Extensor Grip Test for Tennis Elbow

Attached is an interesting small trial demonstrating value of the extensor grip test for predicting positive response to bracing for tennis elbow. However, the results are not quite as impressive if you consider the positive likelihood ratio (LR). For any students of EBP reading the blog, see if you can calculate the positive LR. Post your answers as a comment. Looking forward to the responses.

John

Download Struijs-AmJSportsMed-2005-ExtGripTestPredictBracing.pdf

Background: Tennis elbow is a common complaint. Several treatment strategies, such as corticosteroid injections and physical therapy and braces, have been described.
Hypothesis: The extensor grip test has predictive value in assessing the effectiveness of bracing in tennis elbow.
Study Design: Cohort study (prognosis); Level of evidence, 1.
Methods: Patients with tennis elbow complaints were randomized into 3 groups: brace only, physical therapy, and combination brace and physical therapy. The extensor grip test was performed before randomization on all patients. Outcome measures at 6-week follow-up were success rate, severity of complaints, pain, disability, inconvenience during daily life, and satisfaction.
Results: In the brace-only group, significant differences were identified between patients with a positive test result and patients with a negative test result for 3 outcome measures. The success rate in the test-negative group was 23% (5/22) compared to 47% (21/45) in the test-positive group. Mean decrease in pain was 23 (95% confidence interval, –3 to 49) in the test-positive
group compared to 11 (95% confidence interval, –6 to 28) in the test-negative group, and mean satisfaction in the test-positive group was 71 (95% confidence interval, 48 to 94) compared to 51 (95% confidence interval, 24 to 78) in the test-negative group. In the physical therapy and combination groups, no differences were identified between test-positive and test-negative
patients.
Conclusion: The extensor grip test seems valuable as a predictive factor for the effectiveness of bracing as treatment for tennis elbow over the short term.

Comments

[Jon Newman]

If I did it correctly (which is a big if) I got 2.04 suggesting a minimal to small increase in likelihood.

jon

[Melissa Ogle]

If my math is correct (if?), my results suggested a minimal increase in the likelihood with a slightly different value of 1.56.

Melissa

[Herb Silver]

I noticed in this article they "tested" the elbow with the elbow flexed. I have found that it helps to test for reproduction of symptoms with both the elbow flexed (ECRB) and with the elbow extended (ECRL). It certainly makes a difference when using a grip dynomometer and measuring grip strength--pain will be more limiting in one or the other position of the elbow. If you don't document the position of the elbow, grip strength reproduction of pain will have a much greater variance (if you don't hold elbow range constant). I was just wondering if they had tested the elbow both flexed and extended if it would have added more information in this study-maybe tennis elbow straps only work with one of these muscles but not the other--something I have wondered about but never had sufficient patients to test.

[Michael DuPriest]

LR of 1.35 to predict a positive outcome with bracing in the presence of a positive extensor grip test.

This would not give me much confidence to use a counter-force brace.

[John Childs]

ok, we've got ranges from 1.34-2.04 for the +LR. However you slice it, the test only marginally shifts the post-test probability that patients with a positive test are likely to have a successful outcome from bracing. Although the difference in posted estimates are not large enough to change decision-making (remember, "great" is sometimes the enemy of "good enough" when it comes to clinical decicision-making), let's have a look at the calculations:

+success -success
+test 21(a) 24 (b)
-test 5 (c) 17 (d)

Is this what your 2X2 table looked like?

+LR = Sen/(1-Spec), correct?
Sens = a/(a+c)-true positive rate
Spec = d/(b+d)-true negative rate

So, the positive LR = 1.38, with a 95% CI from (1.003, 1.898) - BONUS POINTS IF YOU GOT THIS:)

Best case (upper boundary of the 95% CI), the +LR appraches 2, which offers a marginal increase in post-probability, the actual extent of which actually depends on the pre-test probability).

Worst case, the +LR is 1, meaning it has no value.

So, would EBP dictate that this test become the preferred test for predicting response to bracing?

If you want to get into the details, the test's low specificity (.42) accounts for the low +LR. The test has decent sensitivity (.80), suggesting that when the test is negative, the patient is unlikely to have a succesful outcome (as defined in this study) from bracing. This might be more useful in clinical practice.

Why don't we see these accuracy stats in the study report? Your guess is as good as mine.

Thanks for the responses. Let me know if you think I mis-calculated anything.

John

[Charles Sheets]

For anyone like me who is still working to make all these calculations second nature, here is a site that will make the calculations for you (including the 95% confidence interval) once you set up the 2X2 table:

http://faculty.vassar.edu/lowry/clin1.html#note

I guess now I've lost any excuse for avoiding calculating and assessing likelihood ratios.

Charlie

[PYH]

Dear all,

Based on the formulas by Simel and colleagues(1), the negative LR for the extensor grip test is 0.46 (95% confidence interval (CI): 0.20 – 1.10). Despite the “decent” sensitivity (0.80), the test does not determine the patients who are unlikely to benefit from bracing. Perhaps this is another reason why we do not see the accuracy stats in the AJSM paper.

Cheerios,
Yonghao
Singapore

Reference:

1. Simel DL, Samsa GP, Matchar DB. Likelihood ratios with confidence: sample size estimation for diagnostic test studies. J Clin Epidemiol. 1991;44:763-70.