Plurirneter-V Inclinometer and Assessment of its Intrarater and Interrater Reliability

Transcription

1 A Standardized Protocol for Measurement of Range of Movement of the Shoulder Using the Plurirneter-V Inclinometer and Assessment of its Intrarater and Interrater Reliability Sally Green, Rachelle Buchbinder, Andrew Forbes, and Nicholas Bellamy Objective. To develop a standardized protocol for measurement of shoulder movements using a gravity inclinometer designed for use in clinical trials, and to assess its intm- and intermter reliability in a group of manipulative physiothempists. Methods. After instruction, 6 manipulative physiotherapists independently assessed 8 movements of the shoulder, including total and glenohumeral flexion (TE GHF), total and glenohumeral abduction (TA, GHA), external rotation in neutml (ERN) and abduction (ERA), internal rotation in abduction (IRA), and hand behind back (HBB), in random orderin 6patients with shoulder pain and stiffness according to a 6 X 6 Latin square design using the standardized protocol. The assessments were then repeated. Analysis of vari- Presented in poster format at the 60th national scientific meeting of the American College of Rheumatology, Orlando, FL, October 18-22, Sally Green, GradDip, BAppSci, MAPA, MMPAA, Manipulative Physiotherapist, PhD Scholar, and Andrew Forbes, PhD, MSc, BSc, Senior Lecturer in Biostatistics, Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, Australia; Rachelle Buchbinder, MSc, MB, BS(Hons), FRACP, Cabrini Medical Centre, and Senior Lecturer in Clinical Epidemiology, Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, Australia: and Nicholas Bellamy, MD, MSc, FRCP, FRCPC, FACP, London Health Sciences Centre-Victoria Campus, and Professor of Medicine and Epidemiology, University of Western Ontario, London, Ontario, Canada. Address correspondence to Rachelle Buchbinder, MSc, MB, BS(Hons), FRACP, Department of Epidemiology and Preventive Medicine, Monash University, Prahran, Vic, 3181, Australia. Submitted for publication March 18, 1997; accepted in revised form July 11, by the American College of Rheumatology /98/$5.00 ance was used to partition total variability into components of variance in order to calculate intmclass correlation coefficients (ICCs). Results. The intm- and interrater reliability of the diffemn t movements varied widely Reliability was higher for TF and TA than for the corresponding glenohumeral movements (e.g., intmrater ICCs: TF = 0.80, GHF = 0.65, TA = 0.75, GHA = 0.62). Intermterreliability was higher in the second round suggesting a practice effect (e.g., round 1,2 intermtericcs TF = 0.62, 0.82; TA = 0.62, 0.88; ERN = 0.85, 0.95). Conclusion. The measurement of the active range of Tl? TA, ERN, and HBB, measured by manipulative physiotherapists following the standardized protocol, has intra- and intermter reliability acceptable for use as an outcome measure in clinical trials assessing interventions for shoulder pain. Key words. Shoulder; Reliability; Range of movement; Outcome measurement. INTRODU(XI0N There is currently no uniformity in the way clinical assessments are performed to evaluate treatment efficacy in clinical trials of shoulder disorder. However, most studies include some measure of range of movement of the shoulder. In a recent methodologic review of outcome measures used in clinical trials of shoulder disorders, we found that 90% of studies included at least one shoulder movement (1). Of these, only 38% included a description of their technique of measure- 43

2 44 Green et a1 Vol. 11, No. 1, February 1998 Figure 1. Total shoulder flexion. Figure 2. Glenohumeral flexion. ment, only 14% specified how end of range was determined, and the majority of studies did not address the reliability of the method used [I). Previous studies that have assessed the reliability of measurement of shoulder movement often have not outlined a standardized protocol of measurement, have predominantly measured passive range, and have often been performed on normal shoulders. The results of these studies may not be applicable to symptomatic patients with varying degrees of painful, restricted movement. The aim of this study was to develop a standardized protocol for measurement of active shoulder movement in the context of a clinical trial, with particular reference to minimizing potential sources of observer variation. We utilized the gravity-referenced inclinometer (Plurimeter-V inclinometer; Dr Rippstein, Zurich, Switzerland) to measure total and glenohumeral flexion, total and glenohumeral abduction, internal rotation, and external rotation in both neutral and abduction. Hand behind back, assessed by palpation to the nearest spinal process, was also included. The inter- and intrarater reliability of the standardized protocol was assessed by 6 manipulative physiotherapists. tension and leading with the thumb to ensure consistent rotation. The Plurimeter was placed along the shaft of the humerus perpendicular to the plane of motion (Figure 1). Glenohumeml flexion. The scapula was stabilized by application of downward pressure on the acromion (Figure 2). The arm and the Plurimeter were placed as for total shoulder flexion. Total shoulder abduction. This movement included scapular motion. It was also measured with full elbow extension and leading with the thumb to ensure consistent rotation. The Plurimeter was placed along the shaft of the humerus perpendicular to the plane of motion (Figure 3). Glenoh umeml abduction. The scapula was stabilized by application of downward pressure on the acromion, and the arm and the Plurimeter were placed as for total shoulder abduction (Figure 4). The following movements were measured with the patient lying in the supine position. PATIENTS AND METHODS Development of standardized protocol. A standardized protocol for the measurement of shoulder movement with the Plurimeter-V gravity inclinometer was developed by two of the investigators (SG, RB). The individual movements are illustrated in Figures 1-8. While sitting and ensuring no trunk movement, the following movements were performed. Total shoulder flexion. This movement includes scapular motion. It was measured with full elbow ex- Figure 3. Total shoulder abduction.

3 Arthritis Care and Research Measurement of Shoulder Movement 45 Figure 4. Glenohumeral flexion. External rotation in neutml abduction. The elbow was fixed at 90" flexion, and the forearm was fully pronated. The elbow remained at the patient's side to ensure neutral shoulder abduction. The Plurimeter was placed along the shaft of the radius, perpendicular to the plane of movement (Figure 5). External rotation in abduction. The shoulder was abducted to 90". The elbow was again fixed at 90" flexion with the forearm fully pronated, and the Plurimeter was placed along the shaft of the radius, perpendicular to the plane of movement (Figure 6). Internal rotation in abduction. The shoulder was abducted to about 45" or sufficient abduction to clear the hip. The elbow was fixed at 90" flexion, and the forearm was fully pronated. The Plurimeter was placed along the shaft of the ulna, perpendicular to the plane of movement (Figure 7). An additional measure of internal rotation, hand behind back, was included in the protocol. Despite its use in clinical trials, to our knowledge its reliability has not been assessed previously. It was measured Figure 6. External rotation in abduction. while standing, and the highest anatomical landmark reached by the thumb was recorded (anterior superior iliac spine, posterior superior iliac spine, or spinous process) (Figure 8). All movements were performed actively, and end of range was defined as the point at which the patient fwst experiences pain (Pl) (2). As the angle of incline measured by the inclinometer at the shoulder may be Figure 5. External rotation in neutral. Figure 7. Internal rotation in abduction.

4 46 Green et a1 Vol. 11, No. 1, February 1998 Figure 8. Hand behind back. affected by movement at the elbow or wrist, care was taken to ensure a constant position of all joints. For measurements made with the Plurimeter V, a single reading to the nearest 2" was recorded for each movement. Reliability study. The study took place over a single morning. Six manipulative physiotherapists participated in the study. All had postgraduate training and experience in the measurement of joint range, but none had previously used the Plurimeter V inclinometer. The 6 Plurimeter V inclinometers used in the study were identical (supplied by Electromed Sales and Rentals, Melbourne), and had been calibrated by the supplier. A 1-hour training and standardization session was conducted initially. A methods manual was distributed to each of the participants. This contained explicit instructions including illustrations for performance of the standardized protocol. One of the investigators (SG) described the exact technique and demonstrated each of the measurements, with particular reference to positioning of the patient, stabilization of the body, and positioning of the instrument. While the starting position of the Plurimeter is fixed by gravity, depending on the instrument placement, the starting position may point to 0" or 180". Participants were therefore instructed to ensure that the starting position of the Plurimeter pointed to 0" after placement of the instrument on the patient but prior to each shoulder movement. Participants were instructed to demonstrate each maneuver to the patient and ask the patient to move his or her shoulder until the first sensation of pain (Pl). All participants were experienced in eliciting P1. For the remainder of the session, each of the participants practiced performance of the standardized protocol under supervision of two of the investigators (SG, RB). Six patients, all of whom had varying degrees of pain and stiffness of one or both shoulders (only the most painful shoulder was examined), were recruited from the private physiotherapy practice of one of the investigators (SG). Patients were chosen to represent a wide spectrum of disease and were excluded if they had an irritable condition of the shoulder, defined as shoulder pain increasing with repeated movement and being slow to settle once aggravated. All patients gave informed consent. Baseline data included age, sex, duration of symptoms, and history of trauma or shoulder surgery. Patients were asked to rate the severity of their symptoms on a 5-point categorical scale (1 = not at all severe, 2 = of mild severity, 3 = of moderate severity, 4 = of marked severity, 5 = of maximum severity). The patients were taken through a series of warm-up exercises. For each movement that they would later make, they were asked to inform the examiner when they first felt any sensation of pain and were instructed to disregard any feelings of stretching or tightness. Each manipulative physiotherapist examined each of the 6 patients, independently, in turn, according to a 6 X 6 Latin square design (3). Both the order of the patient examination and the sequencing of the movements assessed were randomized, the latter according to a computer-generated table of random numbers. Assessments were performed in separate examination rooms, and assessors were unable to observe each others' technique. Each movement was measured only once, and there was no time limit placed on the examinations. Patients and assessors were given a 5-min- Ute rest between each examination and a longer (20 minutes) rest between the third and fourth examinations. Assessors were not permitted to discuss the study during these breaks. After each examination, the data were checked for completeness. Following completion of the first set of assessments (round I), both patients and assessors had a 1-hour break. Participants were again not permitted to discuss the study. Following the break, each patient was again assessed by each of the manipulative physiotherapists using the same 6 X 6 Latin square design, but the sequencing of tests was rerandomized (round 2). The rests between examinations and data checking were identical to round 1. Statistical analysis was conducted using SAS Version 6.10 (4). Descriptive statistics calculated included the mean and standard deviation for each movement for each of the 6 patients. These were calculated separately for rounds 1 and 2. A comparison of the patients' means gives an indication of the spectrum of disease, and comparison of the means between rounds 1 and 2 gives an indication of any major changes that may have occurred between rounds. Comparison of the

5 Arthritis Care and Research Measurement of Shoulder Movement 47 SD gives an indication of the spread of readings between observers, and comparison of the SD between rounds 1 and 2 gives an indication of whether the spread of readings between observers was different or similar for rounds 1 and 2. The replicated Latin square experimental design employed was able to simultaneously estimate the interobserver and intraobserver reliabilities for each of the movements. Analysis of variance (ANOVA) was used to partition the total variability of ratings into components or sources of variation: variability due to patients, observers, round, all &way interactions of these individual components, and the residual unexplained variability. Although the order (or sequence) of assessment was deliberately balanced in the Latin square design, it was not explicitly included in the variance components, as it was subsumed within the patient by observer interaction term. Variance components for each of the sources were estimated by equating observed and expected mean squares, with all sources considered as random effects (using the SAS VARCOMP procedure). In particular, round was considered as a random effect, as the times of the assessments were chosen from an effectively infinite number of possible assessment times. When variance components were estimated as being less than zero, they were reassigned the value of zero. To better understand the meaning of the variance components, consider the circumstance of many patients being assessed by many observers on many occasions. For each patient there would be an average assessment score, which averages the assessment scores of all observers over all occasions. The variance component due to patients would describe how variable the average assessment scores were between patients by giving an indication of how much patients scores tend to differ, on average. A large variance component due to patients would imply that, on average, patients average scores differed greatly, and a small variance component due to patients would imply that the patients average scores were similar. Similarly, each observer would possess an average score, averaged across all patients and all occasions. The variance component due to observers would then describe the variability between the observer averages. A small variance component due to observers would imply that the observers agree, on average. However, this does not necessarily imply that the observers agree on any individual patient. The meaning of the 2-way interaction variance terms is illustrated by consideration of the patient by round variance component. This considers the change in assessment scores for an individual patient from round 1 to round 2 and describes the variability of these changes across all patients. A large variance component due to the patient by round interaction may imply that there is a large variation in scores of one or more patients between rounds. For example, some patients scores may have increased markedly between rounds 1 and 2, others may not have changed, and some may have decreased. Results are presented as the total variability and the percentage of the total variability attributable to each variance component. Interobserver and intraobserver reliability was calculated from the variance components via appropriate intraclass correlation coefficients (ICCs) according to the following formulae: Interrater reliability = [var(patient) + var(round) + var(patient by round)] + [var(total)] Intrarater reliability = [var(pa tient) + var(observer) + var(patient by observer)] + [var(total)] where var(total) is the sum of all the variance components. The closer the interobserver reliability is to 1, the stronger is the agreement between observers on an individual patient for a given round. The closer the intraobserver reliability is to 1, the stronger the agreement between assessment scores for round 1 and round 2 for a particular observer assessing a particular patient. In addition, interobserver reliabilities were calculated for each round separately using simpler ANOVA models. These models partitioned the variability into components due to patients, observers, order of assessment, and residual error. The interobserver reliability was then estimated as: Interrater reliability = RESULTS Var(patien t) Var(total) Four men and two women participated in the study. The mean age was 57.3 years (range years). The right shoulder was assessed in 5 patients, and the left shoulder was assessed in one patient. The mean duration of symptoms was 16.5 months (range 9-24 months). Three patients had a history of shoulder trauma, and 4 patients had previous shoulder surgery. The mean seventy of symptoms was 2.2 (range 14). Table 1 displays the means and SDs for each movement for each patient in rounds 1 and 2 when the measurements of all 6 observers are combined. There

6 ~~~~ ~ 48 Green et a1 Vol. 11, No. 1, February 1998 Table 1. Summary statistics (mean and standard deviation) for each movement measured in 6 patients in rounds 1 and 2 Patient 1 Patient 2 Patient 3 Patient 4 Patient 5 Patient 6 Round Mean* SD* Mean SD Mean SD Mean SD Mean SD Mean SD TSFt GHF TSA GHA ERN ERA I R A HBBS 1 T8/9 T7-11 T11112 Tll-L2 T11112 Tll-L2 T11/12 T10-L1 T6 T5-8 T10/11 T T10 T8-11 Tll/l2 TIO-LI Tll/l2 TIO-LI T12/L1 T10-L1 T5/6 T5-9 T12 T9-L3 Mean and SD are based on assessment by the 6 observers. t TSF = total shoulder flexion; GHF = glenohumeral flexion; TSA = total shoulder abduction; GHA = glenohumeral abduction; ERN = external rotation in neutral; ERA = external rotation in abduction; IRA = internal rotation in abduction; HBB = hand behind back. * Median and range. was a wide variation in the range of movement between patients. For example, the mean total shoulder flexion of patients in round 1 varied between 63.7" and 149.7O (Table 1). In general, the mean measurements were similar for rounds 1 and 2. However, patient 2 demonstrated a marked improvement in range of movement between rounds 1 and 2, particularly for flexion, abduction, and external rotation. Table 2 displays the results of the ANOVA. As expected, a large proportion of the total variance was explained by the variance between patients. However, a large proportion of the variance was also attributable to the interaction between the patient and round, indicating that some patients' scores differed between rounds while others did not. On inspection of the data, there appeared to be a marked improvement in patient 2 for some movements between rounds 1 and 2 (see Table 1). Repeat analysis, excluding patient 2, demonstrated little variance attributable to the patient/ round interaction. After exclusion of patient 2, most of the variance was attributable to between patient differences and residual error (Table 2). There was minimal variance attributable to the observers, and none of the variance was attributable to round. The possibility of a systematic bias between observers was examined by calculating mean scores of observers across subjects. However, this analysis demonstrated no such bias. The interobserver and intraobserver ICCs calculated for the 8 shoulder movements, for all 6 patients, ex- cluding patient 2, are shown in Table 3. The intraobserver agreement was much greater after exclusion of patient 2 from the analysis, and interrater agreement was marginally improved. Both intraobserver and interobserver agreement was greater for total shoulder flexion and abduction than the corresponding glenohumeral movements (Table 3). Agreement was greater for external rotation measured in neutral than external rotation measured in abduction. Interobserver agreement for internal rotation in abduction was modest (ICC = 0.47) but was better when measured by hand behind back (ICC = 0.75). Intraobserver agreement for hand behind back measurement was high (ICC = 0.90). DISCUSSION We have developed a standardized protocol for the measurement of active shoulder movements using the Plurimeter-V inclinometer in patients with shoulder pain. All patients were put through a series of warmup exercises prior to the study to minimize the effect of repeated examination, which we thought may have resulted in either an aggravation of pain and decrease in range of movement or stretching of the joint and improvement in range of movement. We elected to take a single reading for each movement for the same reason. During the standardization session, particular atten-

7 Arthritis Care and Research Measurement of Shoulder Movement 49 Table 2. Percentage of total variance attributable to patient, observer, round, and interactions between patient and round, round and observer, patient and observer, and residual error, and total variance for all patients and for all patients excluding patient 2 Round Patient Observer Patient and Patient and Residual Total Round and round observer observer error variance TSF* 36t * GHF TSA GHA ERN ERA IRA HBB * TSF = total shoulder flexion; GHF = glenohumeral flexion; TSA = total shoulder abduction; GHA = glenohumeral abduction; ERN = external rotation in neutral; ERA = external rotation in abduction: IRA = internal rotation in abduction; HBB = hand behind back. t All patients. * Excluding patient 2. Table 3. Intraobserver and interobserver intraclass correlation coefficients (ICCs) calculated for observations on 6 patients and all patients excluding patient 2 TSF* GHF TSA GHA ERN ERA IRA HBB Interobserver Intra- ICC overall Interobserver Interobserver observer (rounds 1 ICC ICC ICC and 2) (round 1) (round 2) 0.49t 0.80* * TSF = total shoulder flexion; GHF = glenohumeral flexion; TSA = total shoulder abduction; GHA = glenohumeral abduction; ERN = external rotation in neutral; ERA = external rotation in abduction; IRA = internal rotation in abduction; HBB = hand behind back. t All patients. * Excluding patient 2. tion was given to the position of the patient, positioning of the instrument on the patient, stabilization of the joint to be measured, and determination of the endpoint of the movement. These factors have been previously identified as major potential sources of variation (5). While all 6 manipulative physiotherapists involved in the study were experienced in measurement of range of movement, none had previous experience with the Plurimeter-V inclinometer. Although standardization took place prior to the study, a higher interrater reliability was found in round 2 when compared to round 1, suggesting the presence of a practice or learning effect. This suggests that reliability of our standardized protocol may be improved further by practice. As reduction in the variance of measurement has the effect of reducing sample size requirements in clinical trials, this is an important finding. Measurement of shoulder motion can be performed in many ways. The 3 most commonly used methods include visual estimation and use of either the standard 2-armed protractor-like goniometer (universal goniometer) or the gravity-referenced inclinometer. Table 4 summarizes the reliability studies of shoulder movement measurement that have utilized these methods. Croft et al found that visual estimation of shoulder elevation was reliable when measured by a group of trained primary care physicians who recorded the angle of passive elevation onto a diagram which was then

8 50 Green et a1 Vol. 11, No. 1, February 1998 Table 4. Previous studies of reliability of shoulder movement utilizing visual estimation, universal goniometer, or hydrogoniometer Andrews and Croft et a1 Riddle et a1 Pandya et a1 Bovens et a1 Clarke et a1 Bohannon Bower 1994 (6) 1987 (7) 1985 (8) 1990 (9) 1974 (10) 1989 (21) 1982 (12) Active or passive Passive Passive Passive Active Passive Passive Passive movement Population Shoulder pain Shoulder pain Muscular Normal 10 normal 25 stroke 4 normal & 1 (n = 6) (n = 50, dystrophy women patients painful Observers 6 primary care 2 groups) Random pairs 5 physical 3 physicians 2 orthopedic 2 physical shoulder - physicians of 16 physi- therapists surgeons therapists Instrument* m cal therapists UG UG UG HG HG HG Movements assessed Interrater reliability TA, ERN TF,TA,ERA, TA ERN TA,ERN,TR ERA GHA, ERN, IRA ERA, IRA Intrarater reliability - TF, TA,ERA, TA ERN TA,ERN,TR ERA GHA, ERN, IRA ERA, IRA VE = visual estimation; UG = universal goniometer; HG = hydrogoniometer; TA = total shoulder abduction; ERN = external rotation in neutral; TF = total shoulder flexion; ERA = external rotation in abduction; IRA = internal rotation in abduction; TR = total rotation; GHA = glenohumeral abduction. measured by an independent observer using a protractor (6). However, visual estimation of external rotation by the same observers was not reliable, even when the 6 observers simultaneously witnessed the movement of a single subject. Croft et a1 also compared the interobserver agreement for passive elevation to the point of restriction versus passive elevation to the point of first pain and found agreement better when passive elevation was measured to the point of restriction (6). However, range of active movement where endpoint is defined as the onset of first pain may be a more acceptable measure to the patient, particularly in the context of clinical trials where repeated measures may need to be performed. The universal goniometer requires placement of the center of the instrument and alignment of its two arms. Its limitations include its inability to measure rotary movements and its requirement to be held with two hands. The latter results in an inability to use one hand to stabilize the proximal component of the joint and so isolate single joint movements. Riddle et al examined the intrarater and interrater reliability of random pairs of 16 physical therapists measuring passive shoulder movements in 100 patients with shoulder pain with the universal goniometer (7). While the intrarater reliability for all movements was high (ICCs varied between 0.87 to 0.99) (7), lower interrater reliability could probably be accounted for by the lack of standardization of the placement of the goniometer and positioning of the patient. As part of a study assessing the reliability of goniometric measurements in children with Duchenne muscular dystrophy, Pandya et al (8) also found that the interrater reliability of 5 experienced physical therapists measuring passive total abduction using the universal goniometer was not as good as the intrarater reliability (ICC = 0.67 and 0.84, respectively) (8). In the only previous study that has assessed the reliability of active shoulder movements, Bovens et al(9) assessed the reliability of measurement by 3 experienced physicians of external rotation in neutral in normal shoulders using the universal goniometer (9). Intrarater and interrater reliability was 0.76 and 0.63, respectively (9). However, when measurements were taken in different groups of athletes over time, the interrater reliability using the same 3 observers varied greatly ( ). The gravity-referenced inclinometer is calibrated on the basis of gravity, and therefore the starting position of the movement to be measured is ked, minimizing placement error. It can be applied in sitting, and can be used with one hand, leaving the other hand free to stabilize the joint. The reliability of passive shoulder movements using the inclinometer has been assessed in 3 previous studies (10-12). Clarke et al reported the interrater reliability of the inclinometer for the measurement of passive total abduction, external rotation in neutral, and total rotation in 10 normal women examined by 2 observers, and both the inter- and intrarater reliability of the same measurements measured in 2 normal women by two observers twice a day for 4 days (10). They reported a measurement error of less that 7%. The results of this study have subsequently been cited in clinical trials using passive range of movement as an outcome measure as evidence of its

9 Arthritis Care and Research Measurement of Shoulder Movement 51 reliability (13,14). While we used similar methods for measuring total abduction and external rotation in neutral, the description of their measurement protocol failed to mention whether the observers were blinded to each other s examination, the time lapse between the examinations, and whether the goniometer, which was fitted with Velcro fasteners, remained on the patient or was removed between examinations. Furthermore, the definition of the end of range in patients with normal shoulders may not be generalizable to patients with painful shoulders. Andrews and Bohannon demonstrated high intraand interrater reliability of passive external rotation of the shoulder in 25 hemiplegic patients using the inclinometer (ICC = 0.94, 0.96, respectively) (11). External rotation was measured in 45 of abduction, and end of range was defined as onset of pain reported by the patient. The high reliability may be explained by the measurement of passive motion and standardized protocol for measurement. Bower reported high intra- and interrater reliability of the hydrogoniometer in measuring passive glenohumeral abduction, external rotation in neutral abduction, external rotation in abduction, and internal rotation in abduction to the end of range in one symptomatic and 4 asymptomatic persons (12). The observers were not blinded to each other s technique, and the goniometer remained strapped to the subject s arm, therefore eliminating variation due to placement of the instrument. For the purpose of describing reliability, ICCs of >0.75 are considered excellent, are considered fair to good, and are considered poor (15). According to these criteria, we determined that intrarater reliability (when patient 2 was excluded) was excellent for total shoulder flexion and abduction and all rotation measurements. Interrater reliability was excellent for total shoulder abduction and external rotation in neutral and was fair to good for the other measurements. The reliability of hand behind back, which is routinely measured as a determinant of disability of the shoulder (16), has not previously been assessed to our knowledge. We found that it had both excellent intrarater and interrater reliability, Although we only examined 6 patients, there was a wide spectrum of disease, as determined by both the range of movement, and severity and disability indices (data not shown). This suggests that our results may be generalizable, although this requires replication of our study by others. Other areas of rheumatology are moving toward recommendations for uniform outcome measurement in therapeutic clinical trials (17). For example, core sets of outcome measures have been recommended for use in clinical trials of rheumatoid arthritis (18), and similar goals are being set for other diseases such as osteoarthritis and osteoporosis (18). This will improve our ability to interpret study results and compare results across studies. Outcome measures are being chosen on the basis of evidence that they are valid, reliable, and responsive to change (17). In addition, an instrument s reliability influences other trial considerations. For example, high intrarater reliability with a corresponding low variance allows smaller sample sizes, and high interrater reliability may mean that it is possible to use more than one outcome assessor. For clinical trials of shoulder disorders, one or more active movements of the shoulder would probably fulfill the requirements of a suitable outcome measure. We have demonstrated that total shoulder flexion and abduction, external rotation in neutral, and hand behind back may be the most reliable active movements of the shoulder. Their inclusion in an individual trial or a core set of measures will also depend on demonstration of validity and responsiveness to change for the shoulder disorders under study. We are indebted to the following manipulative physiotherapists who participated in the study: Debra Benger, Caroline Bills, Craig Bosworth, Julie Nitschke, Lyn Watson, and Andrew Wilson. We also thank Electromed Sales and Rentals, Melbourne, Australia, who supplied the Plurimeter-V inclinometers for the study. REFERENCES 1. Buchbinder R, Green S, Glazier R, Forbes A. A systematic review of interventions for shoulder pain. BMJ. In press Maitland G. Vertebral manipulation. 4th ed. London: Buttenvorths; Cochran WG, Cox GM. Experimental designs. 2nd ed. New York: John Wiley & Sons; SAS Institute. SAS Cary (NC): SAS Institute; Lea R, Gerhardt JJ. Current concepts review: range of motion measurements. J Bone Joint Surg Am 1995;77- A: Croft P, Pope D, Boswell R, Rigby A, Silman A. Observer variability in measuring elevation and external rotation of the shoulder. Br J Rheumatol 1994;33: Riddle DL, Rothstein JM, Lamb RL. Goniometric reliability in a clinical setting: shoulder measurements. Phys Ther 1987;67: Pandya S, Florence J, King W, Robison J, Oxman M, Province M. Reliability of goniometric measurements in patients with Duchenne muscular dystrophy. Phys Ther 1985;65: Bovens AM, van Baak MA, Vrencken JC, Wijnen JA,

10 52 Green et a1 Vol. 11, No. 1, February 1998 Verstappen FT. Variability and reliability of joint measurements. Am J Sports Med 1990;18: Clarke GR, Willis LA, Fish WW, Nichols PJR. Assessment of movement at the glenohumeral joint. Orthopaedics 1974;7: Andrews A, Bohannon R. Decreased shoulder range of motion on paretic side after stroke. Phys Ther 1989;69: Bower K. The hydrogoniometer and assessment of gleno-humeral joint motion. Aust J Physiother 1992;28: Adebajo AO, Nash P, Hazleman BL. A prospective double blind dummy placebo controlled study comparing triamcinolone hexacetonide injection with oral diclofenac 50 mg TDS in patients with rotator cuff tendonitis. J Rheumatol 1990;17: Binder A, Parr G, Hazleman B, Fitton-Jackson S. Pulsed electromagnetic field therapy of persistent rotator cuff tendonitis. Lancet 1984; Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33: American Orthopaedic Association. Guides to the evaluation of permanent impairment. Chicago: American Orthopaedic Association; Felson DT, Anderson JJ, Boers M, Bombardier C, Chernoff M, Fried B, et al. The American College of Rheumatology preliminary core set of disease activity measures for rheumatoid arthritis clinical trials. Arthritis Rheum 1993;36: Tugwell P, Boers M, for the OMERACT Committee. Developing consensus on preliminary core efficacy endpoints for rheumatoid arthritis clinical trials. J Rheumatol 1993;20:555-6.