close

Вход

Забыли?

вход по аккаунту

?

Reduced joint count indices in the evaluation of rheumatoid arthritis.

код для вставкиСкачать
613
REDUCED JOINT COUNT INDICES IN THE
EVALUATION OF RHEUMATOID ARTHRITIS
M. J. EGGER, D. A. HUTH, J. R. WARD, J. C. READING, and H. J. WILLIAMS
Two types of summary measures of joint disease
were evaluated in 2 controlled clinical trials in rheumatoid arthritis patients. One measure was based solely on
the clinicalhiologic judgment approach; the other combined this methodology with statistical approaches using
reliability and factor analyses. “Signal joint” indices,
summarizing disease activity in 2-5 key joints, were
found to be insensitive to deterioriation of nonsignal
joints. Therefore, they are not recommended as replacements for the complete articular survey in rheumatoid
arthritis. A reduced version of the complete articular
survey was found to be desirable based on its validity,
reliability, accuracy, precision, and sensitivity to active
drugs. Its sensitivity can be slightly higher or lower than
that of the full joint survey.
The assessment of joint involvement in rheumatoid arthritis has been the single most important variable by which an individual patient’s disease is evaluated. The patient is seeking relief of pain and
functional loss and the physician is attempting to
achieve the same objectives in addition to controlling
inflammation. However, there is no single, uniform
-_____
From the University of Utah School of Medicine, Salt Lake
City.
Supported by NIH contract N01AM1-2204 from the National Institute of Arthritis, Diabetes and Digestive and Kidney
Diseases.
M. J. Egger, PhD: Department of Family and Community
Medicine, Division of Biostatistics; D. A. Huth, BS: Department of
Family and Community Medicine, Division of Biostatistics; J. R.
Ward, MD: Department of Medicine; J. C. Reading, PhD: Department of Family and Community Medicine, Division of Biostatistics;
H. J. Williams, MD: Department of Medicine.
Address reprint requests to M. J. Egger, PhD, Family and
Community Medicine, Division of Biostatistics, University of Utah
Medical Center, Salt Lake City, UT 84132.
Submitted for publication January 30, 1984; accepted in
revised form December 14. 1984.
Arthritis and Rheumatism, Vol. 28, No. 6 (June 1985)
measure used in day-to-day management of rheumatoid arthritis or in therapeutic studies for evaluation of
particular treatments. Indices of arthritis activity
which involve counting or scoring every joint for
involvement by pain, tenderness, and/or swelling have
been developed by the Cooperating Clinics Committee
of the American Rheumatism Association (ARA) ( l ) ,
by Lansbury and Haut (2), by Ritchie et a1 (3), and by
the Cooperative Systematic Studies of the Rheumatic
Diseases group (CSSRD) (4-7). These complete joint
assessments are time-consuming to perform and suffer
from limited reliability in joints that are difficult to
examine for swelling, such as the glenohumeral and
metatarsophalangeal (MTP) joints.
An alternative approach would involve a more
concise and limited description of joint inflammation.
Ward et a1 (4) have explored the use of a small number
of representative “signal” joints (SJ) which are chosen
by the clinician to characterize an individual patient’s
disease. SJ indices and other summary measures have
been a recent source of substantial discussion (8-1 1).
In order to examine the utility and reliability of
various joint count indices, we examined joint counts
and summative scores, based on SJ counts and on the
“reduced joint survey” (RJS), a 36-joint subset of the
full joint survey, by a secondary analysis of 2 controlled clinical trials. These abbreviated measures
were compared with traditional complete joint surveys, and recommendations for the uses and limitations of reduced joint counts were developed.
METHODS
Data sources. Scores on the proposed indices, traditional CSSRD joint counts and scores, and a modified
Ritchie index (4) were calculated from data gathered during 2
controlled clinical trials of rheumatoid arthritis treatments:
614
azathioprine (AZA) versus D-penicillamine (DP) (7) and
auranofin versus parenteral gold sodium thiomalate (GST)
versus placebo (6). Both studies were prospective, randomized, double-blind trials by the multi-institutional Cooperative Systematic Studies of the Rheumatic Diseases group. In
the AZA-DP study, 206 patients who had been treated
unsuccessfully with gold received either 1.25-1.5 mg/kg/day
AZA or 10-12 mg/kg/day DP with appropriate placebo for 24
weeks. In the 21-week study of gold therapy, 208 patients
received either auranofin (3 mg orally, twice daily) plus
placebo injections, weekly intramuscular injections of GST
(10 mg on the first injection, 25 mg on the second injection,
and 50 mg on each of 19 subsequent injections) plus placebo
tablets, or both placebo tablets and placebo injections.
Active disease at entry into each study was defined
as a minimum of 6 swollen joints capable of response to an
active drug and at least 2 of the following: (a) 2 9 joints
tender on pressure, (b) at least 45 minutes of morning
stiffness, (c) Westergren erythrocyte sedimentation rate of
228 mm/hour. To minimize the effects of interobserver
variability (12), patients’ joint counts were performed by the
same clinician at the first and last visits. Traditional CSSRD
joint counts and scores (5) were calculated for those patients
who continued in the AZA-DP study for at least 24 weeks or
in the gold study for at least 20 weeks. All 4 active drugs
have been shown to cause significantly greater improvement
than placebo in randomized, controlled clinical trials (57,13).
Conceptual taxonomy. A taxonomy of summary measures of articular response is presented in Figure l. Condensation of the information from joint surveys has been
achieved in the past by 3 methods: (a) clinical/biologic
judgment (lo), as in the Ritchie index (3), (b) statistical
analyses, such as discriminant analysis, as was done in the
indomethacin study conducted by the Cooperating Clinics of
the ARA (14), and (c) simple summation of the number of
active, responsive joints or their paidtenderness or swelling
scores, as was done in the original 7-day variability study by
the Cooperating Clinics of the ARA (1) and extended by
the CSSRD (5 ).
Simple summation effectively gives each joint equal
weight in quantifying a patient’s joint activity. The clinical/
biologic judgment approach chooses joints and assigns them
weights based on a subjective judgment of each individual
joint’s biologic importance (10). The statistical approach
chooses and weights joints to maximize a statistical criterion.
Used alone, the statistical approach produces measures of articular responses which tend to lack face validity
or intuitive reasonableness (see refs. 8,10,15). For example,
the coefficients and subsets of joints produced by a reliability
or factor analysis may well be statistically optimal but
clinically nonsensical, e.g., 10 x first proximal interphalangeal joint scores + 0.5 x third metacarpophalangeal joint
scores + 25.7 x wrist scores. The use of weights from
discriminant analysis is further flawed by poor criterion
validity (16). Because of the absence of a standard criterion
for joint improvement, these weights maximize the ability to
identify which patients were in which treatment group, and
thus will differ depending on which treatments are under
study.
EGGER ET AL
FULL JOINT SURVEY
ASSESSMENT OF PAINITENOERNESS AND/OR SWELLING
TRAOlilONAL
%T%riTS
11 1
TR.4Dli~DN4L
TRAOl~lONdL
CSSRD
JDlNTCDUNTSI5i
CSSRD
JOINT SCDRESRI
A SUBSET O f THE FULL JOINT
SURVEY IS MEASURED
BY STATISTICAL 4NALYSIS
BY “CLINICALIBIOLOGIC JUDGMENTAL
[APPROACH” IlNTUlllVELY REASONABLE1
I
IMATHEMATlCALLV OPTIMAL: BUT
NOT NECESYIRILI CLINICALLY
REASONABLE1
I
DISCRIMINANT
4NbLYSIS 4MONG
DRUG GROUPS 1 1 4
SIGN4L JOINTS 141
Rlr!*/E
INOEII3I
1
GENERAL12ABILITI
COEFFICIENTS
ANALYSIS fl2l
FACTOR 4NALYSIS
36-JOINT RJS SURVEY
Figure 1. Measures of articular response. Measures shown in italics
are discussed in this report. Numbers in parentheses are reference
numbers. ARA = American Rheumatism Association; CSSRD =
Cooperative Systematic Studies of the Rheumatic Diseases; RJS =
reduced joint survey.
A discriminant analytic summary measure was considered for this study, but was dropped because of the above
considerations and the results of a pilot study which produced large variances. However, the RJS measure discussed
in the next section combines the statistical and clinical/
biologic judgment approaches.
Index definitions. The present report compares traditional CSSRD indices and a modified Ritchie index with the
proposed SJ and RJS indices. Traditional CSSRD joint
counts and scores (5) assess painhenderness and swelling
separately, as opposed to the original ARA joint count (1)
which combined these and other indications of joint activity.
Traditional CSSRD joint counts are the total number of
active, responsive joints (excluding joints with mechanical
problems prohibiting complete response to active drugs).
Traditional CSSRD joint scores summate the total severity
of the above joints, with individual joints scored from 0
(uninvolved) to 3 (most severe) (5). CSSRD defines “active”
joints as those currently showing disease activity, as opposed to “involved” joints which have been affected, but are
not necessarily active at the present time. “Responsive”
joints are involved, active joints which are considered to be
capable of responding to an active drug.
The Ritchie index (3) measures tenderness, not
swelling, and pools groups of joints such as all metacarpophalangeal (MCP) joints in a single score, so that it summates 26 tenderness scores. Our modified Ritchie index
omits evaluation of joints of the cervical spine and includes
separate calculations of painkenderness and swelling.
In contrast to the Ritchie index, in which a fixed
subset of joints is evaluated, the SJ method described by
Ward et a1 (4) entails the use of 2-5 actively involved and
JOINT COUNT IN RA
615
ed equally in separate RJS indices for pain/tenderness and
swelling. SJ and RJS counts and scores were defined analogously to the traditional CSSRD joint counts and scores, as
the count of affected joints and the sum of joint severity
ratings, respectively.
Standardization of indices. To standardize the range
of values across indices, each index was divided by the total
number ofjoints utilized by that index. For example, consider a patient with 10 involved joints out of 68, of which all 10
are tender, 7 of the 10 are swollen, and 1 of the swollen joints
is nonresponsive. Suppose the joint swelling scores of the 9
tender, responsive joints are 3, 2, 3, 1, 2, 2, 0, 0, and 0, and
the first and fourth joints had been chosen as signal joints.
The traditional CSSRD joint swelling count and cumulative
joint swelling score would be calculated as 6 and 13, respectively, with analogous signal joint count and score of 2 and 4.
The standardized indices used for this study would be 6/68 =
0.09 and 13/68 = 0.19 for the traditional measures and 212 =
1.0 and 412 = 2.0 for the signal joint measures. Note that
standardized joint counts will never be larger than 1.O, and
standardized joint scores will never be larger than 3.0.
potentially responsive joints, representing no more than 30%
of a patient’s disease, chosen by the individual physician to
characterize a particular patient’s response to drug therapy
(4). Clinicians chose SJs at the first and last visits in the
AZA-DP and gold studies. Painltenderness and swelling
were evaluated by separate indices.
RJS indices were defined by a combination of the
statistical and clinical/biologic judgment approaches: multiple factor and reliability analyses (8) led to identification of 5
chief groupings of joints which seemed to improve or deteriorate together, as well as a list of joints which did not
contribute to the reliability of the summary joint scores. The
S joint groupings were the MCP joints of the hand, interphalangeal (IP) joints of the hand, MTP joints of the feet, IP
joints of the feet, and large joints such as wrists, elbows,
ankles, and knees. Joints judged to provide little statistical
and clinical information about improvement were eliminated
from the survey. The 36 joints which remained were: the
distal IP joint of the thumb, individual proximal IP and MCP
joints of the hand, individual MTP joints of the feet, and the
wrists, elbows, ankles, and knees. These joints were weight-
Table 1. Response indices by treatment at beginning and end of auranofin versus gold sodium thiomalate (GST) trial in rheumatoid arthritis
patients*
Placebo (n
Variable
Painhenderness
RJS count
First visit
0.60
f
0.05
=
23)
Auranofin (n
A (adjusted)
First visit
-0.02
0.61
=
GST (n
37)
38)
First visit
A (adjusted)
0.04
0.59 t 0.04
-0.23 t 0.04
0.99 2 0.01
-0.29 2 0.06
0.96 t 0.02
-0.32 t 0.06
0.01 t 0.04
0.48
f
0.04
-0.12
?
0.03
0.45
?
0.04
-0.18
f
0.03
f
0.08
-0.41
f
0.06
f
0.05
?
0.04
A (adjusted)
=
-0.15
?
p,
ANCOVA
(K-W)t
0.003
(0.027)
0.094
(0.096)
<O.OOl
(0.007)
0.005
(0.075)$
0.017
(0.032)
0.002
(0.037)
SJ count
1.00 t 0.00
Traditional count
0.49
RJS score
0.95 t 0.09
-0.08
f
0.08
0.93
f
0.08
-0.29
f
0.06
0.87
SJ score
2.08
f
0.11
-0.47
f
0.14
1.95
f
0.10
-0.85
f
0.11
1.93 f 0.11
-1.00 t 0.11
Traditional score
0.75
f
0.08
-0.03 t 0.06
0.73
f
0.08
-0.22 t 0.05
0.66 t 0.07
-0.31
Modified Ritchie score
1.03 f 0.08
-0.44 2 0.05
<0.0001
(0.003)
0.03
0.083
(0.177)
0. I37
(0.150)
1.73
(0.227)
0.119
(0.120)
f
0.05
-0.13 t 0.07
0.01
f
0.07
1.04 f 0.09
-0.28 t 0.05
0.96
f
0.07
?
0.05
Swelling
RJS count
0.50
f
0.04
-0.10
f
0.04
0.59 t 0.04
-0.10 2 0.04
0.53 t 0.04
-0.20
SJ count
0.91
?
0.04
-0.11
f
0.06
1.00 f 0.004
-0.18
0.04
0.96 t 0.03
-0.25 2 0.04
Traditional count
0.32
f
0.03
-0.07
f
0.03
0.40
f
0.03
-0.07 t 0.03
0.36 t 0.03
-0.13
f
0.03
RJS score
0.74
f
0.08
-0.10
f
0.06
0.82
?
0.06
-0.19
f
0.04
0.75
-0.30
f
0.04
SJ score
1.56
f
0. I4
-0.22
f
0. I 1
1.86
f
0.07
-0.60
f
0.08
1.72 t 0.08
-0.76 t 0.08
Traditional score
0.47 t 0.05
-0.12 t 0.04
0.54
f
0.04
-0.12
f
0.03
0.50
-0.20 t 0.03
Modified Ritchie score
0.72
-0.11
0.86 2 0.05
f
0.07
f
0.05
f
-0.19 2 0.04
f
f
0.07
0.06
0.77 t 0.05
-0.32
f
f
0.04
<0.001
(<O .oo1)
0. I59
(0.148)
0.005
(0.007)
* Joint counts and scores were divided by the total number of joints in the survey, yielding “per joint” counts and scores. “Per joint” indices
were then averaged for all patients add adjusted for first visit by analysis of covariance (ANCOVA). Figures are group means 2 SE. At entry,
all groups showed statistical equivalence for each response index. RJS = reduced joint survey; SJ = signal joint.
t P values are for the test of differences among groups by ANCOVA, with I’ values for Kruskal-Wallis (K-W) nonparametric tests in
parentheses.
$ P calues for ANCOVA and Kruskal-Wallis test are on opposite sides of 0.05.
616
EGGER ET AL
Standardized SJ counts for swelling will be less than 1.0
when not all signal joints exhibit swelling or when a previously swollen signal joint has ceased swelling. Standardized
indices identify the average contribution per joint to a
patient’s joint count or score, as if all joints were equally
diseased. Analogous counts and scores were constructed for
the pain/tenderness dimension of joint inflammation.
Statistical methods. The initial sample of 414 (206 +
208) patients entered into the 2 controlled clinical trials of
treatments of rheumatoid arthritis was reduced (by dropout,
changes in observers, and other protocol violations), to 134
subjects in the AZA-DP study and 161 in the gold study. In
55 and 63 patients in the respective studies, signal joints
were not identified at the first visit, because of late institution of the signal joint procedure relative to the rest of the
protocol. This left 23 patients in the placebo group, 37
receiving auranofin, 38 receiving GST, 38 receiving AZA,
and 41 receiving DP, for a total of 177 patients in the 2
studies.
In an attempt to utilize the patients lost due to the
above-described administrative error, signal joints were chosen once at patient entry and independently at patient exit.
The difference in a priori versus post hoc choice of signal
joints was analyzed to assess the effect of including those
patients who had signal joints chosen at trial end but not at
entry. Signal joints identified post hoc were found to be
qualitatively different from those identified at entry. Therefore, the following analysis includes only those 177 patients
who had signal joints chosen at entry and in whom there was
no violation of protocol.
The frequency at which a joint was chosen as the
signal joint in the AZA-DP and gold studies was tabulated.
Unusually over- or under-favored joints were tested against
the proportion expected from random choice based on the
distribution of active, responsive joints, the assumption that
signal joints represent 30% of a patient’s diseased joints, and
a binomial formula.
Evaluation of the proposed indices was performed
using standard reliability/validity criteria, specifically, the
definitions of Tugwell and Bombardier (15). Reliability analyses used to develop the RJS measure included alpha
coefficients, point biserial correlations for individual joints
versus the full survey, and factor analysis; these are detailed
elsewhere (8). Convergent construct validity (15), or agreement with other accepted measures, was examined using
Pearson correlations of proposed and traditional indices.
Discriminant validity, or issues of precision and power to
detect an active drug, was compared for the proposed and
Response indices by treatment at beginning and end of azothioprine versus D-penicillamine trial in rheumatoid arthritis patients*
Table 2.
D-penicillarnine (n
Azathioprine (n = 38)
=
41)
First visit
A (adjusted)
First visit
A (adjusted)
0.64 t 0.04
-0.23 t 0.05
0.66 t 0.04
-0.20 2 0.04
SJ count
0.98 t 0.02
-0.35 t 0.06
0.93
2
0.03
-0.30 2 0.06
Traditional count
0.54 t 0.04
-0.19
0.56
2
0.04
-0.17 2 0.04
RJS score
0.98 t 0.07
-0.40 t 0.08
1.07 t 0.09
-0.39 2 0.07
SJ score
2.01 t 0.12
-0.88 t 0.12
1.93 t 0.11
-0.87 2 0. I 1
Traditional score
0.82 t 0.07
-0.33 t 0.07
0.89 t 0.08
-0.32
Modified Ritchie score
1.12
It
0.08
-0.41 t 0.07
1.09 t 0.08
-0.42 2 0.07
0.53
2
0.03
-0.20 t 0.03
0.55 t 0.04
-0.19 2 0.03
SJ count
0.99 t 0.01
-0.31 t 0.05
0.95 t 0.03
-0.23 2 0.05
Traditional count
0.40 t 0.03
-0.15 t 0.02
0.41 2 0.03
-0.16 2 0.02
RJS score
0.84 t 0.08
-0.38 t 0.05
0.84 t 0.07
-0.38
SJ score
2.00 t 0.10
-0.91 t 0.10
1.79 t 0.11
-0.80 2 0.10
Traditional score
0.61 t 0.06
-0.28 t 0.03
0.60
0.05
-0.28 2 0.03
Modified Ritchie score
0.91 t 0.07
-0.34 t 0.05
0.81 2 0.06
-0.37 t 0.05
Variable
Pain/tenderness
RJS count
Swelling
RJS count
~
~~~~~~~~
~
2
0.04
2
C
?
0.07
0.05
P,
ANCOVA
(K-W)-F
0.666
(0.776)
‘0.560
(0.384)
0.742
(0.802)
0.914
(0.8 10)
0.95 I
(0.779)
0.886
(0.922)
0.935
(0.731)
0.965
(0.848)
‘0.296
(0.202)
0.930
(0.709)
0.954
(0.8 10)
0.433
(0.278)
0.978
(0.87 I )
0.667
(0.735)
~
* Joint counts and scores were divided by the total number ofjoints in the survey, yielding “perjoint” counts and scores. “Perjoint” indices
were then averaged for all patients and adjusted for first visit by analysis of covariance (ANCOVA). Figures are group means 2 SE. At entry,
all groups showed statistical equivalence for each response index. RJS = reduced joint survey: SJ = signal joint.
t P values are for the test of differences among groups by ANCOVA, with P values for Kruskal-Wallis (K-W) nonparametric tests in
parentheses.
JOINT COUNT IN RA
617
traditional joint indices by analysis of covariance (ANCOVA), with verification by the Kruskal-Wallis test to
control for violations of assumptions. Precision was further
described by tabulating coefficients of variation ( 17) for
differences between drug groups and ratios of F-statistics
from ANCOVA. To permit comparison of the proposed
measures with the measures actually used in the drug
studies, descriptive statistics and differences between first
and last visit adjusted by ANCOVA are given in Tables 1 and
2 , showing only those patients for whom both traditional and
proposed indices could be computed.
The problem of multiple inference, i.e., performance
of so many statistical tests at significance level alpha that
some. of the results are significant by chance alone, arose in 3
separate areas: the evaluation of clinician preferences for
signal joints, the correlations which establish convergent
construct validity, and the drug efficacy comparisons. A
blanket significance level of a = 0.05 or a = 0.01 was not
sufficient to control for the probability of spurious significance in this study. Therefore, the significance level is
controlled within each table so that < I spuriously significant
result is expected for each research question. For clinician
preference of signal joints, alpha is dropped to 0.005; for
correlations, it is dropped to 0.001. Exact P values are
reported for drug efficacy comparisons in Tables I and 2 ,
since these are what would appear in a clinical report.
The relative precision of traditional versus proposed
indices is calculated as the ratio of squared coefficients
of variation of differences between groups, with improvement at last visit corrected for severity at first
visit by ANCOVA. This ratio is approximately the
sample size inflation factor necessary to achieve the
same significance level and power if traditional indices
were replaced by either SJ or RJS indices. Table 3
shows that both SJ and RJS indices had precision
comparable with traditional measures. Our modified
Ritchie index also performed quite favorably. Both
proposed indices exhibited instances of substantial
savings in sample size, and there were cases in which
traditional measures reported a difference between
groups which was severely underestimated by the SJ
o r RJS index.
Convergent construct validity (17) of SJ and
RJS indices was fair to excellent compared with the
complete traditional joint count and score and a modified Ritchie index, as measured by Pearson correlations. RJS indices were more highly correlated with SJ
and with the full joint survey indices than the latter 2
measures were correlated with each other. All correlations were statistically significant, except at study
entry, where SJ counts had low correlation with other
indices as an artifact of uniform involvement of signal
joints.
RESULTS
Reliability and validity. Assessment of discriminant validity, or precision, and sensitivity of these
reduced indices to an active drug is shown in Table 3.
Table 3. Sample size inflation factors for reduced joint count surveys versus traditional indices, in
controlled trials of treatments for rheumatoid arthritis*
Gold treatment trial
Painitenderness
SJ countitraditional count
SJ scoreitraditional score
SJ score/Ritchie index
RJS countitraditional count
RJS scoreitraditional score
RJS score/Ritchie index
Swelling
SJ countitraditional count
SJ scoreitraditional score
SJ score/Ritchie index
RJS countitraditional count
RJS scoreitraditional score
RJS scoreiRitchie index
Placeboauranofin
PlaceboGST
AuranofinGST
2.02
1.44
2.53
1.44
1.30
2.28
3.34
I .56
3.20
1.21
1.17
2.40
16.321
I .93
0.94
0.98
2.62
2.40
0.002$
0.OOlf
0.21
12.1 If
0.12f
27.35t
0.55
0.15
2.02
1.61
2.76
0.72
0.94
1.61
0.88
0.23
0.71
0.70
0.86
2.61
0.66
0.67
0.77
3.31
5.15
Summary
3.09
I .56
3.16
I .23
1.18
AzathioprineD-penicillamine
trial
0.31
5.38
1.74
0.58
I .77
1.77
0.006f
0.0009f
0.30
4.04i
0.22f
54.61f
* Values are ratios of squared coefficients of variation (CV’ = SE’iA’), where A is the estimated index
difference between treatment groups as calculated by analysis of covariance (ANCOVA). For
example, the number of patients needed to compare painitenderness in the placebo and auranofin
groups at a = 0.05 with 90% power using the signal joint count is 2.02 times the number of patients
required if the traditional joint count were used. The summary column is the ratio of F-statistics from
ANCOVA in the gold study. GST = gold sodium thiomalate; SJ = signal joint; RJS = reducedjoint
survey.
t The estimated index difference (A) is less than its standard error.
$ Estimated As are close to 0.
EGGER ET AL
618
We judged that the index based on signal joints
would comprehensively summarize the intuitive clinical concept of “disease activity” if the clinicians’
choices of SJs met the following criteria: 1) Choice of
SJs must include diseased joints which are easily
observed and which constitute good signals of a patient’s disease progression or response to treatment.
2) Clinicians’ choices of SJs must be sufficiently
predictable to assure reproducibility of summary
scores by trained physicians on similar patients in
similar experiments. Therefore, we tabulated the frequency of choice of each joint as a signal joint and
tested to see which joints were significantly preferred
(see Methods). Wrists, knees, ankles, elbows, and the
first through third MCP joints were favored, controlling for the frequency with which joints were involved
and responsive. Thus, the numbers represent clinician
preference, not simply the disease process. Only the
wrists achieved statistically significant clinician preference ( a = 0.005),with a statistically significant lack
of preference for IP joints of the hands and feet.
The overall ratio of signal joints to responsive
joints was 0.11 in both studies, indicating that the
“30%-of-disease” instructions for choice of SJs could
not always be followed. This will tend to inappropriately reduce the statistical significance for joints which
were preferred 20-40% of the time. As a consequence,
the large joints and first through third MCP joints were
probably more preferred than the statistical model
showed. Joints on the right side tended to be slightly
more favored than joints on the left side, perhaps
reflecting the fact that more patients were right-handed. In the 2 studies, clinicians chose a mean and
median of 4joints as SJs. The frequent choice of large
joints suggests that without other instructions, clinicians will choose SJs that represent the worst of a
patient’s disease. In fact, SJs did tend to be more
severely afflicted than the “average” responsive joint
at the first visit (see Tables 1 and 2).
DISCUSSION
This study investigated 2 “streamlined” indices
of joint disease that combine statistical and clinical/
biologic judgment approaches to data condensation in
2 controlled clinical trials of treatments for rheumatoid
arthritis. With signal joints chosen by the clinician, SJ
counts and scores primarily summarize the patient’s
disease in large joints and are moderately sensitive to
the effects of active drugs. They are less precise than
indices from the full joint survey for paidtendernem,
but more precise for swelling. RJS counts and scores,
based on a 36-joint survey, eliminate joints which
provide little information on the overall disease process. Their precision is more nearly representative of
the full joint survey. Since this study incorporated
multiclinic variability, the indices should be even more
precise when utilized at single institutions.
One advantage of SJ and RJS indices in clinical
trials is cost-effectiveness. They would reduce clinician time required for completion of case report forms,
as well as costs of data storage and analysis. They
eliminate the necessity to estimate scores for joints
that are difficult to evaluate, such as the IP joints of the
foot, the hips, acromioclavicular joints, and temporomandibular joints. This benefit carries over to clinical
practice.
While the signal joint method is flexible in
allowing the clinician to tailor a measure of significant
improvement for each patient at any stage of the
disease, it is limited in its capacity to describe deterioration, particularly when non-signal joints deteriorate.
The SJ approach allows clinicians to choose the studied joints, causing the results of the study to be
dependent on the characteristics of the clinicians involved and their method of clinical practice. Although
clinicians in these studies were fairly consistent in
choosing as signal joints the large joints representing
the worst of a patient’s disease, the use of SJs could
conceivably reduce the reproducibility of a clinical
trial. A more serious concern is that although in this
investigation and previous work (4) clinicians favored
SJs characterizing large joint disease, factor and reliability analyses indicate that it may be equally important to follow the progress of disease in the hands and
feet. Thus, SJs, as presently defined, cannot be recommended as a substitute for the full joint survey.
In contrast, RJS indices reflect the range of
joint disease progression. Their validity has been well
established. They share the inter- and intra-observer
reliability of the full joint survey, since for both
measures, observers score joints individually. Their
precision has been assessed in 2 controlled clinical
trials of active drugs and is comparable with, though
occasionally reduced from, that of the full joint survey. Index precision in the AZA-DP trial should not
be equated with statistical significance of the drug
comparisons since both drugs were active. Thus, RJS
indices can be utilized reasonably by both the clinician
and researcher in quantifying the status of joint disease
in rheumatoid arthritis.
JOINT COUNT IN RA
REFERENCES
1. The Cooperating Clinics Committee of the American
Rheumatism Association: A seven-day variability study
of 499 patients with peripheral rheumatoid arthritis.
Arthritis Rheum 8:302-335, 1965
2. Lansbury J, Haut DD: Quantitation of the manifestations of rheumatoid arthritis. 4. Area of joint surfaces as
an index to total joint inflammation and deformity. Am J
Med Sci 232:150-155, 1956
3. Ritchie DM, Boyle JA, McInnes JM, Jasani MK, Dalakos TG, Grieveson P, Buchanan WW: Clinical studies
with the articular index for the assessment of joint
tenderness in patients with rheumatoid arthritis. Q J
Med 37:393-406, 1968
4. 'Ward JR, Niethammer TA, Egger MJ: Can we just
measure signal joints? Should ring size and walking time
be analyzed only in selected patients? Controversies in
the Clinical Evaluation of Analgesic-Anti-Inflammatory-Antirheumatic Drugs. Edited by HE Paulus, GE
IEhrlich, E Lindenlaub. New York, FK Schattauer Verlag, 1980
5. 'Williams HJ, Ward JR, Reading JC, Egger MJ, Grandone JT, Samuelson CO, Furst DE, Sullivan JM, Watson MA, Guttadauria M, Cathcart ES, Kaplan SB, Halla
.IT, Weinstein A, Plotz PH: Low-dose D-penicillamine
therapy in rheumatoid arthritis: a controlled, doubleblind clinical trial. Arthritis Rheum 26:581-592, 1983
6. 'Ward JR, Williams HJ, Egger MJ, Reading JC, Boyce E,
.Altz-Smith M, Samuelson CO Jr, Willkens RF, Solsky
MA, Hayes SP, Blocka KL, Weinstein A, Meenan RF,
IGuttadauria M, Kaplan SB, Klippel J: Comparison of
;auranofin, gold sodium thiomalate, and placebo in the
.treatment of rheumatoid arthritis: a controlled clinical
trial. Arthritis Rheum 26:1303-1315, 1983
7. Paulus HE, Williams HJ, Ward JR, Reading JC, Egger
MJ, Coleman ML, Samuelson CO Jr, Willkens RF,
Guttadauria M. Alarcon GS, Kaplan SB, MacLaughlin
619
EJ, Weinstein A, Wilder RL, Solsky MA, Meenan RF:
Azathioprine versus D-penicillamine in rheumatoid arthritis patients who have been treated unsuccessfully
with gold. Arthritis Rheum 27:721-727, 1984
8. Egger MJ, Ward JR, Reading JC, Williams HJ, Coleman
ML: Evaluation of summary indices of response in
controlled clinical trials (abstract). Controlled Clin Trials 3:154, 1982
9. Buchanan WW: Assessment of joint tenderness, grip
strength, digital joint circumference and morning stiffness in rheumatoid arthritis. J Rheumatol 9763-765,
1982
10. Buchanan WW, Tugwell P: Traditional assessments of
articular diseases. Clin Rheum Dis 9515-529, 1983
11. Tugwell P, Bombardier C: Practical issues in clinical
trials as applied to current state of the art in rheumatology trials. Br J Rheumatol (suppl) 22:9-13, 1983
12. Eberl DR, Fasching V, Rahlfs V, Schleyer I, Wolf R:
Repeatability and objectivity of various measurements
in rheumatoid arthritis: a comparative study. Arthritk
Rheum 19:1278-1286, 1976
13. Urowitz MB, Hunter T, Bookman AAM, Gordon DA,
Smythe HA, Ogryzlo MA: Azathioprine in rheumatoid
arthritis: a double blind study comparing full dose to half
dose. J Rheumatol 1:274-281, 1974
14. Cooperating Clinics of the ARA Committee: A threemonth trial of indomethacin in rheumatoid arthritis, with
special reference to analysis and inference. Clin Pharmacol Ther 8: 11-37, 1967
15. Tugwell P, Bombardier C: A methodologic framework
for developing and selecting endpoints in clinical trials. J
Rheumatol 9:758-762, 1982
16. Bombardier C, Tugwell P: A methodological framework
to develop and select indices for clinical trials: statistical
and judgmental approaches. J Rheumatol 9:753-757,
1982
17. Snedecor GW, Cochran WG: Statistical Methods. Seventh edition. Ames, The Iowa State University Press,
1980, p 37
Документ
Категория
Без категории
Просмотров
4
Размер файла
664 Кб
Теги
evaluation, joint, arthritis, indices, count, reduced, rheumatoid
1/--страниц
Пожаловаться на содержимое документа