Piloting Patient Reported Outcomes Measurement Information System for Pain Studies
by
William Tyler Hayden
A Senior Honors Project Presented to the
Honors College
East Carolina University
In Partial Fulfillment of the
Requirements for
Graduation with Honors
by
William Tyler Hayden
Greenville, NC
May 2015
Approved by:
Dr. Sharon Gordon and Dr. Raymond Dionne
Department of Foundational Sciences, School of Dental Medicine
I hereby declare I am the sole author of this thesis. It is the result of my own work and is not
the outcome of work done in collaboration, nor has it been submitted elsewhere as
coursework for this or another degree.
Signed:
Date:____________________
William T. Hayden
2
Piloting Patient Reported Outcomes Measurement Information System for Pain Studies
William T. Hayden
Department of Foundational Sciences, School of Dental Medicine, East Carolina University
Abstract: Pain is a therapeutic challenge as well as a public health problem that affects over
116 million American adults; reduces quality of life; and is estimated to cost up to $635
billion annually [1]. There is a growing recognition that health care outcomes will be
improved by matching proven effective treatments with knowledge of patients’ unique
characteristics to optimize efficacy and safety. This project aims to assess the validity of the
Patient Reported Outcomes Measurement Information System (PROMIS) for pain. PROMIS
is a computerized system measuring patient-reported outcomes across a wide range of
chronic conditions but has not yet been studied for acute pain. In this pilot study, PROMIS
questionnaires will be given to healthy volunteers and, in a future study, the responses will be
compared to those of patients with acute pain following the removal of wisdom teeth. We
hypothesize that self-reported responses to PROMIS domains of physical, mental, and social
health in patients with acute pain will differ from those of healthy volunteers. We predict that
we can categorize patients into sub-groups based off PROMIS domains hypothesized to be
related to acute pain. PROMIS item banks to be studied include pain intensity, pain
interference, fatigue, anxiety, depression, the ability to participate in social roles and
activities, physical function, and sleep disturbance. In this pilot study, a preliminary PROMIS
study will be designed and tested for its ability to determine a healthy population and
functionality in clinical research for pain. Validation of PROMIS for acute pain will allow
for a more comprehensive phenotyping in future acute pain studies known as deep
3
phenotyping. Combined with genomic data and quantitative sensory testing, PROMIS can
help eliminate observer-based perceptions of patients’ pain and allow for more specific drug
therapy.
4
Acknowledgments
I would sincerely like to thank the people and organizations that made this study possible: I
am especially grateful to my project advisors, Dr. Sharon Gordon and Dr. Raymond Dionne,
for their knowledge and guidance. I am also thankful for the support and direction from Dr.
Benjamin Putnam, Debra Peaden, and Gerard Camargo. Additionally, I would like to thank
the East Carolina School of Dental Medicine and the East Carolina Honors College for
resources, encouragement, and assistance.
5
Table of Contents
Introduction
8
Research Problem and Background Analysis
8
Relevant Literature
9
Methods
15
Study Design
15
Collection of Data and Information
15
Analysis of Data
17
Results
18
Discussion
22
References
25
6
List of Figures and Tables
Figure 1: Sample Score Report For PROMIS Pain Questionnaires
16
Table 1: Participant Data
19
Figure 2: Data Plots Comparing Gender
21
7
Introduction:
Research Problem and Background Analysis: Pain is a therapeutic challenge as well
as a public health problem that is estimated to affect over 116 million American
adults, reduces quality of life, and is estimated to cost up to $635 billion annually
[1]. Growing recognition of the need for evidence-based, individual-centered
treatment strategies raises expectations that health care will be improved by
matching proven effective treatments with knowledge of patients’ unique
characteristics to optimize efficacy and safety. Essential to the goal of matching
treatments to patients to enhance analgesic drug development and therapy is
identification of intermediate phenotypes that capture the mechanistic complexity,
genetic expression and epigenetic changes of hundreds of ongoing processes and
mediators that influence treatment efficacy and safety and may form the basis for
differential responses to drug therapy. The ability to identify functional variants in
the genomic responses to pain and therapeutics at the sub -group and patient levels,
however, has been limited to date by lack of thorough phenotyping for patients with
acute pain [2].
The need for a more comprehensive understanding of human phenotypes has
spawned a new method of phenotyping studies referred to as “deep phenotyping.”
Deep phenotyping for pharmacogenomic studies requires both breath and depth to
better interpret the complexities of genomic variations that may underlie individual
differences in pain report. One approach to address this complexity is to use
quantitative testing of clinical features to identify more homogeneous subsets within
a group of patients with a given diagnosis or characteristic. Variations in
8
quantitative measures may identify intermediate phenotypes that are genetically less
complex yet have potentially stronger signals closer to the site of gene action. In
pain, quantitative testing is often termed “quantitative sensory testing”, or QST.
Another approach is to focus on patient-reported outcomes. Understanding how
patients report their pain experience is essential to matching p roven and effective
treatments to their disease status [3]. In this experiment, patients will report their
pain experience through the Patient Reported Outcomes Measurement Information
System (PROMIS). This set of questionnaires comprises many domains rela ted to
pain and will help clinicians understand the effect of pain on a patient’s life. Before
genomic testing and QST, PROMIS measures must be validated for pain studies.
While individual questionnaires related to pain have been validated, they have not
been assessed in relation to other patient experiences, thus further research is
needed.
The goal of the study was to pilot the Patient Reported Outcomes Measurement
Information System (PROMIS) for pain studies. We hypothesized that self-reported
responses to PROMIS domains of physical, mental, and social health in patients
with pain will differ from those of healthy volunteers. Additionally, we predict ed
that we can categorize patients into sub-groups based off PROMIS questionnaires
related to pain.
Relevant Literature: Pain is a bodily sensation that serves as a warning signal of
actual or potential tissue damage. Overall, pain can be classified into three distinct
neurobiological categories. First, pain classified as nociceptive is a protective
system used to minimize contact with noxious stimuli. Secondly, pain can be
9
adaptive and protective by heightening sensory sensitivity after tissue damage. This
type of pain assists in healing by reducing further damage and promoting recovery
and is known as inflammatory pain. Finally, pain can be maladaptive and result in
abnormal functioning to the central nervous system. This type of pathological pain
is characterized as a diseased sate caused by neuropathic or dysfunctional damage to
the nervous system. While pain is uncomfortable, it is in many ways essential for
maintaining bodily integrity. Lack of pain due to genetic mutations or excessive use
of drug therapy can be harmful because pain can lose its protective role [4].
Clinically, pain can be diagnosed as either acute or chronic. While this study
focuses on acute pain in patients following the removal or third molars, there is an
underlying relationship between acute and chronic pain. Acute pain is typically
defined as pain that has a quick onset, can possibly be severe, and lasts a relatively
short time. Chronic pain is a disease-state itself and can last over a longer period
than acute pain. Chronic pain poses severe health problems to patients and many
times, patients may have coexisting chronic pain conditions. By better understanding
the mechanism acute pain and how acute pain transitions into chronic pain,
researchers and clinicians can better target drug therapy and patient -centered
treatment.
Nociceptors are specialized sensory receptors that detect noxious stimuli and
transform it into electrical signals that are conducted to the central nervous system.
The process of nociception involves multiple steps that are carried out by specific
molecules. Some of these molecules increase pain sensitivity while others inhibit
pain sensitivity. Genetic variation of these molecules alters the accuracy of the
10
signal to and from the brain. Overall, five events are needed for a nociceptor to relay
pain information to the central nervous system; signal transduction, action potential
generation, transmission of the action potential to the central nervous system,
second-order neuron activation to transmit the signal to the thalamus, and thirdorder neuron transmission of the signal to the cerebral cortex, where the nociceptive
stimulus is perceived as pain. Each process is controlled by a different set of
receptor proteins. By understanding the pain mechanisms existing in the human
body, researchers can begin to understand how genetic variation, pain relief drugs,
and other factors contribute to the pain experience [5].
While genomic-wide association studies (GWAS) been proven useful for
identifying hundreds of loci associated with diseases and conditions, the translation
of the results to clinical practice has met some difficulty [6]. As the cost of
genomic and proteomic methods have declined and technological advancements
continue, it is becoming routine to use genomic and proteomic scanning for
discovery research. A major barrier to the identification of novel targets for
investigation is the identification of phenotypes as a result of inter -individual
variations. In relation to medical and clinical studies, phenotype is d efined in terms
of a deviation from normal morphology, physiology, or behavior [2]. The challenge
in phenotyping individuals with the same clinical diagnosis is that the individuals
might have entirely different sets of risk alleles, creating heterogeneous groups of
potentially overlapping disorders on a genetic level. To deal with the complexity of
these heterogeneous groups, one approach is to incorporate quantitative measures to
identify homogeneous subsets in a given diagnosis [7].
11
Deep phenotyping is defined as the precise and comprehensive analysis of
phenotypic abnormalities in which the individual components of the phenotype are
observed and described [2]. Not only does deep phenotyping involve abnormalities
or variations from the norm, but also patient responses to types of treatment.
Combined with existing genomic data and quantitative measures, deep phenotyping
has the potential to identify clinical complexity and therapeutic implications of
disease subtypes, as well as reduce the inaccuracy and misclassification of disease
outcomes [2, 3]. Deep phenotyping should be engaged more for the following
reasons: First, emphasis on clinical outcomes is necessary but not sufficient for the
advancement of medicine. Second, analytical and biological variati ons tend to dilute
statistical power and strength of association. Finally, phenotypes may vary across
the lifespan of an individual [8]. Deep phenotyping allows for individual variations
in pain reports to be better understood based on quantitative measures and genetic
influences that occur during the pain experience. Individual gene mutations can now
be associated with clinical diagnoses and responses to various treatments can be
related to molecular pathologies.
Conventional measures of disease status do not fully capture the ways that
diseases and their treatment affect individuals [9-11]. QST and patient self-report of
subjective states will be combined with the identification of pain related biomarkers
to help phenotype subgroups of acute pain. The Patient-Reported Outcomes
Measurement Information System (PROMIS) enables efficient and interpretable
clinical trial research and clinical practice application of patient -reported outcomes.
PROMIS is funded by the National Institute of Health (NIH) to improve the tools for
12
measuring patient-reported health outcomes in clinical research by making them
more valid, reliable and generalizable to patient needs [12].
PROMIS is a computerized system measuring patient-recorded outcomes across a
wide range of chronic diseases and demographic characteristics using item banks to
measure key symptoms and health concepts. PROMIS allows clinical researchers to
access a common repository of items and computerized adaptive tests that can be
adapted to research needs. PROMIS comprises several domains and those of interest
to this study are: pain intensity, pain interference, fatigue, emotional distress
(anxiety and depression), ability to participate in social roles and activities, physical
function, and sleep disturbance [13]. Published reports demonstrate that
psychometric measures are valid and reliable in the PROMIS pain interference and
behavior item banks in healthy individuals as well as in specific health problems. In
the population sample used in the development of the P ROMIS pain panel, acute
versus chronic pain were not distinguished, thus additional research is needed to
examine the PROMIS pain behavior items in these different pain types [14]. In
addition, health-related quality of life, particularly when experienced with comorbid
disease, is substantially less than those without a chronic medical condition
diagnosis. Specific health problems evaluated in the development of the pain bank
included: cancer, rheumatoid arthritis, osteoarthritis, sleep disorders, anxiety a nd
depression, and chronic pain [15]. PROMIS also incorporates the use of item
response theory (IRT) and computerized adaptive testing (CAT) into the
development of item banks. IRT is a psychometric method that produces scores associated
with answers to the questionnaires. These scores provide the computer software, CAT, with
13
the information it needs to select the most appropriate follow-up question. The use of IRT
and CAT in clinical research has shown to reduce the number of questions for a participant to
answer and improve the development of questionnaire domains while maintaining the desired
degree of statistical precision [16]. Unique to this study is the adaptation of PROMIS
measurements for acute pain. PROMIS domains related to pain experiences were
developed for chronic conditions. To better understand the mechanisms of pain and
how acute pain evolves into chronic pain, a more thorough understanding of acute
pain experiences and its immediate effects on patient behavior is required. Thus, this
study explored the possible uses of PROMIS in pain studies by first designing an
experiment for healthy volunteers and in future studies for patients with acute pain.
14
Methods:
Study Design: The study was designed using the PROMIS Assessment Center. The
questionnaires chosen for the study were pain interference, depression, anxiety,
fatigue, physical function, sleep disturbance, pain behavior, and the ability to
participate in social roles. The study was designed using IRT testing software CAT
[16]. Additionally, participants were asked for demographic data related to age,
ethinicity, gender, and race.
Collection of Data and Information: The questionnaires were designed to be taken
on an iPad which was done for efficiency. IRT reduced the time participants needed
to complete the questionnaires while mainintaining a high degree of calibration. The
Assessment Center gathered participant data and generated a score report as seen in
Figure 1.
15
Figure 1: Sample Score Report For PROMIS Pain Questionnaires
Score
SE
Pain Behavior
54
2
Sleep Disturbance
56
3
Fatigue
49
3
Anxiety/Fear
50
3
Depressive
Symptoms/Sadness
53
3
Pain Interference
39
6
Better
Physical Function
Score
SE
61
3
Average
Better
Average
Worse
Worse
Figure 1. The diamond represents the estimated score. For each of the areas above, a score
of 50 is average for the United States general population. Most people will score between
40 and 60 and almost all people will score between 30 and 70. The Standard Error (SE) is a
statistical measure of variance and represents a “margin of error” around the estimated
score. The lines on either side of each diamond reflect the likely range of the actual score.
16
Analysis of Data: Participant data was analyzed using the statistical software SPSS
Statistics Version 22 manufactured by IBM. A descriptive analysis of participant
responses to questionnaires conducted and male and female gender was compared.
Descriptive statistics were used in this analysis to determine patterns that exists
between male and female participants based off of gender, other demographic data,
and responses to the questionnaires related to pain.
17
Results:
The sample (n=7) was divided between three male and four female particpants.
The data for the study is displayed in Table 1.
18
Table 1: Participant Data
Ability to
Participate in Social
Roles
Pain
Behavior
Sleep
Disturbance
Physical
Functioning
51.2
55.4
48.9
49.8
50.7
44.3
62.1
53.8
55.8
60.7
48.5
51.2
60.8
46.3
53.1
47.8
35.3
54.3
47.5
35.3
60.3
59.1
Pain
Interference
Gender
Age
44.7
52.8
2
26
49.6
52.7
38.7
2
39
56.1
51.2
49.9
51.1
1
50
57.5
47.4
56
49.4
38.7
1
20
46.3
63.5
33
37.9
34.2
38.7
2
33
51.2
54.3
46.1
48.5
51
45.7
46.4
1
50
35.3
57.1
63.5
56.1
36.6
38.9
38.7
1
23
Fatigue Anxiety Depression
Table 1. Participant data for the eight questionnaires as well as the a ge and gender (female=1, male=2) of participants.
Data shown is the mean score for each questionnaire answered by particpants. The average score for the population norm,
the average American person, is 50.
Plots of the data were generated comparing male and female participants. Figure 2
displays the plots from this study.
Figure 2: Participant Data Comparing Gender
Figure 2. Particpant data to the questionnaires comapring responses based on
gender. For each plot, the line represents the population no rm or the score of the
average American, 50. Each plot reprents the range of scores, the median, and the
error for responses to each questionnaire.
Discussion:
Reponses to the questionnaires related to the pain experience were gathered and
analyzed using descriptive statistics. The participants were compared based on
gender differences because females tend to report higher levels of pain than males
do [17]. Overall, questionnaires related to the pain experience confirmed that a
healthy population was tested. For example, in Figure 2, median score of each
gender was healthier than the average American for each questionnaire except for
pain behavior (female), fatigue (female), sleep disturbance (male), and anxiety
(female).
Overall, PROMIS was an efficient and useful resource for this study. The use of
IRT and CAT proved to shorten the time participants spent on the questionnaire
while revealing data related to how each participant compared to the average
American citizen. Validity studies of PROMIS CAT and short-form questionnaires
showed that CAT proved to offer greater correlation and precision when the same
number of questions were answered. Based on the data, the population tested in this
study was primarily healthy compared to the average American . While some mean
scores were slightly unhealthy compared to the population norm, further testing will
be needed to determine if this trend exists in the data. Additionally, the participant
scores from this study fell within the range or scores for a validity stu dy conducted
by PROMIS except for anxiety, depression, and the ability to participate in social
roles for the male population. While the scores were outside the range of the
PROMIS study, participants were still healthier than the population norm score of
50 [18]. This study only had participant population of n=7 while the PROMIS
22
validity study was conducted with thousands of participants . The low participant
population might acccount for the large variance and range of responses for some
participants. To determine if any trends exist or if these variations are a reflection of
the population pool used in this study, more participants will be need. Further
increasing the population size can uncover specific mean differences between
comparable groups such as gender, age, or ethnicity. Increasing the sample size
would help increase the chance of significance by better representing the population
mean. While preliminary results for the use of PROMIS in pain studies have shown
that the tool can be useful for identifying healthy populations related to pain, further
research must be done on the pain population.
Further steps in this study include increasing the sample size and then repeating
the methods used in the study on an acute pain population. It is estimated that a
population of approximately 50 particpants will be needed to determine trends and
any significane from the study. Testing the pain population and comparing the
demographic data to a healthy population would help reveal differences in how the
pain experience affects the lives of those with acute pain. PROMIS helps to
understand the physical, mental, and social impacts of pain on one’s life and
researching population with acute pain can help create subgroups of the pain
population based on individual experiences. To further understand the acute pain
experience, genomic testing and QST will be conducted. Understanding the genetic
variations in the pain experience and the individual reaction to pain, clinicians can
begin the process of deep phenotying. Deep phenotyping can identify homogeneous
subsets of the population that researchers and clinicians can use to improve the
23
treatment of patients through more targeted and efficient drug therapy. This study
proved important for developing a protocol for using PROMIS in pain studies.
24
References:
1. Institute of Medicine. (2011). Relieving pain in America: A blueprint for
transforming, prevention, care, education and research. The National
Academies Press.
2. Robinson, P. N. (2012). Deep phenotyping for precision medicine. Human
Mutation, 33(5), 777-780.
3. Haring, R., & Wallaschofski, H. (2012). Diving through the "-omics": The
case for deep phenotyping and systems epidemiology. Omics : A Journal of
Integrative Biology, 16(5), 231-234.
4. Woolf, C. J. (2010). What is this thing called pain? The Journal of Clinical
Investigation, 120(11), 3742-3744.
5. Lema, M., Voscopoulos, C. (2010). When does acute pain become chronic?.
British Journal of Anaesthesia, 105(1 Suppl):69-85.
6. Manolio, T. A. (2010). Genomewide association studies and assessment of
the risk of disease. The New England Journal of Medicine, 363(2), 166-176.
7. Almasy, L. (2012). The role of phenotype in gene discovery in the whole
genome sequencing era. Human Genetics, 131(10), 1533-1540.
8. Tracy, R. P. (2008). 'Deep phenotyping': Characterizing populations in the era
of genomics and systems biology. Current Opinion in Lipidology, 19(2), 151157.
9. Hansson, P. (2010). Yet another questionnaire is born! Pain, 150(2), 219.
10. Morley, S., & Vlaeyen, J. W. (2010). 50 years on, Henry Beecher's
"measurement of subjective responses". Pain, 150(2), 211-212.
11. Wilson, I. B., & Cleary, P. D. (1995). Linking clinical variables with health related quality of life. A conceptual model of patient outcomes. JAMA: The
Journal of the American Medical Association, 273(1), 59-65.
12. Fries, J. F., Bruce, B., Bjorner, J., & Rose, M. (2006). More relevant, precise,
and efficient items for assessment of physical function and disability: Moving
beyond the classic instruments. Annals of the Rheumatic Diseases, 65(3), 1621.
13. Amtmann, D., Cook, K. F., Jensen, M. P., Chen, W. H., Choi, S., Revicki, D.,
et al. (2010). Development of a PROMIS item bank to measure pain
interference. Pain, 150(1), 173-182.
14. Revicki, D. A., Chen, W. H., Harnam, N., Cook, K. F., Amtmann, D.,
Callahan, L. F., et al. (2009). Development and psychometric analysis of the
PROMIS pain behavior item bank. Pain, 146(1-2), 158-169.
15. Rothrock, N. E., Hays, R. D., Spritzer, K., Yount, S. E., Riley, W., & Cella,
D. (2010). Relative to the general US population, chronic diseases are
associated with poorer health-related quality of life as measured by the
patient-reported outcomes measurement information system (PROMIS).
Journal of Clinical Epidemiology, 63(11), 1195-1204.
26
16. Fries, J. F., Bruce, B., & Cella, D. (2005). The promise of PROMIS: Using item
response theory to improve assessment of patient-reported outcomes. Clinical and
Experimental Rheumatology, 23(5 Suppl 39), S53-7.
17. Ruau, D., Liu, L. Y., Clark, J. D., Angst, M. S., & Butte, A. J. (2012). Sex differences
in reported pain across 11,000 patients captured in electronic medical records. The
Journal of Pain : Official Journal of the American Pain Society, 13(3), 228-234.
18. Jorner, J. B., Kosinski, M., & Ware, J. E. (2005). Computerized adaptive testing
and item banking. In P. M. Fayers & R. D. Hays (Eds.), Assessing Quality of Life
(pp. 95-112). Oxford: Oxford University Press.
27