Range of shoulder motion in patients with adhesive capsulitis; Intra-tester reproducibility is acceptable for group comparisons

During the last few decades, increased focus has been placed on evidence-based medicine in the orthopedic disciplines. This development has made it essential to utilize instruments and methods that can detect clinically relevant changes. In shoulder research, measurements of range of motion (ROM) have been used as outcome measures in the vast majority of reported trials [1].

Shoulder ROM measurements can be divided into methods where passive or active movements are measured. In short, shoulder passive ROM is a measure of how far the observer can move the arm of the patient, while shoulder active ROM is a measure of how far the patient can move the arm himself.

The choice of an evaluative method is based on many considerations. What is crucial is high reproducibility, indicating little variation between measurements of the same quantity for the same individual. Measurement errors of shoulder ROM have been reported in several studies. Acceptable reproducibility of passive shoulder ROM measurements was reported in a study by Clarke et al. in 1974 [2], investigating healthy volunteers with a gravity-dependent goniometer. More recently, other researchers [3, 4] have reported larger measurement errors when actual shoulder patients have been included. In one study [3], inter-observer differences between single observations of less than 20–25 degrees could not be distinguished from measurement error even with a well-adapted method. Such findings indicate that measurement errors for passive ROM are so large that important "true" changes may be misinterpreted. It has been argued that passive ROM is more difficult to measure reliably than active ROM [5]. Regarding shoulder measurements, however, there is limited evidence that measurement errors for active ROM are any smaller than for passive ROM.

Estimates of reproducibility of active ROM vary considerably. Some researchers investigating measurement errors for active ROM in shoulder patients have reported absolute measurement errors that are comparable to [6], or somewhat larger than [7, 8], the results mentioned above [3, 4] for PROM. Another group reported much larger measurement errors than this [9]. The conflicting results indicate that differences in procedures, study design or investigated population may have a large impact on reproducibility estimates. Gajdosik and Bohannon stated in a review article in 1987: "...results suggest a tremendous need for studies of the reliability of measuring ROM among the different patient types..." [5]. This statement may still be true today.

Adhesive capsulitis is a common cause of shoulder pain, estimated to affect 2% in the normal population [10]. It is characterized by a usually spontaneous onset of shoulder pain accompanied by progressive restriction in active and passive ROM of the glenohumeral joint. The objective of this study is to estimate intra-observer reproducibility of ROM measurements in affected and non-affected shoulders of such patients.

As part of preparing for a randomized clinical trial (RCT) investigating patients with adhesive capsulitis, we developed a specific protocol for measuring range of motion in the shoulder. The reproducibility of these measurements was studied within the framework of the RCT project. The regional ethics committee granted ethical approval. The procedures followed protocol and complied with the Helsinki Declaration as revised in 1983 and current national ethical standards for such studies.

Results for reproducibility are given in Table 1 for PROM and in Table 2 for AROM. Observation 2 tended to display slightly lower scores than observation 1. According to paired t-tests, the difference was significant for P. ABD and A. ABD for the non-affected shoulder. In our analyses, we used a one-way ANOVA model to estimate absolute and relative reliability, and a possible and unexpected bias (approximately 1°–2°) was not corrected for by this method. However, we also calculated the indices using a two-way model (data not shown), where a possible session effect was included. This resulted in estimates that were quite similar to the ones obtained by the one-way model.

For PROM, estimated SDDs range from 12° to 20° for the individual movements and from 31° to 37° for the combined movement (C. PROM). For approximately 50% of pairs of PROM observations, test-retest difference was 5° or below (individual movements). SDD estimates for AROM ranged from 13° to 28° for the individual movements and from 44° to 52° for the combined movement (C. AROM). Estimated ICCs were in the area of 0.60–0.90 for both active and passive movements.

Figures 6 and 7 show the C. PROM and C. AROM difference between the two observations, plotted against the corresponding mean value of the two observations for each patient. Results for the affected and the non-affected side are presented within the same diagrams. Reproducibility seems to be independent of the magnitude of the score for the affected as well as the non-affected shoulder.

We have investigated the intra-observer reproducibility of measurements obtained with a standardized protocol for measuring range of motion in patients with adhesive capsulitis. Thirty-two patients were included in the study. This low number means that figures presented are only "rough" estimates of the reproducibility that can be expected in similar settings.

We observed a tendency for the scores of the second observation to be lower than those of the first observation. We have no good explanation for this possible bias, but would like to point out that any "true" within-patient changes in ROM during the one-week interval will lead to over-estimation of measurement error in a study of this type.

Even though we made a great effort to determine starting points and end-points consistently, this was probably a key source of error. The determination of these points involved deciding when the adequate patient/arm position was reached, the stabilization of arm position, placement of the inclinometer, and finally reading the display. These actions are all likely to add variability to measurements.

When measuring PROM, we stabilized scapula during movements, and used initiation of scapular movement to help determine end-points of glenohumeral motion. This procedure was not problem-free, and especially for flexion, it was sometimes difficult to determine when the scapula started to move. For all movements, patient muscle use could interfere with the passive movements, making representative scores difficult to obtain. Although we sometimes experienced problems with the technique, we share the view that it is essential to restrict scapulothoracic motion in order to more accurately reflect the range of motion of the glenohumeral joint [2, 26, 36]. There are also indications that stabilizing the scapula may produce more reliable results for some movements [16]. Furthermore, strict control of the scapula may be necessary in order to determine end-points of PROM with a method that does not cause excessive pain for the patient. The alternative might have been to determine end-point as the "point of first pain" [27] or as the point where pain makes further movement intolerable [14]. However, it is problematic to use pain as an end-point indicator in patients who experience pain more or less constantly, as many of the patients in our study did. It is our view that this would also affect validity when interpreting PROM values. When reported pain is used to indicate end-of-range, would we be measuring changes in the connective tissues surrounding the joint, or changes in patient pain status? Few clinical studies of patients with adhesive capsulitis have indicated how PROM end-of-range was determined. It will be advantageous if future researchers are more specific when reporting these procedures.

Testing positions are, of course, important when measuring ROM, and reproducibility is probably affected by the choice of testing positions [20]. In this study, we measured shoulder rotation from a 45° abducted position. It would have been more practical to measure rotation at 90° abduction, using positions described by Clarke [2] and others [16, 17, 26, 27, 29]. However, the rigid shoulders of our study population made it necessary to use a narrower angle.

Test positions and poor fixation of scapular movements may result in compensatory movements and thereby increased variation of measurements, in particular for active movements. Patients use various techniques to maximize their performance, some of which may be problematic to correct. Reliable (and valid!) measurements of these movements probably require very strict instructions to the patients. In many settings, however, there may be problems with the application of such methods.

We used a plastic plate in order to reduce "wobble" when applying the inclinometer. When experimenting without such a plate, we noted larger differences between subsequent measurements. We suspect that this method could be refined further, perhaps by using a longer plate, or stiffer material, or by improving plate attachment to the arm. Alternatively, attachment of the inclinometer itself to the arm during the entire movement, in a method similar to Clarke's [2, 29], may be superior to the method we used.

For passive movements, smallest detectable difference (on the 95% level) was estimated to be approximately 15° for the individual movements. In a clinical setting, this indicates that differences between two single observations exceeding 15° can be trusted to represent a ''true'' difference. For some movements of active ROM, SDD (95%) was about 25°. Relative reproducibility estimates (ICC) in our study were 0.61–0.93 for the various movements. In general, reproducibility seems satisfactory for group comparisons, but may be insufficient for individual comparisons in many settings. However, confidence intervals are wide and indicate that larger studies are necessary for definite statements to be made. Riddle et al., investigating shoulder patients [19], reported intra-observer correlation of 0.87–0.99, but absolute measurement errors were not given. This makes comparisons problematic [34]. With the exception of this single study, measurement errors (absolute or relative) reported in comparable studies investigating intra-observer reproducibility in shoulder patient populations [4, 7, 9, 27, 28, 37] are not smaller than the ones reported in our study of patients with adhesive capsulitis.

There was a tendency that SDD for AROM was larger than for PROM in this study. This may not be due to larger "measurement errors", strictly interpreted. It may be due to the patients' exhibiting truly more variable AROM than PROM. When measuring AROM, patients were asked to move their arm as far as possible in the requested direction, allowing for necessary scapular movement. The largest measurement errors in the study were noted for movements of active abduction and flexion, both of which are movements involving potentially large contributions by other joints. Furthermore, active ROM depends not only on the joints, but also very much on the overall function of the neuromuscular system. In addition to pain [14], AROM measurements may be seriously affected by motivation, learning effects or fear avoidance, all very fluctuating parameters.

One must consider that the patients in our study were selected to some degree. Because we wanted to study patients with adhesive capsulitis, we only included patients whose passive ROM could be adequately observed. There are undoubtedly some patients with painful shoulders whose ROM can only with difficulty be reliably and validly measured. This may be associated with pain [14], but the relationship is unclear [3]. One might speculate that other factors such as muscle tone, motivation or fear avoidance could play an important role. If this situation is suspected when ROM measurements must be made in patients, the use of anesthesia might be considered. Studying reproducibility under such conditions was beyond the scope of this project.

In this study, we employed a measure of "combined" total ROM. The corresponding SDD for PROM was 31° – 37°, meaning that differences larger than 8°–9° on average for four single movements are probably not due to measurement error. For combined total AROM, SDD was 44° – 52°, with corresponding figures of 11° – 13°. Average random error for each movement is reduced when several movements are combined like this [38]. Future researchers may want to investigate whether "combined" shoulder ROM is a useful parameter in various clinical settings.

The present study of shoulder ROM measurements in patients with adhesive capsulitis indicates that reproducibility is acceptable for group comparisons, while scores for individual patients need to be interpreted with caution. We advise future researchers to carefully consider measurement protocols and study designs when investigating range of motion in patients with adhesive capsulitis.