OMERACT Rheumatoid Arthritis Magnetic Resonance Imaging Studies. Exercise 3: an international multicenter reliability study using the RA-MRI Score.

Abstract

We examined inter-reader agreement of the revised OMERACT 5 Rheumatoid Arthritis MRI Score (RAMRIS v3). Magnetic resonance (MR) images of 10 sets of metacarpophalangeal (MCP) joints 2-5 and 8 sets of rheumatoid arthritis (RA) wrists [1.5 T, coronal and axial T1 and T2 spin-echo, +/- fat saturation (FS), +/- intravenous gadolinium (Gd)] were scored for (1) synovitis using a global score (0-3) and a direct measurement of synovial thickness (mm) and (2) three bone lesions: erosions, defects and edema, (score 0-10 by the volume of the lesion as a proportion of the "assessed bone volume" by 10% increments). Six readers from 5 multinational centers performed all scoring. Three statistical methods were used to analyze the data: (1) single-measure fixed effects intraclass correlations (sICC) and average-measure fixed effects ICC (avICC), (2) percentage exact and close agreement, and (3) the smallest detectable difference (SDD). The sICC were moderate to good (between 0.60 and 0.91) for half of the joint sites for the 2 synovitis scoring methods, and for bone erosions and bone edema. After adjusting for 6 readers, the avICC was very good to excellent (0.80-0.98) for two-thirds of the joint sites by lesion, excluding bone defects that performed relatively poorly, primarily because few readers scored these lesions. The aggregated scores with the best reliability were those with a wide range of scores, high ICC, low SDD, and low percentage SDD (< 33%). The metacarpophalangeal (MCP) bone erosion (sICC 0.58, avICC 0.89, %SDD +/- 27), wrist bone erosion scores (0.72, 0.94, +/- 31%), the wrist synovitis global (0.74, 0.94, +/- 32%), and synovial maximal thickness (0.6, 0.94, +/- 32%) met these conditions. MCP joint synovitis global (0.76, 0.95, +/-35%), MCP joint bone edema (0.63, 0.91, +/- 34%), and wrist bone edema (0.78, 0.95, +/- 38%) performed marginally less well. Bone defects performed poorly (MCP joint 0.18, 0.46, +/- 56%; wrist 0.06, 0.24, +/- 55%). The revised OMERACT 5 RAMRIS has acceptable inter-reader reliability for measures of disease activity (synovitis global and bone edema scores) and damage (bone erosion score). Whether the score is sensitive to change will be determined by its performance in longitudinal and intervention studies.

Full text

Page 1

The Journal of Rheumatology 2003; 30:6
1366
2002-945-1
From the Department of Rheumatology, St. George Hospital, University
of New South Wales, Sydney, Australia; Department of Molecular
Medicine, Auckland School of Medicine, University of Auckland,
Auckland, New Zealand; The Danish Research Center of Magnetic
Resonance and Departments of Rheumatology at the Copenhagen
University Hospitals at Hvidovre, Herlev, and Rigshospitalet,
Copenhagen, Denmark; Academic Unit of Musculoskeletal and
Rehabilitation Medicine, University of Leeds, Leeds, UK; Mayne Nickless
Sydney Imaging Group, Sydney, Australia; Synarc Inc., San Francisco,
CA, USA; Department of Radiology, Leeds General Infirmary, Leeds, UK;
Department of Radiology, University of California, San Francisco, CA,
USA.
M. Lassere, MBBS (Hons), Grad Dip Epi, PhD, FRACP, FAFPHM, Staff
Specialist in Rheumatology, Senior Lecturer in Medicine, St. George
Hospital; F. McQueen, MD, FRACP, Senior Lecturer in Rheumatology,
University of Auckland; M. Østergaard, MD, PhD, DMSc, Professor in
Rheumatology/Arthritis, Copenhagen University Hospitals at Hvidovre,
Herlev, and Rigshospitalet; P. Conaghan, MBBS, FRACP, Senior Lecturer
in Rheumatology, Academic Unit of Musculoskeletal and Rehabilitation
Medicine, University of Leeds; R. Shnier, MBBS, FRACR, National
Director of Diagnostic Imaging, Mayne Nickless Sydney Imaging Group;
C. Peterfy, MD, PhD, Chief Medical Officer, Synarc; M. Klarlund, MD,
PhD, Senior Registrar in Rheumatology, Copenhagen University
Hospitals at Hvidovre, Herlev, and Rigshospitalet; P. Bird, BMed (Hons),
Grad Cert MRI, FRACP, Research Fellow, St. George Hospital;
P. O’Connor, MBBS, MRCP, FRCR, Consultant Skeletal Radiologist,
Academic Unit of Musculoskeletal and Rehabilitation Medicine,
University of Leeds; N. Stewart, MB, ChB, FRACR, Radiologist,
Department of Molecular Medicine, University of Auckland; P. Emery,
MA, MD, FRCP, FACR, FRCR, Professor of Rheumatology, University of
Leeds; H. Genant, MD, FACR, FRCR, Professor of Radiology, Medicine
and Orthopaedics, Department of Radiology, University of California,
San Francisco; J. Edmonds, MBBS, MA, FRACP, Professor of
Rheumatology, University of Leeds.
Address reprint requests to Dr. M.N. Lassere, Department of
Rheumatology, St. George Hospital, Gray Street, Kogarah, 2217
NSW, Australia. E-mail: lasserem@sesahs.nsw.gov.au
OMERACT Rheumatoid Arthritis Magnetic Resonance
Imaging Studies. Exercise 3: An International
Multicenter Reliability Study Using the RA-MRI Score
MARISSA LASSERE, FIONA McQUEEN, MIKKEL ØSTERGAARD, PHILIP CONAGHAN, RON SHNIER,
CHARLES PETERFY, METTE KLARLUND, PAUL BIRD, PHILIP O’CONNOR, NEAL STEWART, PAUL EMERY,
HARRY GENANT, and JOHN EDMONDS
ABSTRACT. We examined inter-reader agreement of the revised OMERACT 5 Rheumatoid Arthritis MRI Score
(RAMRIS v3). Magnetic resonance (MR) images of 10 sets of metacarpophalangeal (MCP) joints
2–5 and 8 sets of rheumatoid arthritis (RA) wrists [1.5 T, coronal and axial T1 and T2 spin-echo, ±
fat saturation (FS), ± intravenous gadolinium (Gd)] were scored for (1) synovitis using a global score
(0–3) and a direct measurement of synovial thickness (mm) and (2) three bone lesions: erosions,
defects and edema, (score 0–10 by the volume of the lesion as a proportion of the “assessed bone
volume” by 10% increments). Six readers from 5 multinational centers performed all scoring. Three
statistical methods were used to analyze the data: (1) single-measure fixed effects intraclass correla-
tions (sICC) and average-measure fixed effects ICC (avICC), (2) percentage exact and close agree-
ment, and (3) the smallest detectable difference (SDD). The sICC were moderate to good (between
0.60 and 0.91) for half of the joint sites for the 2 synovitis scoring methods, and for bone erosions
and bone edema. After adjusting for 6 readers, the avICC was very good to excellent (0.80–0.98) for
two-thirds of the joint sites by lesion, excluding bone defects that performed relatively poorly,
primarily because few readers scored these lesions. The aggregated scores with the best reliability
were those with a wide range of scores, high ICC, low SDD, and low percentage SDD (< 33%). The
metacarpophalangeal (MCP) bone erosion (sICC 0.58, avICC 0.89, %SDD ± 27), wrist bone erosion
scores (0.72, 0.94, ± 31%), the wrist synovitis global (0.74, 0.94, ± 32%), and synovial maximal
thickness (0.6, 0.94, ± 32%) met these conditions. MCP joint synovitis global (0.76, 0.95, ±35%),
MCP joint bone edema (0.63, 0.91, ± 34%), and wrist bone edema (0.78, 0.95, ± 38%) performed
marginally less well. Bone defects performed poorly (MCP joint 0.18, 0.46, ± 56%; wrist 0.06, 0.24,
± 55%). The revised OMERACT 5 RAMRIS has acceptable inter-reader reliability for measures of
disease activity (synovitis global and bone edema scores) and damage (bone erosion score). Whether
the score is sensitive to change will be determined by its performance in longitudinal and interven-
tion studies. (J Rheumatol 2003;30:1366–75)
Key Indexing Terms:
MAGNETIC RESONANCE IMAGING RELIABILITY
SMALLEST DETECTABLE DIFFERENCE
RHEUMATOID ARTHRITIS
Personal, non-commercial use only. The Journal of Rheumatology Copyright © 2003. All rights reserved.

Page 2

Developing a magnetic resonance imaging (MRI) measure-
ment system to evaluate the various facets of rheumatoid
arthritis (RA) disease activity and damage is a multistep
process. The following require consideration: which joints
[all metacarpophalangeal (MCP) joints, selected wrist
joints, dominant hand, etc.]; what anatomic structures or
tissues (bone, cartilage, synovium, tendon, etc.); what
features (bone edema, bone erosions, cartilage volume,
synovial thickening, etc.); how to quantify these features
(grade, score, count, or calculate areas or volume of
anatomic structures and lesions); whether to weight this
quantity and if so, how (implicitly, explicitly); whether and
how to aggregate (Boolean or arithmetic operators) the
quantified features into a component score. Moreover, at
each step sources of variability, reliability, validity, respon-
siveness, and feasibility1should be evaluated in field trials
using appropriate methods of statistical analysis.
However, many of the MRI measurement methods devel-
oped for use in RA have not been rigorously evaluated. A
recent review of 68 peer reviewed publications of the MRI
literature on measurement methods developed for RA, from
40 research groups2, found that all (but one) reported the
MRI variables, but only 35% of published studies were eval-
uated for reliability, and of these, few used optimal statis-
tical methods of analysis. Responsiveness faired only
slightly better (37% of studies). However, almost all studies
(88%) evaluated some form of validity.
In 1998, after a meeting at OMERACT 4, a MRI working
group of rheumatologists, radiologists, and clinical epidemi-
ologists was established to begin a rigorous, prospective,
transparent, and data-driven research process to develop and
evaluate new MRI scoring methods for use in RA. In 1999
at several international meetings preliminary scoring
methods were developed [Rheumatoid Arthritis MRI Score
version 1 (RAMRISv1) and version 2 (RAMRISv2)], and
their interreader reliability was tested (Study Exercises 1
and 2), presented at OMERACT 5, and published3. In light
of the results, the group proposed a revised MRI scoring
system (version 3) and recommended a standardized
protocol for MR image acquisitions4.
This article presents the results of Exercise 3, which
continues the iterative process of testing the reliability of the
RAMRIS. It was hoped that with further training, reader
calibration, standardization of imaging protocols, and with
more precise definitions of lesions and their measurement,
we could improve the performance of the revised MRI
Score (version 3).
MATERIALS AND METHODS
Design
MR images of 10 sets of 2nd to 5th MCP joints (from Leeds) and 8 sets of
RAwrists (from Auckland) were scored in 5 centers in 5 different countries
by readers with various levels of MRI experience. All MR sets were read
between August and December 2000. The 5 centers were Auckland, New
Zealand (FM, NS), Hvidovre, Denmark (MØ, MK), Leeds, United
Kingdom (PC, POC, PE), San Francisco, USA (CP, HG), and Sydney,
Australia (ML, PB, RS, JE).
Readers
There was no prespecification regarding the reading process. In Sydney, the
2 readers [PB, rheumatologist (Sydney 1) and RS, radiologist (Sydney 2)]
independently read both sets. In Leeds 2 readers (POC, radiologist and PC,
rheumatologist) read both sets, the results reflecting the consensus of both
readers. In Copenhagen, 2 readers (MO and MK, rheumatologists) read
both sets, the results reflecting the consensus of both readers. In Auckland
one reader (FM, rheumatologist) independently read the MCP set, and 2
readers (FM and NS, radiologist) read the wrist set, the results reflecting the
consensus of both readers. Finally, in San Francisco one reader (CP, radiol-
ogist) independently read both sets. In summary, there were 6 sets of results
from 5 centers for both the MCP joint and wrist sets. All readers partici-
pated in the OMERACT 5 discussions and all but one reader took part in
the previous exercises (Exercises 1 and 2). All statistical analyses were
conducted with and without the results of this additional reader. There was
no systematic difference depending whether 5 or 6 reader results were
analyzed, therefore the results that include all 6 readers are provided in
detail. However, the 5-reader results are available on request.
MRI Acquisition Specification and RA Patients
MR images of the 2nd to 5th MCP joints were obtained on a 1.5 Tesla MR
scanner (Phillips). The field of view was 10 cm and was centered on the
2nd to 5th MCPjoints. The imaging protocol comprised first, coronal (slice
thickness 3 mm, no gap) and axial T1 weighted spin echo images, followed
by coronal fat saturated T2 weighted images. After intravenous injection of
gadolinium-DTPA contrast, the axial T1 weighted sequence was repeated,
followed by a fat saturated coronal T1 weighted sequence. MR images of
the dominant wrist were obtained using a 1.5 Tesla MR scanner (GE Signa
Horizon) with a dedicated wrist coil (Medical Devices). The field of view
was 8 cm and included the distal radioulnar, radiocarpal, and midcarpal
joints as well as the metacarpal bases. The imaging protocol comprised
first, coronal (slice thickness 3 mm, no gap) and axial (slice thickness 3
mm, 1 mm gap) T1 sequences, followed by axial fat suppressed fast spin
echo T2, then coronal fat suppressed T1 sequences after injection of
gadolinium (Nicomed Omniscan). The mean disease duration for the 10
MCP joint set and the 8 wrist joint set was 12 months. All patients fulfilled
the American College of Rheumatology (ACR) 1987 criteria for RA.
Scoring of MRI
Structures included — definition of lesions. The revised OMERACT
RAMRIS (version 3). An erosion was defined as a bone defect with sharp
margins, visible in 2 planes (when 2 planes are available) with a cortical
break seen in at least one plane. A bone defect was defined as a sharply
marginated area of trabecular loss without a visible cortical break. Bone
edema could occur alone or surround a “defect” or “erosion” and was
defined as a lesion with ill defined margins that was neither erosion nor
defect and had high signal intensity on T2 weighted sequences. Synovitis
was the area in the synovial compartment that showed enhancement of a
thickness greater than the width of the joint capsule after gadolinium.
Cartilage was not scored because at Exercise 2 the demarcation of this
tissue in the small joints of the wrist and hand was found to be too unreli-
able3.
Scoring of bone lesions. A bone erosion lesion was scored from 0 to 10 by
the volume of the erosion as a proportion of the “assessed bone volume” by
10% increments judged on all available images. For the carpal bones, the
“assessed bone volume” was the whole bone. For long bones, the “assessed
bone volume” was from the cortex of the articular surface (or its best esti-
mated position if absent) to a depth of 1 cm. Bone defects and bone edema
similarly were each scored 0–10 by the volume of the defect or of edema,
as for erosion.
Bone erosions, bone defects, and bone edema were measured at 15 sites
on the wrist image set and 8 sites (proximal and distal half of each 2nd to
Lassere, et al: MRI Exercise 3
1367
Personal, non-commercial use only. The Journal of Rheumatology Copyright © 2003. All rights reserved.

Page 3

5th MCP joint) on the MCP joint image set (see Figures 1 and 2 for scoring
templates). Therefore the score for each MCP joint was 0–20, and the
aggregated score for all 4 joints was 0–80 (where a score of zero indicates
no erosions and a score of 80 indicates no bone). Bone defects and edema
were scored similarly. In the wrist, 15 sites were scored: the base of 1st to
5th, the 8 carpal bones (hamate, capitate, trapezoid, trapezium, triquetrum,
pisiform, lunate, scaphoid), and the distal radius and distal ulnar. Therefore
the aggregated score for wrist was 0–150 for bone erosions, 0–150 for
defects, and 0–150 for edema.
Scoring of synovitis. Synovitis was determined by gadolinium enhance-
ment of the synovial compartment by 2 methods. In method 1, a global
score of 0 to 3 was assigned, where 0 was normal with no synovial
enhancement or enhancement no thicker than the joint capsule. Score of
1 to 3 was by thirds of the presumed maximum volume of enhancing
tissue in the synovial compartment. This global score was assigned to
the 4 MCP joint sites, giving an aggregated score of 0 to 12. In the wrist,
the global score was assigned at 3 sites: the radioulnar joint; the radio-
carpal joint; the intercarpal-carpometacarpal joints, giving an aggre-
gated score of 0 to 9. In method 2, the maximum thickness of enhancing
tissue in the slice showing the most thickening was directly measured in
millimeters. This was measured at all 4 MCP joints on the axial view;
in the wrists it was measured perpendicular to the cortical surface in the
coronal view at the scaphoid and triquetrum; and in the axial view at the
radioulnar joint and along the curved dorsal surface of the 1st and 2nd
carpal rows.
Statistical Methods
The data were analyzed (1) individually by joint and lesion (i.e., synovitis
global 2nd MCP, proximal 2nd MCP erosion, proximal 2nd MCP defect,
etc.) to determine how agreement differed by joint and by lesion, and (2) as
aggregated scores, analogous to the methods used to score radiographs,
such as the Sharp radiographic score5.
Descriptive statistics of each lesion (mean, minimum, maximum, stan-
dard deviation, median, 25th and 75th percentiles) for individual joints and
aggregated scores were calculated by reader and across readers. Reliability
was comprehensively evaluated with 3 statistical methods: intraclass corre-
lation coefficient (ICC), percentage close/exact agreement, and smallest
detectable difference (SDD). Three methods were used because each
method entails certain assumptions that can produce biased results
depending on the distribution of the scores under evaluation.
The first statistical method was the single measure fixed effects ICC
(sICC) as described by Shrout and Fleiss6. ICC are a relative measure of
agreement. The sICC and its 95% confidence interval6is similar to the
quadratic weighted kappa for ordinal scale measures, where the weighted
kappa is agreement beyond chance agreement. A second ICC, the average
measure fixed effects ICC (avICC), corrects for the number of readers, so
this was also provided7.
One shortcoming of ICC is that if there is limited variation in the
features scored, then the ICC value will be low despite trivial differences
between reader scores. ICC values are biased towards high coefficients (1.0
is perfect reliability) if the data vary over a wide range. Another disadvan-
The Journal of Rheumatology 2003; 30:6
1368
2002-945-3
Figure 1. MCP joint scoring template.
Personal, non-commercial use only. The Journal of Rheumatology Copyright © 2003. All rights reserved.

Page 4

tage of ICC is that they are not robust to the effects of outliers and can be
affected considerably by a few large agreements.
To compensate for these problems we used a second statistical method,
the percentage close agreement (PCA) and percentage exact agreement
(PEA). PCA is based on a distribution of the degree of difference between
readers. This is the percentage of occasions that different readers’ scores
fall within a certain distance of each other8. When calculating PCA, “close”
is defined by judging what is meaningful for the measure and data set
concerned. We set the PCA as within ± 1 interval for all lesions, including
all aggregated lesions. Therefore percentage close agreement should be
considered within the context of the actual score range.
The third statistical method was the smallest detectable difference9,10,
which is derived from the limits of agreement method11. Random error is
quantified using an absolute metric. The SDD, unlike the ICC, is biased
toward smaller values (SDD of 0 is perfect agreement, and there is no
convention that anchors the upper limit) if the lesions are measured over a
narrower range of values. The SDD is expressed in the same units of
measurement as calculated for all aggregated scores. It is determined from
the residual error variance of repeated measures analysis of variance
(ANOVA)12,13. The SDD was also expressed as a percentage of the highest
actual score to permit comparison of the reliability of MRI scoring with
radiographic and clinical measures. Single factor repeated measures
ANOVA was used to investigate whether “reader” was a significant source
of variability. The statistical programs used were Stata 7.014, SPPS 6.07, and
ICC.EXE8.
RESULTS
Descriptive statistics (mean, range, standard deviations) of
aggregated scores by lesion for each reader are shown in
Figure 2. Wrist scoring template. CMC: carpometacarpal.
Lassere, et al: MRI Exercise 3
1369
Personal, non-commercial use only. The Journal of Rheumatology Copyright © 2003. All rights reserved.

Page 5

Table 1. The full spectrum of the range was scored for the
synovitis global lesions, particularly the MCP joints, less so
the wrist joints, whereas bone erosions scored in the first
quarter of the range for both MCP joints and wrists. A p
value < 0.05 indicates that at least one reader differs from
the others and that “reader” is a significant source of
systematic variation in this dataset. Only for bone defects
was the repeated measures ANOVA consistently not signifi-
cant.
Averaging across readers (results not shown but available
on request), the 2nd MCP joint consistently had the highest
synovitis global score and the 4th MCP joint had the lowest.
Repeated measures ANOVA did not show reader to be a
significant source of variation for any MCP joint synovitis
global score. The 2nd and 3rd MCP joints had higher scores
for synovial maximal thickness, and reader was a significant
source of variation only for 2nd MCP. The 2nd, 3rd, and 5th
MCP consistently had higher erosion scores than the 4th
MCP, and proximal scores were always greater than distal
scores. Reader was a significant source of variability for 2nd
and 5th MCP joints but not for 3rd and 4th MCP joints, irre-
spective of whether the joints were analyzed by their prox-
imal or distal site. Defect scores were low for all joints, and
there was no significant variability. The 2nd MCP joint
consistently scored highest for bone edema, and proximal
sites scored higher than distal sites. Reader was a significant
source of variability for all but one site.
Average synovitis global scores at the 3 wrist sites did
not differ, although synovial maximal thicknesses seen on
coronal views were greater than those on axial. Reader vari-
ation was significant for all global scores and for 3 of the 4
maximal thickness scores. The 1st and 4th metacarpal bases,
hamate, triquetrum, lunate, and scaphoid had the highest
bone erosion and bone edema scores, followed by the capi-
tate, trapezoid, trapezium, distal ulnar, distal radius, 2nd, 3rd
and 5th metacarpal bases, and pisiform. There was little
difference among all bones for “defects.” Reader variation
for bone erosions was significant for 1st metacarpal base,
the capitate, trapezoid, triquetrum, lunate, scaphoid, distal
ulnar, and radius; however, for bone edema it was signifi-
cant at only the 1st metacarpal base, capitate, and
triquetrum. It was not significant for defects.
The sites and lesions that performed best because
joint/lesion showed a wide range of scores and because
“reader” was not a significant source of variation on
ANOVA were as follows: all MCP joints for synovitis
global; 3rd MCP proximal site, 4th metacarpal base, and
hamate for bone erosions; 4th metacarpal base, hamate,
lunate and scaphoid for bone edema.
Table 2 shows the interreader fixed ICC and percentage
agreement by joint/site and by lesion. Both the sICC and the
avICC results are provided as well as the percentage exact
agreement, or percentage close agreement within one
interval. Bone defects usually scored zero at most sites,
therefore as expected the ICC perform poorly and the
percentage agreement was high. For the remaining joint/site
by lesions the sICC was moderate to good (0.60–0.91) for
half the joint sites by lesion. However, if the number of
readers is taken into consideration, more than two-thirds of
the joint sites by lesion (excluding bone defects) had very
The Journal of Rheumatology 2003; 30:6
1370
2002-945-5
Table 1. Means (minimum and maximum) of aggregated scores of MCP joints and wrists by different readers/centers.
AU 1AU 2 UKDK NZ USAANOVA
MCP joints
Synovitis global score (0–12)
Synovial maximal thickness, (mm)
Bone erosions proximal + distal (0–80)
Bone defects proximal + distal (0–80)
Bone edema proximal + distal (0–80)
Wrists
Synovitis global score (0–12)
Synovial maximal thickness (mm)
Bone erosions (0–150)
Metacarpal bases (0–50)
Carpus (0–80)
Radius + ulnar (0–20)
Bone defects total (0–150)
Metacarpal bases (0–50)
Carpus (0–80)
Radius + ulnar (0–20)
Bone edema total (0–150)
Metacarpal bases (0–50)
Carpus (0–80)
Radius + ulnar (0–20)
4.7 (0–8)
11.1 (0–21)
8.2 (2–15)
0.5 (0–2)
3.2 (0–11)
6.5 (1–10)
12.9 (6–20)
10.4 (6–25)
0 (0–0)
7.9 (0–23)
6.7 (1–12)
11.9 (1–17)
5.8 (3–9)
0.1 (0–1)
11.3 (3–19)
6.6 (0–12)
13.2 (0–20)
4.7 (1–7)
0.4 (0–2)
3.3 (0–9)
5.7 (0–12)
9.8 (4–19)
4.7 (2–7)
0.6 (0–3)
4.5 (0–12)
6.7 (1–11)
8.6 (1–15)
8.3 (4–17)
0 (0–0)
4.3 (0–14)
0.027
0.020
0.000
0.108
0.000
5.5 (2–9)
21.3 (10–31)
13.5 (3–32)
3 (0–10)
8.1 (2–18)
2.4 (0–7)
1.5 (0–3)
0.3 (0–2)
1.3 (0–3)
0 (0)
10.13 (1–24)
2 (0–7)
6.5 (1–11)
1.6 (0–6)
7.6 (5–9)
23.4 (11–38)
17.1 (9–41)
4 (1–12)
11.4 (7–25)
1.8 (0–4)
0.25 (0–2)
0 (0)
0.3 (0–2)
0 (0)
10.2 (0–52)
2.9 (0–17)
6.5 (0–31)
0.7 (0–4)
4.9 (3–9)
9.8 (4–18)
13.8 (4–38)
3.8 (0–16)
9 (4–19)
1 (0–3)
0.13 (0–1)
0 (0)
0.1 (0–1)
0 (0)
24.3 (7–55)
8.1 (0–33)
14.1 (6–27)
2 (0–6)
3.6 (0–7)
16 (0–28)
2.8 (0–8)
1 (0-4)
1.5 (0–3)
0.3 (0–1)
1.6 (0–4)
0.6 (0–4)
0.9 (0–3)
0.1 (0–1)
14.1 (0–51)
4 (0–16)
9.4 (0–32)
0.8 (0–3)
4.6 (2–6)
9.8 (4–12.5)
6.6 (1–14)
1.8 (0–4)
3.8 (1–7)
1 (0–3)
2.8 (0–5)
0.9 (0–3)
1.3 (0–3)
0.3 (0–1)
10.9 (1–42)
2.9 (0–16)
6.9 (1–22)
1.1 (0–5)
3.6 (2–5)
19.9 (13–27)
15.8 (5–32)
4.3 (0–9)
9.3 (3–17)
2.3 (0–6)
0 (0–0)
0 (0)
0 (0)
0 (0)
8.6 (0–48)
2.9 (0–18)
4.8 (0–26)
1 (0–4)
0.000
0.000
0.000
0.021
0.000
0.000
0.000
0.090
0.011
0.220
0.002
0.056
0.015
0.099
Values in the 2nd to 7th columns are mean; max-min in parentheses. Values in 8th (right) column are p values. ANOVA: analysis of variance. AU 1: Australia
Reader 1, AU 2: Australia reader 2; UK: United Kingdom, DK: Denmark, NZ: New Zealand, USA: United States of America.
Personal, non-commercial use only. The Journal of Rheumatology Copyright © 2003. All rights reserved.