p Values Don’t Have a Shot With Injections for Frozen Shoulder

One shot or three? Read this post, and you decide.

Editor’s Note: A randomized clinical trial misinterpreted its own findings, and we want to be sure you won’t modify your practice based on its claims. The error made in that study comes up frequently, so we’ll share it as a case study in what to watch out for. — SSL

What’s the Claim?

An RCT about the treatment of adhesive capsulitis of the shoulder compared one ultrasound-guided rotator interval corticosteroid injection with a three-injection approach (rotator interval, intra-articular, and subacromial bursa, all ultrasound-guided), and claimed the multiple-injection approach delivered better:

  • VAS pain scores at four weeks (3.1 ± 1.2 versus 4.3 ± 1.6) and eight weeks (2.2 ± 1.2 versus 3.4 ± 1.2 [all p values < 0.05]; no differences at 12 weeks)
  • Outcomes scores (ASES score at 12 weeks 88 ± 6 versus 81 ± 7 [p < 0.001]; Constant-Murley score at 12 weeks 79 ± 5 versus 74 ± 8 [p = 0.01])
  • Flexion and abduction by 10° to 15° (out of 180°, as they measured it), but there were no differences in rotation

The biggest issue, and one to always watch for when an article floods you with p values, is effect size. Patients perceive effect sizes, not p values. Here, the effect sizes were too small to care about.

How’s It Stack Up?