„ KNEE
Assessment of radiolucent lines around the
Oxford unicompartmental knee replacement
SENSITIVITY AND SPECIFICITY FOR LOOSENING
S. Kalra,
T. O. Smith,
B. Berko,
N. P. Walton
From Norfolk and
Norwich University
Hospital Trust,
Norwich, England
„ S. Kalra, MA, MBChB,
FRCS(Tr&Orth), Specialist
Registrar
„ B. Berko, MBChB, Research
Fellow
„ N. P. Walton, FRCS(Tr&Orth),
Consultant Orthopaedic
Surgeon
Norfolk and Norwich University
Hospital, Colney Lane, Norwich
NR4 7UY, UK.
„ T. O. Smith, MSc, BSc(Hons),
MCSP, University Lecturer
Faculty of Medicine and Health
Sciences
University of East Anglia,
Norwich NR4 7TJ, UK.
Correspondence should be sent
to Mr S. Kalra; e-mail:
sanjaykalra20@gmail.com
©2011 British Editorial Society
of Bone and Joint Surgery
doi:10.1302/0301-620X.93B6.
26062 $2.00
J Bone Joint Surg [Br]
2011;93-B:777-81.
Received 14 October 2010;
Accepted after revision 4
February 2011
VOL. 93-B, No. 6, JUNE 2011
The Oxford unicompartmental knee replacement gives good results in patients with
symptomatic osteoarthritis of the medial compartment. Previous studies have suggested
that the presence of radiolucent lines (RLLs) does not reflect a poor outcome in such
patients. However, the reliability and validity of this assessment have not been determined.
Our aim was to assess the intra- and interobserver reliability and the sensitivity and
specificity of the assessment of RLLs around both tibial and femoral components using
standard radiographs.
Two reviewers assessed the radiographs of 45 patients who had loosening of the tibial or
femoral component confirmed at revision surgery and compared them with those of a series
of 45 asymptomatic patients matched for age and gender.
The results suggested that, using standard radiographs, tibial RLLs were 63.6% sensitive
and 94.4% specific and femoral RLLs 63.9% sensitive and 72.7% specific for loosening.
Overall intra- and interobserver reliability was highly variable, but zonal analysis showed
that lucency at the tip of the femoral peg was significantly associated with loosening of the
femoral component.
Fluoroscopically guided radiographs may improve the zonal reliability of the assessment
of RLLs, but further independent and comparative studies are required. In the meantime,
the innocence of the physiological RLLs detected by standard radiographs should be viewed
with caution.
The Oxford medial phase-3 unicompartmental knee replacement (UKR; Biomet, Swindon,
United Kingdom) comprises a keeled tibial
tray with a flat upper surface and a pegged
spherical femoral component. It uses an
unconstrained mobile bearing which is congruent with the femoral component throughout the range of movement. Both tibial and
femoral components are cemented. The
implant has been advocated for use in painful
isolated anteromedial osteoarthritis1 with an
intact anterior cruciate ligament and correctable varus deformity (< 15°).2 The Oxford
group have recently reported good results at a
mean of 5.6 years for 1000 UKRs.3
Peri-prosthetic radiolucent lines (RLLs)
after joint replacement surgery are usually
considered to be indicative or predictive of
loosening of the component. The presence of
RLLs associated with the Oxford medial UKR
is well documented, but their aetiology and
significance are poorly understood.
In 1984, Tibrewal, Grant and Goodfellow4
reported an incidence of tibial radiolucency of
96% in 80 phase-1 UKRs at one year. In
2006, Pandit et al5 described an incidence of
70% in 688 phase-3 UKRs at five years.
Despite these radiological abnormalities,
other series have shown a good survival and
clinical outcome at ten and 20 years.6-11
Goodfellow et al2 subdivided tibial RLLs into
two subtypes: physiological and pathological.
Physiological RLLs were classified as those
< 2 mm thick, well defined, with an accompanying parallel radiodense line. By contrast,
pathological RLLs were > 2 mm thick, poorly
defined, and not accompanied by radiodensity. Gulati et al12 reported an incidence of
62% of physiological RLLs around the tibial
component and concluded that the presence
of these lines was neither related to symptoms nor indicative or predictive of loosening.
We are unaware of any previous studies
which have investigated RLLs around the
femoral component. The incidence, distribution and significance of these remain
unknown.
Recently, a much lower incidence of RLLs
has been reported with the cementless Oxford
medial UKR,13 which is thought to represent
777
778
S. KALRA, T. O. SMITH, B. BERKO, N. P. WALTON
bone incorporation within the prosthesis. While several
hypotheses have been proposed to explain the RLLs
observed with the cemented prosthesis,4 the favoured
theory is formation of a layer of fibrocartilage as a result
of the compressive loading which occurs under the tibial
plateau.14
The importance of fluoroscopically guided radiographs
in the assessment of the tibial component has been emphasised in numerous studies,4,12,15 but their accuracy and
validity, both alone and in comparison with standard
radiographs, has never been formally investigated. No
independent group has yet investigated the radiological
assessment of the Oxford UKR. Consequently, it is
unclear whether the use of fluoroscopically guided radiographs is common practice and their general adoption for
the assessment of tibial RLLs in UKR is debatable.
Fluoroscopically guided radiographs are not recommended for assessment of the femoral component.
Our aim therefore was to evaluate two issues related to
the radiological assessment of UKR. We wished firstly to
determine how common is the use of fluoroscopically
guided radiographs in joint replacement centres for assessing UKR and secondly, to evaluate the reliability, sensitivity and specificity of tibial and femoral RLLs using
standard radiographs.
Patients and Methods
National survey. We carried out a telephone survey of all
acute hospital trusts in the NHS which have an orthopaedic department, in order to determine how often they use
fluoroscopically guided radiographs. The NHS website
was accessed to identify all 110 eligible hospitals. Each
hospital’s Superintendent Radiographer (or equivalent)
was contacted through the hospital switchboard and was
asked about local protocols for the post-operative radiological assessment of patients with UKR. In addition, they
were directly questioned about the use of fluoroscopically
guided radiographs in this group of patients.
Radiological assessment. Our study was carried out in a
University Teaching Hospital with a prospective database
which included over 1200 operations performed over the
last 12 years using the Oxford UKRs, mostly by a single
surgeon. Since 2000, 949 phase-3 UKRs had been logged.
For each, the database included the clinical outcome
score, range of movement and complications. The incidence of revision surgery was also recorded including the
intra-operative findings in each case.
In total, 92 patients (9.7%) required revision surgery.
Table I summarises the reasons for revision. In our study,
our revision series included only those 45 cases in which
loosening of either the tibial or femoral component had
been confirmed intra-operatively.
We identified 90 patients who had an Oxford phase-3
UKR performed between January 1, 2000 and December
1, 2008. Of these, 45 had required revision because of sus-
Table I. Reasons for revision surgery in the 92 revisions of 949 Oxford phase-3 procedures. Only those
45 with loosening of a component were included into
our revision series
Reason
Number of cases
Femoral loosening
Tibial loosening
Bearing dislocation
Unexplained pain
Progression of lateral disease
Fracture
Infection
Total
35
10
14
10
10
8
5
92
pected loosening which had subsequently been confirmed
at revision and in whom details of the loose component
were recorded (Table I). As Table II shows, these patients
were approximately matched by age, gender and knee
with 45 patients who had a UKR performed with an
uncomplicated follow-up and who did not require revision.
The clinical details of both groups are presented in
Table II. There were 43 men and 47 women with a mean
age of 69.5 years (51 to 91) at the time of radiological
examination. In the revision series, ten patients had
revision for loosening of the tibial component and 35 for
loosening of the femoral component as determined intraoperatively. The radiographs of the revision group were
the last taken before revision, at a mean follow-up of
26.8 months (3 to 90). Those of the non-revision group
were approximately matched for period of follow-up
(26.2 months (1 to 71)).
All radiographs were reviewed by two assessors (SK,
BB) who were blinded as to whether the patient had or
had not undergone revision for loosening after the radiographs were taken. By this means it was possible to assess
the sensitivity and specificity of each RLL measurement
without assessor bias affecting the results. Both assessors
had read the previous literature regarding RLL subtypes
and the assessment of tibial RLLs.12 Both the tibial and
femoral components were assessed for each patient by
each assessor. One assessor (SK) then repeated his review
of both the tibial and femoral implants after a period of
three months.
All the radiographs were reviewed on the institute’s Picture Archiving and Communication System and digitally
measured. Tibial RLLs were assessed using the same zonal
method described previously by Gulati et al12 (Fig. 1a). On
anteroposterior (AP) view, the tibial component was
divided into six zones. Each zone was assessed for the presence of RLLs and the subtype recorded. For femoral RLLs,
we developed a system of dividing the lateral cross-table
view of the femoral component into six zones (Fig. 1b).
Each zone was assessed for the presence of RLLs and the
subtype recorded.
THE JOURNAL OF BONE AND JOINT SURGERY
ASSESSMENT OF RADIOLUCENT LINES AROUND THE OXFORD UNICOMPARTMENTAL KNEE REPLACEMENT
779
Fig. 1b
Fig. 1a
The zonal assessment for radiolucent lines. Figure 1a – anteroposterior radiograph showing the six zones of the tibial component originally
described by Gulati et al.12 Figure 1b – lateral cross-table radiograph showing the six zones of the femoral component.
Table II. Details of the 90 patients included in our series
Revision
(n = 45)
Non-revision
(n = 45)
Total
(n = 90)
Male (%)
Female (%)
Mean age (range) at radiography in years
23 (51.1)
20 (44.4)
22 (48.9)
25 (55.6)
69.6 (52 to 91) 69.3 (51 to 84)
43 (47.8)
47 (52.2)
69.5 (51 to 91)
Side (%)
Left
Right
18 (40.0)
27 (60.0)
36 (40.0)
54 (60.0)
Frequency of intra-operative tibial loosening (%)
10 (22.2)
Frequency of intra-operative femoral loosening (%) 35 (77.8)
The non-revision series had a mean Oxford knee
score16 of 30 (19 to 47) and a mean SF-1217 score of 38
(18 to 60) at the time of the radiological examination.
Statistical analysis. The clinical details including gender,
side of implant, age at radiological examination, and frequency of tibial or femoral loosening were assessed using
descriptive statistics consisting of mean, range values and
frequences reported as percentages.
An assessment of the difference in the frequency of
occurrence of physiological and pathological RLLs was
compared between the revision and non-revision series at
each RLL zone using Fisher’s exact test, and for all zones
using the chi-squared test. This was carried out for both
tibial and femoral implants in both the revision and nonrevision series. A p-value < 0.05 was considered to be
statistically significant
The sensitivity and specificity of the femoral and tibial
RLL assessments were also made as described by Bland.18
Interobserver reliability was assessed by comparing the
femoral and tibial RLL assessments between the two
assessors and intra-observer reliability by the assessments
of one assessor on two separate occasions. These were
VOL. 93-B, No. 6, JUNE 2011
18 (40.0)
27 (60.0)
10 (11.1)
35 (38.9)
calculated using a weighted kappa analysis,19 and interpreted using the agreement recommendations of Landis
and Koch.20
Results
National survey. Of the 110 eligible hospitals, 109 were
contacted. The Nuffield Orthopaedic Centre, Oxford, was
not contacted since it had previously reported the fact that
it used fluoroscopically guided radiographs.12 The senior
radiographer from each of the remaining 109 stated that
the hospital used standard weight-bearing AP and crosstable lateral radiographs in the follow-up of UKR. None
reported the use of fluoroscopic guidance.
Radiological assessment
Comparison of tibial radiolucency. There was a greater frequency of pathological tibial lucency in the non-revision
compared with the revision group (Table III). However, this
was not statistically significant either overall (p =0.37) or
when individual zones were compared (between p = 0.62 and
p = 0.79 for the six zones, Fisher’s exact test).
There was a higher incidence of tibial physiological
lucency in the revision compared with the non-revision
780
S. KALRA, T. O. SMITH, B. BERKO, N. P. WALTON
Table III. Radiological assessment of pathological tibial lucency
Radiolucent zone
Revision
Non-revision
p-value*
1
2
3
4
5
6
Total
3
6
4
5
5
5
28
19
17
17
4
10
9
76
0.79
0.70
0.79
0.73
0.73
0.62
0.37†
* p-value calculated by Fisher’s exact test unless otherwise stated
† chi-squared test
Table IV. Radiological assessment of physiological tibial lucency
Radiolucent zone
Revision
Non-revision
p-value*
1
2
3
4
5
6
Total
15
9
2
2
1
3
32
8
5
3
2
3
2
23
0.98
0.54
0.59
0.66
0.64
0.76
0.52†
* p-value calculated by Fisher’s exact test unless otherwise stated
† chi-squared test
Table V. Radiological assessment of pathological femoral lucency
Radiolucent zone
Revision
Non-revision
p-value*
1
2
3
4
5
6
Total
7
9
7
9
6
12
50
3
3
0
2
2
6
16
0.35
0.22
0.65
0.58
0.48
0.10
0.69†
* p-value calculated by Fisher’s exact test unless otherwise stated
† chi-squared test
Table VI. Radiological assessment of physiological femoral lucency
Radiolucent zone
Revision
Non-revision
p-value*
1
2
3
4
5
6
Total
8
15
10
14
8
20
75
4
6
2
7
7
12
38
0.38
0.34
0.03
0.08
0.90
0.55
0.01†
* p-value calculated by Fisher’s exact test unless otherwise stated
† chi-squared test
group (Table IV). However, there was no significant difference in the frequency of physiological lucency between
the revision and non-revision groups either overall
(p = 0.52) or in individual zones (p = 0.54 to p = 0.98,
Fisher’s exact test).
Comparison of femoral radiolucency. There was a greater
frequency of pathological lucency in the revision compared with the non-revision group for each femoral zone
(Table V). Statistical analysis showed that this trend was
not significant between the revision and non-revision
series either overall (p = 0.69) or for individual zones
(p = 0.10 to p = 0.65, Fisher’s exact test).
Table VI shows that while there was a greater number
of physiological femoral RLLs in the revised compared
with non-revised patients, this difference was not significantly different for zones 1, 2, 4, 5 and 6 (p = 0.08 to
p = 0.90, Fisher’s exact test). However, there was a statistically significant difference between the groups in detecting RLLs in femoral zone 3 (p = 0.03, Fisher’s exact test)
as well as an overall difference between the revision and
non-revision groups (p = 0.01, chi-squared test).
Diagnostic accuracy. The sensitivity and specificity of the
tibial RLLs for detecting loosening of the component were
63.6% and 94.4%, respectively. The femoral RLLs were
63.9% sensitive and 72.7% specific. The results of the
interobserver reliability indicated that the assessment of
tibial pathological loosening ranged from fair to substantial agreement between the two reviewers. However, the
interobserver reliability was, at best, moderate for femoral
pathological RLLs indicating that this method of assessing
RLLs may vary between assessors. Similarly, the assessment of intra-observer reliability indicated poor to good
agreement between radiological assessments for both
tibial and femoral components suggesting a lack of consistency between the first and second time that they were
studied by the same assessor.
Discussion
Our findings indicate that while the sensitivity and specificity of RLLs using standard radiographs appear to be
acceptable, their intra- and interobserver reliability is
poor.
Although the Oxford group have suggested that RLLs
should be assessed using fluoroscopically guided radiographs,12 the national survey conducted in our study has
indicated that this is not routine practice in the United
Kingdom. Instead, standard non-fluoroscopically guided
radiographs are generally obtained.
Our findings agreed with those of previous studies
reporting a high incidence of RLLs around the tibial component of the Oxford UKR.4,5,12 We have also shown a
high incidence of RLLs around the femoral component.
The aetiology of these is unknown and their reported lack
of significance in predicting or indicating loosening of the
components remains unexplained.4,12
We acknowledge that our small sample size may have
been insufficient to detect a statistically significant radiological difference between those components which were
found to be loose at revision and those which were not,
except in femoral zone 3 (at the tip of the peg of the femoral component). However, physiological RLLs were present in greater numbers in the revision group, thereby
bringing their lack of significance into question, as suggested by Gulati et al.12 Conversely, pathological RLLs
were also present in large numbers in the non-revision
THE JOURNAL OF BONE AND JOINT SURGERY
ASSESSMENT OF RADIOLUCENT LINES AROUND THE OXFORD UNICOMPARTMENTAL KNEE REPLACEMENT
group, although the difference in incidence was not statistically significant (p = 0.05).
The difference shown in femoral zone 3 may be a representation of the pistoning phenomenon which has been
described in a knee with a loose femoral component during flexion and extension of the knee.21 Lucency in this
area may represent a genuinely loose component according to our results. Therefore we recommend that particular attention should be paid to femoral zone 3 when
reviewing radiographs.
Our results suggest that it is difficult to distinguish
between physiological and pathological RLLs around
both tibial and femoral components using standard radiographs. With respect to the tibial component, this may be
because their characteristic features are obscured by a
non-parallel x-ray beam, thereby causing RLLs to be
missed unless fluoroscopic control is used, as suggested by
the Oxford group.12
By contrast, the geometry of the femoral component
may prevent the acquisition of a good view of the interface
even if the x-ray beam is parallel. The Oxford group have
reported an inter- and intra-observer correlation of more
than 0.98 for physiological tibial RLLs when using
fluoroscopically guided radiographs.12
It therefore follows that fluoroscopically guided radiographs may improve the agreement between assessors and
therefore make it easier to distinguish between pathological
and physiological RLLs, increasing the reliability of diagnosis and the monitoring of progression of loosening. When
standard radiographs are used, which is standard procedure
in most hospitals, a RLL around the femoral peg is likely to
represent loosening of the femoral component. Further studies are required to evaluate the role of fluoroscopically
guided radiographs in distinguishing between physiological
and pathological RLLs, as are comparative studies to show a
clear improvement in the assessment of UKR between
fluoroscopically guided and standard radiographs.
Listen live
Listen to the abstract of this article at
www.jbjs.org.uk/interactive/audio
Supplementary material
A table detailing the intra- and interobserver reliability of the assessment of radiological pathological
loosening of each zone of the femoral and tibial components is available with the electronic version of this article
on our website at www.jbjs.org.uk
VOL. 93-B, No. 6, JUNE 2011
781
The authors would like to thank Mr M. Glasgow, as the surgeon who performed
most of the operations.
No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article.
References
1. White SH, Ludkowski PF, Goodfellow JW. Anteromedial osteoarthritis of the
knee. J Bone Joint Surg [Br] 1991;73-B:582-6.
2. Goodfellow JW, O’Connor JJ, Dodd CA, Murray DW. Unicompartmental arthroplasty with the Oxford knee. Oxford: Oxford University Press, 2006:117-28.
3. Pandit H, Jenkins C, Gill HS, et al. Minimally invasive Oxford phase 3 unicompartmental knee replacement: results of 1000 cases. J Bone Joint Surg [Br] 2011;93B:198-204.
4. Tibrewal SB, Grant KA, Goodfellow JW. The radiolucent line beneath the tibial
components of the Oxford meniscal knee. J Bone Joint Surg [Br] 1984;66-B:523-8.
5. Pandit H, Jenkins C, Barker K, Dodd CA, Murray DW. The Oxford medial unicompartmental knee replacement using a minimally-invasive approach. J Bone Joint Surg
[Br] 2006;88-B:54-60.
6. Murray DW, Goodfellow JW, O’Connor JJ. The Oxford medial unicompartmental
arthroplasty: a ten-year survival study. J Bone Joint Surg [Br] 1998;80-B:983-9.
7. Svärd UC, Price AJ. Oxford medial unicompartmental knee arthroplasty: a survival
analysis of an independent series. J Bone Joint Surg [Br] 2001;83-B:191-4.
8. Price AJ, Short A, Kellett C, et al. Ten-year in vivo wear measurement of a fully
congruent mobile bearing unicompartmental knee arthroplasty. J Bone Joint Surg [Br]
2005;87-B:1493-7.
9. Emerson RH Jr, Higgins LL. Unicompartmental knee arthroplasty with the Oxford
prosthesis in patients with medial compartment arthritis. J Bone Joint Surg [Am]
2008;90-A:118-22.
10. Price AJ, Dodd CA, Svärd UG, Murray DW. Oxford medial unicompartmental knee
arthroplasty in patients younger and older than 60 years of age. J Bone Joint Surg [Br]
2005;87-B:1488-92.
11. Price AJ, Waite JC, Svard U. Long-term clinical results of the medial Oxford unicompartmental knee arthroplasty. Clin Orthop 2005;435:171-80.
12. Gulati A, Chau R, Pandit HG, et al. The incidence of physiological radiolucency following Oxford unicompartmental knee replacement and its relationship to outcome. J
Bone Joint Surg [Br] 2009;91-B:896-902.
13. Pandit H, Jenkins C, Beard DJ, et al. Cementless Oxford unicompartmental knee
replacement shows reduced radiolucency at one year. J Bone Joint Surg [Br] 2009;91B:185-9.
14. Simpson DJ, Price AJ, Gulati A, Murray DW, Gill HS. Elevated proximal tibial
strains following unicompartmental knee replacement: a possible cause of pain. Med
Eng Phys 2009;31:752-7.
15. Mukherjee K, Pandit H, Dodd CA, Ostlere S, Murray DW. The Oxford unicompartmental knee arthroplasty: a radiological perspective. Clin Radiol 2008;63:116976.
16. Dawson J, Fitzpatrick R, Murray D, Carr A. Questionnaire on the perceptions of
patients about total knee replacement. J Bone Joint Surg [Br] 1988;80-B:63-9.
17. Ware J Jr, Kosinski M, Keller SD. A 12-item short-form health survey: construction of scales and preliminary tests of reliability and validity. Med Care 1996;34:22033.
18. Bland M. An introduction to medical statistics. Third ed. London: Oxford University
Press, 2006:275-9.
19. Altman DG. Practical statistics for medical research. London: Chapman & Hall,
1991:403-9.
20. Landis JR, Koch GG. The measurement of observer agreement for categorical data.
Biometrics 1977;33:159-74.
21. Monk AP, Keys GW, Murray DW. Loosening of the femoral component after unicompartmental knee replacement. J Bone Joint Surg [Br] 2009;91-B:405-8.