Send letter to journal:
Re: Meta-analysis confidence intervals

Dear Editor

Hegedus et al [1] used a fixed-effects model to pool diagnostic odds ratios (DOR) for the Neer and the Hawkins-Kennedy tests for impingement. In Figure 2 and in Figure 3 they display the natural logarithms of the DOR for the individual and pooled results (the figures are inadvertently labeled as DOR rather than as the logarithms). In each diagram, the null value (natural logarithm of the DOR =0) is included in the confidence intervals of the pooled results, while it is not included in the confidence intervals of any of the individual studies. The authors report that the 95% confidence interval for the pooled DOR crosses 1 (corresponding to a log-odds of 0) indicating that neither test has diagnostic utility for impingement.

In a fixed-effects model, a weighted average is taken of the odds ratios of the individual studies. Different calculation methods may arrive at slightly different confidence intervals for the pooled odds ratio, but the standard error of the weighted sum is smaller than the standard errors of the individual studies. It often happens that the confidence intervals for individual studies may include the value for the null hypothesis, while the pooled estimate confidence interval excludes that value; this is one reason that meta-analyses are undertaken. If the converse occurs, a computational mishap is likely to have occurred. Pooling several statistically significant odds ratios will not yield a non-significant odds ratio. In the present meta-analysis, extracting the data from the four studies and pooling the results may yield a DOR of insufficient magnitude to discriminate between cases and non-cases of impingement, but the calculation yielding a DOR whose confidence interval includes 1 should be re-examined.

References

Send letter to journal:
Re: Comment regarding Hegedus et al and comparing investigations into OST's

Dear Editor

Firstly I think the authors of this article are to be congratulated for condensing a significant amount of information to a digestible article of interest to clinicians and investigators alike. The investigators point out that many examination techniques fail to repeat the initial success of their investigation during subsequent study by other authors. One aspect of this oft-cited point which has evaded attention may be failure to perform the physical examination in the same manner. For example, in the original description of the Active Compression Sign[1], the OST is described thus:

‘The patient was asked to forward flex the affected arm 90° with the elbow in full extension. The patient then adducted the arm 10° to 15° medial to the sagittal plane of the body. The arm was internally rotated so that the thumb pointed downward. The examiner then applied a uniform downward force to the arm. With the arm in the same position, the palm was then fully supinated and the maneuver was repeated.’ (p. 610).

The Active Compression sign has been re-investigated by Parentis et al[2] who describe the test somewhat differently as follows:

‘To perform this test, the shoulder is forward flexed to 90° and adducted 10° to 15°. The shoulder is then fully internally rotated, and the patient is asked to resist downward pressure on the arm by the examiner. The test is then repeated with the shoulder in the fully externally rotated position’[2] (p. 410)

Similarly Parentis et al[3] describe the Active Compression sign:

‘The active compression test was performed with the patient standing. The shoulder was flexed forward to 90° and adducted 10° while the elbow was fully extended. The shoulder was then fully internally rotated. A downward force was applied, and the patient was asked to resist. This motion was then repeated with the shoulder externally rotated.’[3] (p. 266)

Significantly in the original description the shoulder position is maintained constant whilst the degree of forearm supination is the only alteration in re-performing the second part of the test whilst the two subsequent investigations make no such distinction altering shoulder position in the second portion of the test. Intuitively the Active Compression Sign has utility in the examination of long head of the biceps origin injury since the patient’s pain behaviour is described by altering only the forearm pronation/supination and therefore perhaps altering the amount of tension on the long head of the biceps but little else at the shoulder joint. This would seem to have similarity with a number of other examination techniques directed toward biceps anchor pathology such as The Biceps Load Test[4], The Biceps Load II[5], the ‘New Pain Provocation Sign’[6], and the ‘SLAP prehension’ test[7]

Comparing the original examination[1] with subsequent tests[2, 3] which do not accurately reproduce the test is simply not ‘comparing apples with apples’. I’d suggest that a vital addition to the checklist of diagnostic accuracy is ensuring accurate performance of the test as originally described before between-test examination can be conducted.

References

1. O'Brien, S.J., et al., The active compression test: a new and effective test for diagnosing labral tears and acromioclavicular joint abnormality. Am J Sports Med, 1998. 26(5): p. 610-3.

2. Parentis, M.A., et al., An anatomic evaluation of the active compression test. J Shoulder Elbow Surg, 2004. 13(4): p. 410-6.

3. Parentis, M.A., et al., An evaluation of the provocative tests for superior labral anterior posterior lesions. Am J Sports Med, 2006. 34(2): p. 265-8.

4. Kim, S.H., K.I. Ha, and K.Y. Han, Biceps load test: a clinical test for superior labrum anterior and posterior lesions in shoulders with recurrent anterior dislocations. Am J Sports Med, 1999. 27(3): p. 300-3.

5. Kim, S.H., et al., Biceps load test II: A clinical test for SLAP lesions of the shoulder. Arthroscopy, 2001. 17(2): p. 160-4.

6. Mimori, K., et al., A new pain provocation test for superior labral tears of the shoulder. Am J Sports Med, 1999. 27(2): p. 137-42.