Scand J Rheumatology 8: 101-105, 1979
CLINICAL ASSESSMENT OF DISEASE ACTIVITY IN RHEUMATOID ARTHRITIS T. Msrk Hansen. S. Keiding, S. Lilholt Lauritzen,' R. Manthorpe, S. Freiesleben Ssrensen and A. Wiik
Scand J Rheumatol Downloaded from informahealthcare.com by University of Auckland on 12/06/14 For personal use only.
From the Medical Department TA, Rigshospitalet, and the 'Institute of Mathematical Statistics, University of Copenhagen, Copenhagen, Denmark
ABSTRACT. A new experimental design was developed to study the value of clinical parameters of disease activity in patients with rheumatoid arthritis. Ten patients with classical rheumatoid arthritis were examined by five senior doctors in a department of medical rheumatology. In spite of a n attempt to make the clinical examination as uniform as possible the inter-observer variation among the doctors was greater than the variation among the patients, for the following parameters: joint pain a t rest, joint tenderness and joint swelling. An acceptable inter-observer variation in relation to patient variation was found for 1) a combined registration of joint pain at rest or on movement, 2) duration of morning stiffness, 3) grip strength, 4) subjective wellbeing as indicated on a visual analogue scale, 5) fingertip palm distance, and 6) maximum flexionextension in elbows, wrists and knees. The variation from morning to afternoon and from day to day was negligible. I t is concluded that registration of elaborated articular scores is useless in the daily routine in rheumatological departments when different doctors examine the patients.
It is essential to apply objective clinical criteria in the quantitation of disease activity in order to follow progression and regression in chronic diseases such as rheumatoid arthritis (RA). This is important for the assessment of disease activity and of the response to a particular treatment in the individual patient as well as for the evaluation of the efficacy of new treatments in clinical trials (5). The determination of various articular indices is often quite time-consuming. This raises the question of the validity of the registration of clinical parameters. The purpose of the present investigation was to evaluate the value of certain clinical parameters in RA, when they were measured by the senior doctors in a rheumatological department. Variations among doctors were compared with variations among the patients. The day-to-day variations and the variations from morning to afternoon were also considered.
MATERIALS AND METHODS Ten in-patients, 6 women and 4 men aged 30-71 years with classical RA according to the criteria of the Committee of the American Rheumatism Association (2) and with a duration of the disease of at least I year, were examined by five senior doctors according to the following scheme: Doctor no. Day 2
Day 1
Patient 1
2 3 4 5
Moming
Afternoon
Moming
Afternoon
1 2 3 4
2 3 4
4 5
3 4
2 3 4 2
2 2 5
5
5 1
6 7 8 9
1 2 3 4
3 4 5 1
10
5
2
1
1
5
1
5
4 3
3
1
The design of the experiment was made so as to minimize the effect of repeated examinations on the same patients. Each patient was examined only once by the same doctor and only once at the same time on each day. All patients were examined four times by four different doctors. The examinations were performed in the mornings between 8.30 and 10.00 and in the afternoons between 1.30 and 3.00 on 2 consecutive days. In an attempt to make the clinical evaluation by the 5 doctors as uniform as possible, a detailed written instruction was prepared before the investigation took place. Clinical parameters The joint examinations included the following joints: temporo-mandibular joints, cervical spine, thoraco-lumbar spine, shoulder, elbow, wrist, metacarpal-phalangeal, proximal interphalangeal, hip, knee, ankle, mid-tarsal, and metatarsal-phalangeal joints. The following joints were treated as single units: The temporomandibular Sccmd J Rheiimntology 8
Scand J Rheumatol Downloaded from informahealthcare.com by University of Auckland on 12/06/14 For personal use only.
102
T . M0rk Hansen et (11.
joints. the joints of the cervical and thoraco-lumbar spine. the metacarpal-phalangeal and proximal interphalangeal joints of each hand, the mid-tarsal and the metatarsalphalangeal joints of each foot. 1. Pain ut rest. The number of joints with pain at rest were recorded. 2. Puin on movement. The number of joints with pain on movement were recorded. All the joints with pain at rest were included as having pain on movement. 3. Pain by pressure. A firm pressure was applied over the joint margins. The pressure should not result in pain in normal joints, but be firm enough to prevent minor attempts at withdrawal. The tenderness was scored from 0 to 3 : where 0 equals no pain, 1 equals complaint of pain. 2 equals pain and wincing and 3 equals wince and withdrawal. When the joint margins could not be felt (the spine and hips), pain on passive movement was recorded. 4 . Number of joints tt*ith pain on pressure itithout yruding .
5 . Joint swjelling. Joint swelling was recorded in the following joints: elbows, wrists. metacarpal-phalangeal, proximal interphalangeal. knees, ankles, and metatarsalphalangeal joints. A score from 0 to 3 was used. 0 equalled no swelling: I , swelling just palpable; 2. swelling palpable and visible but with normal contours preserved: 3 . swelling distorting the normal joint contours. 6 . Number of joints itith src,elling t+~ithoutgradir7g 7 . Morning stiffness. The patients’ information on the time in minutes which was required until the maximum relief of morning stiffness was obtained was recorded. The value was given as an average of the last 7 days. 8 . Grip strength. Grip strength was recorded for both hands with a Martin Vigorimeter. The value was a summation of the best of 3 measurements for each hand. expressed in kp/cm2. 9 . Subjective well-being indicated on a tisual analogue scide. The patient was asked to indicate on a 100 mm long scale the degree of well-being with respect to his rheumatism during the last week. 10. Fingertip-palm distcince. The minimum distance from the tip of the 3rd finger to the palm was measured when the patient clenched his first. The result was given in mm as the sum of both hands. 11. Muximumfle.rion-e.rrension was recorded in the elbows, wrists. and knees, using a protractor. Stati.stiru1 methods
The experimental design was composed of complete designs and balanced incomplete block designs (BIBDs) so as to make i t possible from a relatively modest number of patients and doctors to distinguish the various main sources of variation (3). The model for the experiment was such that a result of a measurement was the realization of a random variable Xinit where i-doctor, i = l . ....5 . j-patient. j = 1. .... 10. d-day. d= 1.2. /-morning or afternoon, t= I , 2. All combinations of j , d , t occurred, whereas i was obtained from j . d and r by the design. We assumed that X,,,, was composed as X i j ,= q , l + ~ j d + y c , , + ~ Ji=i(j), d f r i.e. an effect q,lthat was specific to a given doctor on a given day. and effect that was specific to a patient on a given day. Y , ~ ,specific to the day and the time of day, i u r r i d J Rheumoroiogy 8
and. finally, a random effect E ~ which ~ , ~was ~ to be interpreted as measurement error. Note that the error term eurlfincludes such things as an interaction between doctor and patient, since no patient was examined more than once by the same doctor. Different E ~ , , ~ , ’ Swere assumed to be independent, with expectation equal to zero and variance u2. The design is a BIBD in several ways: For each fixed day we have a BIBD with patients as 10 blocks of size 2 and doctors as 5 treatments. We have a BIBD with patients on different days as 20 blocks of size 2 and doctors as 5 treatments. We have a BIBD with patients as 10 blocks of size 4 and doctors as 5 treatments. Further, the design is complete in patients. days. and time of day. Using these facts and standard linear methods in the analysis of factorial experiments and BIBDs (3, we were able to break down the variation into the following s ? = R , - effect of time of day S: = R 2 - day-today variation of time-of-day effect s:=&, R 3 -. residual variation S L;-? -l R 1 - day-today variation within doctors s 25-Tc - 1 R 5 - day-today variation within patients s g = t R e - variation among doctors s;=; R,- variation among patients
The most important feature of this breakdown of the variation is that any one of the variations so computed is not distorted by any other variation. For example, the day-today variation among the doctors is eliminated when computing the variation among doctors, etc. These were calculated by the following formulas R 1-40 -L (Zj(Xij1 1 +xij21-xu 1 2 - x i j 2 2 ) ) 2 R ? = & Gj(xijl1- x i j z ! -Xij1z+xij22))* R , = S S - $ ( S S p l + S S p 2 + 5 SSL!+5 SSL2)-R1-R2
Clinical assessment of disease activity in R A
103
Table I . Vm-icinces, means, overall standard deviations, and overcill coefficients of variance of cliriicul joint examinations in 10 patients with rheumatoid arthritis examined by 5 doctors Number of joints Pain at rest or on Pain on movement pressure
Pain at rest or on movement or Joint on pressure swelling
12.2 8.1 0.4 29.3 10.7 64.8 58.2 4 3.52
3.99 2.03 0.03 2.95 4.56 5.98 74.68
8.35 1.23 0.63 8.28 11.13 54.53 73.56
11
11
4.63 2.03 5.63 4.20 3.10 14.38 54.19 14
88
17
Scand J Rheumatol Downloaded from informahealthcare.com by University of Auckland on 12/06/14 For personal use only.
Pain at rest Variance Residual Sf Morninglafternoon s: Day/day S: Daylday (doctor) s: Day/day (patient) s: Doctors s: Patients S: Mean Overall standard deviation Overall coefficient of variance (%)
Li=&.j
1.87
2.86 26
2.52 0.90 0.90 8.73 3.61 89.88 17.79 7 1.89
1.%
Pain on pressure 0-3
Joint swelling 0-3
27.8 8.1 14.4 146.9 24.8 633.2 259.8 18 6.84
10.5 13.2 0.6 19.5 16.5 187.8 79.5 12 3.72
27
14
38
31
RESULTS
t X i j d t , Ki=ZjsUi)Zd, r Xu,,/
k(i)=set of patients treated by doctor i R,=FSSL
RT=+(ssp-hGi,j,d , t X i W / ) ' ) the residual variation, thus had an expectation equal to cr', the variance of the measurement error. All the other s: values had expectation equal u2plus a positive quantity depending on the force of the corresponding effect. The coefficients of variance were estimated as
s;,
ri=4Osi/(4 , j ,d , X U I ) .
The variances, the means, the overall standard deviations, and the overall coefficients of variance are given in Tables 1 and 2. The coefficients of variance make it easier to compare the different variables. The variations from morning to afternoon and from day to day were of the same magnitude as the residual variance and were negligible in comparison with the inter-individual variation among doctors and patients. For this reason a better estimate of the variance of the measurement error was determined from a weighted average of the sf for i= 1, ...,5 .
Table 11. Variances, means, overall standard deviations, and overall coefficients of variance of clinical variables in 10 patients with rheumatoid arthritis examined by 5 doctors
Variance sf Residual s: Morning/afternoon S% Daylday S2 Day/day (doctor) s: Day/day (patient) s% Doctors s: Patients Mean Overall standard deviation Overall coefficient of variance (%)
Morning stiffness
Grip strength
Wellbeing
Fingertip palm distance
Maximum flexionextension
129 181 60 1 109 757 333 30 167 86 min 20 23
0.0076
33 63 240 121 20 126 2 712 41 mm 7 17
36 84 176 29 94 39 610 f l mm 8 73
440 640 23 598 1 208 192 71 744
0.0055
0.0141 0.0052 0.0083 0.0188 0.1367 0.21 kg/cmz 0.0088 42
590" 27 5 Scand J Rtieumarology 8
Scand J Rheumatol Downloaded from informahealthcare.com by University of Auckland on 12/06/14 For personal use only.
104
T . M m k Hcrnseri et trl.
The overall coefficients of variance of meaaurement error were 88% for pain at rest and 73% for fingertippalm distance, while for all the other variables the coefficients ranged from 5 R to 42 R. The variations among doctors were of the same magnitude or greater than the variations among patients for the following parameters: pain at rest, pain on pressure. joint swelling without grading. and pain and joint swelling with grading. The following parameters showed greater variation among patients than among doctors: A registration of pain at rest or on movement, morning stiffness, grip strength, well-being on a visual analogue scale. fingertippalm distance. and maximum flexionextension of joints.
DISCUSSION
I n the present investigation a selection has taken place of both patients and doctors. The patients all had classical rheumatoid arthritis according to the criteria of the American Rheumatism Association, the disease had lasted for at least I year. and t h e patients had all been referred to a rheumatological department for treatment because of active disease. The doctors were senior doctors with experience and interest in rheumatology. This may explain the low intra-patient and intra-observer variance for all the measurements [daylday (doctor), day/day (patient) tables l and 31. Most of the investigated parameters had a regular pattern of variation indicating that the analyses performed were adequate. In the case of morning stiffness there was almost no variation at all between the different observations on the same patient. This is possibly due to the registration of the average value for the last 7 days and not an exact single-day value. Also. a tendency to approximate the values to half or whole hours may have affected the measured variation. The statistical results for this parameter should therefore be accepted with the greatest precaution. In the parameters fingertippalm distance and pain at rest there appeared to be so many zeros among the observations that it is questionable whether the analysis describes the data adequately. Also. the overall coefficients of variance of measurement error were relatively high for these variables. In such cases, a more non-parametric statistical approach would be suitable. The drawback
of this kind of approach is that it would not enable us to distinguish the sources of variation. For routine clinical use it is imperative, though not necessarily sufficient for the validity of a measured variable, that the variation among the examiners is smaller than the variation among the patients. This requirement was only fulfilled for pain at rest or on movement, morning stiffness, grip strength, well-being. fingertippalm distance, and maximum flexion-extension. For this reason, it is not recommendable for a routine use, to include a registration of joint pain at rest only, joint tenderness, orjoint swelling in an articular score, because of the high inter-observer variation. The high interobserver variation of the measurement of joint tenderness may be due to variation in the location and the force of the applied joint pressure. The registration of joint pain at rest relies on the patients' information alone and it may be influenced by the individual doctors' method of questioning and their interpretation of the patients' answers. This does not rule out, however, that the more elaborated articular scores may be of some value in clinical studies, providing the examinations are done by the same observer throughout the study ( 1 , 6). This study supports the recommendations of Lansbury (4) to include registration of grip strength and joint pain on passive movement in any activity index. Furthermore, it may be of value to use the visual analogue scale to register the patients' "wellbeing" in the same way as it has been used to register pain (7). Although an elaborate registration of joint symptoms in patients with rheumatoid arthritis is desirable, the effort is not justified in daily routine because of the individualism of doctors.
ACKNOWLEDGEMENT Thiw work was supported by the Danish Medical Research Council (grant no. 552495).
REFERENCES I . Camp, A. V.: An articular index for the assessment of rheumatoid arthritis. Orthopaedics (Oxford) 4: 39, 1971. 2. Committee of the American Rheumatism Association. Arthritis Rheum2: 16, 1959. 3. Kempthorne, 0.:The design and analysis of experiments. Wiley, New York 1952.
Clinical assessment of disease activity in R A
assessment ofjoint tenderness in patients with rheumatoid arthritis. Quart J Med37: 393, 1968. 7. Scott. J. & Huskisson, E. C.: Graphic representation ofpain. Pain2: 175, 1976. Submitted for publication May 10, 1978
T. Mark Hansen Medical Department TA Rigshospitalet DK-2200 Copenhagen, Denmark
Scand J Rheumatol Downloaded from informahealthcare.com by University of Auckland on 12/06/14 For personal use only.
4. Lansbury, J.: Methods for evaluating rheumatoid arthritis. In Arthritis and Allied Conditions (eds. J. L. Hollander & D. J. McCarty), 8th ed., pp. 269. Lea & Febiger, Philadelphia 1972. 5. Lee, P., Sturrock, R. D., Kennedy, A. & Dick, W. C.: The evaluation of antirheumatic drugs. Curr Med Res Opin 1: 427, 1973. 6. Ritchie, D. M., Boyle, J. A., McInnes, J. M., Jasani, M. K., Dalakos, T. G., Grieveson, P. & Buchanan, W. W.: Clinical studies with an articular index for the
105
Scand J Rheumatology 8
CLINICAL ASSESSMENT OF DISEASE ACTIVITY IN RHEUMATOID ARTHRITIS T. Msrk Hansen. S. Keiding, S. Lilholt Lauritzen,' R. Manthorpe, S. Freiesleben Ssrensen and A. Wiik
Scand J Rheumatol Downloaded from informahealthcare.com by University of Auckland on 12/06/14 For personal use only.
From the Medical Department TA, Rigshospitalet, and the 'Institute of Mathematical Statistics, University of Copenhagen, Copenhagen, Denmark
ABSTRACT. A new experimental design was developed to study the value of clinical parameters of disease activity in patients with rheumatoid arthritis. Ten patients with classical rheumatoid arthritis were examined by five senior doctors in a department of medical rheumatology. In spite of a n attempt to make the clinical examination as uniform as possible the inter-observer variation among the doctors was greater than the variation among the patients, for the following parameters: joint pain a t rest, joint tenderness and joint swelling. An acceptable inter-observer variation in relation to patient variation was found for 1) a combined registration of joint pain at rest or on movement, 2) duration of morning stiffness, 3) grip strength, 4) subjective wellbeing as indicated on a visual analogue scale, 5) fingertip palm distance, and 6) maximum flexionextension in elbows, wrists and knees. The variation from morning to afternoon and from day to day was negligible. I t is concluded that registration of elaborated articular scores is useless in the daily routine in rheumatological departments when different doctors examine the patients.
It is essential to apply objective clinical criteria in the quantitation of disease activity in order to follow progression and regression in chronic diseases such as rheumatoid arthritis (RA). This is important for the assessment of disease activity and of the response to a particular treatment in the individual patient as well as for the evaluation of the efficacy of new treatments in clinical trials (5). The determination of various articular indices is often quite time-consuming. This raises the question of the validity of the registration of clinical parameters. The purpose of the present investigation was to evaluate the value of certain clinical parameters in RA, when they were measured by the senior doctors in a rheumatological department. Variations among doctors were compared with variations among the patients. The day-to-day variations and the variations from morning to afternoon were also considered.
MATERIALS AND METHODS Ten in-patients, 6 women and 4 men aged 30-71 years with classical RA according to the criteria of the Committee of the American Rheumatism Association (2) and with a duration of the disease of at least I year, were examined by five senior doctors according to the following scheme: Doctor no. Day 2
Day 1
Patient 1
2 3 4 5
Moming
Afternoon
Moming
Afternoon
1 2 3 4
2 3 4
4 5
3 4
2 3 4 2
2 2 5
5
5 1
6 7 8 9
1 2 3 4
3 4 5 1
10
5
2
1
1
5
1
5
4 3
3
1
The design of the experiment was made so as to minimize the effect of repeated examinations on the same patients. Each patient was examined only once by the same doctor and only once at the same time on each day. All patients were examined four times by four different doctors. The examinations were performed in the mornings between 8.30 and 10.00 and in the afternoons between 1.30 and 3.00 on 2 consecutive days. In an attempt to make the clinical evaluation by the 5 doctors as uniform as possible, a detailed written instruction was prepared before the investigation took place. Clinical parameters The joint examinations included the following joints: temporo-mandibular joints, cervical spine, thoraco-lumbar spine, shoulder, elbow, wrist, metacarpal-phalangeal, proximal interphalangeal, hip, knee, ankle, mid-tarsal, and metatarsal-phalangeal joints. The following joints were treated as single units: The temporomandibular Sccmd J Rheiimntology 8
Scand J Rheumatol Downloaded from informahealthcare.com by University of Auckland on 12/06/14 For personal use only.
102
T . M0rk Hansen et (11.
joints. the joints of the cervical and thoraco-lumbar spine. the metacarpal-phalangeal and proximal interphalangeal joints of each hand, the mid-tarsal and the metatarsalphalangeal joints of each foot. 1. Pain ut rest. The number of joints with pain at rest were recorded. 2. Puin on movement. The number of joints with pain on movement were recorded. All the joints with pain at rest were included as having pain on movement. 3. Pain by pressure. A firm pressure was applied over the joint margins. The pressure should not result in pain in normal joints, but be firm enough to prevent minor attempts at withdrawal. The tenderness was scored from 0 to 3 : where 0 equals no pain, 1 equals complaint of pain. 2 equals pain and wincing and 3 equals wince and withdrawal. When the joint margins could not be felt (the spine and hips), pain on passive movement was recorded. 4 . Number of joints tt*ith pain on pressure itithout yruding .
5 . Joint swjelling. Joint swelling was recorded in the following joints: elbows, wrists. metacarpal-phalangeal, proximal interphalangeal. knees, ankles, and metatarsalphalangeal joints. A score from 0 to 3 was used. 0 equalled no swelling: I , swelling just palpable; 2. swelling palpable and visible but with normal contours preserved: 3 . swelling distorting the normal joint contours. 6 . Number of joints itith src,elling t+~ithoutgradir7g 7 . Morning stiffness. The patients’ information on the time in minutes which was required until the maximum relief of morning stiffness was obtained was recorded. The value was given as an average of the last 7 days. 8 . Grip strength. Grip strength was recorded for both hands with a Martin Vigorimeter. The value was a summation of the best of 3 measurements for each hand. expressed in kp/cm2. 9 . Subjective well-being indicated on a tisual analogue scide. The patient was asked to indicate on a 100 mm long scale the degree of well-being with respect to his rheumatism during the last week. 10. Fingertip-palm distcince. The minimum distance from the tip of the 3rd finger to the palm was measured when the patient clenched his first. The result was given in mm as the sum of both hands. 11. Muximumfle.rion-e.rrension was recorded in the elbows, wrists. and knees, using a protractor. Stati.stiru1 methods
The experimental design was composed of complete designs and balanced incomplete block designs (BIBDs) so as to make i t possible from a relatively modest number of patients and doctors to distinguish the various main sources of variation (3). The model for the experiment was such that a result of a measurement was the realization of a random variable Xinit where i-doctor, i = l . ....5 . j-patient. j = 1. .... 10. d-day. d= 1.2. /-morning or afternoon, t= I , 2. All combinations of j , d , t occurred, whereas i was obtained from j . d and r by the design. We assumed that X,,,, was composed as X i j ,= q , l + ~ j d + y c , , + ~ Ji=i(j), d f r i.e. an effect q,lthat was specific to a given doctor on a given day. and effect that was specific to a patient on a given day. Y , ~ ,specific to the day and the time of day, i u r r i d J Rheumoroiogy 8
and. finally, a random effect E ~ which ~ , ~was ~ to be interpreted as measurement error. Note that the error term eurlfincludes such things as an interaction between doctor and patient, since no patient was examined more than once by the same doctor. Different E ~ , , ~ , ’ Swere assumed to be independent, with expectation equal to zero and variance u2. The design is a BIBD in several ways: For each fixed day we have a BIBD with patients as 10 blocks of size 2 and doctors as 5 treatments. We have a BIBD with patients on different days as 20 blocks of size 2 and doctors as 5 treatments. We have a BIBD with patients as 10 blocks of size 4 and doctors as 5 treatments. Further, the design is complete in patients. days. and time of day. Using these facts and standard linear methods in the analysis of factorial experiments and BIBDs (3, we were able to break down the variation into the following s ? = R , - effect of time of day S: = R 2 - day-today variation of time-of-day effect s:=&, R 3 -. residual variation S L;-? -l R 1 - day-today variation within doctors s 25-Tc - 1 R 5 - day-today variation within patients s g = t R e - variation among doctors s;=; R,- variation among patients
The most important feature of this breakdown of the variation is that any one of the variations so computed is not distorted by any other variation. For example, the day-today variation among the doctors is eliminated when computing the variation among doctors, etc. These were calculated by the following formulas R 1-40 -L (Zj(Xij1 1 +xij21-xu 1 2 - x i j 2 2 ) ) 2 R ? = & Gj(xijl1- x i j z ! -Xij1z+xij22))* R , = S S - $ ( S S p l + S S p 2 + 5 SSL!+5 SSL2)-R1-R2
Clinical assessment of disease activity in R A
103
Table I . Vm-icinces, means, overall standard deviations, and overcill coefficients of variance of cliriicul joint examinations in 10 patients with rheumatoid arthritis examined by 5 doctors Number of joints Pain at rest or on Pain on movement pressure
Pain at rest or on movement or Joint on pressure swelling
12.2 8.1 0.4 29.3 10.7 64.8 58.2 4 3.52
3.99 2.03 0.03 2.95 4.56 5.98 74.68
8.35 1.23 0.63 8.28 11.13 54.53 73.56
11
11
4.63 2.03 5.63 4.20 3.10 14.38 54.19 14
88
17
Scand J Rheumatol Downloaded from informahealthcare.com by University of Auckland on 12/06/14 For personal use only.
Pain at rest Variance Residual Sf Morninglafternoon s: Day/day S: Daylday (doctor) s: Day/day (patient) s: Doctors s: Patients S: Mean Overall standard deviation Overall coefficient of variance (%)
Li=&.j
1.87
2.86 26
2.52 0.90 0.90 8.73 3.61 89.88 17.79 7 1.89
1.%
Pain on pressure 0-3
Joint swelling 0-3
27.8 8.1 14.4 146.9 24.8 633.2 259.8 18 6.84
10.5 13.2 0.6 19.5 16.5 187.8 79.5 12 3.72
27
14
38
31
RESULTS
t X i j d t , Ki=ZjsUi)Zd, r Xu,,/
k(i)=set of patients treated by doctor i R,=FSSL
RT=+(ssp-hGi,j,d , t X i W / ) ' ) the residual variation, thus had an expectation equal to cr', the variance of the measurement error. All the other s: values had expectation equal u2plus a positive quantity depending on the force of the corresponding effect. The coefficients of variance were estimated as
s;,
ri=4Osi/(4 , j ,d , X U I ) .
The variances, the means, the overall standard deviations, and the overall coefficients of variance are given in Tables 1 and 2. The coefficients of variance make it easier to compare the different variables. The variations from morning to afternoon and from day to day were of the same magnitude as the residual variance and were negligible in comparison with the inter-individual variation among doctors and patients. For this reason a better estimate of the variance of the measurement error was determined from a weighted average of the sf for i= 1, ...,5 .
Table 11. Variances, means, overall standard deviations, and overall coefficients of variance of clinical variables in 10 patients with rheumatoid arthritis examined by 5 doctors
Variance sf Residual s: Morning/afternoon S% Daylday S2 Day/day (doctor) s: Day/day (patient) s% Doctors s: Patients Mean Overall standard deviation Overall coefficient of variance (%)
Morning stiffness
Grip strength
Wellbeing
Fingertip palm distance
Maximum flexionextension
129 181 60 1 109 757 333 30 167 86 min 20 23
0.0076
33 63 240 121 20 126 2 712 41 mm 7 17
36 84 176 29 94 39 610 f l mm 8 73
440 640 23 598 1 208 192 71 744
0.0055
0.0141 0.0052 0.0083 0.0188 0.1367 0.21 kg/cmz 0.0088 42
590" 27 5 Scand J Rtieumarology 8
Scand J Rheumatol Downloaded from informahealthcare.com by University of Auckland on 12/06/14 For personal use only.
104
T . M m k Hcrnseri et trl.
The overall coefficients of variance of meaaurement error were 88% for pain at rest and 73% for fingertippalm distance, while for all the other variables the coefficients ranged from 5 R to 42 R. The variations among doctors were of the same magnitude or greater than the variations among patients for the following parameters: pain at rest, pain on pressure. joint swelling without grading. and pain and joint swelling with grading. The following parameters showed greater variation among patients than among doctors: A registration of pain at rest or on movement, morning stiffness, grip strength, well-being on a visual analogue scale. fingertippalm distance. and maximum flexionextension of joints.
DISCUSSION
I n the present investigation a selection has taken place of both patients and doctors. The patients all had classical rheumatoid arthritis according to the criteria of the American Rheumatism Association, the disease had lasted for at least I year. and t h e patients had all been referred to a rheumatological department for treatment because of active disease. The doctors were senior doctors with experience and interest in rheumatology. This may explain the low intra-patient and intra-observer variance for all the measurements [daylday (doctor), day/day (patient) tables l and 31. Most of the investigated parameters had a regular pattern of variation indicating that the analyses performed were adequate. In the case of morning stiffness there was almost no variation at all between the different observations on the same patient. This is possibly due to the registration of the average value for the last 7 days and not an exact single-day value. Also. a tendency to approximate the values to half or whole hours may have affected the measured variation. The statistical results for this parameter should therefore be accepted with the greatest precaution. In the parameters fingertippalm distance and pain at rest there appeared to be so many zeros among the observations that it is questionable whether the analysis describes the data adequately. Also. the overall coefficients of variance of measurement error were relatively high for these variables. In such cases, a more non-parametric statistical approach would be suitable. The drawback
of this kind of approach is that it would not enable us to distinguish the sources of variation. For routine clinical use it is imperative, though not necessarily sufficient for the validity of a measured variable, that the variation among the examiners is smaller than the variation among the patients. This requirement was only fulfilled for pain at rest or on movement, morning stiffness, grip strength, well-being. fingertippalm distance, and maximum flexion-extension. For this reason, it is not recommendable for a routine use, to include a registration of joint pain at rest only, joint tenderness, orjoint swelling in an articular score, because of the high inter-observer variation. The high interobserver variation of the measurement of joint tenderness may be due to variation in the location and the force of the applied joint pressure. The registration of joint pain at rest relies on the patients' information alone and it may be influenced by the individual doctors' method of questioning and their interpretation of the patients' answers. This does not rule out, however, that the more elaborated articular scores may be of some value in clinical studies, providing the examinations are done by the same observer throughout the study ( 1 , 6). This study supports the recommendations of Lansbury (4) to include registration of grip strength and joint pain on passive movement in any activity index. Furthermore, it may be of value to use the visual analogue scale to register the patients' "wellbeing" in the same way as it has been used to register pain (7). Although an elaborate registration of joint symptoms in patients with rheumatoid arthritis is desirable, the effort is not justified in daily routine because of the individualism of doctors.
ACKNOWLEDGEMENT Thiw work was supported by the Danish Medical Research Council (grant no. 552495).
REFERENCES I . Camp, A. V.: An articular index for the assessment of rheumatoid arthritis. Orthopaedics (Oxford) 4: 39, 1971. 2. Committee of the American Rheumatism Association. Arthritis Rheum2: 16, 1959. 3. Kempthorne, 0.:The design and analysis of experiments. Wiley, New York 1952.
Clinical assessment of disease activity in R A
assessment ofjoint tenderness in patients with rheumatoid arthritis. Quart J Med37: 393, 1968. 7. Scott. J. & Huskisson, E. C.: Graphic representation ofpain. Pain2: 175, 1976. Submitted for publication May 10, 1978
T. Mark Hansen Medical Department TA Rigshospitalet DK-2200 Copenhagen, Denmark
Scand J Rheumatol Downloaded from informahealthcare.com by University of Auckland on 12/06/14 For personal use only.
4. Lansbury, J.: Methods for evaluating rheumatoid arthritis. In Arthritis and Allied Conditions (eds. J. L. Hollander & D. J. McCarty), 8th ed., pp. 269. Lea & Febiger, Philadelphia 1972. 5. Lee, P., Sturrock, R. D., Kennedy, A. & Dick, W. C.: The evaluation of antirheumatic drugs. Curr Med Res Opin 1: 427, 1973. 6. Ritchie, D. M., Boyle, J. A., McInnes, J. M., Jasani, M. K., Dalakos, T. G., Grieveson, P. & Buchanan, W. W.: Clinical studies with an articular index for the
105
Scand J Rheumatology 8
