Roth and Shainhouse1 reported in the ARCHIVES a well-designed randomized controlled trial on efficacy and safety of a topical diclofenac solution in the treatment of primary osteoarthritis of the knee. For clinical implementation we argue that additional information is needed.
In the sample size calculation, the authors state that they wanted to detect a clinically significant difference of 2 (SD of 4.5) in the change in Western Ontario McMaster Universities LK3.1 OA Index (WOMAC) pain dimension scores. To detect this difference, they needed a sample size of at least 80 in each treatment arm, but they decided to use a sample size of 200 patients (100 per treatment arm) to allow for a nonevaluable rate of 20%. In the trial, however, they included 326 patients, leading to 322 evaluable patients for the intent-to-treat analysis (163 in the topical diclofenac group and 159 in the vehicle-control group). Without any explanation the study size has almost been doubled compared with the original sample size calculation. Subsequently, the authors detected a statistically significant difference of 1.6 in change in the WOMAC pain score in favor of the topical diclofenac group, a smaller difference than originally stated as clinically significant.
Bierma-Zeinstra SMA, Belo JN, Thomas S. Efficacy and Relevance of Topical Diclofenac for Osteoarthritic Pain. Arch Intern Med. 2005;165(5):588–589. doi:10.1001/archinte.165.5.588-b