52.8
1.43
-1.33
31.89
35.1
0.29
-0.30
33.68
66.1
0.88
-0.97
29.50
74.8
1.14
-1.14
28.07
56.3
0.50
-0.50
ms
ns
0.59
24
0.58
68
0.58
47
0.58
39
0.59
88
mT
mT
5.3
20.2
1.6
3.4
2.7
8.6
4.1
13.1
1.5
3.9
mW/g
mW/g
1.21
0.826
1.54
1.01
1.49
1.02
0.616
0.438
1.16
0.707
≤1
≤1
1.2
1.2
≤1
≤1
≤1
≤1
≤1
≤1
≤1
1.2
1.6
1.6
1.1
≤1
1.5
2.0
1.7
1.4
≤1
1.1
1.3
1.1
≤1
Front side
Pulse height, 5 mma
Pulse height, extp.b
Max dB/dt, up slopeb
Max dB/dt, down slopeb
Back side
Pulse height, 5 mma
Pulse height, extp.b
Max dB/dt, up slopeb
Max dB/dt, down slopeb
Pulse form
Half width
Half rise time
DC B-field
at 5 mma distance
extrapolatedb
Spatial peak SAR
SAR (1g)
SAR (10g)
Ratio to ICNIRP limit
217 Hz
1st harmonic: 433 Hz
2nd harmonic: 650 Hz
3rd harmonic: 867 Hz
4th harmonic: 1083 Hz
a
Signal 5 mm from surface (probe touching surface) at maximal B-field point
(217 Hz or DC).
b
Extrapolated signal on surface at maximal B-field point.
Table 3.1: Mobile phone field measurement results.
48
3.5
CHAPTER 3. ELF MF OF MOBILE PHONES
Generic Exposure Test Signal
Although the statistical power of the sample is limited, the derived values provide good basis for selecting the worst-case ELF MF exposure
scenario. The selection criteria were the maximized B-field strength
and maximized spectral power derived from the mobile phone measurements. This led to the following proposed worst-case ELF test
signal as shown in Figure 3.4.
Figure 3.4: Generic worst-case ELF test signal generated from the
measurement results of five GSM phones (1) pulse height: 75 µT (max
of 5 phones), (2) pulse slope up: 1.4 T/s (maximum of five phones),
(3) pulse slope down: -1.3 T/s (minimum of five phones) and (4) pulse
half width: 0.584 ms (average of five phones).
3.6
Discussion and Conclusion
The five tested phones showed considerably different ELF MF waveforms and amplitudes as well as surface SAR values. Whereas the
3.6. DISCUSSION AND CONCLUSION
49
pulse width is similar for all phones and corresponds to the GSM RF
pulse width of 0.58 ms, the pulse shapes show individual characteristics with respect to their half rise times, which vary between 24 and
88 ns. All phones show the maximum ELF MF on the back side with
extrapolated pulse heights between 35 and 75 µT. At this location,
four out of the five tested phones exceeded the ICNIRP reference levels for several harmonics of 217 Hz. The maximum exceedance by a
factor of two was detected at 650 Hz. The ELF MF on the front side
of the phones tested were by a factor of 2 - 6 times smaller and varied
between 8 and 20 µT. The field distribution showed a localized field
maximum near the battery, whereby the field is dominantly polarized
perpendicular to the phone surface. Hence, the current distribution
on the phone can be approximated by a horizontal current loop in the
battery region. In contrast, the maximum DC fields were detected
near the phone loudspeaker and reached levels up to 20 mT (corresponding ICNIRP reference level: 40mT). The SAR measurements
revealed spatial peak SAR values between 0.6 - 1.5 mW/g (1g) and
0.4 - 1.0 mW/g (10g). No correlation between the spatial peak SAR
and the peak B-field was seen.
It is important to note that the ICNIRP basic restrictions for
ELF magnetic fields are primarily focused on induced current density.
Mathematical modeling of the human body and extrapolation were
applied to derive the reference levels for the magnetic field strength.
Since the current density is a function of the spatially averaged B-field
times the cross section of the induced current paths, localized B-fields
will lead to considerably fewer induced currents than the anticipated
ICNIRP worst-case of a homogeneously distributed B-field across the
entire human body. Therefore, the reported B-field excessive values
do not necessarily correspond to violations in current density; for the
latter, they are rather expected to cause no violation of its limit value.
However, further investigations are needed to assess the induced currents resulting from ELF mobile phone exposure in detail.
The suggested generic worst-case exposure test signal can be used
in future bioexperiments evaluating the possible effects of ELF or
combined RF/ELF fields from mobile phones.
Chapter 4
Add-on ELF MF
Exposure Setup
Proposal for Studies
with Human Volunteers
4.1
Introduction
Some scientists are concerned about the possible impact of combined
exposure to RF and ELF fields. Magnetic ELF components are mainly
generated by supply currents in the phone (see also Chapter 3). The
device with the largest power consumption is the front-end amplifier.
Consequently, the corresponding ELF magnetic field has a spectrum
similar to the pulse envelope of the RF signals. The results of the
measurements show frequency components as expected by analyzing
the GSM signal at 216.7 Hz and 8.3 Hz as well as higher harmonics.
When the magnetic fields are extrapolated from the measurement
distance at 5 mm to the phone’s surface, they reach values between
35 µT and 75 µT on the back and between 8 µT and 20 µT on the front
side of the mobile phone. The spatial peak SAR and the peak B-field
show no correlation; the spatial peak SAR values of the tested phones
51
52
CHAPTER 4. ELF MF SETUP ADD-ON PROPOSAL
range between 0.6 to 1.5 mW/g for 1 g and 0.4 to 1.0 mW/g for 10 g.
Based on the measurements, a worst-case signal was proposed
which includes the maximum values for the pulse height, pulse slopes
and pulse half width of the measured waveform shapes. This worstcase signal can be used in biomedical laboratory studies investigating
the possible effects of ELF or combined RF/ELF fields from mobile
phones.
This report presents an ELF add-on to an RF exposure setup,
which enables the exposure of human volunteers with the proposed
ELF worst-case signal in addition to the RF signal.
4.2
Setup Design
The proposed design for the RF/ELF exposure system is based on an
update of the RF exposure system as described in [60]. A sketch of
the configuration is given in Figure 4.1. However, similar considerations can be used to apply the exposure system as an add-on to other
systems for human exposure, e.g., as the sXh-900 system presented in
Chapter 5.
Two square coils with a side length of 1 m are arranged in parallel at a distance of 0.88 m. The coils are mounted exteriorly on two
wooden racks which are placed on both sides of the test person. Hence,
the head of the test person and the RF setup (which is based on two
patch antennas) are located between the coils. The coil configuration
does not exactly meet the Helmholtz condition (distance between coils
equals coil radius) but was chosen in order to keep the coil size moderate and still provide a homogeneous B-field distribution across the
human head. No interaction between the B-field and the RF setup is
present, because the RF setup is built only with non-magnetic materials. The coil wires are shielded by copper foil in order to prevent
RF coupling into the ELF current circuit. The coil setup does not
interfere with the RF exposure, because the coils are located at a sufficiently far distance and behind the patch antenna back planes. Each
coil is formed by ten windings with a 1 mm diameter copper wire resulting in an ohmic resistance of 1.88 Ω. To minimize vibrations, the
wires are glued together with epoxy resin.
4.3. B-FIELD STRENGTH AND DISTRIBUTION
53
Figure 4.1: Front view of a combined ELF and RF exposure setup.
The ELF add-on is shown in dark gray.
4.3
B-Field Strength and Distribution
The B-field exposure from the coil setup was analyzed to determine
the B-field vs. current efficiency and the homogeneity of the resulting
B-field distribution. The equations according to [61] were used to
calculate the vector components of the B-field. The resulting field
distribution at pulse maximum is shown in Figure 4.2 with the head
area indicated as a box in the center of the figure.
A homogeneous distribution is present, although the Helmholtz
condition is not exactly met. To evaluate the homogeneity in the
head area, the average B-field together with minimum, maximum and
standard deviation were determined. Hereby, the head area was assumed to be located in the center of the coils, with the dimensions
of 15 cm (±3cm) x 0.2 m x 0.3 m (width x depth x height). A
current of 1 A generates a 10.8 µT average B-field with a spatial inhomogeneity (standard deviation) of 2.1 % (maximal deviation: 2.8 %,
54
CHAPTER 4. ELF MF SETUP ADD-ON PROPOSAL
Figure 4.2: Simulated B-field of rectangular loops. The center box
represents the head area.
minimal deviation: -4.3 %). The effect of varying head width was further investigated, since it determines the effective coil distance and
may represent a source of inter-subject variability. Two worst-cases
with a minimum head width of 12 cm and a maximum head width of
18 cm were analyzed, and the resulting B-field efficiency and standard
deviation were calculated (Table 4.1). No significant changes from the
reference configuration (deviations < 3.3 %) were observed.
4.4
Current Requirements
The current for the coil setup is generated by an arbitrary function generator, combined with a voltage controlled high-current
source. In order to produce the proposed worst-case generic test
signal, a B-field amplitude of up to 75 µT with pulse slopes of
1.4 T/s / -1.3 T/s (rise/fall) must be achieved. The current amplifier
must fulfill the following power requirements in order to generate the
B-fields of Table 4.2.
• The current I = 6.9 A to achieve the 75 µT B-field,
• the current flank dI/dt = 1.23 · 105 A/s to produce the 1.4 T/s
B-field slope and
4.5. SUMMARY ELF ADD-ON SETUP
B-field efficiency
Bavg /I
B-field efficiency
deviation from 15 cm
Spatial B-field
inhomogeneityb
Maximal deviationc
a
b
c
55
Head width
15 cm
Head width
18 cm
Head width
12 cm
10.8 µT/A
10.5 µT/A
11.1 µT/A
Reference
(-3.3%)a
(+2.7%)a
2.1 %
3.2 %
1.5 %
4.0 %
5.1 %
1.8 %
Relative deviation to the reference configuration with head width 15 cm.
Standard deviation of B-field in head volume.
Maximal B-field deviation in head volume.
Table 4.1: B-field efficiency and uniformity for varying head width.
• the voltage U = 21 V from U = L dI
dt across the coil setup.
The inductance L value of 73 nH for the coil setup was determined
numerically by calculating the magnetic flux and using the relation
L = 2 dΦ
dt , where Φ is the flux through one of the coils (2 coils in series).
4.5
Summary ELF Add-on Setup
Table 4.2 summarizes the specifications for the ELF add-on coil
setup, including the current requirements for the amplifier. The suggested add-on upgrades the existing RF exposure setup for combined
ELF/RF studies in the context of mobile phone health risk assessment.
56
CHAPTER 4. ELF MF SETUP ADD-ON PROPOSAL
Side length of square loops
Distance between loops
Number of windings
Coil resistance
Coil inductance
B-field efficiency
Spatial B-field inhomogeneity
Current requirements for pulse peak
Maximum current flank
Voltage requirements
∗
1m
0.88 m (± 3 cm, head with variation)
10
R = 1.88 Ohm
L = 73 nH
∗
Bavg
/I = 10.8 µT/A
2.1%
I = 6.9 A
dI
5
dt = 1.23 · 10 A/s
U = 21 V
Bavg : Average B-field in head volume
Table 4.2: ELF add-on exposure setup specifications.
Part III
Exposure Systems and
Studies
57
Chapter 5
Flexible Exposure Setup
for Experimental
Provocation Studies at
884 MHz
5.1
Abstract
Experimental provocation studies allow the investigation of subjective
symptoms in humans caused by exposure to mobile communication
frequencies. Presented is the system for the exposure of the heads
of volunteers in the experimental provocation study performed at the
Karolinska Institute. The setup was based on a new, low-weight,
stacked micropatch antenna fixed on a headset. The radiation pattern of the antenna was optimized to unilaterally expose regions of the
brain. A hanging construction balanced the weight of the headset and
allowed head rotations without changing the exposure pattern in the
head. A controllable heat load was realized by a laser diode, connected
via an optical fiber to a photo diode that was heated when absorbing
the laser radiation. The photo diode was packed inside a thermally
conductive ceramic plate (cross section 10 mm2 ) taped to the ear lobe
59
60
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
of the test person. A fully computer-controlled signal unit regulated
the GSM modulated RF exposures of two test persons. The exposure was controlled by measurements of forward and reflected power;
the temperature load was controlled by an RF immune thermistor
sensor housed within the ceramic plate. The information on the exposure condition (RF or sham exposure) was encoded and allowed
double-blinded exposure protocols. The experimental dosimetry was
conducted with a DASY near-field scanner and revealed an antenna
efficiency of 0.52 - 0.57 W/kg per Watt input power at 10g-averaged
spatial peak SAR. The non-DTX exposure was set to 10g-averaged
peak spatial SAR of 1.95 W/kg. The applied random temporal change
of 11 s to 5 s for non-DTX to DTX exposure signals resulted in a timeaveraged 10g-averaged peak spatial SAR of 1.4 W/kg. The peak spatial SAR for the gray matter was 1.8 W/kg averaged over a cube of
1 g. The averaged values for gray matter was 0.2 W/kg and for the
thalamus 0.18 W/kg.
5.2
Introduction
The Department of Health Science at the Karolinska Institute in
Stockholm, Sweden, conducted a human experimental provocation
study in the context of health risk assessment of low-level exposure to
the RF of mobile phones. Examined were the effects of the GSM wireless communication signal at 884 MHz on subjective symptoms and
physiological reactions [62]. The study followed standardized exposure
protocols and was initially conducted within the PERFORM studies (PERFORM C) of the 5th Framework Program of the European
Union. During the course of the development the project was separated from the PERFORM studies and renamed to “Mobile Phones
and Direct Health Effects”. However, within this thesis it is still referred to as PERFORM C. The results contributed to the conclusions
of the WHO about potential health effects from mobile phone use.
This chapter presents the flexible exposure setup which was used in
the Karolinska Institute study between November 2004 and January
2006. The exposure setup satisfied the ethical standards for human
studies, the exposure level met the ICNIRP guideline and the exposure
duration was in the order of a frequent mobile phone user.
5.3. REQUIREMENTS
61
The aim of the study was to establish whether exposure to radiofrequency fields from mobile phone use during the day has any acute effects on self reported symptoms (headache, vertigo, skin irritation and
sensation of heat). Further examined were potential biological correlations between the subjective symptoms (skin temperature, serum
levels of stress related hormones, electronystagmography, breath rate,
thoraco-abdominal coordination in breathing and tidal volume) and
an influence on the cardiovascular system with regard to blood pressure, heart rate, heart rate variability, and sleepiness and performance
on a subsequent night’s sleep. Therefore the exposure was developed
to be consistent with worst-case exposure occurring in real situations,
with an extended duration.
The following specific hypotheses were tested: Exposure to radiofrequency fields caused by mobile phone use leads to (1) a significant increase in systolic blood pressure and heart rate, (2) a decreased
response time in simple serial reaction test and vigilance test, (3) a
change in brain electric oscillations in particular during cognitive processes and (4) whether it leads to a change in brain electric oscillations
during subsequent sleep.
5.3
Requirements
There were several requirements for the exposure system of the proposed study. They followed the general underlying principle design:
As long as no interaction mechanism is established, the objective of
the study design should be to provide maximized significance of negative findings. If positive effects are found, a subsequent analysis of
all parameters shall be used to assess the actual health risk. Exposure should represent the worst case with respect to exposed tissues,
exposure level and signal characteristics [10].
This required that the exposed area cover the entire cortex of the
exposed hemisphere with an absorption depth reaching subcortical
structures and can be achieved through antennas with large radiator
cross section. The exposure should be unilateral, with an exposure
strength maximized according to occupational limits.
Figure 5.1 shows an exposure from a patch antenna in comparison
to a generic telephone exposure in the tilt and touch positions. The
62
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
[
dB]
Figure 5.1: The exposure pattern of a patch antenna covers the various
possible exposures which occur from the holding positions of mobile
phones. Shown are the exposed areas of a patch antenna and of a
generic mobile phone (GP) in the tilt and touch positions.
patch antenna enables unilateral exposure and has been successfully
applied in several previous studies [60] [63]. A comparable concept
was pursued for this setup design.
The following design requirements were given through the study
design [62] and through the requirements for well-designed exposure
systems for human studies [10]. They were commonly agreed upon
and finalized during the PERFORM C steering committee meeting in
December 2003. Furthermore it was decided that the exposure should
be similar to the Zurich studies [64] for comparison, but not replicate
of the Zurich Study.
The setup design requirements were:
• Intersubject exposure variability must be minimal; effects of
head size and head movements on SAR values and input
impedance must be small. Therefore highly localized exposures
are not suitable, so that the distance antenna to head is larger
than 50 mm.
5.3. REQUIREMENTS
63
• Unilateral RF exposure from the left head side.
• Double-blind protocols, neither the volunteer nor the conducting
experimenter should know if an exposure takes place or a sham
exposure is applied.
• Signal unit must allow complex modulation schemes to apply
the defined exposure signal.
• Computer controlled unit allows complex modulation, doubleblind protocols, monitoring and easy handling.
• Monitoring and feedback-regulation of relevant exposure parameters to optimize the exposure control to be implemented.
• Human safety must be ensured through hardware measures that
makes impossible the overexposure of subjects.
• Comfortable setup design; the mechanical setup should have no
influence on the performance of the subjects and should allow
exposure times of three hours or more.
• Reflections within the exposure room should be kept below
20 dB. Two parallel exposures are planned in adjacent rooms.
It must be ensured that they do not influence each other and
that no other source is present.
• Temperature load due to power losses and contact between headset and skin should be simulated. The ear lobe must always be
heated, for sham or real exposure, so that the test person has
the same sensation as if a mobile phone is pressed against the
ear and head.n as if a mobile phone is pressed against the ear
and head.
• Signal characteristics should have maximized spectral power signal schemes and should cover all ELF components of the GSM
technology. A cocktail of modulation components should be
considered, e.g., non-DTX, DTX, handover and power control.
Proposed was the additional application of ELF magnetic fields as
described in the previous chapter; however, it was decided by the
study steering committee not to examine their influence.
64
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
A key component of the setup is the antenna, which had to fulfill the
following requirements:
• Antenna should operate around 900 MHz and expose the human
head with 2 W/kg (peak).
• Test persons should be able to move their heads in order to
perform simple tasks like reading, talking, etc. The exposure
unit must be mounted on the head. This requires the antenna
to be compact and very lightweight.
• A significantly large area of the test person’s head should be
exposed. The aim is to achieve a uniform distribution on the
sagittal planes of the head, keeping about the same rate of SAR
fall-off in the coronal (frontal) planes. This simulates a worstcase exposure scenario and should cover the exposure of all different types of mobile phone antennas. The antenna is therefore
required to be larger than usual mobile phone antennas to expose one full side of a head hemisphere.
• The antenna should not be sensitive to absorbers in the near
field, i.e., no major shift in the reflecting power when the head
of the test person moves. The antenna should produce a good,
broadband S11 input reflection coefficient around 900 MHz.
• EEG of the test persons are measured during the RF exposure. In order to avoid or at least minimize any coupling effects,
the E-field of the antenna should be vertically polarized, which
allows mounting of the EEG cables perpendicular to the field
polarization.
5.4. EXPOSURE SYSTEM DESIGN
5.4
65
Exposure System Design
Taking all boundary conditions of the study and the setup requirements [62] [10] into account, the following system for exposure of
humans (named sXh-900) was developed. An antenna exposed the
left hemisphere of the test person’s head with a signal in the GSM900 MHz region as in the Zurich studies (for an analogy comparison).
The antenna was connected via a flexible and light headset to the test
person’s head, enabling slow movements; in addition a spot at the left
ear lobe was heated to cause a heat sensation. The whole system was
secured against any kind of overexposure and fully self controlled. The
exposure system was designed for double-blind protocols and exposed
or sham-exposed two subjects in parallel.
This required the development of:
• new antenna, which is lightweight and exposes a large area of
the test persons head,
• controllable RF immune heat load,
• light headset and
• self-controlled exposure system with an easy user interface.
Additionally, suitable exposure rooms needed to be prepared and the
conduction of the study ensured.
5.4.1
Low Weight Stacked Micropatch Antenna
No commercially suitable antenna was found; therefore a new antenna
was developed to fulfill the requirements. Various antenna designs
were evaluated [65] [66] [67] [68] - the most suitable seemed the enhanced (stacked) micropatch antenna design. The antenna design, dimensions, materials and further mechanical parameters are presented
here and compared to the commercially available antenna from Huber
and Suhner (Huber & Suhner SPA 920/65/9/0/V) which was used in
the Zurich studies [64] [60] [63].
Micropatch antennas have several advantages: low weight, small
volume, ease of fabrication, low-profile, enable linear or cross polarization and can have multiple radiating frequencies, making them very
66
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
attractive for a wide range of applications. In their simplest form
micropatch antennas consist of a radiating patch on one side of a dielectric substrate and a ground plate on the other side, with the patch
having various shapes [69] [70] [71]. Major limitations of micropatch
antennas in their basic form are: low gain, low power-handling capability and a small bandwidth of usually less than 1%. Especially the
narrow bandwidth is a major limiting factor, but it can be overcome by
employing different techniques, i.e. use of thick substrates, aperture
coupling, electromagnetic coupling, parasitic elements, multi-layer design, multi-resonator designs. Depending upon the technique used,
the bandwidth can be increased up to 70% [72] [73].
A stacked electro-magnetically coupled micropatch antenna (SMP)
with a center rectangular patch and thick substrates was chosen as
the design of the setup antenna. This design has advantages to other
techniques: easy and reproducible fabrication, low cost with high performance, large bandwidth, high gain, sufficient power handling capacities, relatively light weight and dimensions which can be increased
to the desired size for exposing one head hemisphere.
The developed stacked patch antenna consists of three copper
plates: ground, patch and stack panel. The plates are separated by the
lightweight, low loss substrate ROHACELL HF51, with a dielectric
constant of close to one (Röhm GmbH, Darmstadt, Germany). The
layers are fixed together with Araldit Rapid two component epoxy glue
(Vantico AG, Basel, Switzerland). The front and side areas are covered with a black polypropylene foil of 0.5 mm thickness. Figure 5.2
shows the design, dimensions and photos of the developed antenna.
The antenna consists of a bent ground plate of 190x130x0.1 mm3
extended at a 90◦ angle at both long sides by 30 mm. The rectangular patch is separated from the ground plane with a 14 mm thick
substrate and fed centrally from one long end. The patch at the feed
point is shorted to increase the matching with 50 Ω. The patch has
the dimensions 157-162x80x0.1 mm3 . The rectangular stack plate is
positioned on the other side of a second 14 mm thick substrate on top
of the patch. The dimensions of the stack panel are 125x115x0.1 mm3 .
The radiation from the micropatch antenna occurs from the fringing fields between the periphery of the coaxially fed patch and the
ground plate. The stack on top of the antenna is electromagnetically
coupled and radiates at its edges. The length of the rectangular patch
5.4. EXPOSURE SYSTEM DESIGN
Figure 5.2: Design and dimensions of the SMP antenna.
67
68
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
was chosen to be slightly smaller than λe /2 (half wavelength in the
dielectric medium), so that the fundamental TM10 mode is excited.
The ground plate is increased along its long ends to enhance the front
to back radiation ratio. Simulations showed that an increase at its
long ends is sufficient, since the extension is parallel to the E-field polarization. The ground plate extensions are bent on both sides by 90◦ ,
minimizing the space requirements. The dimensions and positions of
the patch and stack are optimized so that the resonance frequencies
are close to each other to yield a broad bandwidth.
Numerical and experimental evaluations were conducted to characterize the antenna and to validate the numerical antenna model.
The numerical evaluation was performed with the FDTD based simulation platform SEMCAD Version 1.8b24 on a Pentium4 Windows XP
based workstation (SPEAG, Switzerland). The experimental evaluations were undertaken in an anechoic chamber using the DASY4 measurement system (SPEAG, Switzerland) with the latest high-precision
E- and H-field probes (EF3DVP6, SN 4004 and H3DV6, SN 6065).
The performance results show the SMP antenna fulfills the requirements of the setup antenna and has no major drawbacks in comparison
to the Huber & Suhner SPA 920 antenna (SPA). Table 5.1 summarizes
the antenna parameters derived from simulations and measurements
of the antenna SMP and antenna SPA. The advantages and disadvantages of the SMP antenna in comparison to the commercial SPA
antenna are listed in Table 5.2.
Furthermore, the numerical antenna model has the same characteristic radiation performance as the real antenna and is suitable for
simulations with anatomical head phantoms in the final dosimetry.
The validation revealed no significant differences in the free-field results; differences were only detected at the half-power beam width
(HPBW) and impedance: the simulation model has a smaller halfpower beam width, and the impedance differs between the model and
the real antenna. However, the field comparisons in E- and H-fields
are most important: the SMP SEMCAD model and SMP antenna
showed good agreement in both E- and H-field values with deviations
of less than 16 % at the various distances (see also Figure 5.3).
5.4. EXPOSURE SYSTEM DESIGN
69
Figure 5.3: Comparison of the numeric and measured E- and H-fields
of the SMP antenna shows a deviation of less than 16%. The simulation model represents the real antenna.
70
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
Parameter
Electrical:
Frequency Range
Bandwidth
Return Loss (900 MHz)
Impedance (900 MHz)
Front-to-Back-Ratio
Gain
Horizontal HPBW
Vertical HPBW
Maximum Input
Power (T<25◦ C)
Polarization
Mechanical:
Dimensions
Weight
SMP (Meas.)
SMP (Sim.)
SPA Specs
MHz
MHz
dB
Ohm
dB
dBi
818-948
130
35
(51+j2)
20
◦
95
80
>50
819-945
126
16
(45 + j14)
20
7.6
86
73
890 - 960
70
18
50
15
9.0
65
65
>100
linear,
vertical
linear,
vertical
◦
W
linear,
vertical
mm3
g
190 x 130 x 28
110
250 x 210 x 36
500
Table 5.1: Performance summary of antennas SMP and SPA.
Property comparison SMP to SPA antenna:
+ light weight (just 110 g), therefore mountable on a headset.
+ better front-to-back power ratio, less disturbance of 2nd setup.
+ larger HPBW, allows the antenna to be mounted closer to head.
o large frequency range and bandwidth.
o impedance is matched to 50 Ω for both antennas at 900 MHz.
o sufficient input power capabilities.
o linear, vertical polarization.
- slightly lower gain (potential higher amp costs for same field strength).
- needs to be built, not just ordered.
Table 5.2: Advantages, equivalent properties and disadvantages of the
SMP antenna in comparison to the SPA antenna.
5.4. EXPOSURE SYSTEM DESIGN
5.4.2
71
Heat Load
Battery heating or contact pressure - when the user presses the mobile phone against his/her skin - could lead to a thermal skin sensation (skin surface temperature increase due to RF exposure is negligible). Several studies estimated a surface skin temperature increase
to 39 - 40◦ C [74][75][76].
To simulate this sensation on the skin surface, the study design
planned a heat load at the ear lobe of the exposed side, which applies
a 2◦ C local temperature increase. During an exposure session the
heat load was always on, for both sham and RF exposure through the
radiation of the antenna. Important was the presence of a heat load
and not the size of the heat simulator [77], therefore a relatively small,
heated area at the ear lobe was sufficient to cause the sensation. Of
further importance was that the heat load is RF immune; it must not
disturb or be disturbed in the RF field.
The heat load at the ear lobe was realized as follows: A laser diode
module generated a maximum output power of 1 W at a wavelength
of 810 nm (Model PoF-LD, Photonic Power Systems Inc.). The output power was regulated via a voltage control signal from the data
logger. The laser beam was coupled into an optical fiber and terminated at a matching photo diode (PPC-5MT, Photonic Power Systems
Inc.), which dissipated the radiation by generating heat. The photo
diode was glued together with a T1V3 temperature probe (SPEAG,
Switzerland) inside a MACOR glass-ceramic plate (Eperon, Switzerland), which has a high thermal conductivity of 1.5 W/Km. The Loctite Hysol 9496 A&B glue with a thermal conductivity of 1.7 W/Km
was used for gluing. This enabled a homogeneous temperature distribution at the ceramic plate surface.
The laser/photodiode heat load together with the SPEAG temperature probe enabled a feedback-loop-control of the heat applied on the
ceramic surface. When fixed at the ear lobe, the ceramic needed about
two minutes to reach the steady state temperature and subsequently
remained stable within 0.3◦ C. Figure 5.4 shows the RF immune heat
load which simulated the temperature increase of the skin surface on
the exposed side of the head.
72
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
Data
Control
optical
pwr ctrl
Laser Diode
810nm, max 1W
ear
lobe
ceramic plate
(12mm x 8 mm)
with photo diode
& temp probe ear tape
RF immune
temperature meas.
optical fibre (d<62um)
Figure 5.4: Heat dissipation from a laser/photodiode combination was
used to heat a ceramic surface. The ceramic was fixed at the left ear
lobe with adhesive tape. A RF immune temperature probe inside the
ceramics enabled a feedback loop control. When heated the ceramic
needed about two minutes to reach the steady state temperature and
remained stable within 0.3◦ C.
5.4.3
Headset
The headset positions the heat load and antenna at a defined, constant distance from the test person’s head. Previous setup designs for
human exposure studies forced the test person to hold their head in a
certain position. Thereby a spacer construction assisted to positioning of the antenna and head at a certain distance from each other.
The test person had to hold his/her head without movements during
the exposure time with high level of discomfort for longer exposure
times. Other setup designs allowed free head movements. However,
this often resulted in scenarios with high exposure variability or lack
of an exposure gradient between the head hemispheres.
The new headset overcomes these drawbacks. It allows basic head
movements (area ± 40 cm), enabling the test person to move his/her
head and perform simple tasks like reading, talking, etc. without
changing the position of the antenna relative to the head. The headset
was designed to be adjustable in width and nose position, respecting
the various head sizes of test persons [78]. Together with additional
foam padding, a comfort level was achieved which allows exposure
times of several hours.
5.4. EXPOSURE SYSTEM DESIGN
73
The headset was optimized with respect to the following parameters: (1) antenna vs. head position, optimizing the radiation
pattern (SAR ratio left/right head hemisphere, SAR ratio exposed
side/thalamus), and (2) comfort of headset (weight, adjustability,
foam padding). Additionally, the headset allows the positioning of
the heat load on the exposure side in a comfortable way.
The position of the antenna to the head has a great influence on
the exposure pattern; the antenna to head distance also influences the
mechanical comfort. The position was optimized in a simulation matrix by moving the antenna up/down and closer/farther away in relation to the head. The simulations were performed using the validated
SMP antenna model and a human anatomical head model HREF-1,
placing the SMP antenna at different locations on the left head side.
The simulation results delivered the spatial peak SAR (1g and 10g) of
left and right head hemispheres, the antenna efficiency, and also the
organ specific SAR, e.g., of the thalamus (located in the center of the
brain and often used as a reference) and also other parts of the brain
like the cortex. Figure 5.5 shows a selection of simulations of various
heights and distances between the antenna and head.
The findings include: the further away the antenna, the more homogeneous is the exposure of the full head hemisphere. The absorption gradient inside the head is less steep; a positioning of the antenna
at a short distance from the head leads to high absorption gradients.
The advantage of a high gradient is that the exposure of the left and
right head hemispheres can be distinguished; however, a too steep
gradient represents an unrealistic exposure scenario. To categorize
the exposure distribution in the brain, a ratio between the directly
exposed head hemisphere and non-exposed head hemisphere was calculated, as well as the ratio of the SAR at the cortex to the spatial
peak SAR of the exposed hemisphere for different exposure scenarios. An additional criterion was that the exposure pattern on the skin
represent a realistic but worst-case exposure pattern. Furthermore
the simulations showed that the closer the antenna, the higher the
SAR for the same input power and therefore the higher the antenna
efficiency (antenna efficiency ranged for distances between 50 mm and
150 mm between 60 % and 20 %, respectively, values varied further
with horizontal and vertical shifts of the antenna).
These exposure conditions were the key criteria for choosing the
74
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
Figure 5.5: SAR distribution on the head skin (A) and on the cortex
(B). The ratio between exposed and non-exposed hemispheres as well
as the ratio between the SAR in the cortex to the SAR in the thalamus
are provided in (B).
5.4. EXPOSURE SYSTEM DESIGN
75
distance and height with respect to the user’s head. The center scenario as shown in Figure 5.5 was chosen for the final headset design
(decision made during a PERFORM C steering committee meeting in
Stockholm, Dec. 2003). Besides well-balanced exposure, additional
criteria were a similar exposure pattern of the brain as in the Zurich
studies and a feasible mechanical location of the antenna to realize a
comfortable headset.
With the positioning of the antenna to head finalized, the next
goals to achieve in the headset design were to make it light in weight
and as comfortable as possible. The final headset was based on a
glasses-style design using lightweight plastic foam material and foam
padding on the frame. The headset was mounted on the ceiling in
the laboratory through a hanging system which allowed movements
in all three room directions using hand pulley blocks. A counterweight
system was installed, so that the complete weight of the headset was
well balanced. Although effort was undertaken to minimize its weight,
the total headset had a weight of 460 g (antenna 110 g, headset 300 g,
heat load 50 g). The headset and its mechanics for adjusting to different head sizes are shown in Figure 5.6.
The developed headset allows easy and comfortable usage and fulfilled its mechanical purpose to position antenna and heat load at a
fixed position with respect to the test persons head. The headset can
be easily used:
• Positioning of the subject: to provide the best comfort for the
subjects, place the chair directly below the hanging headset so
that the head is right below the fixation point of the hanging
system on the ceiling.
• Adjusting the head width: Different head widths can be adjusted in discrete steps by the four screws at the rollover bar
and additionally by the two screws fixing the axis of the nosepiece. To verify the correct width of the headset, it should be
checked that the roll-over bar has a straight shape, is not curved
and that it is slightly pressed against the head without tightening the band at the back side of the headset.
• Adjusting the head length: Adjust the horizontal position of the
headset with respect to the ears by verifying that the curved part
76
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
Figure 5.6: Headset of the exposure system. The upper pictures show
the headset at the Karolinska Institute, the lower picture present parts
of the headset.
5.4. EXPOSURE SYSTEM DESIGN
77
of the two ear frames touches the ears from behind. The axis
of the ear piece can be horizontally adjusted by unscrewing and
fixing in a different screw hole. The nose piece can be adjusted
by rotating its axis and by rotating the soft plastic part at the
end of the nose piece to fit comfortably on the nose. Finally,
the fixation pressure at the head is adjusted by the tightening
strength of the elastic band at the back of the headset.
• Adjusting the weight of the headset: The weight of the headset is
counterbalanced by a bottle filled with sand. Individual comfort
weight can be further adjusted by adding or removing small sand
bags.
• Fixing the heat load: The ceramic heat load is fixed at the ear
lobe by using one or two strips of adhesive tape.
5.4.4
Exposure Room, Monitoring and Control
System
The study design specified the parallel exposure of two test persons
in adjacent rooms. An exposure unit was installed in each room,
the signal generation, controlling and monitoring was managed outside the two exposure rooms, where the control computer and hardware electronics were located. Figure 5.7 shows the system schematics
and Figure 5.8 the control tower of the exposure system. Figure 5.10
shows the exposure rooms.
The computer controlled and monitored the entire setup using
GPIB connections. The signal was produced through an RF signal
generator (Rohde & Schwarz, SML 02B); an arbitrary function generator (Agilent, 33120A) modulated the amplitude as a GSM burst
and a radio frame generator (SPEAG, Digital Control Unit (DCU))
blanked selected radio frames according to the applied digital signal.
In addition, the DCU functioned as a safety watch-dog for unexpected
computer shutdowns. Whenever the software control signal was not
sent within ten seconds, the amplifiers were automatically blanked.
The generated signal was divided and fed into two RF amplifiers for
the two exposure units. Each output signal passed a dual directional
coupler to measure the forward and reflected power via diode de-
78
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
Computer
Data
Control
Digital
Control Unit
Function
Generator
A
RF Signal
Generator
A
water
flow
sensor
B
Amp Water
Cooling Unit
Perform C
Data Böxli
A
B
A
B
temperature,
shutdown,
power level
3dB
Splitter
A
B
Heater
Source
gain
B
RF
Amplifier
water cooling
Terminator
Coupler
Sync signal
for EEG
P.fwd
P.refl
Signal + RF
to antenna of
exposure unit
Temperature
measurement
at exposure unit
Exposure Unit A
EEG
Optical heating
at exposure unit
Identical for unit B
Exposure Unit B
EEG
RF+Signal
RF+Signal
Heating
Heating
Temp Meas
Temp Meas
Figure 5.7: Schematic overview of the exposure system for humans
sXh-900 at the Karolinska Institute.
5.4. EXPOSURE SYSTEM DESIGN
Funct
i
onGener
at
or
Si
gnalGener
at
or
PowerSuppl
yf
or
DCU and Laser
79
Dat
aAcqui
si
t
i
on
Uni
t
Di
gi
t
alCont
r
olUni
t
Laser(
810nm,1W )
Spl
i
t
t
er
,Coupl
er
Ampl
i
f
i
erSyst
em 1
(
wat
ercool
ed)
Ampl
i
f
i
erSyst
em 2
(
wat
ercool
ed)
Figure 5.8: Signal and control unit of the sXh-900 exposure system.
tectors (diode detectors were calibrated to measure the power balance at the feed-point of the corresponding antenna). A data unit
(Agilent, 34790A) was used to collect all sensor signals (forward and
reflected powers, temperatures of ceramic plates at heat loads) and for
the control of the amplifiers, DCUs and laser sources of both exposure
units.
The software controlled the data, so that a fully self-controlled exposure system was realized: the amplifiers were digitally switched on;
the output power was regulated via a gain voltage according to the
power measurements. The output power of the heat load lasers were
controlled by analog voltage and adjusted according to the temperature feedback. Additionally, a digital signal was provided to timely
synchronize exposure and the EEG recording.
The exposure system was designed to be used by non-technical
personnel. The front-end of the control software was easy to use,
the user needed only to select the test person pair, the corresponding
exposure condition (habituation, exposure one, exposure two) and
enter arbitrary text information, before pressing the exposure start
80
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
Figure 5.9: The exposure system control software has an easy graphical user interface. The user just enters the test person pair number, selects the exposure condition (habituation, exposure one, exposure two) and optional further text information. Clicking on the
start/pause/abort button controls the status of the exposure system.
button (Figure 5.9). The software was developed in collaboration with
a software contractor (Peter Sepan, Schaffhausen, Switzerland).
Every ten seconds during the exposure the computer stored all
data and experimental settings necessary for a full reproduction of the
exposure in an encoded file. When the encoded data was mailed to
the Foundation IT’IS for evaluation, a binary feedback was returned,
stating whether the exposure was conducted successfully. The decoded files (including the information on sham or real exposure) were
not shown to any of the Karolinska personnel before the end of the
study, strictly following the double-blind study requirements.
To minimize the potential influence from reflections, the other exposure system or potential external RF fields, RF absorbers (Eccosorb
VHP 8) were placed on three sides around the test persons (up to a
height of 1.80 m). To further minimize and avoid the influence of any
possible leakage of fields, the exposure conditions were set identical
for both participants during a session.
5.4. EXPOSURE SYSTEM DESIGN
81
Figure 5.10: Exposure rooms and headset with fixed antennas and
ceramic plate attached to the left ear lobe for heating.
The low frequency fields and also the background RF fields were
examined prior to the start of the study and subsequently confirmed with new measurements at three different times during the
study. Assessed were low frequency fields (Band 1: 5-2000 Hz and
Band 2: 2-400 kHz) and RFs (up to 2500 MHz) in the two exposure
rooms. The measurements were conducted through an external expert; the conclusive results showed low levels of measured fields (total field strength of less than 0.05 V/m), lower than levels commonly
found in homes. Hence, the requirement of less than 20 dB for the
equivalent psSAR10g for head tissue was met with a great margin.
Besides securing the exposure during the study, great care was also
taken to assure human safety at all times. This was achieved through
the following procedures:
• The output power at the antennas was continuously monitored
and controlled (update rate: every 10 s). The exposure was
aborted immediately and automatically whenever a value was
82
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
out of the safety interval by 3 dB.
• The watch-dog inhibited an independent exposure control without software control. Whenever the software or computer hung
up, the amplifier was automatically blanked by the watch-dog.
• For any malfunction of the software, the exposure level could
not considerably exceed the target SAR. The amplifiers operated
close to their maximum output power and were attenuated to
the desired output levels afterwards.
An overexposure was therefore prevented, and human safety was guaranteed.
5.4.5
Exposure Signal
The applied exposure signal represents a “worst-case” scenario. It included all ELF amplitude modulation components 2, 8, 217, 1736 Hz
of the GSM signal and the dose met the ethical standards for human
studies. The exposure was equivalent to that of frequent cell phone
users and did not exceed the ICNIRP guideline [25]. Details of the applied signal were decided within the PERFORM C steering committee
meeting in February 2004.
The active exposure was an 884 MHz RF signal and simulated
the electromagnetic fields induced in the head during a conversation with a GSM handset. The exposure was compliant with the
definition of the GSM signaling standard (burst width: 0.577 ms,
basic frame: 4.6 ms, [53]) and consisted of a temporal change between
non-DTX/DTX modes (Figure 5.11).
A conversation on the phone usually consists of a dialog, letting
both mobile users talk and listen while making a phone call. In order to have a close-to-reality exposure, the exposure signal refers to
a changing modulation between the non-DTX mode (active during
talking phases) and the DTX mode (active during listening phases).
The non-DTX (every 26th basic frame idle), corresponding to exposure during active talking, was set to a 10g-averaged peak spatial SAR
(psSAR10g) of 1.95 W/kg. The battery-saving DTX mode was active
during listening, i.e., only control signals and comfort noise are transmitted, resulting in a reduced time-averaged psSAR10g of only 12%
5.4. EXPOSURE SYSTEM DESIGN
83
of the non-DTX mode. The temporal change was random with an
average duration of 11 s for non-DTX and 5 s for DTX resulting in a
time-averaged psSAR10g of 1.4 W/kg.
Figure 5.11: Frame structure of the DTX and non-DTX modulations.
Both alternate actively during an actual GSM conversation, depending on whether one is listening or talking. The basic GSM frame for
the uplink has a period of 4.6 ms and contains 576 µs bursts including
12 µs rising and falling edges. Frames of the non-DTX mode, representing the majority of the frames, consist of one active slot per frame,
whereas the DTX frames are simulated by seven active slots.
84
5.5
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
Dosimetry and Validation
The dosimetry for this exposure setup was performed through a combination of numerical field simulations and experimental field sensing
techniques [79] and can be divided into two parts: the preliminary
dosimetry and final dosimetry.
The performed preliminary dosimetry encompassed evaluations of
(1) the antenna efficiency, including effects of EEG electrodes on the
induced fields, (2) setup parameters influencing the spatial peak SAR
(e.g., a tilted antenna) and (3) the exposure pattern of the SPA antenna from the Zurich studies in comparison to the SMP antenna of
the sXh-900 system. The final dosimetry went a step further and
evaluated the findings and uncertainty analysis in more detail (e.g.,
effects from variations). The preliminary dosimetry was performed
during my thesis prior to the start of the study; the final dosimetry
was completed until the end of the study by Clementine Boutry of
Foundation IT’IS. The results are published in [80] and are included
here in the final dosimetry section.
Tools and Techniques
The numerical evaluation was performed with the finite-difference
time-domain based simulation platform SEMCAD (SPEAG, Switzerland) on a Windows workstation. SEMCAD is enhanced with several special features for RF dosimetry and has been widely validated [81] [82]. The numerical results were then validated with the
near-field dosimetric assessment system DASY (SPEAG, Switzerland)
[83] equipped with the latest probe technology (ET3DV6, EF3DV2,
H3DV3). The DASY system includes - in addition to its precisely
controlled 3-axis robotic arm - a homogeneous generic twin phantom,
which is a standardized head model shaped like an average human
head [57]. The twin phantom can be filled with dielectric liquids to
perform SAR measurements.
Two simulation models were used: the homogeneous head model
SAM and the inhomogeneous model HREF-1. The homogeneous
model was used to compare simulation results with measurement results; shape and materials of the simulation model were identical to
the measurement model. The simulation results of the inhomogeneous
5.5. DOSIMETRY AND VALIDATION
85
head model allowed the detailed determination of the SAR in different
head tissues. The head model HREF-1 was derived from 121 magnetic
resonance images (MRI) of a human female subject of age 40. The
MRI scans have a slice separation of 1 mm in the ear region and 3 mm
at the other parts of the head [84]. Functional subregions have been
identified from MRI scans, so that 23 tissues were discretized. The
dielectric values of the 23 tissues were set according to [85]. Figure
5.12 shows the three models used: measurements in the generic twin
phantom with DASY4 (left), SAM simulations (center) and HREF-1
simulations (right) with SEMCAD.
Figure 5.12: Measurements with DASY in the twin phantom (left)
determine the SAR values for the comparison with the numeric results using the homogeneous SAM model (center) and the anatomical
HR-EF1 model (right).
The design of the simulation model was performed with great care.
In addition to the determination of SAR values of certain tissues, the
left/right asymmetries of the two head hemispheres were also to be
determined. Therefore the head model was divided into two evaluation volume halves. The head model was positioned in the center
of the coordinate system with the antenna radiating on the left head
side. The long side of the antenna was parallel to the head axis.
A 3-dimensional field sensor was placed around the antenna and the
head model. The field sensors were extended for simulations with a
larger distance between antenna and head model. 2-dimensional field
sensors were positioned vertical and horizontal to the exposure direction to record the fields around the head. Control sensors were placed
86
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
between antenna and head to control the stability of the simulation
run. To maximize the spatial grid resolution in the area of interest
a graded mesh was used: the grid resolution in the antenna region
was chosen as 1 mm in the direction of the radiation propagation and
1 to 2 mm in other directions. The grid resolution inside the field
sensor was 1 to 2 mm, which led to a fine grid. The grid resolution
around the field sensor was increased up to 10 mm. The antenna was
modeled with voltage edge sources and harmonically excited until a
steady state was reached. The dimensions of the simulation model
represented the exposure setup.
For the dosimetric measurements with DASY, the twin phantom
was filled with head tissue simulating liquid (HSL 900, electric conductivity σ=0.943 S/m, relative permittivity r =41.6, relative permeability µr =1.0, density ρ = 1000 kg/m3). The antenna was placed
below the twin phantom, operating at 884.6 MHz, CW. The distance
and position were exactly those of the setup headset. Measurements
were performed also with EEG electrodes glued to the outer side of
the twin phantom shell at the same locations as they were during the
study (10/20-EEG-system). Additionally, the heat load was placed
on the left ear lobe as well as half of the head set (see also Figure
5.13).
Preliminary Dosimetry
The tasks of the preliminary dosimetry were to set the spatial peak
SAR (10g) of the exposure system and evaluate major influence parameters (to exclude overexposure due to large uncertainties). Examined were the antenna efficiencies of all antennas, influence of EEG
cables and rotation of the antenna. Furthermore the exposure pattern
of the SMP antenna was compared to the SPA antenna of the Zurich
studies.
Five identical SMP antennas were built for the two headsets of the
study and as spares (SN 1002, SN 1003, SN 1004, SN 1005, SN 1006).
All antennas were tested for malfunction (no power loss/change, no
temperature increase, return loss < 10 dB) for 24 hours.
The efficiencies of the five antennas were measured with a DASY4
system using a twin phantom (Figure 5.13). Also examined were the
effects due to the EEG electrodes on the induced fields and therefore
5.5. DOSIMETRY AND VALIDATION
87
Figure 5.13: Dosimetric measurements with DASY4 including EEG
electrodes to determine the antenna efficiencies and an evaluation of
uncertainty parameters.
88
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
Antenna
a
SN
w/o electrodes
[W/kg/W]
1002
1003
1004
1005
1006
0.506
0.509
0.503
Efficiencya
with electrodes
[W/kg/W]
0.561
0.529
0.537
0.573
0.522
Efficiency: 10 g averaged peak spatial SAR per watt input power
Table 5.3: Antenna efficiencies (10g-averaged peak spatial SAR per
watt input power) with and without EEG electrodes.
also on the antenna efficiency. Table 5.3 shows the SAR (10g) efficiencies of the five antennas. The average spatial peak SAR (10g)
is 0.544 W/kg per Watt input power with EEG electrodes and
0.505 W/kg per Watt input power without electrodes. Although the
polarization of the antenna was perpendicular to the EEG electrode
wires, a small effect was visible: the antenna efficiency increased by
5 %. During this examination the electrodes did not have any direct
contact with the liquid. The result required further examinations in
the final dosimetry; at minimum, the influence on the SAR distribution due to the EEG wires needed to be covered in the uncertainty
evaluation (see also the detailed study on artifacts from EEG electrodes during RF exposures in [86]). During the study no influence of
the RF on the EEG signal was reported.
Slight rotations of the antenna can occur, although the headset
defines the orientation of the antenna to the test persons head. Examined was the influence of a 20 degree rotation of the antenna. The
antenna efficiency increased from 0.573 W/kg/W to 0.599 W/Kg/W,
which is a change of 5%. Tested was also the influence of the heat
load, placed between the antenna and the user’s head. The heat load
was designed to be RF immune. This was confirmed: the antenna efficiency changed less than 1 % from 0.573 W/kg/W to 0.579 W/kg/W.
The results were considered in the overall uncertainty analysis of the
5.5. DOSIMETRY AND VALIDATION
89
final dosimetry.
An additional study compared the exposure pattern of the Zurich
studies antenna SPA with the SMP antenna. Both antennas were
placed in the same position towards the DASY4 twin phantom. The
pattern shows good agreement (Figure 5.14); detail tissue parameters
were compared in the simulation model during the final dosimetry.
Figure 5.14: Exposure pattern comparison between the SMP antenna
(left) and Zurich studies antenna (right).
Final Dosimetry
The final dosimetry and uncertainty assessment considered data from
the preliminary dosimetry, the setup performance during the study
and results from simulations. As mentioned, the conclusive final dosimetric assessment was conducted by Clementine Boutry of
Foundation IT’IS and published in [80].
Twenty-three tissues were distinguished including subregions of
the brain. The inter-subject variations were estimated by scaling the
head model by 10 %, which was the maximum variation in brain volume measured by [87]. The positioning variations were assessed for
maximum deviations of 2 mm for the separation antenna/head, 10 mm
for movement to the front and back and 5 mm for up and down. The
90
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
Brain
Tissue
avgSARa
[W/kg] (SD)
psSAR1gb
[W/kg]
Variations
psSAR1g
[%]
Uncertainty
psSAR1g
[%]
Grey matter
White matter
Thalamus
0.20 (0.26)
0.18 (0.21)
0.18 (0.06)
1.78
1.12
0.30
11
11
16
16
15
18
a
Tissue averaged peak spatial SAR
b
1g averaged peak spatial SAR
Table 5.4: Summary of the exposure values induced in the brain tissues corresponding to 10 g averaged peak spatial SAR exposure settings of 1.4 W/kg. Provided are the values for the averaged SAR
plus standard deviation (SD) and the 1g-averaged peak spatial SAR
(psSAR1g). In addition, estimations of the inter-subject variations
as well as the assessment uncertainty are given (source: C. Boutry in
[80]).
variations due to different electrode locations were assumed to be 5 %
as assessed experimentally on the SAM head. The power drift was
determined to be less than 2 %.
The parameters included in the uncertainty analysis of the dosimetry were the uncertainty from (1) the head dielectric parameters, assessed by varying the relative permittivity and conductivity of the
dielectric parts of the head numerical model by 10 %; (2) the head
discretization and segmentation; (3) the effect of electrodes; (4) calibration; and (5) the power calibration.
The validation of the antenna revealed a difference in the nearfield of E- and H-fields between simulation and measurements of less
than 7 % (at the antenna distance). The difference of the dosimetric
evaluations employing the SAM head filled with head tissue simulating
liquid (permittivity 41.5; conductivity 0.97 S/m) was less than 4 %.
Both values were well within the combined uncertainty bounds.
The organ specific dosimetric results are summarized in Table 5.4,
and the SAR distribution on brain and skin is shown in Figure 5.15.
The exposure parameters correspond well to those of the setup
5.5. DOSIMETRY AND VALIDATION
91
used by [64]. At the applied time-averaged psSAR10g value (all head
tissues) of 1.4 W/kg, the peak spatial SAR of the gray matter averaged
over a cube of 1 g of 1.8 W/kg. The all-brain gray matter averaged
values were 0.2 and 0.18 W/kg for the thalamus.
Figure 5.15: Axial, coronal and sagittal views of the anatomical head
model HR-EF1 and their SAR distribution during exposure with the
SMP antenna. The picture on the right shows the SAR distribution
on the surface of the exposed left head side (source: C. Boutry in
[80]).
92
5.6
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
Conclusion
A flexible setup for the exposure of test persons with a GSM signal
at 884 MHz is presented. The Department of Health Science at the
Karolinska Institute in Stockholm, Sweden, successfully conducted a
human provocation study in the context of health risk assessment of
low-level exposure to the RF of mobile phones with this exposure
system. Examined were the effects of the GSM wireless communication signal on subjective symptoms, physiological reactions, alertness,
performance and sleep.
The setup enabled the exposure of the left head hemisphere and
was designed to maximize the exposure of brain tissues as it may occur
during actual usage of GSM phones. The headset system was based on
a low-weight, stacked micropatch antenna fixed on a headset, which
allowed the test person to move/rotate his/her head within a limited
area (± 40 cm) without a change of the exposure distribution. This
allowed flexible and comfortable exposure situations with durations of
several hours. The weight of the headset was compensated through an
adjustable hanging design, so that the test persons did not carry more
than about 100 g of weight. Additionally, a controllable, RF-immune
heat load was realized using optical heating of a ceramic plate placed
at the ear lobe at the exposed side. The computer-controlled signal
unit monitored and controlled the applied RF signal, the heat load
and further system parameters at all times. The system was designed
for double-blind exposure protocols and allowed the recording of EEG
during exposure.
The experimental dosimetry revealed an antenna efficiency for the
10g-averaged spatial peak SAR of 0.52 - 0.57 W/kg per Watt input
power. The test persons were exposed with the GSM signal cocktail
at the 10g-peak spatial SAR of 1.4 W/kg. The gray matter averaged
over a cube of 1g had the peak spatial SAR of 1.8 W/kg. The all-brain
gray matter averaged values were 0.2 and 0.18 W/kg for the thalamus.
The antenna showed a comparable exposure pattern to that used in
the Zurich studies. All simulations were validated with measurements.
The setup as well as the dosimetry satisfied the quality requirements for exposure setups and the ethical standards for human studies. The exposure level met the ICNIRP guidelines; human safety was
ensured at all times.
5.7. STUDY ABSTRACT: PERFORM C
5.7
93
Study Abstract: PERFORM C
Bioelectromagnetics, vol. 29, no. 3, pp. 185-196, 2008
The Effects of 884 MHz GSM Wireless Communication Signals on Headache and Other Symptoms: An Experimental
Provocation Study
Lena Hillert1,2 , Torbjorn Akerstedt3 , Arne Lowden3 , Clairy
Wiholm4,5 , Niels Kuster6 , Sven Ebert6 , Clementine Boutry6 , Scott
Douglas Moffat7,8 , Mats Berg9 and Bengt Birger Arnetz4,5
1
Department of Public Health Sciences, Division of Occupational
Medicine, Karolinska Institutet, Stockholm, Sweden
2
Department of Occupational and Environmental Health, Stockholm
Centre for Public Health, Stockholm, Sweden
3
National Institute for Psychosocial Medicine (IPM), Karolinska Institutet, Stockholm, Sweden
4
Department of Family Medicine and Public Health Sciences, Division of Occupational and Environmental Health, Wayne State University, Detroit, Michigan
5
Department of Public Health and Caring Sciences, Uppsala University, Uppsala, Sweden
6
IT’IS Foundation for Research on Information Technologies in Society, Swiss Federal Institute of Technology (ETH), Zurich, Switzerland
7
Institute of Gerontology, Wayne State University, Detroit, Michigan
8
Department of Psychology, Wayne State University, Detroit, Michigan
9
Department of Medical Sciences, Division of Dermatology, Uppsala
University Hospital, Uppsala, Sweden
Findings from prior studies of possible health and physiological effects
from mobile phone use have been inconsistent. Exposure periods in
provocation studies have been rather short and personal characteristics of the participants poorly defined. We studied the effect of
radiofrequency field (RF) on self-reported symptoms and detection of
fields after a prolonged exposure time and with a well defined study
group including subjects reporting symptoms attributed to mobile
94
CHAPTER 5. FLEXIBLE HUMAN EXPOSURE SETUP
phone use. The design was a double blind, cross-over provocation
study testing a 3 hour long GSM handset exposure versus sham. The
study group was 71 subjects age 18 to 45, including 38 subjects reporting headache or vertigo in relation to mobile phone use (symptom
group) and 33 non-symptomatic subjects. Symptoms were scored on
a 7-point Likert scale before, after 1 1/2 and 2 3/4 hours of exposure.
Subjects reported their belief of actual exposure status. The results
showed that headache was more commonly reported after RF exposure than sham, mainly due to an increase in the non-symptom group.
Neither group could detect RF exposure better than by chance. A belief that the RF exposure had been active was associated with skin
symptoms. The higher prevalence of headache in the non-symptom
group towards the end of RF exposure justifies further investigation of
possible physiological correlates. The current study indicates a need
to better characterize study participants in mobile phone exposure
studies and differences between symptom and non-symptom groups.
Chapter 6
Exposure Systems for
Large-Scale In Vivo
Laboratory GSM/DCS
Risk Assessment Studies
6.1
Abstract
For large-scale laboratory studies two types of efficient wheel exposure units for GSM-900 MHz and DCS-1800 MHz signals were developed. Each wheel exposure unit enabled the simultaneous exposure
of up to 65 mice; four wheel exposure units combine into an exposure system for multi-dose studies. The exposure system allows blind
study protocols and enables an easy daily handling routine (cleaning,
simple and rapid loading/unloading process). All relevant exposure
and environmental parameters are monitored and controlled at all
times; malfunctions are detected automatically. A detailed dosimetry and uncertainty analysis of the exposure systems was performed
with extensive measurements and simulations. The average SAR of
the whole-body as well as for each organ were determined for the
duration of the study, including an assessment of uncertainties and
95
96 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
variabilities.
The Fraunhofer Institute of Toxicology and Experimental
Medicine in Germany and Istituto di Ricerche Biomediche Antoine
Marxer in Italy conducted three bioassay/carcinogenic studies with
these exposure systems in the context of health risk assessment of exposure with mobile communication signals. In these large-scale studies
over 1500 mice were exposed at four dose levels (4 W/kg, 1.3 W/kg,
0.4 W/kg and Sham). The Good Laboratory Practice is compliant
studies were successfully conducted; technical quality control, support and maintenance during the entire time was ensured; 98 % of the
performed exposures were completed successfully.
6.2
Introduction
In the context of the health risk assessment of low-level exposure to
GSM and DCS mobile phone signals, the European project PERFORM A conducted several toxicological/carcinogenic studies with
mice and rats. Within this project four types of exposure systems
were developed.
The following partner laboratories participated in these large
European in vivo studies: Austrian Research Centers Seibersdorf
(ARCS) in Austria, RCC in Switzerland, Fraunhofer Institute of Toxicology and Experimental Medicine (ITEM) in Germany and Istituto di Ricerche Biomediche Antoine Marxer (RBM) in Italy. Furthermore an exposure study with a PERFORM A setup was conducted at Zhejiang University in Hang Zhou, China (different funding). The development, construction and testing of the in vivo exposure setups, as well as the installation and study maintenance and
de-installation were performed through Foundation IT’IS in collaboration with Schmid & Partner Engineering AG.
This chapter presents the mouse exposure systems used at Fraunhofer ITEM in Germany and RBM in Italy. The setups developed for
the exposure of rats are not presented here. Details about the exposure setups and studies with rats can be found in [88] [89] [90] [91] and
their study results in [92] [93] [94] [95]. The results of the exposure
studies with mice can be found at the end of this chapter.
6.3. OBJECTIVES AND REQUIREMENTS
6.3
97
Objectives and Requirements
It is a challenging task to develop in vivo exposure systems. Among
the major drawbacks of previous studies were the often poorly defined
and characterized exposure conditions [96]. Well-defined exposures in
animals having body sizes in the range of a wavelength are a difficult task; the uniformity of the SAR distribution inside the animals
requires special attention. Furthermore the setups must fulfill a number of constraints for biological in vivo experiments to be compliant
with the regulatory requirements of in vivo studies and the guidelines
of the partner laboratories.
6.3.1
Study Outline and Objectives
The PERFORM A mouse studies were planned as large scale studies
and foresaw the exposure of over 1500 mice (see also study plans [97]
and [98]). Three exposure setups were developed, two for the exposure
of mice at GSM 900 MHz and one for exposure at DCS 1800 MHz.
Each exposure setup had four different dose groups: high, medium,
low and sham dose group. Additional mice were held in a separate
rooms under the same living conditions (cage control group).
The mouse exposure setups were used to conduct:
• Two classical combined chronic toxicity and carcinogenicity
two-year bioassays with B6C3F1 mice at the Fraunhofer ITEM
in Hanover, Germany, at both GSM frequency bands, GSM-900
and DCS-1800.
• A co-carcinogenicity study using Eµ-Pim 1 transgenic mice at
RBM in Ivrea, Italy, at the GSM-900 frequency band. The
co-carcinogenicity study at RBM was set up as a replication experiment of the study carried out in Adelaide, Australia, which showed a two-fold increase in lymphoma cancer
as reported in [99]. (Compare also the study performed by
Utteridge et al. [100]).
The studies are outlined as long-term (life-time) studies. The mice
are all of the same age; at the beginning of the study in November
2001 they were 12 weeks old. Fraunhofer ITEM in Germany conducted the study with B6C1F1 mice, a commonly used mouse hybrid
98 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
strain in toxicity studies. RBM in Italy conducted the study with
Eµ-PIM 1 mice with express elevated levels of the Pim-1 transgene
in their lymphoid tissues, and as a result are predisposed to develop
lymphomas.
Prior to the main exposure period, the mice were acclimatized to
the housing conditions in the animal room and trained to become
accustomed to their exposure tubes for a restraint time of up to two
hours. During a four week period the mice were also trained using
the complete setup and exposed with the original GSM/DCS signal
as during the main study. This way the mice and personnel became
accustomed to the daily loading/exposure/unloading routine. During
this pre-study period the animals were observed daily and their health
condition checked.
The main study at Fraunhofer ITEM exposed all animals to
GSM-900 MHz and DCS-1800 MHz signals for two hours per day,
five days per week for 24 months. At RBM the mice were exposed with GSM-900 MHz for one hour a day, seven days per
week for 18 consecutive months. Both institutes kept an additional
group unexposed which served as the cage control. In addition
Fraunhofer ITEM conducted at the beginning of the main study (in a
third shift) 1-week and 6-week-studies in which the influence of the radiation on the micronucleus in B6C3F1 mice was examined. In all performed studies the animals were subject to pathological examinations,
necropsy, tissue and slide preparations, and complete histopathological evaluations.
The studies were performed blind to all project members. The
personnel at the laboratories were not able to assign dose groups to
wheel exposure units, and the technical personnel of IT’IS did not
have access to the animal group identifier and other data at the animal laboratory, including the results from the histopathology. The
key codes and identifier were not disclosed until completion of the
histopathological evaluation. The still blinded raw data were given
to the representatives of the sponsors prior to disclosure of the data.
The studies were performed in compliance with the principles of Good
Laboratory Practice (GLP).
6.3. OBJECTIVES AND REQUIREMENTS
6.3.2
99
Exposure and Environmental Requirements
The project PERFORM A started with the following performance
requirements, in order to establish defined exposure conditions in the
exposed mice:
• SAR (whole-body): The whole-body SAR shall be applied
in three different dose levels separated by a factor of three.
(The highest dose was set to a 10g-averaged whole-body SAR
of 4 W/kg as determined by pre-studies on thermal stress in the
animals. See also Chapter 2).
• SAR (organ): The average SAR of each organ shall also be
determined for the duration of the exposure.
• Homogeneity: The SAR distribution inside the animal resulting from the exposure should be as homogeneous as possible.
This value is referred to as uniformity of exposure within the
animal.
• Variability: The whole-body SAR and the distribution of the
local SAR should be as similar as possible in each animal. The
resulting mouse to mouse variation and entire life should be
smaller than ±2 dB.
• Exposure: blinded, self-detection of malfunctions.
Furthermore it was required to provide technical quality control,
support and maintenance during the period of experiments. This
required high reliability of the setups, a prepared rapid repair concept
in case of a malfunction (not to endanger the daily exposure) and a
solid control and monitoring of the exposures.
Further requirements:
• Setups shall enable the exposure of 1500 mice: 1170 B6C3F1
mice in 10 groups (at 900 and 1800 MHz: high, med, low, sham
and cage control group) and 500 µPim 1 mice in 5 groups (all at
900MHz: high, med, low, sham and cage control group).
• usage of non-toxic materials only (preferably stainless steel and
polycarbonate),
100 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
• no metal parts in the vicinity of the animal,
• completely sealed during exposure with respect to the applied
RF,
• well defined light conditions,
• easy to clean,
• easy and fast loading/unloading,
• 24h monitoring of all relevant environmental parameters (exposure, temperature, humidity, air flow, oxygen level),
• easy-to-use and self-detection of any malfunction (double sensors, watchdogs, etc.),
• blinded setup design,
• setup should allow GLP conform studies,
• negligible ambient ELF and RF fields in the laboratory,
• reasonable setup cost,
• setup needs to be designed to be operated by non-engineering
personnel and
• technical quality control, support and maintenance during the
entire project time must be ensured.
6.4
Setup Design
The task was to develop, construct, install and characterize mouse
exposure setups for 900 MHz and 1800 MHz, as well as to guarantee
steady technical quality control, support and maintenance during the
entire project, fulfilling the requirements of the study proposal.
After a short numerical study of various design concepts it
was decided to base the setup design on the concept of the
Adelaide study [99] and extend it for use within the PERFORM A
project. The Adelaide study used a mice exposure system developed
6.4. SETUP DESIGN
101
by Motorola [45], also referred to as the Ferris wheel setup. This
Ferris wheel setup has proven to be user-friendly and efficient; however, some modifications were performed to comply with the requirements of the PERFORM A study design and to improve the concept
from design findings during the Adelaide study. Furthermore the use
of the same exposure concept also has the advantage of achieving more
comparable results, especially since the RBM study was intended as
a verification study of the findings from Adelaide.
The PERFORM A mouse exposure system consists of four identical cylindrically-shape exposure units (wheels), each enabling the
whole-body exposure of up to 65 mice and a signal generation, control and monitoring unit. Each wheel exposure unit was adjusted to
a different exposure level, resulting in four dose levels: high, medium,
low and sham. Two types of wheels were developed for the experiments operating in the uplink mid band of the GSM-900 MHz frequency band (i.e., 902 MHz), the other at the uplink mid band of the
DCS-1800 MHz system (i.e., 1747 MHz) and apply a complex GSM
signal into the resonant structure of the exposure setup.
The PERFORM A mouse setup design differs from the Adelaide
Ferris wheel as follows:
• The wheel was enlarged in order to accommodate the larger
group size of the PERFORM A study, i.e., 65 instead of 40 animals per wheel.
• The location of the animals with respect to the short cut was
optimized to maximize uniformity of the exposure.
• The original stoppers had to be replaced, since during the prestudy they were found to be unpractical for fixing the animals
during daily operations. The new stoppers are made of lossy
material in order to reduce the effect of higher modes for small
animals.
• The relatively large holes for loading/unloading of the mice are
closed by metallic covers in order to increase the uniformity and
minimize leakage at 900 MHz and 1800 MHz. Electric contact
is provided by a 3-point fixation which presses the metal cover
towards the metal plate.
102 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
• Non-matched antennas were used: conical antenna for the
902 MHz and a bi-conical antenna for the 1747 MHz setups.
Matching was achieved by a three-stub tuner located in the water resistant black box on the backside of the wheels.
• The ventilation system was simplified.
• Temperature, humidity and oxygen sensors were integrated inside the ventilation canal.
• Two field sensors (short monopoles with Schottky-detectors)
were integrated in order to provide a redundant feedback system.
• Solid materials were used, i.e., stainless steel for the metallic
parts of the wheel, polycarbonate for the mice tubes. Only the
outer tubes and ventilation system are made of Plexiglas, and
the antennas are of brass. However, this was not seen as critical
by the participating laboratories, since the animals would not
come in contact with these parts. Special consideration was
given to easy cleaning.
Figure 6.1 shows the exposure room with the exposure system for
4 x 65 mice at Fraunhofer ITEM. The four exposure wheels are absolutely identical, and it is therefore impossible for the personnel to
recognize the different exposure doses. Outside the exposure rooms,
user-friendly touch screens were installed, enabling easy exposure control. Furthermore a status color signal outside of the room showed
the status of the exposure system. The hardware control rack with
signal generator, data logger and computers was placed in a storage
room above the exposure system, unaccessible to the personnel. The
systems for 900 and 1800 MHz were identical to the user; only the exposure wheels were slightly different. The system at RBM followed the
same concept as the ITEM 900 MHz setup: control of four exposure
wheels from the outside of the exposure room.
6.4. SETUP DESIGN
103
Figure 6.1: System for simultaneous exposure of 4 x 65 mice. The
four exposure wheels are identical, making it impossible for the user
to identify the exposure dose. The system is operated via an userfriendly touch screen outside the exposure room. The system state was
additionally indicated via a status color signal. The control rack with
signal generator, data logger and computer was placed in a storage
room above the exposure room.
104 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
6.4.1
Mechanical and Electrical Design
Figure 6.2 presents pictures of the PERFORM A system details and
Figure 6.3 shows the mechanical wheel design. The wheel exposure
units weigh more than 100kg and are made only out of stainless steel
and polycarbonate (except the antenna, which is made out of brass).
The exposure units were outlined as a resonant structure to maintain
the power requirements at an affordable level (i.e., using 200 W amplifiers instead of >2 kW for the expected amount of mice exposed in
parallel).
Each wheel consisted of two parallel circular stainless steel metal
plates (radius 775 mm) with a separation distance of 117 mm. A conical antenna in the 902 MHz setups (or a bi-conical antenna in the
1747 MHz setups) was placed in the center between the plates. The
antenna was fed from the backside of the setup from a water-resistant
black box containing the electronics for the environmental sensors (humidity, temperature and oxygen level) and field sensors and a tuner
(Emitec Triple Stub Tuner, 1878B) for matching the impedance of the
antenna. Encased between the plates were 65 cylindrical polycarbonate tubes (inner diameter 40 mm at wheel radius 700 mm) arranged
radially around the antenna. The front metal plate of the setup had
holes at the positions of the tubes so that they could be easily accessed. At the back plate were 65 smaller holes (diameter 10 mm),
which led to the ring channel of the ventilation system. Around the
edges of the plates, parallel metal bars (at wheel radius 755 mm) connected the two plates and functioned as a shortcut to make the setup
structure resonant.
The 902 MHz wheels additionally contained high-permittivity plastic bricks (Eccostock, K=15, size 70 x 15.88 x 120 mm3 , starting at
radius 675 mm) between neighboring tubes to increase the electrical
distance and therefore suppress distortion from neighboring animals.
Due to the different wavelength this was not necessary to apply at the
1800 MHz wheels. The whole wheel was installed on a movable wagon
to enable easier cleaning.
6.4. SETUP DESIGN
105
Figure 6.2: (A) cavity reflector bars, (B) conical or biconical antenna
in center, (C) temperature, humidity and oxygen sensors located in
the air ring channel, (D) RF amplifier and ventilation system with
air hoses located above each wheel exposure unit, (E) rear view with
water-sealed electronics box, air ring channel and RF power cable
connector, (F) and (G) sides and front view of the wheel exposure
units.
106 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
Figure 6.3: Schematics of the mouse wheel exposure unit. The antenna (900 MHz conical or 1800 MHz bi-conical) is placed in the center
and made out of brass. The mouse tube centers are at radius 700 mm,
shortcut bars are at radius 755 mm from the center of the antenna;
inner distance between the circular plates is 117 mm. 65 mouse tubes
and 253 shortcut bars are radially orientated in equidistance with symmetry axis at the bottom center of the wheel. The wheel is mounted
slightly tilted at three sides. The cavity is made entirely of stainless
steel and polycarbonate materials.
6.4. SETUP DESIGN
107
Ventilation System
A ventilation system for each wheel - consisting of a ventilator, a
ventilator box, air hose and ring channel - guided constant airflow
from the ventilator to the front side of each of the 65 tubes (see also
Figure 6.2 D and E). The air entered the tube at the nasal region
of the animal. The mice were placed around the exposure antenna
in tapered polycarbonate tubes with adjustable backstops (stopper).
The animal’s snout protruded from the anterior end of the tube, which
was connected to the ventilated part of the exposure unit by way of
a push fit, so that the air entered the nasal region of the animal. The
ventilated part of the exposure unit was operated at slightly positive
pressure with respect to the surrounding air. This ensured continuous
airflow passing through the animal’s breathing zone. The airflow was
adjusted to about 1 l/min. The necessary air flow was monitored in
two ways: the ventilator speed was monitored and also the oxygen
level in the ring channel. This ensured that the mice had enough air
during their presence in the setup.
Easy Handling
Since a large number of animals were simultaneously irradiated, the
exposure systems design was developed to be compact and easy to use
for loading and unloading the animals. The design and materials of
the setups were chosen for easy accessibility, cleaning and disinfection
of all parts of the setup. The mice tubes were made of polycarbonate
for sterilizing by an autoclave; the ventilation system consisted of a
transparent plastic so that any pollutant in the removable ring channel
could be seen and cleaned out. The realized concept enabled fast
loading and unloading of the setup and easy cleaning.
Mouse Restrainer Exposure Tubes
Newly developed restrainer tubes - which were devoid of metal parts
- provided the housing for the mice during the exposure. During
the loading procedure, each mouse was put into a mouse tube. The
ability of each mouse to move within its tube was minimized with the
adjustable backstop. A metal lid at the end of the mouse tube ensured
electric sealing of the setup after tube insertion to minimize losses and
108 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
to increase efficiency. Suitable for lifetime studies, the tubes come in
two different sizes, for young animals and for older animals. The
smaller tubes have an inner diameter of 31 mm, and the larger tubes
have an inner diameter of 39 mm. Figure 6.4 shows the developed
mouse tubes.
Figure 6.4: Left: mouse exposure tube with cap and stopper as used
during the study; right: Eµ-Pim 1 mouse in a small exposure tube
with a first version of the stopper.
Environmental and System Parameter Recording
The environmental parameters oxygen (Pewatron AG, FCX-MV CH),
humidity and temperature (Flach Elektronik AG, EE06-FT1A1) were
recorded at all times, even when no exposure takes place. These parameters were sensored in the ventilation ring canal close to the mice
on the backside of each wheel and therefore represent the actual condition for the animals (see also Figure 6.2 C and E). The sensor values
were managed in the water-resistant black box on the back side of
the wheels and sent via the data logger to the computer, where the
data were processed, stored and displayed. Furthermore the current
demand of all ventilators was recorded and a water flow sensor supervised the amplifier water-cooling circuit. Any measured values which
drifted outside the specified range triggered the system control to report a warning or alarm.
Exposure Control by Field Sensors
Since direct power measurements are not particularly reliable for exposure prediction in the case of resonant structures, the implemented
6.4. SETUP DESIGN
109
feedback system was based on field sensors. Short monopoles (length
optimized for desired dynamic range) with a Schottky diode (Advanced Control Components, ACSP-2663NZC15) were used as E-field
sensors. During exposure the field strength was measured quasicontinuously at two different places inside the exposure wheel. This
way a redundant feedback system was achieved.
A control loop system enforced a stable electromagnetic field inside the setup. Exposure levels were automatically adjusted if any
drift/mismatch between the actual field strength and the target field
strength occurred, ensuring stable exposure conditions inside the
setup. The two E-field sensors in the setup were calibrated using
the method of transfer calibration. The reference point was the center
between the plates right above the sensor location. A DASY3mini system (SPEAG, Switzerland) with the free-space E-field probe EF3DP6
(SN4004) was used for the field measurements.
6.4.2
Signal Generation, Monitoring and Control
of Exposure
The exposure signal and sensor signals were generated, distributed,
controlled and monitored by electronic equipment (concept in
Figure 6.5). The communication between the controlling PC and the
devices used a GPIB system. Digital signals were mainly driven and
read by a multifunction module (Agilent, type 34907A) of the data
logger (Agilent, Type 34970A). The amplifier (LS Elektronik, Sweden)
and Digital Control Unit (SPEAG, DCU) used standardized TTL signals.
The exposure signal was generated by a digital GSM/DCS signal
generator (Rohde & Schwarz, SMIQ02B). In order to enable switching
between DTX and non-DTX modes, a custom-made DCU for generation of the GSM frame structures was developed and manufactured
by Schmid & Partner Engineering AG. It was frame-triggered by the
RF signal generator.
In order to reduce costs, multi-usage of the amplifier was achieved
by generation of a four channel up-link GSM/DCS signal (time slots:
0, 2, 4, 6) of different amplitudes and utilization of a high-speed, highpower switch to sequentially assign each time-slot to another wheel.
110 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
Amplifier Gain and Temperature
Data
Acquisition
Unit
Amp On/Off
DTX/nonDTX
Watchdog
Alarm, etc.
Digital
Control
Unit
TTL Clock
High
Frame
Clock
Computer
Shutdown
RF Signal
Generator
Power
Level
Med.
Low
Switch
Amplifier
Two Field, Oxygen,
Temperature and
Humidity Sensors
Sham
Sensors
Antenna
RF Signal & Power,
Vent. System and
Room Light
Vent.
Tuner
Figure 6.5: Schematics of the PERFORM A exposure system: controlled through the computer the RF signal generator together with
the digital control unit (for DTX/non-DTX) generated the exposure
signal, which was then amplified. The amplifier was water-cooled to
minimize environmental noise and had a built-in high-speed, highpower switch, which distributes the signal to the corresponding exposure wheels. One slot of the eight slot TDMA signal of the RF signal
generator was used for each wheel, the slots in between were used for
switching to the next amplifier output channel, which was connected
to a different dose group wheel. Potential power reflections from the
antenna (i.e., due to unmatched tuning) were redirected into highpower terminators. The gain of the amplifier was controllable; the
E-field sensors in each wheel gave feedback of the applied power. Additionally, environmental and further system parameters were sensed
and returned via the data logger to the controlling computer. The
computer screen and the color signal light showed the status of the
system and displayed a warning or error in case of any malfunction.
6.4. SETUP DESIGN
111
An easy-to-use software was developed allowing fast, reliable and
highly automated handling of the exposure setup (details of the software design in [101]). The user just has to enter the average mouse
weight of all mice and for archival reasons an exposure group in case
of several daily exposure shifts. Figure 6.6 shows the graphical user
screen as it was displayed on the touch screen monitor of each system.
Figure 6.6: Easy system control via touch screen user interface: the
user just has to provide the current average mouse weight, exposure
shift group and inform the system when the loading/unloading is completed. The software and hardware is self-controlled and reports any
malfunction. The software was developed by Dr. Walter Oesch [101].
The software automatically controls the field strength and monitors the environmental sensor values. Small signal drifts are automatically compensated; large deviations from target values and any
malfunctions are indicated through a (red) status signal light, reported
to the user screen and also automatically via email to the technical
support to the Foundation IT’IS.
All hardware communication is logged, enabling full reconstruction
of every exposure occurring during the whole study. For maximized
data security, all recorded data were saved encoded at three different
112 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
physical locations: the experiment computer, a local file server and at
a file server at the Foundation IT’IS, to which the data were transmitted automatically each night. The data of the previous day was
evaluated at IT’IS, and the blinded evaluation results (ok or not-ok
feedback) were sent back to the study director. In the case of a system malfunction, IT’IS as well as the study director was automatically
informed via e-mail.
A blind study design requires that the user does not recognize details of the exposure system (i.e., dose groups). All technical data
evaluation and repairs were therefore only handled through the Foundation IT’IS. Key codes and identifier were not disclosed until completion of the biological (histopathological) evaluation. The blind study
was enabled through the data encryption and the identical wheel exposure unit design.
6.4.3
Exposure Signal
The wheel exposure units were designed to expose mice to the complex
GSM and DCS signals. The European Telecommunication Standard
Institute (ETSI) provides the technical standards for mobile communication at GSM 900 and DCS 1800 [102]. Both use the same digital
modulation, but use different carrier frequencies; DCS 1800 is occasionally also referenced as GSM 1800.
Therefore the same modulation signal on different carrier frequencies was applied for the two exposure systems. The carrier frequencies
for the exposure systems were chosen in the middle of the uplink band
at 902 MHz for the GSM 900 signal and at 1747 MHz for the DCS 1800
signal.
The exposure signal with its frequency and power modulation
cocktail was carefully chosen, broadly discussed and internationally
presented at conferences, before its final design was decided (see i.e.,
[7] [103] [104]). The applied signal is a 3-phase signal cocktail and
simulates all elements of an exposure as occurring during usage of a
mobile phone in the environment:
• Phase I: In this first phase (GSM Basic), the exposure condition
was as it occurs during talking, i.e., one active slot per basic
frame, with each 26th basic frame idle (Figure 6.7).
6.4. SETUP DESIGN
113
• Phase II: The second phase (GSM Talk) simulated a conversation, i.e., a random change between the non-DTX (average time
active: 2/3) and DTX (average time active: 1/3) modes.
• Phase III: The third phase (GSM Environment) simulated the
exposure during a conversation while moving in the environment. This included GSM features such as non-DTX, DTX,
power control, and handovers, etc., according to their statistical occurrence. The statistical parameters were derived from
measurements performed by [105] and [106].
At Fraunhofer ITEM all three phases with a duration of 40 minutes each were applied, leading to an exposure time of 120 min. At
the replication study at RBM only the first phase (GSM Basic) was
applied for a duration of 60 min.
The peak SAR level for all three phases is constant; however, the
resulting average SAR level varies: 100 % (GSM Basic), 70 % (GSM
Talk) and 26 % (GSM Environment). Figure 6.8 shows the applied
signal at Fraunhofer ITEM, the graphic was derived from actual exposure data.
The exposure signal itself was generated by a computer-controlled
vector signal generator (SMIQ 02B, Rohde & Schwarz, Germany)
combined with a custom-made frame-controlling unit (SPEAG,
Switzerland) and amplified by a custom-made water-cooled highpower amplifier (LS Elektronik, Sweden).
114 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
Non-DTX
4.6 ms
4.6 ms
26 frames
104 frames, 480 ms
DTX
Figure 6.7: Intermediate multiframe structure of the non-DTX and
DTX modes (composed of 104 basic frames with a duration of 480 ms).
One basic frame consists of eight time slots with a total duration
of 4.6 ms. In non-DTX mode, up to 100 out of 104 frames are active,
whereas in DTX mode transmission reduces to twelve active frames.
6.4. SETUP DESIGN
Phase I:
GSM Basic
Phase II:
GSM Talk
Phase III:
GSM Environment
DTX OFF (avg. duration: 11ms)
100
DTX Power Factor [%]
115
DTX OFF
80
60
DTX
40
20
DTX ON
Power Control Factor [%]
0
DTX ON (avg. duration: 5.5ms)
100
80
60
40
20
Power
Control
0
100 %
SAR Factor [%]
100
80
DTX and
Power Control
SARavg
70 %
60
40
26 %
20
0
0
20
40
60
Time [min]
80
100
120
88
89
90
Time [min]
Figure 6.8: Three phases of exposure signal: GSM Basic,
GSM Talk and GSM Environment.
The peak SAR level
for all three phases is constant; however, the resulting average SAR level varies: 100 % (GSM Basic), 70 % (GSM Talk) and
26 % (GSM Environment).
116 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
6.5
Dosimetry and Validation
The aim of the exposure system was to apply a well defined dose with
little variation; therefore each exposure system was optimized with
respect to the dosimetric performance within the given boundaries.
Especially in multi-dose studies it must be assured that each dose level
is clearly separated. This section shows the dosimetry and uncertainty
assessment that were performed; details about the applied dosimetric
methodology can be found in [15] and [88].
6.5.1
Dosimetric Method, Tools and Models
One of the goals was to establish a constant whole-body SAR exposure
over the study period. During the beginning of the exposure system
development it was decided to base the exposure feedback system on
E-field sensors since direct power measurements are not reliable for
exposure prediction in resonant structures. The antenna output power
was controlled through a power control function which provides the
relation between the target average whole-body SAR, the measured Efield at the field sensor positions and average mass of the mice loaded
into the exposure unit. This power control function was established by
numerical and experimental means and is given through the following
relation:
2
SARW B = |Einc | · η(m)
(6.1)
Where the absorption efficiency η(m) is a function of animal size
(weight). η(m) was determined by numerical means and verified with
dummy measurements as part of the preliminary dosimetry. The final
dosimetry included a more extensive experimental verification of η(m).
Since the dimensions of the wheel exposure unit were too large to
simulate in total with the available computers, the numerical evaluation was performed using sector models of the exposure setup, consequently the experimental part played a major role in the performed
dosimetry. The numerical evaluation enabled a rigorous uncertainty
and variability evaluation of the dosimetric parameters.
6.5. DOSIMETRY AND VALIDATION
117
Dosimetry, Uncertainty and Variations
The applied dosimetry, uncertainty and variability assessment was
based on extensive experimental measurements and numerical evaluations.
The exposure systems were calibrated on-site at the partner laboratories, having the advantage to include influences of transport and
installation. For these on site dosimetric measurements a new dosimetric method based on temperature measurements was developed
and applied. The experimental examinations were performed using
mouse dummies of various sizes and in different loading scenarios,
representing the range of animals weights occurring during the experiments. The target was the determination of the whole-body averaged SAR and corresponding E-field sensor readings, for the rela2
tion SARW B /|Einc | . Additional measurements were also performed
at the laboratories of the Foundation IT’IS for a general analysis of
the wheel setup performance and for worst-case scenarios (see also
Chapter 7).
The numerical dosimetry was performed in the following steps:
detailed numerically optimized models (with respect to accuracy and
numerical efficiency) of the two physical exposure setups at 900 MHz
and 1800 MHz were generated. For validation, these numerical setups
were compared with the performances of the detailed measurements
with dummy tubes of different sizes. Afterwards detailed anatomical mouse models were placed in the numerical setup, representing
averaged positions and postures in the restraining tubes. Numerical
results from scaled anatomical models represented the entire lifespan
of mice were further refined using different anatomical mouse models
to simulate different ages.
The uncertainty/variation considerations incorporate further aspects influencing the assessment of the SAR values. This includes
experimentally examination findings, setup performance parameters
from the actual study and results determined with numerical evaluations like variations of physical parameters [variations in body weight
(lifetime), anatomy (male/female, species)], mouse positions and postures [sector position (four positions), position in the restrainer tube
(front, back), posture (stuffed vs. non-stuffed), stopper contact (full
contact, no contact)], as well as the influence of the dielectric proper-
118 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
ties of the tissue parameters (σ, ).
The final dosimetry, variability and uncertainty assessment combined all results: (1) extensive dosimetric measurements, (2) variations from exposure data of the conducted study at the partner laboratory and (3) the results from the extensive numerical simulations.
This allowed the detailed assessment of the whole-body averaged SAR,
spatial peak SAR in the whole animals, organ-averaged SAR value and
spatial peak SAR in all organs.
Such a detailed dosimetry and uncertainty/variability assessment
is new for in vivo setups and was performed within the PERFORM A
project for the first time. It required extensive work to conduct such
a full dosimetric evaluation; my focus was the experimental part,
while the numerical parts were performed by Dr. Veronica Berdiñas,
Anja Klingenböck and Dr. Jürg Fröhlich. Presented here are the conclusive results for the exposure setup; details on the numerical evaluation can be found in [107] and [48]. A brief overview is given in here
for completeness.
To perform such a detailed dosimetry and uncertainty assessment
requires specialized measurement equipment, the development of representative mouse dummies for experimental evaluation, the development of a dosimetric method and the development of high resolution
numerical mouse phantoms of different sizes, genders, species.
Standard Tools: SEMCAD, DASY4 and EASY4
The numerical evaluation of the setup was performed using the FDTD
simulation platform SEMCAD Version 1.8 (SPEAG, Switzerland).
SEMCAD is a powerful dosimetric simulation tool; it allows efficient
3D real-time modeling of complex structures consisting of hundreds of
polyhedra; graded meshes with local resolution independent of animal
orientation and segmentation are automatically generated. SEMCAD
was extended for the automatic numerical evaluation of the averaged,
maximum, minimum SAR values for the whole-body and for every
organ and tissue.
The experimental evaluations were conducted using the nearfield scanner DASY4 (SPEAG, Switzerland) equipped with the latest SPEAG probes. Details on DASY can be found in [83], the
field/dosimetric probes are described in [108] and [109].
6.5. DOSIMETRY AND VALIDATION
119
All temperature measurements were conducted using the 4-channel
data acquisition system EASY4 (SPEAG, Switzerland) equipped with
SPEAG thermal probes. This enabled temperature measurements in
an RF environment with automatic recording of the temperature over
long time periods.
Figure 6.9: DASY4 examination of the wheel during the evaluation.
Temperature Method for Dosimetric Whole-Body SAR Determination
A simple and reliable dosimetric method was developed and used to
conduct efficient on-site dosimetric analysis and calibrations at the
animal laboratories. The temperature method is described in detail
in Section 7.6.2 and allows the assessment of the exposure uniformity
of whole-body average SAR as a function of position within the wheel
as well as the absolute determination of the whole-body averaged SAR
(SARW B ).
120 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
Figure 6.10: Direct and sophisticated temperature method for dosimetric whole-body SAR determination in mouse dummies using
EASY4 systems and RF immune temperature probes.
Experimental Models
Four different homogeneous mouse dummies were generated to represent different sizes or ages of the exposed mice over the life-time
study. The dummy sizes correspond to mice of the sizes: 20 g, 31 g,
37 g and 46 g. Figure 6.11 shows a direct comparison of a small/young
mouse with a 17 ml dummy.
6.5. DOSIMETRY AND VALIDATION
121
The mouse dummies consist of a plastic tube with a conically
tapered end at one side and a removable fitting cap at the other.
The plastic tubes (No. 1611, Semadeni, Switzerland) are made out of
polypropylene with a volume of up to 40 ml. The tube has an outer
diameter of 32 mm and a length of up to 90 mm. All dummies differ
only in length. The lengths of the tubes were shortened to generate
the different dummy sizes, so that the dummy is always completely
filled with as little air inside as possible. The dummy tubes were
filled with tissue simulating liquid. Different liquids were used for
each frequency, as the dielectric parameters are frequency dependent.
For further information about the experimental models used, see also
Section 7.6.4.
Figure 6.11: Left: young mouse in a small restrainer tube; right: small
dummy in a mouse tube.
Numerical Models
The numerical models, tools and simulations were generated and used
within the project [107] but not within this thesis. However, an
overview is provided since the numerical results are part of the dosimetry, uncertainty and variability assessment.
For the numerical evaluation new high resolution anatomical
mouse models were generated, since previous dosimetric studies have
shown that available models with axis resolutions of larger than 1 mm
are insufficient to achieve reliable estimates of the absorption [110].
Therefore high resolution pictures of the entire animal were generated.
For this purpose animals were placed in an exposure tube and frozen
in microtome liquid. The frozen blocks containing the animal were
moved over an adapted electro planer and tissue pictures were taken
with a high resolution scanner. Graphical processing of the segmentation models of these pictures generated high resolution anatomical
122 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
3D mouse models which were used in the numerical evaluation.
The following three anatomical mouse models were generated, each
with 46 different tissues: (1) mouse OF1 (female, 19.7 g, slice separation 0.52 mm), (2) mouse OF1 (male, 37.2 g, slice separation 0.74 mm)
and (3) mouse Pim1 (male, 51.9 g, slice separation of 0.48 mm).
Figure 6.12 shows on the left side a discretized mouse model with
its tissues and on the right side the three anatomical mouse models.
OF1(
Femal
e)
OF1(
Mal
e)
PI
M1(
Mal
e)
Figure 6.12: Left: mouse model with discretized tissues; right: the
three anatomical mouse models OF1 male, female and Pim1 male.
In addition to the three anatomical mouse models, four scaled
mouse and two stuffed mouse models were generated: mouse scaled
OF1-29 (female, 29.1 g), mouse scaled OF1-31 (male, 30.9 g), mouse
scaled Pim1-61 (male, 60.9 g) and mouse scaled Pim1-69g (male,
69.1 g); the two stuffed mouse models were: mouse stuffed OF1-19
(female, 19.6g, slice separation 0.36 mm) and mouse stuffed OF1-37
(male, 37.1 g, slice separation 0.67 mm). Furthermore - to examine
anatomical differences - the three anatomical mice were all scaled to
the same weight of approximately 31 g.
Due to its large dimensions, it was only possible to simulate sectors of the wheel exposure unit. Figure 6.14 shows the models used for
the 900 MHz and the 1800 MHz setups. Figure 6.13 shows examples
of simulation models: a homogeneous dummy for comparison to measurements on the left side and a stuffed mouse model in an exposure
tube on the right side. Further details about the mouse models used
and the numerical simulations performed can be found in [107].
6.5. DOSIMETRY AND VALIDATION
123
Figure 6.13: Left: 37 ml liquid filled dummy in a large tube; right:
25.2 g stuffed mouse with stopper in a small tube (source [107]).
900MHzSet
upSect
orModel(
wi
t
hhi
ghper
mi
t
t
i
vi
t
ybr
i
cks)
1800MHzSet
upSect
orModel
Figure 6.14: Numerical sector models of the 900/1800MHz setups
(source [107]).
124 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
6.5.2
Initial Setup Challenges
At installation of the mouse exposure systems the occurrence of higher
modes was detected. The challenge during the experimental evaluation of the PERFORM A prototype wheel was an initial strong coupling effect between neighboring mice of as large as 8 dB at 900 MHz,
which was solved/minimized through the introduction of high permittivity dielectric blocks between the outer mouse tubes. This initial
prototype wheel was not examined for higher modes; the revised prototype became available in February 2001. At that time the study plan
for the start of the main study was finalized and the large amount of
mice ordered; the installation of the exposure systems was scheduled
for May. Due to the time pressure, the second prototype was mainly
confirmed for the effectiveness of the high-permittivity blocks, before
the series of the exposure system wheels were built. The higher modes
were then discovered during installation of the exposure systems at
the partner laboratories.
In personal communication from the developer of the original
Ferris wheel setup, significant higher modes were not detected for the
original exposure unit [45] [111] [112]. To examine the differences,
a comparison of the Ferris wheel setups was decided, which is presented in Chapter 7. Higher modes are excited by any asymmetrical
changes; the magnitude is a function of asymmetries of the mechanical
construction as well as of animal weight, posture and position within
the tubes. Higher modes lead to a significant non-uniformity of the
whole-body exposure with respect to position as well as an increased
uncertainty of the absolute SAR assessment.
Immediate modifications of the original design were developed to
reduce the resulting uncertainties as much as possible. The constraints
were affordability, realizability within three months and that the modifications could be undertaken on-site. The wheel modifications were
completed in October 2001, prior to the start of the main study. Of
great concern was also that the field sensor based feedback concept
failed due to the dependence of the field at the sensor location in the
presence of higher modes. Initially the E-field sensors were placed at
two different radius and angular positions (positions 14 and 23 counting clockwise from the top), each of which lay at an E-field maximum
of the standing wave. The sensor positions were changed.
6.5. DOSIMETRY AND VALIDATION
125
The following improvements were introduced:
• A dosimetry method was developed which is fast and suitable
for analyzing the wheel exposure units accurate enough to conduct a SAR calibration on-site (see temperature method in
Section 7.6.2).
• The original stoppers were replaced by stoppers made of PFTE,
enriched with 25% carbon. This helped to reduce the quality
factor of the wheel for small mice.
• The sensor positions were moved as close as possible to the antenna in order to reduce sensitivity to reflected fields.
• An rotation schema for the mouse exposure was proposed to
leverage the position dependent effect on the SARW B over time.
An on-site SAR calibration of all wheels via the new temperature
method was conducted at the partner laboratories. The power transfer
function was modified accordingly: the originally numerical assessed
absorption efficiency η(m) (see Equation 6.1) for the symmetric wheel
was extended for the new sensor positions and the existence of nonsymmetric higher modes and replaced through the new absorption
efficiency η 0 :
η 0 (mmouse ) = ν · ηsim (mmouse )/ηsim (mdummy )
(6.2)
Whereby ν is the measured SAR averaged over all considered
2
mouse positions divided by the weighted averaged value of Einc
at
the two new sensor positions. ηsim (mdummy ) is the numerically assessed efficiency for the dummies.
The pre-study was performed at selected and marked mouse positions in the upper third of the wheels. The exposure conditions remained acceptable, since most of the mouse positions were filled with
dummy mice. Their dosimetric values can be found in reports [113]
and [114]. Presented here is the dosimetry and uncertainty assessment
for the modified exposure system.
126 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
6.5.3
Results of
Variations
Dosimetry,
Uncertainty
and
The final dosimetry consisted of the assessment of the whole-body
average as well as the organ average specific absorption rate SAR
for the different exposure phases. The ultimate target was to have
constant and uniform whole-body and organ-specific SAR during the
entire exposure period. Due to physical restrictions this target could
not be achieved. Hence the objective was to provide an exposure
which is as constant and as uniform as achievable and to provide a
comprehensive dosimetric analysis. The results are presented here.
Dosimetry, uncertainty and variability assessment were conducted using the new methodology as described in [15]. Data
were included from the numerical evaluations conducted by
Dr. Veronica Berdinãs [107], my experimental setup evaluations and
my setup performance summary of the systems at the animal laboratories [115] [116] [117]. Further considered were influences from the
study: a rotational scheme with a weekly rotation of the mouse positions was applied, so that each mouse was exposed at all positions
in the wheel during the course of the life-time study. The variability
analysis respects these influences and differs in parameters which have
an instant and which have a life-time influences. The parameters for
the variability assessment are listed in Table 6.1 and were considered
according to [47] with k=1. Table 6.2 shows the parameters considered in the uncertainty evaluation (also according to [47], but with
k=2).
The combined results are summarized in the following Tables 6.3
to 6.8 and include:
• Whole-body averaged SAR for the entire exposure group and
exposure period as well as the estimation of the uncertainty of
this value,
• development over time of the weekly whole-body SAR for the
entire exposure group,
• assessment of the organ averaged SAR for the entire exposure
group and the estimation of the uncertainty of this value,
• uncertainty of the exposure separation between groups,
6.5. DOSIMETRY AND VALIDATION
127
• assessment of the instant single animal variation of the wholebody and organ exposure with respect to the whole-body average SAR,
• assessment of the single animal variation of the whole-body and
organ exposure averaged over the entire exposure period with
respect to the weekly whole-body average SAR,
• assessment of the spatial peak SAR (5mg, 0.5mg) as well as its
variation.
Parameter: Variation Assessment
Wheel sensor linearity
Sensor value dependence (weight)
Variation of weight
Effect of neighbors (weight)
Effect of neighbors (position in tubes)
Deviation from E-inc
Weight function SAR/E2
Wheel homogeneity
Position in wheel
Posture
Position in tubes
Influence stopper
Sweat/urine
Instant
Life-Time
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Table 6.1: Parameters considered in the variability analysis of the
mouse exposure system. Due to the applied weekly rotational schema
during the study, less parameters become relevant for the variability
during life-time.
128 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
Parameter: Uncertainty Assessment
Transfer calibration (calibration E-field probe,
position E-field probe (±10 mm), sensor linearity)
Setup model
Anatomical model
Dielectric parameters (epsilon (±10%), sigma (±10%))
Grid resolution
Influence of stopper
Table 6.2: Parameters considered in the uncertainty analysis of the
mouse exposure system.
Dosimetry and Uncertainty of 900 MHz Mouse Exposure
System
Table 6.3 provides the dosimetry for the weekly whole-body averaged
SAR for the entire exposure group, including the uncertainty. The
whole-body averaged SAR development over time of the entire group
was kept constant during the entire period of exposure. The spatial
peak SAR averaged over 5 mg and 0.5 mg are summarized for different
anatomical models in Table 6.4. Table 6.5 shows the average ratio of
the average organ SAR to whole-body SAR within the wheel. The
targeted exposure separation by a factor of three between the dose
groups could be maintained with an uncertainty of less than 1 dB.
6.5. DOSIMETRY AND VALIDATION
129
900 MHz Exposure System
Exposure
Dose
high
medium
low
sham
a
b
Phase I
[W/Kg]
SARW B
Phase II
[W/Kg]
Phase III
[W/Kg]
Uncertainty
SD (k=2)
[dB]
4.0
1.3
0.4
0
2.8
0.93
0.31
0
1.04
0.35
0.11
0
±2.6
±2.6
±2.6
0
Variations
SDa
SDb
[dB]
[dB]
±2.2
±2.2
±2.2
0
±1.2
±1.2
±1.2
0
Instant standard deviation.
Lifetime averaged standard deviation.
Table 6.3: Dosimetry for the weekly whole-body averaged SAR for
the entire exposure group, including the value uncertainty for the
900 MHz exposure. The phases correspond to the three applied signal
scenarios (see Section 6.4.3).
900 MHz Exposure System
Tissue
SAR5mg
SAR0.5mg
a
b
Spatial Peak SAR / SARW B
(Group&Lifetime Avg)
[dB]
Uncertainty
SD (k=2)
[dB]
17.9
19.3
±6.0
±5.7
Variations
SDa
SDb
[dB]
[dB]
±4.4
±3.3
±2.6
±1.8
Instant standard deviation.
Lifetime averaged standard deviation.
Table 6.4: Estimation of spatial peak SAR for the 900 MHz exposure
(source: [107]).
130 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
900 MHz Exposure System
Tissue
Bladder
Blood
Bone Marrow
Bones
Brain
Cartilages
Eyes
Fat
Gland, Lacrimal
Glands
Heart
Kidneys
Large Intestine
Liver
Lungs
Muscles
Nerves
Oesophagus
Pharynx
Skin
Small Intestine
Spinal Cord
Spleen
Stomach
Tongue
Trachea
a
b
SARorgan /SARW B
(Group&Lifetime Avg)
[dB]
-0.6
1.6
-4.6
-7.5
-2.1
-1.7
-1.6
-7.2
-1.6
0.6
0.8
0.7
0.2
0.2
1.8
0.1
-2.9
0.5
-0.9
-0.7
2.7
-1.2
0.6
0.0
0.2
-0.9
Uncertainty
SD (k=2)
[dB]
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
4.0
2.7
3.4
2.8
2.6
2.7
2.6
3.0
2.6
2.9
3.0
2.7
2.8
2.7
2.8
2.6
3.8
3.2
2.7
2.8
2.9
2.6
3.2
2.9
2.8
2.6
Variations
SDa
SDb
[dB]
[dB]
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
3.6
2.9
4.2
2.8
2.8
2.5
2.4
2.7
2.3
3.8
3.8
3.3
3.3
2.6
3.6
2.6
2.6
2.7
2.3
2.7
3.2
2.7
3.5
3.3
2.7
2.7
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
1.6
2.1
3.2
2.1
1.3
1.7
1.3
1.5
1.3
1.8
2.2
1.6
1.7
1.7
1.7
1.7
1.8
2.0
1.3
1.9
1.9
1.7
2.4
2.3
1.3
1.3
Instant standard deviation.
Lifetime averaged standard deviation.
Table 6.5: Organ averaged SAR exposures as a group average and with
consideration of location and temporal variations for the 900 MHz
exposure (source: [107]).
6.5. DOSIMETRY AND VALIDATION
131
Dosimetry and Uncertainty of 1800 MHz Mouse Exposure
System
Table 6.6 provides the dosimetry for the weekly whole-body averaged
SAR for the entire exposure group, including its uncertainty. The
whole-body averaged SAR development over time of the entire group
was kept constant during the entire period of exposure. The spatial
peak SAR averaged over 5 mg and 0.5 mg are summarized for different
anatomical models in Table 6.7. Table 6.8 shows the average ratio of
the average organ SAR to whole-body SAR within the wheel. The
targeted exposure separation by a factor of three between the dose
groups could be maintained with an uncertainty of less than 1 dB.
1800 MHz Exposure System
Exposure
Dose
high
medium
low
sham
a
b
Phase I
[W/Kg]
SARW B
Phase II
[W/Kg]
Phase III
[W/Kg]
Uncertainty
SD (k=2)
[dB]
4.0
1.3
0.4
0
2.8
0.93
0.31
0
1.04
0.35
0.114
0
±2.2
±2.2
±2.2
0
Variations
SDa
SDb
[dB]
[dB]
±1.6
±1.6
±1.6
0
±0.8
±0.8
±0.8
0
Instant standard deviation.
Lifetime averaged standard deviation.
Table 6.6: Dosimetry for the weekly whole-body averaged SAR for
the entire exposure group, including the value uncertainty for the
1800 MHz exposure. The phases correspond to the three applied signal
scenarios (see Section 6.4.3).
132 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
1800 MHz Exposure System
Tissue
SAR5mg
SAR0.5mg
a
b
Spatial Peak SAR / SARW B
(Group&Lifetime Avg)
[dB]
Uncertainty
SD (k=2)
[dB]
8.9
10.1
±2.3
±3.2
Variations
SDa
SDb
[dB]
[dB]
±2.3
±2.8
±1.7
±2.1
Instant standard deviation.
Lifetime averaged standard deviation.
Table 6.7: Estimation of spatial peak SAR for the 1800 MHz exposure
(source: [107]).
6.5. DOSIMETRY AND VALIDATION
133
1800 MHz Exposure System
Tissue
Bladder
Blood
Bone Marrow
Bones
Brain
Cartilages
Eyes
Fat
Gland, Lacrimal
Glands
Heart
Kidneys
Large Intestine
Liver
Lungs
Muscles
Nerves
Oesophagus
Pharynx
Skin
Small Intestine
Spinal Cord
Spleen
Stomach
Tongue
Trachea
a
b
SARorgan /SARW B
(Group&Lifetime Avg)
[dB]
-2.9
5.2
-6.1
-7.5
1.0
1.5
-1.4
-8.6
-0.8
3.1
4.3
-1.4
0.3
1.0
4.5
0.5
-0.6
4.0
0.8
-2.7
1.8
1.3
-4.3
-1.1
0.6
3.1
Uncertainty
SD (k=2)
[dB]
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
3.0
2.3
3.2
2.3
2.3
2.5
2.4
2.3
2.4
2.7
2.3
2.8
2.6
2.6
2.5
2.2
3.1
3.1
2.9
2.3
2.5
2.5
3.0
2.4
3.3
2.3
Variations
SDa
SDb
[dB]
[dB]
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
3.5
2.3
3.2
2.3
3.1
3.0
2.9
2.0
2.9
2.8
2.7
2.9
2.2
2.2
2.5
2.1
2.7
2.4
2.8
2.0
2.1
2.3
2.3
2.7
2.6
2.8
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
1.6
1.4
2.4
1.7
2.0
2.0
2.1
1.2
2.1
1.5
1.1
2.1
1.1
1.3
1.3
1.3
2.0
1.9
2.1
1.2
1.1
1.7
1.1
1.8
2.1
2.2
Instant standard deviation.
Lifetime averaged standard deviation.
Table 6.8: Organ averaged SAR exposures as a group average and with
consideration of location and temporal variations for the 1800 MHz
exposure (source: [107]).
134 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
6.6
Setup Performance during Study
The PERFORM A mouse exposure systems for 900 MHz and
1800 MHz operated successfully during the 18- and 24-month studies at RBM and Fraunhofer ITEM. All systems had an on-time of
more than 98 %. Figures 6.15 and 6.16 show the summarized sensor
readings of the three systems, and Table 6.9 lists the occurring incidents and failures, which required technical maintenance. A detailed
report which includes the performance of each wheel can be found in
[115], [116] and [117].
The average SAR offset level for each wheel from its target value
was less than 1 dB at all systems. The measured deviations in SAR
remained below 0.5 dB. On April, 29th 2002 at Fraunhofer ITEM
900 MHz the only incident occurred where the maximum deviation
from the target SAR was exceeded. The RF switch in the amplifier failed and for a brief moment (<1 minute) group A in wheel 4
was exposed with 6.9 dB more SAR than the targeted amount until
the system detected the failure and automatically shut down. Apart
from this single event, all other technical incidents were automatically
detected. The SAR offset and deviations of the exposure system performance were considered in the overall SAR analysis and uncertainty
determination of the previous section.
The environmental sensor values recorded an average temperature
for the exposure session of (22±3) ◦ C, an oxygen level of (20.4±0.5) %
and a humidity range between 21 % and 76 % during the entire study
time. There were some incidents where EMI issues disturbed the
environmental sensor readings; however due to the redundant system
design this was uncritical. The oxygen sensor readings were backed up
through a speed sensor of the ventilators; temperature and humidity
were additionally monitored from the partner laboratories. The ventilator speed was constant at all times, the mice were always supplied
with enough oxygen and the temperature and humidity was within
the target zone at all times.
The RBM 900 MHz exposure study was performed between
October, 24th 2001 and June, 12th 2003. The exposure took place
seven days per week. Exposed were two groups in two shifts with a
GSM basic signal for a duration of 60 min. On nine exposure days no
exposure was possible due to uncritical system failures; this resulted
6.6. SETUP PERFORMANCE DURING STUDY
135
in an on-time of more than 98 % during the 18-month study.
The Fraunhofer ITEM 900 MHz exposure study was performed
between November, 1st 2001 and October, 31st 2003. The Fraunhofer
ITEM 1800 MHz exposure study was performed between November,
8th 2001 and November, 7th 2003. Both exposures took place on
five days per week. Exposed were in each system two groups of mice
per wheel in two shifts applying all three phases (each 40 min) of
the exposure signal for a duration of 120 min. The 900 MHz system
showed failures on 13 days, such that 18 of the scheduled 26 exposure
shifts could not be performed or showed failures. Though this system
showed the largest number of technical incidents, it had an on-time
of 98% during the 24-months. The 1800 MHz system worked well
and had only one unplanned non-exposure day due to a failure. The
1800 MHz system had an on-time of more than 99 % during the 24month study.
Figure 6.15: Summary of field and environment sensor recordings for
all exposure wheels at RBM.
136 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
Figure 6.16: Summary of all field and environment sensor recordings
for all exposure wheels at Fraunhofer ITEM 900 (upper graphs) and
Fraunhofer ITEM 1800 (lower graphs).
6.6. SETUP PERFORMANCE DURING STUDY
Date
First
Exp.
Second
Exp.
137
Remark about Incident
900 MHz Exposure System at
11.11.2001
no exp
no exp
12.11.2001
no exp
no exp
13.11.2001
no exp
no exp
15.12.2002
no exp
no exp
16.12.2002
no exp
no exp
17.12.2002
no exp
no exp
01.05.2003
no exp
no exp
02.05.2003
no exp
no exp
03.05.2003
no exp
no exp
RBM
DCU failure
DCU failure
IT’IS service
Amplifier problem
Amplifier problem
IT’IS service
Amplifier failure
Amplifier failure
IT’IS service
900 MHz Exposure System at
07.11.2001
ok
45 min
25.01.2002
55 min
ok
28.01.2002
no exp
no exp
26.04.2002
high dev. ok
Fraunhofer ITEM
EMI at T+H sensor wheel 1
EMI at O2 sensors in wheel 3
EMI at sensors
RF switch failure lead to dev from target
level for group A in wheel 1: -1.6 dB
and wheel 4: +2.6 dB.
Broken RF switch in amplifier lead to dev
for group A wheel 1: -4.7 dB from target
level and wheel 4: +6.9 dB.
Repair by IT’IS
EMI problems, O2-sensor, wheel 3
EMI problems, O2-sensor, wheel 3
O2-sensor, wheel 3 switched off.
29.04.2002
no exp
no exp
30.04.2002
29.07.2002
30.07.2002
31.07.2002
- 1.11.2002
25.06.2003
09.07.2003
10.07.2003
11.07.2003
15.07.2003
no exp
ok
30 min
ok
no exp
25 min
ok
ok
no exp
60 min
115 min
no exp
115 min
no exp
ok
ok
no exp
ok
Ventilator defect.
Power supply O2-sensor
Power supply O2-sensor
Power supply O2-sensor
Power supply O2-sensor
wheel
wheel
wheel
wheel
4
4
4
4
defect.
defect.
defect.
defect.
1800 MHz Exposure System at Fraunhofer ITEM
6.12.ok
ok
EMI of temperature sensor wheel 1,
17.12.01
no effect on exposure.
26.06.2003
no exp
no exp Ventilator failure, replaced.
Table 6.9: List of days with exposure failures during the PERFORM A
study.
138 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
6.7
Conclusion
The PERFORM A mouse studies at 900 MHz and at 1800 MHz were
successfully conducted with the presented exposure systems. In these
large-scale studies over 1500 mice were exposed in four doses (4 W/kg,
1.3 W/kg, 0.4 W/kg and sham). The technical quality control, support
and maintenance during the entire time was ensured, with 98 % of
the performed exposures completed successfully. The system worked
reliably and malfunctions were detected automatically.
The exposure systems were successfully blinded; the laboratory
personnel did not identify the different dose groups at any time. Only
non-toxic materials were used (stainless steel and plastic) for the construction of the setups. The design considered the light requirements
of the mice and eased the daily handling routine, which included cleaning and a simple and rapid loading/unloading process. The exposure
system was successfully designed to be operated by non-engineering
personnel. All relevant exposure and environmental parameters (field,
temperature, humidity, air flow and oxygen level) were monitored.
The systems were sufficiently sealed: only negligible ambient ELF
and RF fields in the laboratory were detected; laboratory personnel
and mice in other dose group wheels were safe. GLP conformal studies
are supported by the setup.
The dosimetry and uncertainty analysis of the exposure systems
were performed with extensive measurements and simulations. The
average SAR of the whole-body as well as for each organ was specified for the duration of the exposure. The dosimetric parameters
reported also include the uncertainty of the exposure separation between groups, instant single animal variation of the whole-body and
organ exposure with respect to the whole-body average SAR, single
animal variation of the whole-body and organ exposure averaged with
respect to the weekly whole-body average SAR and the spatial peak
SAR (5mg, 0.5mg) as well as their uncertainties.
During the experimental evaluation phase it was detected that the
wheel design has a drawback due to the occurrence of higher modes.
A newly developed temperature method allowed an on-site dosimetry
and fast experimental characterization of the setup. This enabled the
application of countermeasures for maintaining the SAR distribution
inside the animal as even as possible. The Ferris wheel design has the
6.7. CONCLUSION
139
disadvantage of any multi-mode setup: higher modes can be excited
by the smallest asymmetries (Q of the cavity is high, small absorbing cross section). Measures such as lowering the Q value, reducing
the interaction between animals, and higher-mode stirrers considerably improve the performance of this kind of setup. The inter-animal
variation of the averaged exposure over lifetime is further lowered by
following a strict rotational schema in the exposure protocol.
The uncertainty/variability analysis examined the uniformity
within the animal model and revealed further interesting findings:
the variation of the average organ SAR values within the body can
reach up to 8.6 dB from the whole-body SAR; the spatial peak SAR
of the whole-body averaged SAR can reach up to 17.9 dB at 5 mg and
19.3 dB at 0.5 mg. The simulations also showed that the variations
within animals are highly influenced through the exposure scenario.
Depending on the biological endpoint under investigation, high variation of the averaged organ SAR can significantly influence the outcome
of the statistical analysis. In vivo health risk assessment studies of
electromagnetic fields should therefore always include organ-specific
information of the specific absorption rate and thermal load.
Depending on the aim of the study, wheel setups are still very
useful within their SAR uniformity limitations. They have a high
loading volume, high SAR efficiency due to the cavity design and deliver a good cost/SAR ratio. As presented in Chapter 7, wheel setups
can achieve a high uniformity as demanded by the requirements for
in vivo whole-body exposure studies through applying certain design
concepts (small units, high permittivity bricks, rotational scheme, homogeneous load). However, the dosimetry of an exposure setup should
always include an analysis of the whole-body and also an analysis of
the detailed organ-specific SAR, including uncertainties and variations.
For the outcome of the PERFORM A mouse study results, the
occurring higher modes were not relevant; no negative effect findings
were detected, and the dosimetry specified the whole-body and organ
specific SAR including uncertainties for the study. The performed
study was successfully conducted and completed.
140 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
6.8
Study Abstracts: PERFORM A
The described in vivo exposure setups were used in the large scale
risk assessment exposure study PERFORM (A1 and A4) as well as
an additional Micronucleus study. PERFORM A consists of 6 experiments divided into 5 parts, conducted under GLP according to the
EU directives 83/571/EU Apx 3 and 91/507/EU:
• PERFORM-A1: Two combined toxicity / carcinogenicity studies of 900 MHz GSM & 1800 MHz DCS exposures in B6C3F1
mice (NTP-like). This study was performed at Fraunhofer
ITEM (Hanover, Germany).
• PERFORM-A2: Two combined toxicity / carcinogenicity studies of 900 MHz GSM & 1800 MHz DCS exposures in Wistar rats
(NTP-like). This study was performed at RCC (Liestal, Switzerland).
• PERFORM-A3: Evaluation of 900 MHz GSM wireless communication signals on DMBA-induced mammary tumors in
Sprague Dawley rats. This study was performed at Research
Center Seibersdorf (Seibersdorf, Austria).
• PERFORM-A4: Evaluation of 900 MHz GSM exposure on Lymphoma induction in E-PIM 1 transgenic mice. This study was
performed at RBM (Ivrea, Italy).
• PERFORM-A5: Exposure system design, construction &
dosimetry. The setup development, dosimetry and maintenance
as described in the previous sections was performed at the Foundation IT’IS (ETHZ, ARCS, Aristotle University of Thessaloniki).
• ZHEJIANG: Evaluation of 900 MHz GSM wireless communication signals on DMBA-induced mammary tumors in Sprague
Dawley rats. This study was performed at MRL, Animal Center
& Dept. of Pathology, Zhejiang University, China.
The carcinogenicity studies were the first NTP studies addressing
the emerging health questions which have arisen in Europe and around
6.8. STUDY ABSTRACTS: PERFORM A
141
the world due to the use of wireless communications technologies and
provided important contributions to the health hazard assessments
of ICNIRP, IARC and WHO. The studies were conducted between
2001 and 2004 (A1, A4) and 2005 (A2, A3, Zhejiang). The studies
were supported through the following co-founders: EU 5th Framework
Program, BBW, MMF, GSMA, Foundation IT’IS.
The performed studies led to a series of conference proceedings,
reports and paper publications about the PERFORM A mouse setups.
Here is a selection: [118], [119], [120], [121], [122] [48] [123], [119], [113]
[114] [124] [125].
6.8.1
Fraunhofer ITEM (Germany)
Radiation Research, vol. 164, no. 4, pp. 431-439, 2005
Effects of 1-week and 6-week GSM/DCS Radiofrequency
Exposure on Micronucleus Formation in B6C3F1 Mice
B.D. Görlitz, M. Müller, S. Ebert, H. Hecker, N. Kuster,
and C. Dasenbrock
The aim of the study was to provide findings on possible induction of micronuclei in erythrocytes of the peripheral blood and bone
marrow, in keratinocytes and spleen lymphocytes of mice exposed to
radiofrequency (RF) for 2 hours per day over a period of 1 week and
a period of 6 weeks, respectively. The applied signal simulated the
exposure from GSM 900 and DCS 1800 handsets, including the lowfrequency amplitude-modulation components as occur during speaking (GSM Basic), listening (DTX) and moving within the environment
(handover, power control). The carrier frequency was set to the center
of the system’s uplink band, i.e., 902 MHz for GSM and 1747 MHz for
DCS. Uniform whole-body exposure was achieved by restraining the
mice in tubes at fixed positions in the exposure setup. Mice were exposed to slot-averaged whole-body SARs of 33.2, 11.0, 3.7 and 0 mW/g
during the 1-week study and of 24.9, 8.3, 2.8 and 0 mW/g during the
6-week study. The exposure levels for the 1-week and the 6-week study
were determined in a pre-test to confirm that no thermal effect could
influence the genotoxic endpoints. During both experiments and for
142 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
both frequencies, no clinical abnormalities were detected in the animals. Cells of the bone marrow from the femur (1-week study), erythrocytes of the peripheral blood (6-week study), keratinocytes from
the tail root and lymphocytes from the spleen (both studies) were
isolated on slides and stained for micronucleus analysis. Per animal,
2000 cells each were scored in erythrocytes and in keratinocytes. In
spleen lymphocytes, 1000 binucleated lymphocytes were scored for
each animal. The RF exposure had no influence on the formation of
red blood cells. After one week of exposure, the ratio of polychromatic to normochromatic erythrocytes was unchanged in the treated
groups compared to the sham-exposed groups. Furthermore, the RF
exposure of mice induced no increase in the number of micronuclei in
erythrocytes of the bone marrow or peripheral blood, in keratinocytes,
nor in spleen lymphocytes, compared to the sham control.
Bioelectromagnetics, vol. 28, no. 3, pp. 173-187, 2007
Carcinogenicity Study of GSM and DCS Wireless Communication Signals in B6C3F1 Mice
T. Tillmann, H. Ernst, S. Ebert,
S. Rittinghausen, and C. Dasenbrock
N. Kuster,
W. Behnke,
The purpose of this study using a total of 1170 B6C3F1 mice was
to detect and evaluate possible carcinogenic effects in mice exposed
to radio-frequency-radiation (RFR) from Global System for Mobile
Communication (GSM) and Digital Personal Communications System (DCS) handsets as emitted by handsets operating in the center
of the communication band, that is, at 902 MHz (GSM) and 1747 MHz
(DCS). Restrained mice were exposed for 2 h per day, 5 days per week
over a period of 2 years to three different whole-body averaged specific
absorption rate (SAR) levels of 0.4, 1.3, 4.0 mW/g bw (SAR), or were
sham exposed. Regarding the organ-related tumor incidence, pairwise
Fisher’s test did not show any significant increase in the incidence of
any particular tumor type in the RF exposed groups as compared to
the sham exposed group. Interestingly, while the incidences of hepatocellular carcinomas were similar in EMF and sham exposed groups, in
both studies the incidences of liver adenomas in males decreased with
6.8. STUDY ABSTRACTS: PERFORM A
143
increasing dose levels; the incidences in the high dose groups were statistically significantly different from those in the sham exposed groups.
Comparison to published tumor rates in untreated mice revealed that
the observed tumor rates were within the range of historical control
data. In conclusion, the present study produced no evidence that
the exposure of male and female B6C3F1 mice to wireless GSM and
DCS radio frequency signals at a whole body absorption rate of up to
4.0 W/kg resulted in any adverse health effect or had any cumulative
influence on the incidence or severity of neoplastic and non-neoplastic
background lesions, and thus the study did not provide any evidence
of RF possessing a carcinogenic potential.
6.8.2
RBM (Italy)
Radiation Research, vol. 168, no. 3, pp. 316-326, 2007
Carcinogenicity Study of 217 Hz-Pulsed 900 MHz Electromagnetic Fields in Pim1 Transgenic Mice
G. Oberto, K. Rolfo, P. Yu, M. Carbonatto, S. Peano, N. Kuster,
S. Ebert, and S. Tofani
In an 18-month carcinogenicity study, Pim 1 transgenic mice were
exposed to pulsed 900 MHz (pulse width: 0.577 ms; pulse repetition
rate: 217 Hz) radiofrequency (RF) radiation at a whole-body specific
absorption rate (SAR) of 0.5, 1.4 or 4.0 W/kg [uncertainty (k=2):
2.6 dB; lifetime variation (k=1): 1.2 dB]. A total of 500 mice, 50 per
sex per group, were exposed, sham-exposed or used as cage controls.
The experiment was an extension of a previously published study in
female Pim 1 transgenic mice conducted by Repacholi et al. (Radiat. Res. 147, 631640, 1997) that reported a significant increase in
lymphomas after exposure to the same 900 MHz RF signal. Animals
were exposed for 1 h/day, 7 days/week in plastic tubes similar to those
used in inhalation studies to obtain well-defined uniform exposure.
The study was conducted blind. The highest exposure level (4 W/kg)
used in this study resulted in organ-averaged SARs that are above
the peak spatial SAR limits allowed by the ICNIRP (International
Commission on Non-ionizing Radiation Protection) standard for envi-
144 CHAPTER 6. LARGE SCALE IN VIVO EXPOSURE SETUP
ronmental exposures. The whole-body average was about three times
greater than the highest average SAR reported in the earlier study by
Repacholi et al. The results of this study do not suggest any effect of
217 Hz-pulsed RF-radiation exposure (pulse width: 0.577 ms) on the
incidence of tumors at any site, and thus the findings of Repacholi
et al. were not confirmed. Overall, the study shows no effect of RF
radiation under the conditions used on the incidence of any neoplastic
or non-neoplastic lesion, and thus the study does not provide evidence
that RF radiation possesses carcinogenic potential.
Chapter 7
SAR Uniformity in
Ferris Wheel Setups
7.1
Abstract
In vivo exposure setups, known as “Ferris wheel” setups, comprise
a cylindrical electromagnetic cavity of two circular parallel plates in
which rodents are exposed at the periphery by an antenna in the center of the cavity. Setups of different sizes ranging from 8 animals to
65 animals have been used for exposing mice in two-year bioassays as
well as in experiments only requiring short-term exposure. This concept provides a mono-mode structure vertically between the plates
but not radially. The objectives of this study were to evaluate the
effects of the multi-mode nature of the setup on the instant, weekly
and life-averaged dosimetry. Three setups of 8, 40 and 65 animals
with the corresponding radii of 166 mm, 480 mm, 670 mm were experimentally and numerically compared, all operating at 900 MHz. The
instant exposure variations were large for inhomogeneous (realistic
and worst-case) loading scenarios, i.e., as large as 7.4 dB (SD 2.1 dB)
between the highest and the lowest whole-body SAR. Only for homogeneous loading with adult mice were the variations significantly
lower (variations ≤ 2 dB with SD 1 dB). The variations in whole-body
exposure depend on the cavity Q, size, asymmetry of constructions
145
146
CHAPTER 7. SAR UNIFORMITY IN FERRIS WHEELS
and loading, as well as on the interactions between animals. However,
the inter-animal variation of the averaged exposure over lifetime can
be kept reasonably low by following a strict rotational scheme over
time.
7.2
Introduction
Ferris wheel (FW) exposure setups have been proposed by Motorola
[45] as an affordable and efficient solution for exposing mice at wholebody specific absorption rate (SAR) values of up to several W/kg.
Ferris wheel exposure setups comprise a radial electromagnetic
cavity formed by parallel circular plates joined around the perimeter
by an array of shorting posts. The cavity is excited in the center,
whereas the animals are placed round the periphery at a defined radius. The FW concept is very effective with respect to:
• space (vertical, narrow cylinders housing an entire group),
• radio frequency power (resonant cavity),
• number of exciters and amplifiers (only one exciter and amplifier
per wheel).
The most significant disadvantage of the setup is its multimode
cavity. Locally, the incident field is a freespace-like TEM plane wave,
as long as circumferential or longitudinal higher order mode excitation
is not very significant. However, due to the rather high Q of the system
even slight asymmetries may strongly excite higher order modes.
7.3
Objectives
The objectives of this comparison study were to evaluate the effect
of multi-modes on the instant, weekly and lifetime whole-body exposure as a result of asymmetries occurring during the experiments
caused by movements and different weights of the mice and to provide
the detailed dosimetry including uncertainty and variations for such
setups.
7.4. EXAMINED FERRIS WHEELS
7.4
147
Examined Ferris Wheels
Figure 7.1 shows the three Ferris wheel setups examined. They were
developed and used in various health risk assessment studies:
FW-O: The original Ferris wheel setup developed by Motorola Corporation was used in Adelaide, South Australia, as part of the National Health and Medical Research Council Electromagnetic Energy
Program.
FW-A: The large 65 animal wheel developed by the IT’IS Foundation
was used in the PERFORM A program at Fraunhofer ITEM (Germany) and RBM (Italy) as well as in the Micronucleus Study [113]
sponsored by CRADA.
FW-B: The Mini Wheel Setup developed by the IT’IS Foundation
was used at NRPB (Great Britain) within the PERFORM B program.
Figure 7.1: Examined Ferris wheel exposure setups.
FW-O: Motorola Ferris Wheel, FW-A: IT’IS PERFORM A Ferris Wheel,
FW-B: IT’IS Mini Wheel.
7.5
Ferris Wheel Setup Design
Each Ferris wheel consists of a resonant structure of two parallel, circular metal plates (see Figure 7.2). Around the edges of the plates,
parallel metal bars connect the two plates and function as an electromagnetic short. The bars are such that they provide full reflections
but also allow sufficient light for the animals. The structure is excited
148
CHAPTER 7. SAR UNIFORMITY IN FERRIS WHEELS
in the center by a monopole or dipole structure (E-field normal to the
plates, H-field circularly around the center). The antenna is matched
to the resonant system, e.g., by external or internal tuning elements.
Encased between the plates are symmetrically positioned cylindrical
polycarbonate tubes arranged radially around the antenna. The animals are restrained in tubes by a stopper which prevents the animal
from turning around or moving forward or backward. These tubes
are the same or similar as used in inhalation studies for which a large
database of NTP studies is available. The animals are inserted into
the tubes between the plates, providing a well-defined position of the
animal within the wheel. This was achieved in the experiments of
FW-A and FW-B but not in FW-O, in which the animals could easily
turn around, i.e., walk forwards and backwards. A defined airflow of
1 l/min was provided from one side, e.g., through a ventilation system
consisting of a single ventilation tube per mouse or a ring channel
with the corresponding overpressure.
Metallic Bars
as Short-cut
Holder
Tube
Ra
d
1
ius
Rad
Antenna
Front View
Metallic
Plates on
Front- and
Backside
ius 2
Side View
Figure 7.2: Principle design – features front and side view of a Ferris
wheel.
The separation between the centers of the tubes were FW-O:
69mm, FW-A: 65mm, FW-B: 85mm. Initial studies showed strong
electromagnetic coupling between the mice if they were of different
Mouse Dummy and
7.5. FERRIS WHEEL SETUP DESIGN
149
weights. In order to achieve greater separation between the mice without changing the dimensions of the wheel, high permittivity bricks
(=12) were positioned between the tubes. Additionally, the Q value
was lowered by introducing PFTE restrainer stoppers enriched with
25% carbon. Differences in key design parameters are listed in Table
7.1.
Wheel 0
Wheel A
Wheel B
Operating frequency
900 MHz
902 MHz
905 MHz
Max. animals per wheel
40
65
8
Distance between plates
Radius R1: plate center
to location of mice
Radius R2: plate center
to short
Short:
- distance between polls
- diameter
100 mm
440 mm
117 mm
670 mm
120 mm
111.5 mm
480 mm
755 mm
166 mm
18.25 mm
6.25 mm
10 mm
8 mm
10.5 mm
8 mm
conical
antenna
conical
antenna
Exposure control by
measurement of:
capacitively
coupled toploaded monopole antenna
changing
capacity
forward and
reflected power
external
3-stub tuner
two redundant
E.inc sensors
external
3-stub tuner
two redundant
E.inc sensors
Environmental sensors
none
temperature,
humidity, O2
none
copper
copper-clad
laminate PCB
polycarbonate
copper
stainless
steel
polycarbonate
copper
stainless
steel
polycarbonate
Antenna type
Tuning
Materials:
- antenna
- plates
- tubes
Table 7.1: Key design parameters of the three Ferris wheel setups.
150
CHAPTER 7. SAR UNIFORMITY IN FERRIS WHEELS
7.6
Material and Methods
The dosimetric examinations of the Ferris wheel setups consisted of
experimental and numerical parts. The experimental part examined
the averaged whole-body SAR of the mice in the wheel and the individual whole-body averaged SAR (SAR uniformity), including its
uncertainty. The numerical examinations were performed by FDTD
simulations using high resolution mouse models to validate and assess
the uncertainty of the individual whole-body and organ-specific SAR.
The same experimental method was applied to all three Ferris wheels.
7.6.1
Experimental Investigations
All measurements were conducted in an air-conditioned anechoic
chamber at the stable room temperature of 22 ±1 ◦ C. The examined
exposure setups were loaded and operated with different mouse dummies (plastic tubes filled with dielectric liquid representing real mice).
The wheels were standing upright as during the study to have the same
mechanical forces on the metallic plates and consequent possible deformations. The background temperature was monitored, as well as
the horizontal and vertical temperature distributions in the surroundings of the wheel. The temperature of the dummies was recorded and
measured using RF-immune thermistor probes (T1V3LA, SPEAG,
Switzerland; noise level <0.005 ◦ C; absolute accuracy 0.1 ◦ C).
Usually, an exposure study uses mice of the same strain, similar
mass and same age. However, the body mass changes over the course
of the study. During a long-term study like, e.g., PERFORM A, mice
grew from about 20g (age 4 weeks) to about 40g (after two years).
Variations in mass occur between the mice, leading to differences of
up to ± 40% from the average mouse mass. Figure 7.3 shows the
body mass developement of male B6C3F1 mice as they were used in
the FW-A during the Perform A study.
The SAR uncertainty and variability assessment therefore requires
examinations of the wheels with homogeneous and inhomogeneous
loading scenarios. These were performed using four different dummy
sizes, which correspond to the average mass of light/young mice of
a few weeks of age (20g), average adult female mice (31g), average
7.6. MATERIAL AND METHODS
151
adult male mice (37g) and heavy/old mice (46g). The applied different
loading situations are summarized in Table 7.2.
Wheel loading
A
B
C
D
E
- 20g dummies
mouse)
- 31g dummies
mouse)
- 37g dummies
mouse)
- 46g dummies
Assessment of
(light/young
(average female
(average male
- wheels loaded with various
mouse weight classes (best case
loading scenario)
- aging during the lifespan/study
duration
- heat time constant τ
(heavy/old mouse)
- 20g dummies randomly shifted
along body axis
- 31g dummies randomly shifted
along body axis
- mouse movements, loose
stopper
- imprecise loading
G
- 31g dummies, every 4th (wheel O)
or every 5th (wheel A) dummy
was replaced alternating with a
20g or 46g dummy
- strong weight variations
with corresponding change
in SAR for a wheel loaded
with adult female mice
H
- 37g dummies, every 4th (wheel O)
or every 5th (wheel A) dummy
was replaced alternating with a
20g or 46g dummy
- strong weight variations
with corresponding change
in SAR for a wheel loaded
with adult male mice
I
- equal amount of 20g, 31g, 37g and
46g dummies. Locations were
chosen randomly.
- worst case weight variations
with corresponding change in
SAR
J
- loading of wheel A without bricks
like in scenario G
- influence and improvement due
to the high permittivity bricks
F
Table 7.2: Loading scenarios to experimentally examine the uncertainty and variability of the Ferris wheel setups.
152
CHAPTER 7. SAR UNIFORMITY IN FERRIS WHEELS
Figure 7.3: Body mass development of male mice during a two year
study. Body mass increases during the course of the study with variations up to ±-40% from the average body mass.
7.6.2
Temperature Method for SAR Uniformity
Assessment
A temperature method enables the determination of the wheel averaged whole-body SAR as well as the individual deviation at each
position in the cavity. When a liquid dummy is exposed to RF, its
temperature change is determined through the applied RF radiation
and through the heat transfer with a time constant tau. By applying
the lumped-heat-capacity method ([126]: uniform temperature distribution in dummy; heat flux inside outside) the liquid temperature
follows the differential equation:
7.6. MATERIAL AND METHODS
153
dT
−(Tliquid − Troom ) SARW B
=
+
dt
τ
cliquid
(7.1)
Two processes can occur:
1. RF is on: liquid temperature increases until an equilibrium state
is reached at which the heat transfer due to the RF exposure
and the heat loss due to convection equal out.
2. RF is off: liquid cools down through heat convection due to the
lower environmental temperature.
The SAR values result from the solution of the differential equation
for long exposure times (t τ ) of the heating curve equation when
the system is in an equilibrium temperature state:
SARW B = cliquid
∆T
τ
(7.2)
with cliquid as the heat capacity of the liquid with the time constant and ∆T as the liquid temperature increase. The determination
of SARW B requires stable environmental temperatures, absolute temperature measurements of the dummy liquid and room temperature as
well as the time constant (determined by fitting the theoretical curve
to the recorded temperature change in the dummy liquid during the
heating or cooling process).
Table 7.3 shows an uncertainty analysis of the applied method as
used. The analysis results in three combined standard uncertainties:
the uncertainty of the relative SAR uniformities for a homogeneous
and an inhomogeneous loading scenario are 5.2 % (0.22 dB) and 11.3 %
(0.46 dB) respectively and the uncertainty for the absolute SAR determination is 13.0 % (0.53 dB).
154
CHAPTER 7. SAR UNIFORMITY IN FERRIS WHEELS
Error Description
Tolerance
Prob.
Distr.
Standard
Uncertainty
Temperature
T. probe accuracy
T. distribution in dummy
Room T. change over time
T. distribution in room
2.0%
3.0%
4.5%
3.6%
N
R
R
R
2.0%
1.7%
2.6%
2.1%
Dummy
Variation in filling volume
5.00%
R
2.9%
Time constant τ
τ determination
10.0%
N
10.0%
Liquid parameter
Uncertainty of heat capacity
Uncertainty of Uncertainty of σ
5.0%
7.1%
1.0%
N
R
R
5.0 %
4.1%
0.6%
Results in combined standard uncertainty (K=1) for
- relative SAR uniformity (homog. loading)
- relative SAR uniformity (inhomog. loading)
- absolute SAR determination.
5.2%
11.3%
13.0%
Table 7.3: Estimation of the combined standard uncertainties for the
relative SAR uniformity at a homogeneous and inhomogeneous loading scenario and for the absolute SAR determination (conducted according to [47]). Probability Distribution N: normal and R: rectangular.
7.6.3
Measurement Procedure
1. Load the Ferris wheel with dummies.
2. Place RF immune temperature sensor in at least one dummy
and record the liquid temperature; also monitor that the environment temperature does not change over time.
3. Start the RF exposure of the dummies; choose the RF power
level leading to a temperature increase in the dummies of about
10◦ C.
7.6. MATERIAL AND METHODS
155
4. After some hours of exposure the liquid temperature reaches
equilibrium state, at which the additional heat due to the RF
power is equal to the heat convection.
With RF on, start the measurement of the liquid temperatures
in the dummies consecutively:
(a) Take one dummy out of the Ferris wheel.
(b) Shake or stir the liquid to ensure an equal temperature
distribution.
(c) Measure the liquid temperature.
(d) Put the dummy back into its original place.
(e) Continue with the next dummy.
The uncertainty analysis for the temperature measurements for
the assessment of the whole-body SAR according to NIST Technical
Note 1297 [47] revealed a combined standard uncertainty (k=1) of
13%.
7.6.4
Dummy Characteristics
Special care was taken in selecting appropriate dummies representing
the mice. Plastic tubes with a conical tip at one end and flat at the
other end (No. 1611, Semadeni, Switzerland) represented the shape
of a mouse in the restrainer tube; see also Figure 7.4. The tube length
was adjusted to the filling amount, assuring the dummy was always
completely filled. The lengths of the dummy tubes were in agreement
with measurements of real mouse lengths in restrainer tubes. The
dielectric parameters of the tissue simulating liquid yield comparable
whole-body SAR in dummies and in mice as determined by FDTD
simulations (SEMCAD Vers. 1.8, SPEAG, Switzerland).
The criterion for the tissue simulating liquid was a comparable
whole-body SAR in dummies and in mice. This was determined by
FDTD simulations (SEMCAD Vers. 1.8, SPEAG, Switzerland) by
exposing liquid dummies of the four weight classes and corresponding
high resolution anatomical mouse models [122] in a sector model of
a Ferris wheel setup while changing the liquid parameters (ε, σ) until the normalized whole-body averaged SAR agrees. A liquid with
156
CHAPTER 7. SAR UNIFORMITY IN FERRIS WHEELS
ε = 27.9 and σ = 0.66 suits best for all four dummy sizes with a maximum deviation of 17 % in whole-body SAR. (This equals a dummy
liquid consisting of 30 % deionized water and 70 % sugar, experimentally verified with a probe and network analyzer (Hewlett Packard,
USA, accuracy ∆ = 5 % and ∆σ = 10 %)).
The tissue simulating liquid consisted of 30 % deionized water and
70 % sugar, with the dielectric parameters: = 27.9 and σ = 0.66.
The dielectric parameters of the liquid vary slightly when heated:
ε increases linearly at a rate of ∆ε = 1 per 5◦ C over the measured
range from 22◦ C room temperature to 43◦ C. In the same temperature
range σ stays constant at the same level. The specific heat constant
cliquid of the dummy liquid was determined with calorimetric measurements to cliquid = 2840J/(kgK). Several measurements of cliquid
led to the same results within 5 % deviation.
Figure 7.4: Mouse tube and conical dummy filled with a tissue simulating dielectric liquid.
7.6.5
Numerical Evaluation
The numerical evaluations were performed within the PERFORM A
mouse setup dosimetry and uncertainty assessment by
Dr. Veronica Berdiñas. Details of the results are listed in [107];
the method followed the concept described in [15].
The FDTD simulations were performed with three different highresolution anatomical mouse models (strains OF1, PIM1) and four
scaled mouse models to examine the effect on (1) body weight (lifetime), (2) anatomy (male/female, strain), (3) sector position (four
positions), (4) position in the tube (front/back), (5) posture (restrained / non-restrained), (6) stopper contact (full contact, no con-
7.7. RESULTS
157
tact), (7) dielectric properties of tissue parameters (, σ) and discretization. In total more than 200 simulations were conducted to
obtain the detailed dosimetry, its uncertainty and the instant and
life-time variations.
7.7
Results
The polar plots in Figures 7.5 and 7.6 present the SAR patterns which
occur for the different loading scenarios. Displayed are the deviations
of the individual whole-body SAR for each exposure location to the
wheel averaged SAR. The polar plots reveal characteristic patterns
for each wheel.
Deviations between the dummy with the least amount of
SAR and the one with the highest SAR represent the worstcase deviations in SAR uniformity. With homogeneous loading
they range for wheel O between 1.1 and 2.9 dB (SD 1.1 dB), for
wheel A between 1.5 and 4.0 dB (SD 1.2 dB) and for wheel B between
0.4 and 1.4 dB (SD 0.7 dB). An axial shift of the dummies decrease
the SAR uniformity in all three wheels between 0.5 dB (wheel O)
and 1.4 dB (wheel A). The SAR uniformity decreases dramatically for inhomogeneous loading scenarios: wheel O varies up to
7.4 dB (SD 2.1 dB) and wheel A up to 4.1 dB (SD 1.0 dB). In all
wheels lighter animals receive less and heavier animals receive higher
SAR compared to the wheel averaged SAR in the examined setups.
Figure 7.7 presents the influence of the high permittivity bricks
with an inhomogeneous loading scenario for FW-A. Comparing the
loading scenarios of G and J shows that the usage of bricks improves
the SAR uniformity by about 2 dB. Without bricks, the worst-case
uniformity has a maximum deviation of up to 6.1 dB (SD 1.9 dB).
The results of the detailed numerical uncertainty analysis (see Section
6.5.3) demonstrate that in long-term studies the overall variation can
be strongly further reduced by introducing a strict rotational scheme,
i.e., the position of the mice in the wheel is shifted at regular intervals
(e.g., daily, weekly). The lifetime averaged variations are significantly
lower than the occurring instant variations.
158
CHAPTER 7. SAR UNIFORMITY IN FERRIS WHEELS
Figure 7.5: SAR uniformity pattern for homogeneous loading of the
three Ferris wheels. Left side (scenarios A, B, C and D): all positions
filled with either 20g, 31g, 37g or 46g dummies. Right side (scenarios
B and F): all positions filled with 31g dummies in the correct loading
position and shifted along the body axis in a random manner. The
averaged SAR level is marked with a bold line at 0dB.
7.7. RESULTS
159
Figure 7.6: SAR uniformity pattern for inhomogeneous loading of
FW-O and FW-A for loading scenario I: all positions are randomly
filled with equal amounts of 20g, 31g, 37g and 46g dummies. The
wheel averaged SAR level is marked with a bold line at 0dB. High deviations of up 7.4dB occur for this worst-case loading scenario between
the highest and the lowest whole-body SAR.
Figure 7.7: SAR uniformity pattern of FW-A with loading scenario
G (with high permittivity bricks) and scenario J (without bricks). By
introducing high permittivity bricks between the mouse holder tubes,
the uniformity is improved by about 2dB.
160
CHAPTER 7. SAR UNIFORMITY IN FERRIS WHEELS
SAR.sp/
SAR.wba
[dB]
[dB]
Variations
SDb
SDc
[dB]
[dB]
Spatial Peak SAR
SAR 5mg
17.9
SAR 0.5mg
19.3
± 6.9
± 5.7
± 4.4
± 3.3
± 2.6
± 1.8
Organ/Tissue
Bladder
Blood
Bone Marrow
Bones
Brain
Cartilages
Eyes
Fat
Gland, Lacrimal
Glands
Heart
Kidneys
Large Intestine
Liver
Lungs
Muscles
Nerves
Oesophagus
Pharynx
Skin
Small Intestine
Spinal Cord
Spleen
Stomach
Tongue
Trachea
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
a
b
c
-0.6
1.6
-4.6
-7.5
-2.1
-1.7
-1.6
-7.2
-1.6
0.6
0.8
0.7
0.2
0.2
1.8
0.1
-2.9
0.5
-0.9
-0.7
2.7
-1.2
0.6
0.0
0.2
-0.9
Uncertainty
4.0
2.7
3.4
2.8
2.6
2.7
2.6
3.0
2.6
2.9
3.0
2.7
2.8
2.7
2.8
2.6
3.8
3.2
2.7
2.8
2.9
2.6
3.2
2.9
2.8
2.6
3.6
2.9
4.2
2.8
2.8
2.5
2.4
2.7
2.3
3.8
3.8
3.3
3.3
2.6
3.6
2.6
2.6
2.7
2.3
2.7
3.2
2.7
3.5
3.3
2.7
2.7
1.6
2.1
3.2
2.1
1.3
1.7
1.3
1.5
1.3
1.8
2.2
1.6
1.7
1.7
1.7
1.7
1.8
2.0
1.3
1.9
1.9
1.7
2.4
2.3
1.3
1.3
Group and lifetime average.
Instant standard deviation.
Lifetime averaged standard deviation.
Table 7.4: The detailed uncertainty and variability assessment of FWA shows that introducing a strict rotational scheme in long-term studies reduces the lifetime (2.6 dB) averaged variations compared to the
instant (4.4 dB) variations (see also Section 6.5.3, source: [107]).
7.8. DISCUSSION
7.8
161
Discussion
The results show that the SAR uniformity of all three
Ferris wheel setups are rather low with worst-case variations of up
to 7.4 dB (SD 2.1 dB). Possible reasons are, i.e., the occurrence of
higher modes, a non-isotropic antenna pattern, sensitivity of the antenna pattern on the load, a very sharp Q-factor, imprecise manufacturing or asymmetries of the setup due to mechanical forces.
The measurement results show several ways to improve the uniformity: introducing high permittivity bricks between the mouse positions to symmetricise and stabilize the field at the load; increasing
the load mass (increase losses), so that the Q value of the cavity is
lowered, though this lowers the SAR efficiency and therefore increases
the amplification costs.
It was also shown that the more homogeneous the mouse weight
distribution in a Ferris wheel, the better the SAR uniformity; a shorter
wheel radius additionally improves the SAR uniformity: wheel B
showed the best performance. When using smaller exposure units,
mice could be sorted more precisely into equal weight groups prior to
the exposure.
Furthermore, the averaging effect can be used in long-term studies to lower the overall variation by introducing a rotational scheme.
The position of mice is thereby shifted at regular time intervals
(e.g., weekly, daily) in the wheel, so that their average SAR tends
towards the average SAR of the exposure wheel.
Various concepts for whole-body exposure can be found in the
literature. Classic concepts, like circular or rectangular waveguides
and rectangular horns, appear not to deliver the required exposure
uniformity to a larger number of animals [9], [127], [128]. Other,
more modern concepts make use of radial waveguides [129] or sectoral
waveguides [89] and revertebration chambers [88], which seem to have
promising uniformity behavior. However, the Ferris wheel concept
also achieves high uniformity, when all possible improvements (small
units, high permittivity bricks, rotational scheme, homogeneous load)
are applied.
162
CHAPTER 7. SAR UNIFORMITY IN FERRIS WHEELS
7.9
Conclusions
Multiple radial modes occur in Ferris wheel setups, leading to
assymetries in SAR. The instant exposure variations were large
for inhomogeneous (realistic and worst-case) loading scenarios,
i.e., as large as 7.4 dB between the highest and the lowest
whole-body SAR (SD 2.1 dB). If a strict rotational scheme over time
is followed, the inter-animal variation of the averaged exposure over
lifetime can be kept reasonably low, i.e., better than 2.6 dB. In general non-multimode systems are preferred for bioexperiments since
they result in:
• better uniformity
• greater repeatability
• smaller variations.
In addition non-multimode systems greatly simplify the complexity of
the dosimetry and uncertainty analysis.
Three Ferris wheel setups for the whole-body exposure of mice at
900 MHz have been examined with a temperature method, showing
similar disadvantages in their SAR uniformity. Especially inhomogeneous loading scenarios – as they usually occur during long-term
studies – lead to poor SAR uniformity behavior with variations of
up to 7.4 dB (SD 2.1 dB). Only for homogeneous loading with adult
mice are the variations of the individual whole-body SAR significantly
lower (< 2 dB with SD 1 dB).
Depending on the aim of the study, Ferris wheel setups are still
useful within their SAR uniformity limitations. They have a high
loading volume, high SAR efficiency due to the cavity design and provide a good cost/SAR ratio. Additionally, this study delivered several
concepts to improve the uniformity; by applying those concepts (small
units, high permittivity bricks, rotational scheme, homogeneous load)
the Ferris wheel concept achieves the high uniformity demanded by
the requirements for in vivo whole-body exposure studies [9].
Epilogue
The objectives of this dissertation were achieved: the development, characterization and successful operation of optimized exposure systems for in vivo and human studies of the European
PERFORM project with a significant contribution for all major requirements of such systems to provide well-defined, well-controlled,
well-characterized, and artifact-free exposures.
In PERFORM A the Fraunhofer Institute of Toxicology and
Experimental Medicine in Germany conducted two combined chronic
toxicity and carcinogenicity studies with 1170 B6C3F1 mice using
GSM and DCS communication signals and the Istituto di Ricerche
Biomediche Antoine Marxer in Italy performed a co-carcinogenicity
study with 500 Pim1 mice using a GSM signal, which was also set
up as a replication study of the widely discussed Adelaide study [99].
The developed mouse exposure systems worked successfully during the
two years (life-time) laboratory study. The human provocation study
PERFORM C conducted at the Department of Health Science of the
Karolinska Institute in Sweden with close to 80 volunteers exposed to
GSM signals was also successfully completed. All exposure systems
operated fully self-controlled with an auto-detection of malfunctions
and were well-characterized. Dosimetry, uncertainty and variability
were assessed from experimental and numerical evaluations and also
from the system performance data at the laboratories.
The setup design significantly influences the exposure conditions,
and it is a technical challenge to achieve well-defined and artifactfree exposures. The comparison study shows that slight design variations lead to significantly different exposures, especially in multi-mode
systems. The exposure variations for each animal and between ani163
164
EPILOGUE
mals were greatly improved by applying various design enhancements.
A further highlight of this dissertation was the precise determination
of the thermal thresholds of tube-restrained, whole-body RF exposed
mice. Though thermal thresholds were determined before, it was possible to differ between the thermal regulatory and breakdown thresholds caused by RF exposure with a new precision. The knowledge
of these thresholds made the PERFORM A study most relevant by
setting the highest exposure dose as high as possible but below the
threshold of induced known adverse biological affects. In addition to
the exposure dose, the exposure signal is also crucial for well-designed
exposure studies. The RF electromagnetic fields emitted by mobile
phones have been well examined, but little was known about the low
frequency magnetic fields caused through the relatively high amplifier
supply currents. To close this knowledge gap, the B-field strengths
and waveforms of commercial phones were assessed and an exposure
signal proposed to be used in combined ELF/RF exposure studies.
However, the most important outcome of this work comprises the
biological study results which were derived with the technologies realized in this dissertation. Part of my work, though not included
in this thesis, were also design, characterization and operation of the
PERFORM A exposure systems for rats [90]. Austrian Research Centers Seibersdorf (ARCS) and RCC Ltd in Switzerland conducted rat
exposure studies within the PERFORM A project. In addition the
Zhejiang University in China also conducted a study with rats using
the same exposure setup.
Five out of the six PERFORM A laboratory studies with mice and
rats did not show any evidence that exposure had any adverse health
effect [93] [94] [119] [123]. This includes the study at RBM in Italy
conducted with Eµ-PIM 1 mice which have express elevated levels of
the Pim-1 transgene in their lymphoid compartments, and as a result
are predisposed to develop lymphomas. However, the DMBA study
with Sprague-Dawley rats at the Austrian Research Centers Seibersdorf [92] showed some incidents. In this study rats were initially exposed to 7,12-Dimethylbenz[a]anthracen (DMBA), so that they had
an increased likelihood of receiving mammary cancer. The results at
ARCS show significant differences between one or more of the EMFexposed groups and the sham exposed groups for the mammary glands
and their neoplastic lesions; however, the largest differences in this
165
study were noted between the sham exposed group and the cage control group. The ARCS study is therefore considered as showing a
borderline effect, in which the long-term repeated exposure to GSM902 MHz signals affects the DMBA induced mammary tumor response
in rats with an equivocal biological relevance. Though not part of
this thesis, it is interesting to mention that the results obtained in
the PERFORM B replication studies were negative [42] and did not
confirm the previously found effects (e.g., [130] [131] [132] [133]). The
human provocation study PERFORM C focused on a different biological endpoint and examined subjective symptoms of a well-defined
study group exposed to GSM signals for three hours. The study group
included subjects reporting symptoms attributed to mobile phone use.
The results reveal that neither group was able to detect RF exposure
better than by chance. The results also reveal that headaches were
more commonly reported after RF exposure than in the sham exposed control group, mainly due to an increase in the subject group
not reporting symptoms attributed to mobile phone use [80].
The laboratory studies indicate that exposure to RF fields from
mobile phones is unlikely to lead to an increase in cancer in humans.
However, as the widespread duration of the exposure of humans to
the RF-EMF from mobile phones is shorter than the induction time
of some cancers, further studies are required to identify potential
longer-term (beyond ten years) health risks. Relevant in this context will also be the currently conducted, life-time in vivo laboratory study at the National Institute of Environmental Health Science
(NIEHS, Chicago, USA) [134]). As for non-carcinogenic outcomes,
several studies, in which subjects report subjective symptoms, indicate an association between RF exposure and single symptoms (e.g.,
PERFORM C). However, there is still a lack of consistency in the findings. One assumption under discussion is that an adverse non-specific
effect caused by expectation or belief that something is harmful (nocebo effect) may play a role in symptom formation. So far there is
no evidence that individuals, including those attributing symptoms to
RF exposure, are able to detect RF fields [20] [80]. There is some evidence that RF fields can influence EEG patterns and sleep in humans
[60] [135]. However, the health relevance is uncertain and mechanistic
explanation is lacking. Further investigation of possible physiological
correlations and other effects are needed.
166
EPILOGUE
Appendices
167
List of Acronyms
ARCS
B6C3F1
Austrian Research Centers Seibersdorf, Austria
Laboratory mouse; first-generation hybrid strain
by crossing C57BL/6 females and C3H males
DASY
Dosimetric Assessment System
DCS
Digital Cellular System
DCU
Digital Control Unit
DMBA
7,12-Dimethylbenz[a]anthracen
DTX
Discontinuous Transmission (see GSM standard)
DTX mode is active during listening phases
EASY
Four Channel Exposure Aquisition System
EEG
Electroencephalogram
ELF MF
Extremely Low Frequency Magnetic Fields
EMC
Electromagnetic Compatibility
EMF
Electromagnetic Field
EMI
Electromagnetic Interference
ETH
Swiss Federal Institute of Technology
FDTD
Finite-Difference Time-Domain
Fraunhofer ITEM Fraunhofer Institute for Toxicology and
Experimental Medicine, Germany
FW
Ferris wheel (exposure setup concept)
GLP
Good Laboratory Practice
GPIB
General Purpose Interface Bus
GSM
Global System for Mobile Communications
GSMA
GSM Association
HR-EF1
High Resolution European Female 1
169
170
HPBW
ICNIRP
IIS
IT’IS
Karolinska
MF
MMF
MRI
NIST
Non-DTX
Pim1
RBM
RCC
RF
RF-EMF
SAM
SAR
SD
SEMCAD
SHAM
SMP
SPA
SPEAG
Sprague Dawley
sXh
TDMA
TEM
UMTS
WHO
LIST OF ACRONYMS
Half-power beam width
International Commission on Non-ionizing
Radiation Protection
Integrated Systems Laboratory
Foundation for Research on Information
Technologies in Society
Karolinska Institute, Sweden
Magnetic Fields
Mobile Manufacturers Forum
Magnetic Resonance Imaging
National Institute of Standards in Technology, USA
non-DTX mode is active during talking phases
E µ-Pim1 transgenic mice are predisposed to
develop lymphomas
Istituto di Ricerche Biomediche Antoine Marxer, Italy
RCC, Liestal, Switzerland
Radio Frequency
Radio Frequency Electromagnetic Fields
Specific Anthropomorphic Mannequin
Specific Absorption Rate
Standard Deviation
Simulation Platform for EMC, Antenna Design
and Dosimetry
An exposure without signal/agent
Stacked Micropatch Antenna
Huber & Suhner antenna model SPA 920/65/9/0/V
Schmid & Partner Engineering AG, Switzerland
Outbred multipurpose breed of albino rat used
extensively in medical research
System for Exposure of Humans
Time Devision Multiple Access
Transversal Electromagnetic
Universal Mobile Telecommunication System
World Health Organization
Publications
Publications covered in this Thesis
1. S. Ebert, S.-J. Eom, J. Schuderer, U. Apostel, T. Tillmann,
C. Dasenbrock, and N. Kuster. “Response, thermal regulatory threshold, and thermal breakdown threshold of restrained
RF-exposed mice at 905 MHz” Physics in Medicine and Biology
vol. 50, no. 21, pp. 5203-5215, 2005.
2. M. Tuor, S. Ebert, J. Schuderer, and N. Kuster. “Assessment
of ELF magnetic fields from five mobile handsets” Proceedings
of the 27th Annual Meeting of the Bioelectromagnetics Society,
pp. 125-126, 2005.
3. L. Hillert, T. Åkerstedt, A. Lowden, C. Wiholm, N. Kuster,
S. Ebert, C. Boutry, S.D. Moffat, M. Berg, and B. B. Arnetz.
“The effects of 884 MHz GSM wireless communication signals
on headache and other symptoms: an experimental provocation
study” Bioelectromagnetics, vol. 29, no. 3, pp. 185-196, 2008.
4. B-D. Görlitz, M. Müller, S. Ebert, H. Hecker, N. Kuster, and
C Dasenbrock. “Effects of 1-week and 6-week exposure to
GSM/ DCS radiofrequency radiation on micronucleus formation
in B6C3F1 mice”, Radiation Research, vol. 164, no. 4, pp. 431439, 2005.
5. T. Tillmann, H. Ernst, S. Ebert, N. Kuster, W. Behnke, S. Rittinghausen, and C. Dasenbrock. “Carcinogenicity study of GSM
171
172
PUBLICATIONS
and DCS wireless communication signals in B6C3F1 mice” Bioelectromagnetics, vol. 28, no. 3, pp. 173-187, 2007.
6. G. Oberto, K. Rolfo, P. Yu, M. Carbonatto, S. Peano, N. Kuster,
S. Ebert, and S. Tofani. “Carcinogenicity study of 217 Hz-pulsed
900 MHz electromagnetic fields in Pim1 transgenic mice” Radiation Research, vol. 168, no. 3, pp. 316-326, 2007.
7. S. Ebert, V. Berdiñas Torres, J. Fröhlich, and N. Kuster, “Comparison of wheel-like mouse exposure systems at 900 MHz”, Proceedings of the 27th Annual Meeting of the Bioelectromagnetics
Society, Dublin, Ireland, pp.354-355, 2005.
Further Thesis related Publications
8. S. Ebert, R. Mertens, and N. Kuster. “Criteria for selecting
specific EMF exposure conditions for bioexperiments in the context of health risk assessment”, Proceedings of the 14th International Zurich Symposium on Electromagnetic Compatibility
2001, Zurich, Switzerland, pp. 181-182, 2001.
9. N. Kuster, W. R. Adey, S. Ebert, and W. Oesch. “Selection
and implementation of specific EMF exposure conditions for
bioexperiments in the context of health risk assessment”, 2001
International Symposium on Electromagnetics in Biology and
Medicine, Tokyo, Japan, p. 67, 2001.
10. S. Ebert, J. Fröhlich, W. Oesch, U. Frauenknecht, and
N. Kuster. “Optimized in vivo exposure setups for risk assessment studies at mobile communication frequencies 902 MHz and
1747 MHz”, Proceedings of the 23rd Annual Meeting of the Bioelectromagnetics Society, St. Paul, Minnesota, USA, p. 27, 2001.
11. N. Kuster, J. Fröhlich, and S. Ebert. “Exposure setup design
for large-scale NTP-like bioassays”, Asia-Pacific Radio Science
Conference, University of Tokyo, Tokyo, Japan, pp. 245-256,
2001.
173
12. J. Fröhlich, W. Kainz, S. Ebert, T. Samaras, G. Neubauer, and
N. Kuster. “Exposure setups for simultaneous exposure of a
large number of rats for risk asses
13. S. Ebert, J. Fröhlich, W. Oesch, R. Mertens, and N. Kuster.
“Characterization and Dosimetry of the PERFORM A in
vivo exposure system at the mobile communication frequencies 902 MHz and 1747 MHz”, Proceedings of the 24th Annual
Meeting of the Bioelectromagnetics Society, Quebec, Canada,
pp. 109-110, 2002.
14. R. Mertens, S. Ebert and N. Kuster. “Simulating environmental GSM features for use in bioexperiments”, Proceedings of the
24th Annual Meeting of the Bioelectromagnetics Society, Quebec, Canada, 2002.
15. S. Ebert, J.-C. Gröbli, and N. Kuster. “Mobile exposure setup
for human provocative studies at 900 MHz”, Proceedings of the
25th Annual Meeting of the Bioelectromagnetics Society, Maui,
Hawaii, USA, p. 71, 2003.
16. S. Ebert, S. J. Eom, J. Schuderer, C. Dasenbrock, T. Tillmann,
and N. Kuster. “Thermal thresholds of restrained RF exposed
mice at 905 MHz”, IEEE ICES/COST 281 Thermal Physiological Workshop, Paris, France, 2004.
17. S. Ebert, S. J. Eom, J. Schuderer, C. Dasenbrock, T. Tillmann,
and N. Kuster. “Thermal thresholds of restrained RF exposed
mice at 905 MHz”, Proceedings of the 3rd International Workshop on Biological Effects of EMF, Kos, Greece, pp. 505-510,
2004.
18. N. Kuster, J. Schuderer, A. Christ, P. Futter, and S. Ebert.
“Guidance for exposure design of human studies addressing
health risk evaluations of mobile phones”, Bioelectromagnetics,
vol. 25, no. 7, pp. 524-529, 2004.
19. V. Berdiñas Torres, S. Ebert, A. Klingenböck, J. Fröhlich, and
N. Kuster. “Dosimetry and uncertainty assessment of PERFORM A exposure systems”, Proceedings of the 27th Annual
174
PUBLICATIONS
Meeting of the Bioelectromagnetics Society, Dublin, Ireland, p.
96, 2005.
20. S. Ebert, C. Dasenbrock, T. Tillmann, and N. Kuster. “Thermal response and threshold measurements in mice exposed to
905 MHz”, Proceedings of the 27th Annual Meeting of the Bioelectromagnetics Society, Dublin, Ireland, pp. 173-174, 2005.
21. L. Hillert, T. Åkerstedt, N. Kuster, A. Lowden, M. Berg, C. Wiholm, S. Ebert, S. D. Moffat, and B. B. Arnetz. “The effects
of 900 MHz GSM wireless communication signal on subjective
symptoms, physiological reactions, alertness, performance and
sleep, an experimental provocation study”, Proceedings of the
27th Annual Meeting of the Bioelectromagnetics Society, Dublin,
Ireland, pp. 384-385, 2005.
22. S. Ebert, N. Nikoloski, V. Berdiñas Torres, J. Fröhlich,
and N. Kuster. “In vivo exposure setups for large-scale
toxicity/carcinogenicity studies with rats at 900/1800 MHz”,
Progress in Electromagnetics Research Symposium (PIERS),
Hang Zhou, China, p. 618, 2005.
23. S. Ebert, J. Schuderer, and N. Kuster. “Flexible exposure setup
for humans provocative studies at 900 MHz”, Proceedings of the
28th General Assembly of International Union of Radio Science
(URSI), New Delhi, India, 2005.
24. L. Hillert, T. Åkerstedt, N. Kuster, A. Lowden, M. Berg, C. Wilholm, S. Ebert, S. D. Moffat, V. Musabasic, and B. B. Arnetz.
“The effects of 900 MHz GSM wireless communication signal
on subjective symptoms and physiological reactions, an experimental provocation study”, Proceedings of the 4th International
Workshop on Biological Effects of Electromagnetic Field, Crete,
Greece, pp. 16-20, 2006.
25. B. B. Arnetz, T. Åkerstedt, N. Kuster, A. Lowden, M. Berg,
C. Wiholm, S. Ebert, S. D. Moffat, and L. Hillert. “Mobile
phone use and health: self-rated health, neurocognitive function, neurophysiological effects using 900 MHz wireless communication signals. A laboratory-based exposure study”, Progress
175
in Electromagnetics Research Symposium (PIERS), Cambridge,
USA, p. 219, 2006.
26. W. Kainz, W. Oesch, N. Nikoloski, V. Berdiñas Torres, S. Ebert,
M. Frauscher, A. Klingenböck, J. Fröhlich, J.-C. Gröbli, and
N. Kuster. “Performance of the exposure systems developed
for the PERFORM A studies”, Proceedings of the International
Congress of the European BioElectromagnetics Association, Bordeaux, France, p. 53, 2007.
27. P. A. Smith, N. Kuster, S. Ebert, and H. J. Chevalier. “GSM
and DCS wireless communication signals: combined chronic toxicity/carcinogenicity study in the Wistar rat”, Proceedings of
the International Congress of the European BioElectromagnetics Association, Bordeaux, France, p. 54, 2007.
28. T. Tillmann, H. Ernst, S. Ebert, N. Kuster, W. Behnke, S. Rittinghausen, J. Buschmann, and C. Dasenbrock. “PERFORMA1: Carcinogenicity study of GSM and DCS wireless communication signals in B6C3F1 mice”, Proceedings of the International Congress of the European Bio Electromagnetics Association, Bordeaux, France, p. 55, 2007.
29. K. Rolfo, P. Yu, M. Carbonatto, G. Oberto, N. Kuster, S. Ebert,
and S. Tofani. “Carcinogenicity study in Pim1 transgenic mice
exposed to pulsed 900 MHz radiofrequency radiation”, Proceedings of the International Congress of the European BioElectromagnetics Association, Bordeaux, France, p. 56, 2007.
30. L. Hillert, T. Åkerstedt, A. Lowden, C. Wiholm, N. Kuster,
S. Ebert, C. Boutry, and B. B. Arnetz. “Effects of a 900 MHz
GSM exposure on self reported symptoms and blood chemistry,
an experimental provocation study”, Proceedings of the 29th Annual Meeting of the Bioelectromagnetics Society, Kanazawa-shi
Bunka Hall, Japan, pp. 98-99, 2007.
31. P.A. Smith, N. Kuster, S. Ebert, and H. J. Chevalier. “GSM
and DCS wireless communication signals: Combined chronic
toxicity/carcinogenicity study in the Wistar rat”, Radiation Research, 168, 4, pp. 480-492, 2007.
Bibliography
[1] R. B. Goldberg, “Literature resources for understanding biological effects of EM fields”, IEEE Engineering in Medicine and
Biology Magazine, vol. 15, no. 4, pp. 96–101, 1996.
[2] C. K. Chou, “Thirty-five years in bioelectromagnetics research”,
Bioelectromagnetics, vol. 28, no. 1, pp. 3–15, 2007.
[3] K. R. Foster and J. E. Molder, “Are mobile phones safe?”, IEEE
Spectrum, vol. 37, no. 8, pp. 23–28, 2000.
[4] US National Research Council, Risk Assessment in the Federal
Government: Managing the Process, National Academic Press,
Washington DC, USA, 1983.
[5] P. Slovic, The Perception of Risk, Earthscan Publications Ltd.,
2000.
[6] M. H. Repacholi and J. A. J. Stolwijk, “Criteria for evaluating
scientific literature and developing exposure limits”, Radiation
Protection in Australia, vol. 9, no. 3, pp. 97–84, 1991.
[7] N. Kuster, W. R. Adey, S. Ebert, and W. Oesch, “Selection and
implementation of specific EMF exposure conditions for bioexperiments in the context of health risk assessment.”, International Symposium on Electromagnetics in Biology and Medicine,
p. 67, 2001.
[8] K. R. Foster and M. H. Repacholi, “Biological effects of radiofrequency fields: Does modulation matter?”, Radiation Research,
vol. 162, no. 2, pp. 219–225, 2004.
177
178
BIBLIOGRAPHY
[9] N. Kuster and F. Schönborn, “Recommended minimal requirements and development guidelines for exposure setups of bioexperiments addressing the health risk concern of wireless communications”, Bioelectromagnetics, vol. 21, no. 7, pp. 508–14,
2000.
[10] N. Kuster, J. Schuderer, A. Christ, P. Futter, and S. Ebert,
“Guidance for exposure design of human studies addressing
health risk evaluations of mobile phones”, Bioelectromagnetics,
vol. 25, pp. 524–529, 2004.
[11] M. H. Repacholi, “Low-level exposure to radiofrequency electromagnetic fields: Health effects and research needs”, Bioelectromagnetics, vol. 19, no. 1, pp. 1–19, 1998.
[12] M. H. Repacholi and B. Greenebaum, “Interaction of static and
extremely low frequency electric and magnetic fields with living
systems: Health effects and research needs”, Bioelectromagnetics, vol. 20, no. 3, pp. 133–160, 1999.
[13] V. Hansen, “Guidelines for experiments to investigate the effect
of radio frequency electromagnetic fields on biological systems”,
Forschungsgemeinschaft Funk e. V., vol. 18, May 2003.
[14] C. K. Chou, H. Bassen, J. Osepchuk, Q. Balzano, R. Petersen,
M. Meltz, R. Cleveland, J. C. Lin, and L. Heynick, “Radio
frequency electromagnetic exposure: Tutorial review on experimental dosimetry”, Bioelectromagnetics, vol. 17, no. 3, pp.
195–208, 1996.
[15] N. Kuster, V. Berdiñas Torres, N. Nikoloski, M. Frauscher, and
W. Kainz, “Methodology of detailed dosimetry and treatment
of uncertainty and variations for in vivo studies”, Bioelectromagnetics, vol. 27, no. 5, pp. 378–391, 2006.
[16] A. Ahlbom, E. Cardis, A. Green, M. Linet, D. Savitz, and
A. Swerdlow, “Icnirp standing committee on epidemiology: Review of the epidemiologic literature on EMF and health”, Environmental Health Perspectives, vol. 109, no. 6, pp. 911–933,
2001.
BIBLIOGRAPHY
179
[17] Q. Balzano and A. R. Sheppard, “The influence of the precautionary principle on science-based decision-making: questionable applications to risks of radiofrequency fields”, Journal of
Risk Research, vol. 5, no. 4, pp. 351–369, 2002.
[18] “Hochfrequente Strahlung und Gesundheit - Bewertung von wissenschaftlichen Studien im Niedrigdosisbereich”, Tech. Rep.,
Bundesamt für Umwelt BAFU, Sep 2006.
[19] K. R. Foster and R. Glaser, “Thermal mechanisms of interaction
of radiofrequency energy with biological systems with relevance
to exposure guidelines”, Health Physics, vol. 92, no. 6, pp. 609–
620, 2007.
[20] M. Röösli, “Radiofrequency electromagnetic field exposure and
non-specific symptoms of ill health: A systematic review”, Environmental Research, vol. 107, no. 2, pp. 277–287, 2008.
[21] M. Röösli, R. Rapp, and C. Braun-Fahrländer, “Radio and
microwave frequency radiation and health – an analysis of the
literature”, Gesundheitswesen, vol. 65, no. 6, pp. 378–392, 2003.
[22] D. Krewski, B. W. Glickman, R. W. Habash, B. Habbick,
R. Lotz, W. G. Mandeville, F. S. Prato, T. Salem, and D. F.
Weaver, “Recent advances in research on radiofrequency fields
and health: 2001-2003”, Journal of Toxicology and Environmental Health Part B, vol. 10, no. 4, pp. 287–318, 2007.
[23] M. Kundi, “The controversy about a possible relationship between mobile phone use and cancer”, Environmental Health
Perspectives, vol. 117, no. 3, pp. 316–324, 2009.
[24] “IEEE standard for safety levels with respect to human exposure
to radio frequency electromagnetic fields, 3 kHz to 300 GHz”,
IEEE (Std C95.1), 1999.
[25] A. Ahlbom, U. Bergqvist, J. H. Bernhardt, J. P. Cesarini, L. A.
Court, M. Grandolfo, M. Hietanen, A. F. McKinlay, M. H. Repacholi, D. H. Sliney, J. A. J. Stolwijk, M. L. Swicord, L. D. Szabo,
M. Taki, T. S. Tenforde, H. P. Jammet, and R. Matthes, “Guidelines for limiting exposure to time-varying electric, magnetic,
180
BIBLIOGRAPHY
and electromagnetic fields (up to 300 GHz)”, Health Physics,
vol. 74, no. 4, pp. 494–522, 1998.
[26] J. M. Osepchuk and R. C. Petersen, “Historical review of RF
exposure standards and the international committee on electromagnetic safety (ICES)”, Bioelectromagnetics, vol. 6, pp.
S7–S16, 2003.
[27] “World Health Organization international EMF project
(http://www.who.int/peh-emf/project/en/)”.
[28] M. H. Repacholi, “WHO’s international EMF project”, Radiation Protection Dosimetry, vol. 83, no. 1-2, pp. 1–4, 1999.
[29] E. R. Adair and D. R. Black, “Thermoregulatory responses to
RF energy absorption”, Bioelectromagnetics, vol. 6, pp. S17–
S38, 2003.
[30] E. R. Adair, B. W. Adams, and S. K. Hartman, “Physiological
interaction processes and radiofrequency energy-absorption”,
Bioelectromagnetics, vol. 13, no. 6, pp. 497–512, 1992.
[31] E. R. Adair, B. W. Adams, and S. A. Kelleher, “Resonant CW
and pulsed fields of identical average power densities provoke
similar thermoregulatory adjustments by squirrel monkeys”, in
Proceedings of the 15th Annual Meeting of the Bioelectromagnetic Society, 1993, p. 63.
[32] C. J. Gordon and E. C. White, “Distinction between heating rate and total heat absorption in the microwave-exposed
mouse”, Physiological Zoology, vol. 55, no. 3, pp. 300–308, 1982.
[33] C. J. Gordon, “Behavioral and autonomic thermoregulation
in mice exposed to microwave-radiation”, Journal of Applied
Physiology, vol. 55, no. 4, pp. 1242–1248, 1983.
[34] C. J. Gordon, “Influence of heating rate on control of heat-loss
from the tail in mice”, American Journal of Physiology, vol.
244, no. 6, pp. R778–R784, 1983.
BIBLIOGRAPHY
181
[35] C. J. Gordon, “Effects of ambient temperature and exposure
to 2450 MHz microwave radiation on evaporative heat loss in
the mouse”, Journal of Microwave Power, vol. 17, no. 2, pp.
145–50, 1982.
[36] C. J. Gordon, “Body temperature in the mouse, hamster, and
rat exposed to radiofrequency radiation: an interspecies comparison”, Journal of Thermal Biology, 1985.
[37] S.-T. Lu, N. A. Lebda, S.-J. Lu, S. Pettit, and S. M. Michaelson, “Effects of microwaves on three different strains of rats”,
Radiation Research, vol. 110, pp. 173–191, 1987.
[38] C. J. Gordon, “Relationship between autonomic and behavioral
thermoregulation in the mouse”, Physiological And Behavior,
vol. 34, pp. 687–690, 1984.
[39] C. J. Gordon, “Temperature regulation in the mouse and hamster exposed to microwaves in hot environment”, Health Physics,
1985.
[40] M. H. Repacholi, “Health risks from the use of mobile phones”,
Toxicology Letters, vol. 120, no. 1-3, pp. 323–331, 2001.
[41] S. J. Eom, J. Fröhlich, N. Nikoloski, and N. Kuster, “An exposure system for behavioral studies with a small number of mice
at 905 MHz”, in Proceedings of the 25th Annual Meeting of the
Bioelectromagnetic Society, Maui, Hawaii, 2003, p. 17.
[42] R. Bartram, B. Billaudel, A. Bottomley, J.-C. Cassel,
E. Cordelli, B. Cosquer, A. Edwards, S.-J. Eom, J. Fröhlich,
R. Galani, J.-C. Gröbli, N. Kuster, A. Höytö, N. Jones, I. Lagroye, D. Lloyd, G. Lovisolo, J. Luukkonen, c. Marino, J. Moquet, J. Naarala, N. Nikoloski, W. Oesch, A. Pereira de Vasconcelos, R. Pinto, G. Ruffi, J. Schuderer, Z. Sienkiewicz,
M. Sokura, D. Spät, L. Stronati, M. Taxile, A. Testa, B. Veyret,
and P. Villani, “PERFORM B: In vitro and in vivo replication
studies related to mobile telephones and base stations (final report)”, Tech. Rep., Nov 2004.
182
BIBLIOGRAPHY
[43] M. S. Johnson, S. M. Thomson, E. Krol, and J. R. Speakman,
“Limits to sustained energy intake during lactation in the mouse
mus musculus”, The Journal of Experimental Biology, vol. 15,
no. 4, pp. A414–A414, 2001.
[44] G. S. Habicht, “Body temperature in normal and endotoxintreated mice of different ages”, Mechanism of Ageing and Development, vol. 16, no. 1, pp. 97–104, 1981.
[45] Q. Balzano, C. K. Chou, R. Cicchetti, A. Faraone, and R. Y. S.
Tay, “An efficient RF exposure system with precise wholebody average SAR determination for in vivo animal studies at
900 MHz”, IEEE Transactions on Microwave Theory and Techniques, vol. 48, no. 11, pp. 2040–2049, 2000.
[46] S. J. Eom, J. Fröhlich, N. Nikoloski, J.-C. Gröbli, and N. Kuster,
“PERFORM B: Report on mini wheel”, Tech. Rep., Foundation
for Research on Information Technology in Society, 2004.
[47] B. N. Taylor and C. E. Kuyatt, “Guidelines for evaluating and
expressing the uncertainty of NIST measurement results”, Tech.
Rep., National Institute of Standards and Technology, 1994.
[48] V. Berdiñas Torres, S. Ebert, A. Klingenböck, J. Fröhlich, and
N. Kuster, “Dosimetry and uncertainty assessment of PERFORM A exposure systems”, in Proceedings of the 27th Annual Meeting of the Bioelectromagnetic Society, Dublin, Ireland,
2005.
[49] R. K. Porter, “Allometry of mammalian cellular oxygen consumption”, Cellular and Molecular Life Sciences, vol. 58, no.
5-6, pp. 815–822, 2001.
[50] P. S. Dodds, D. H. Rothman, and J. S. Weitz, “Re-examination
of the ”3/4-law” of metabolism”, Journal of Theoretical Biology,
vol. 209, no. 1, pp. 9–27, 2001.
[51] M. S. Johnson, S. C. Thomson, and J. R. Speakman, “Limits
to sustained energy intake ii. inter-relationships between resting
BIBLIOGRAPHY
183
metabolic rate, life-history traits and morphology in mus musculus”, The Journal of Experimental Biology, vol. 204, no. 11,
pp. 1937–1946, 2001.
[52] G. F. Pedersen, “Amplitude modulated RF fields stemming
from a GSM/DCS-1800 phone”, Wireless Networks, vol. 3, no.
6, pp. 489–98, 1997.
[53] G. F. Pedersen and J. B. Andersen, “RF and ELF exposure from
cellular phone handsets: TDMA and CDMA systems”, Radiation Protection Dosimetry, vol. 83, no. 1-2, pp. 131–8, 1999.
[54] T. Linde and K. H. Mild, “Measurement of low frequency magnetic fields from digital cellular telephones”, Bioelectromagnetics, vol. 18, no. 2, pp. 184–6, 1997.
[55] B. Heath, S. Jenvey, and I. Cosic, “Investigation of analogue and
digital mobile phone low frequency radiation spectrum characteristics”, in Proceedings of the 2nd International Conference
on Bioelectromagnetism, Melbourne, Australia, 1998.
[56] K. Jokela, L. Puranen, and A. P. Sihvonen, “Assessment of the
magnetic field exposure due to the battery current of digital
mobile phones”, Health Physics, vol. 86, no. 1, pp. 56–66, 2004.
[57] N. Kuster, R. Kastle, and T. Schmid, “Dosimetric evaluation of
handheld mobile communications equipment with known precision”, IEICE Transactions on Communications, vol. E80-B, no.
5, pp. 645–652, 1997.
[58] J. D. Jackson, Klassische Elektrodynamik, Walter de Gruyter,
Berlin/New York, 1982.
[59] “IEEE standard: Recommended practice for determining the
peak spatial-average specific absorption rate (SAR) in the human head from wireless communications devices: Measurement
techniques”, IEEE (Std 1528), 2003.
[60] R. Huber, J. Schuderer, T. Graf, K. Jutz, A. A. Borbely,
N. Kuster, and P. Achermann, “Radio frequency electromagnetic field exposure in humans: Estimation of SAR distribution
184
BIBLIOGRAPHY
in the brain, effects on sleep and heart rate”, Bioelectromagnetics, vol. 24, no. 4, pp. 262–276, 2003.
[61] M. Misakian, A. R. Sheppard, D. Krause, M. E. Frazier, and
D. L. Miller, “Biological, physical, and electrical parameters for
in vitro studies with ELF magnetic and electric-fields”, Bioelectromagnetics, pp. 1–73, 1993.
[62] B. B. Arnetz, “The effects of 900 MHz GSM wireless communication signals on subjective symtons, physiological reactions,
alertness, performance and sleep; an experimental provocation
study, perform c study proposal”, Tech. Rep., Karolinska Institute, Stockholm, Sweden, 18. June 2002.
[63] J. Schuderer, EMF Risk Assessment: ”In Vitro” Research and
Sleep Studies, PhD thesis, Swiss Federal Institute of Technology,
2003.
[64] R. Huber, V. Treyer, A. A. Borbely, J. Schuderer, J. M.
Gottselig, H. P. Landolt, E. Werth, T. Berthold, N. Kuster,
A. Buck, and P. Achermann, “Electromagnetic fields, such as
those from mobile phones, alter regional cerebral blood flow and
sleep and waking EEG”, Journal of Sleep Research, vol. 11, no.
4, pp. 289–295, 2002.
[65] R. C. Johnson, Antenna Engineering Handbook, McGraw-Hill,
New York, 3rd edition, 1993.
[66] R. Garg, P. Bhartia, I. Bahl, and A. Ittipiboon, Microstrip
Antenna Design Handbook, Artech House, Boston, 1st edition,
2001.
[67] K-L. Wong, Compact and Broadband Microstrip Antennas, John
Wiley & Sons, New York, 1st edition, 2002.
[68] G. Kumar and K. P. Ray, Broadband Microstrip Antennas,
Artech House, Norwood, 1st edition, 2003.
[69] W. L. Stutzman and G. A. Thiele, Antenna Theory and Design,
John Wiley & Sons, 2nd edition, 1998.
BIBLIOGRAPHY
185
[70] C. A. Balanis, Antenna Theory - Analysis and Design, John
Wiley and Sons, 2nd edition, 1997.
[71] J. D. Kraus, Antennas, John Wiley, 2nd edition, 1988.
[72] J. R. James and P. S. Hall, Handbook of Microstrip Antennas,
vol. 1, Peter Peregrinus Ltd., London, 1989.
[73] H. F. Lee and W. Chen, Advances in Microstrip and Printed
Antennas, John Wiley & Sons, New York, 1997.
[74] V. Anderson and J. Rowley, “Measurements of skin surface
temperature during mobile phone use”, Bioelectromagnetics,
vol. 28, no. 2, pp. 159–62, 2007.
[75] A. Straume, G. Oftedal, and A. Johnsson, “Skin temperature
increase caused by a mobile phone: a methodological infrared
camera study”, Bioelectromagnetics, vol. 26, no. 6, pp. 510–9,
2005.
[76] E. B. Elabassi, “Mobile phone use and temporal skin heat sensation”, in Proceedings of the 28th General Assembly of International Union of Radio Science (URSI), New Delhi, India, 2005,
p. KP18.
[77] B. B. Arnetz and L. Hillert, “Personal communication: The size
of the heated surface contact area is not primarily important for
the psychological effect of a heat stimulus”, Dec 2003.
[78] D. J. Tisserand, H. Bosma, M. P. J. Van Boxtel, and J. Jolles,
“Head size and cognitive ability in nondemented older adults
are related”, Neurology, vol. 56, pp. 969–971, 2001.
[79] N. Kuster, Q. Balzano, and J. C. Lin, Mobile Communications
Safety, Springer, 1997.
[80] L. Hillert, T. Åkerstedt, A. Lowden, C. Wiholm, N. Kuster,
S. Ebert, C. Boutry, S. D. Moffat, M. Berg, and B. B. Arnetz,
“The effects of 884 MHz GSM wireless communication signals
on headache and other symptoms: an experimental provocation
study”, Bioelectromagnetics, vol. 29, no. 3, pp. 185–196, 2008.
186
BIBLIOGRAPHY
[81] N. Chavannes, Local Mesh Refinement Algorithms for Enhanced
Modeling Capabilities in the FDTD Method, PhD thesis, Swiss
Federal Institute of Technology, 2002.
[82] A. Christ, Analysis and Improvement of the Numerical Properties of the FDTD Algorithm, PhD thesis, Swiss Federal Institute
of Technology, 2003.
[83] T. Schmid, O. Egger, and N. Kuster, “Automated E-field scanning system for dosimetric assessments”, IEEE Transactions on
Microwave Theory and Techniques, vol. 44, no. 1, pp. 105–113,
1996.
[84] M. Burkhardt and N. Kuster, “Appropiate modeling of the ear
for compliance testing of handheld MTE with SAR safety limits
at 900/1800 MHz”, IEEE Transactions on Microwave Theory
and Techniques, vol. 48, pp. 1921–1934, 2000.
[85] C. Gabriel, S. Gabriel, and E. Corthout, “The dielectric properties of biological tissues: I. Literature survey”, Physics in
Medicine and Biology, vol. 41, no. 11, pp. 22312249, 1996.
[86] M. Murbach, S. Kühn, M. Christopoulou, A. Christ, P. Achermann, and N. Kuster, “Evaluation of artifacts by EEG electrodes during RF exposures”, in Proceedings of the 31th Annual
Meeting of the Bioelectromagnetics Society, Davos, Switzerland,
2009, pp. 62–63.
[87] D. J. Tisserand, P. J. Visser, M. P. J. van Boxtel, and J. Jolles,
“The relation between global and limbic brain volumes on MRI
and cognitive performance in healthy individuals across the age
range”, Neurobiology of Aging, vol. 21, no. 4, pp. 569–576, 2000.
[88] V. Berdiñas Torres, Exposure Systems and Dosimetry of LargeScale In Vivo Studies, PhD thesis, Swiss Federal Institute of
Technology, 2007.
[89] N. Nikoloski, W. Kainz, and N. Kuster, “PERFORM A rat
setup”, in Proceedings of the 24th Annual Meeting of the Bioelectromagnetics Society, Quebec, Canada, 2002.
BIBLIOGRAPHY
187
[90] W. Kainz, N. Nikoloski, W. Oesch, V. Berdiñas Torres,
J. Fröhlich, G. Neubauer, and N. Kuster, “Development of
novel whole-body exposure setups for rats providing high efficiency, national toxicology program (NTP) compatibility and
well-characterized exposure”, Physics in Medicine and Biology,
vol. 51, no. 20, pp. 5211–5229, 2006.
[91] S. Ebert, N. Nikoloski, V. Berdiñas Torres, J. Fröhlich, and
N. Kuster, “In vivo exposure setups for large-scale toxicity/carcinogenicity studies with rats at 900/1800 MHz”, in
Progress in Electromagnetics Research Symposium (PIERS),
Hangzhou, China, 2005, p. 618.
[92] R. Hrubya, G. Neubauer, N. Kuster, and M. Frauscher,
“Study on potential effects of 902 MHz GSM-type wireless communication signals on DMBA-induced mammary tumours in
SpragueDawley rats”, Mutation Research/Genetic Toxicology
and Environmental Mutagenesis, vol. 649, no. 1-2, pp. 34–44,
2008.
[93] P. A. Smith, N. Kuster, S. Ebert, and H. J. Chevalier, “GSM
and DCS wireless communication signals: Combined chronic
toxicity/carcinogenicity study in the Wistar rat”, Radiation
Research, vol. 168, no. 4, pp. 480–492, 2007.
[94] D. Yu, Y. Shen, N. Kuster, Y. Fu, and H. Chiang, “Effects
of 900 MHz GSM wireless communication signals on DMBAinduced mammary tumors in rats”, Radiation Research, vol.
165, no. 2, pp. 174–180, 2006.
[95] P. A. Smith, N. Kuster, S. Ebert, and H. J. Chevalier, “GSM
and DCS wireless communication signals: Combined chronic
toxicity/carcinogenicity study in the Wistar rat”, in Proceedings of the 8th European BioElectromagnetics Association, Bordeaux, France, 2007, p. 54.
[96] M. H. Repacholi and E. Cardis, “Criteria for EMF health risk
assessment”, Radiation Protection Dosimetry, vol. 72, no. 3-4,
pp. 305–312, 1997.
188
BIBLIOGRAPHY
[97] Fraunhofer Institute of Toxicology and Experimental Medicine,
“PERFORM A study plan: Carcenogenicity study of GSM
(902 MHz) and DCS (1747 MHz) wireless communication signals
in B6C3F1 mice. (Fraunhofer ITEM study 01G01035)”, Tech.
Rep., Oct 2001.
[98] Institutio di Ricerche Biomediche ”Antoine Marxer”, “PERFORM A study plan: Lymphoma induction and carcinogenicity study in Pim1 transgenic mice exposed to pulsed 900 MHz
electromagnetic fields. (RBM study R05720)”, Tech. Rep., 2001.
[99] M. H. Repacholi, A. Basten, V. Gebski, D. Noonan, J. Finnie,
and A. W. Harris, “Lymphomas in E µ-Pim1 transgenic mice
exposed to pulsed 900 MHz electromagnetic fields”, Radiation
Research, vol. 147, no. 5, pp. 631–640, 1997.
[100] T. D. Utteridge, V. Gebski, J. W. Finnie, B. Vernon-Roberts,
and T. R. Kuchel, “Long-term exposure of E µ-Pim1 transgenic
mice to 898.4 MHz microwaves does not increase lymphoma incidence”, Radiation Research, vol. 158, no. 3, pp. 357–364, 2002.
[101] W. Oesch, Controlling software for EMF laboratory studies,
PhD thesis, Swiss Federal Institute of Technology, 2006.
[102] “ETSI standard - european digital cellular telecommunication
system (phase 2); radio subsystem link control”, ETSI Standard:
GSM 05.08 (ETS 300 578), 1999.
[103] N. Kuster and W. R. Adey, “Criteria for selecting specific EMF
exposure conditions for bio-experiments in the context of health
risk assessment”, in Proceedings of the 23rd Annual Meeting of
the Bioelectromagnetics Society, St. Paul, USA, 2001, p. 24.
[104] R. Mertens, W. Kainz, and N. Kuster, “Simulating environmental GSM features for use in bioexperiments”, in Proceedings
of the 24th Annual Meeting of the Bioelectromagnetics Society,
Quebec, Canada, 2002, p. 106.
[105] J. Wiart, C. Dale, A. V. Bosisio, and A. Le Cornec, “Analysis of
the influence of the power control and discontinuous transmission on RF exposure with GSM mobile phones”, IEEE Trans-
BIBLIOGRAPHY
189
actions on Electromagnetic Compatibility, vol. 42, no. 4, pp.
376–385, 2000.
[106] R. Mertens, W. Kainz, and N. Kuster, “Proposal for simulating
environmental GSM exposure for the PERFORM and REFLEX
projects”, Tech. Rep., Foundation for Research on Information
Technology in Society, Sep 2001.
[107] V. Berdiñas Torres, J. Fröhlich, A. Klingenböck, and N. Kuster,
“PERFORM A - mouse numerical dosimetry (draft)”, Tech.
Rep., Foundation for Research on Information Technology in
Society, Oct 2004.
[108] K. Meier, M. Burkhardt, T. Schmid, and N. Kuster, “Broadband calibration of E-field probes in lossy media”, IEEE Transactions on Microwave Theory and Techniques, vol. 44, no. 10,
pp. 1954–1962, 1996.
[109] K. Pokovic, Advanced Electromagnetic Probes for Near-Field
Evaluations, PhD thesis, Swiss Federal Institute of Technology,
1999.
[110] F. Schönborn, Risk Assessment to EMF Exposure by Mobile
Phones: Optimization of RF Exposure Setups, PhD thesis, Swiss
Federal Institute of Technology, 2000.
[111] A. Faraone, S. Chebrolu, W. Luengas, M. Ballen, G. Bit-Babik,
M. Kanda, T. Babij, M. L. Swicord, and C. K. Chou, “Dosimetry of the 40-mice 900 MHz ”Ferris-wheel” exposure system”,
in Proceedings of the 6th European Bioelectromagnetics Association, Budapest, Hungary, 2003, p. 37.
[112] A. Faraone, W. Luengas, S. Chebrolu, M. Ballen, G. Bit-Babik,
A. V. Gessner, M. Y. Kanda, T. Babij, M. L. Swicord, and
C. K. Chou, “Radiofrequency dosimetry for the Ferris-wheel
mouse exposure system”, Radiation Research, vol. 165, no. 1,
pp. 105–112, 2006.
[113] B. D. Görlitz, M. Müller, S. Ebert, H. Hecker, N. Kuster, and
C. Dasenbrock, “Effects of 1-week and 6-week exposure to
GSM/DCS radiofrequency radiation on micronucleus formation
190
BIBLIOGRAPHY
in B6C3F1 mice”, Radiation Research, vol. 164, no. 4, pp. 431–
439, 2005.
[114] S. Ebert, W. Oesch, J. Fröhlich, J.-C. Gröbli, and N. Kuster,
“Exposure report - Effects of RF exposure on micronucleus formation in B6C3F1 mice (ITA study 01G01034 and 02G02009)”,
Tech. Rep., Foundation for Research on Information Technology
in Society, Oct 2002.
[115] S. Ebert, W. Oesch, J. Fröhlich, and N. Kuster, “Setup performance report PERFORM A 900 MHz mouse exposure system
at Fraunhofer ITEM”, Tech. Rep., Foundation for Research on
Information Technology in Society, Mar 2005.
[116] S. Ebert, W. Oesch, J. Fröhlich, and N. Kuster, “Setup performance report PERFORM A 1800 MHz mouse exposure system
at Fraunhofer ITEM”, Tech. Rep., Foundation for Research on
Information Technology in Society, Mar 2005.
[117] S. Ebert, W. Oesch, J. Fröhlich, and N. Kuster, “Setup performance report PERFORM A 900 MHz mouse exposure system
at RBM”, Tech. Rep., Foundation for Research on Information
Technology in Society, Mar 2005.
[118] K. Rolfo, P. Yu, M. Carbonatto, G. Oberto, N. Kuster, S. Ebert,
and S. Tofani, “Carcinogenicity study in Pim1 transgenic mice
exposed to pulsed 900 MHz radiofrequency radiation”, in Proceedings of the 8th European BioElectromagnetics Association,
Bordeaux, France, 2007, p. 56.
[119] T. Tillmann, H. Ernst, S. Ebert, N. Kuster, W. Behnke, S. Rittinghausen, and C. Dasenbrock, “Carcinogenicity study of GSM
and DCS wireless communication signals in B6C3F1 mice”, Bioelectromagnetics, vol. 28, no. 3, pp. 173–187, 2007.
[120] S. Ebert, J. Fröhlich, W. Oesch, U. Frauenknecht, and
N. Kuster, “Optimized in vivo exposure setups for risk assessment studies at the mobile communication frequencies 902 MHz
and 1747 MHz”, in Proceedings of the 23rd Annual Meeting of
the Bioelectromagnetic Society, St. Paul, USA, 2001, p. 27.
BIBLIOGRAPHY
191
[121] S. Ebert, J. Fröhlich, W. Oesch, R. Mertens, and N. Kuster,
“Characterization and dosimetry of the PERFORM A in vivo
exposure system at the mobile communication frequencies
902 MHz and 1747 MHz”, in Proceedings of the 24th Annual
Meeting of the Bioelectromagnetics Society, Quebec, Canada,
2002, pp. 109–110.
[122] S. Ebert, V. Berdiñas Torres, J. Fröhlich, and N. Kuster,
“Comparison study of wheel-like mouse exposure systems at
900 MHz”, in Proceedings of the 27th Annual Meeting of the
Bioelectromagnetic Society, Dublin, Ireland, 2005.
[123] G. Oberto, K. Rolfo, P. Yu, M. Carbonatto, S. Peano, N. Kuster,
S. Ebert, and S. Tofani, “Carcinogenicity study of 217 Hz-pulsed
900 MHz electromagnetic fields in Pim1 transgenic mice”, Radiation Research, vol. 168, no. 3, pp. 316–326, 2007.
[124] C. Ziemann, H. Brockmeyer, S. B. Reddy, Vijayalaxmi, T. J.
Prihoda, N. Kuster, T. Tillmann, and C. Dasenbrock, “Absence
of genotoxic potential of 902 MHz (GSM) and 1747 MHz (DCS)
wireless communication signals: In vivo two-year bioassay in
B6C3F1 mice”, International Journal of Radiation Biology, vol.
85, no. 5, pp. 454–464, 2009.
[125] M. Frauscher, Dosimetric Analysis and Uncertainty Assessment
of long-term, large-scale in vivo exposure setups at 902 MHz,
PhD thesis, Technical University of Graz, 2006.
[126] J. P. Holman, Heat transfer, vol. 7th Ed., McGraw-Hill, New
York, 1990.
[127] B. W. Wilson, A. Faraone, D. Sheen, M. L. Swicord, W. Park,
J. Morrissey, J. Creim, and L. E. Anderson, “Space efficient
system for small animal, whole body microwave exposure at
1.6 GHz”, Bioelectromagnetics, vol. 23, pp. 127–131, 2002.
[128] L. Gatta, R. Pinto, V. Ubaldi, L. Pace, P. Galloni, G. A. Lovisolo, C. Marino, and C. Piolo, “Effects of in vivo exposure to
GSM-modulated 900 MHz radiation on mouse peripheral lymphocytes”, Radiation Research, vol. 160, pp. 600–605, 2003.
192
BIBLIOGRAPHY
[129] V. W. Hansen, A. K. Bitz, and J. R. Streckert, “RF exposure
of biological systems in radial waveguides”, IEEE Transactions
on Electromagnetic Compatibility, vol. 41, no. 4, pp. 487–493,
1999.
[130] J.-C. Cassel, B. Cosquer, R. Galani, and N. Kuster, “Wholebody exposure to 2.45 GHz electromagnetic fields does not alter
radial-maze performance in rats”, Behavioural Brain Research,
vol. 155, no. 1, pp. 37–43, 2004.
[131] B. Cosquer, R. Galani, N. Kuster, and J.-C. Cassel, “Wholebody exposure to 2.45 GHz electromagnetic fields does not alter
anxiety responses in rats: a plus-maze study including test validation”, Behavioural Brain Research, vol. 156, no. 1, pp. 65–74,
2005.
[132] J. Naarala, A. Höytö, and A. Markkanen, “Cellular effects of
electromagnetic fields”, Alternatives to Laboratory Animals, vol.
32, no. 4, pp. 355–60, 2004.
[133] A. Höytö, J. Juutilainen, and J. Naarala, “Ornithine decarboxylase activity of l929 cells after exposure to continuous wave or
50 Hz modulated radiofrequency radiation - a replication study”,
Bioelectromagnetics, vol. 28, no. 7, pp. 501–508, 2007.
[134] M. Capstick, N. Kuster, S. Kühn, V. Berdiñas Torres, J. Ladbury, G. Koepke, D. McCormick, J. Gauger, and R. Melnick, “A
radio frequency radiation reverberation chamber exposure system for rodents”, in Proceedings of the 29th General Assembly
of International Union of Radio Science (URSI), 2008.
[135] R. Huber, A. A. Borbely, J. M. Gottselig, N. Kuster, H. P.
Landolt, J. Schuderer, E. Werth, and P. Achermann, “Effects
of electromagnetic fields of mobile phones on the human sleep
EEG”, Sleep, vol. 25, pp. A291–A292, 2002.