CRD Summary: The review assessed the long-term efficacy (at least 4 weeks) of topical therapies for pain control in primary knee osteoarthritis. The authors concluded that topical non-steroidal anti-inflammatory drugs (NSAIDs) provide effective pain relief for knee osteoarthritis. The authors caution against generalising to other NSAID preparations. The small data set on which the review was based, differences in treatment duration, uncertain study quality, and uncertainty about the statistical techniques used suggest that the authors' conclusion should be interpreted with caution.

CRD Commentary: The review addressed a clear question that was supported by clearly defined inclusion criteria, and several relevant sources were searched. The methodology undertaken to select studies and extract the data was not reported, thus the likelihood of reviewer error or bias at these stages cannot be assessed. The authors do not appear to have systematically assessed the validity of the included studies, and there were insufficient details to enable the quality of the studies to be gauged. There is uncertainty as to whether appropriate statistical analyses were undertaken; it is unclear why the authors did not use the standardised mean difference method. The authors did not report whether the reported pooled ES was taken from the fixed-effect or random-effects model; however, the pooled estimates in the presented forest plot from both analyses appear similar. Although the results indicated that diclofenac may be more effective than eltenac, no sensitivity analysis was performed. Only two preparations were involved in the analysis and the authors appropriately highlight caution against generalising to all NSAID preparations. The small data set on which the review was based, differences in treatment duration, uncertainty about statistical techniques, and the absence of information about study quality suggest that the authors' conclusions should be interpreted with caution.

Three of the included studies were supported (partially or fully) by the pharmaceutical industry.