Searches found 86 reports (10160 patients) that met inclusion criteria, 76 of which had dichotomous pain outcomes, including three unpublished reports from a pharmaceutical company. For acute conditions, 37 placebo-controlled trials were analyzed. All but one of these showed better pain reduction with the topical nonsteroidal anti-inflammatory drug (NSAID) than with placebo, for a pooled relative benefit of 1.7 (95% confidence interval [CI] 1.5 to 1.9) and a number-needed-to treat (NNT) of 3.9 (95% CI 3.4 to 4.4). Trials with higher quality scores showed the same results. Smaller-sized trials gave a larger effect size and a smaller NNT than large trials. Ketoprofen, felbinac, ibuprofen, and piroxicam were statistically superior to placebo, with NNTs of 2.6 to 4.2.
For chronic conditions, all 12 placebo-controlled trials showed better pain relief with topical nonsteroidal anti-inflammatory drugs than with placebo. Seven of these studies had statistically significant results. The pooled relative benefit for all 12 studies was 2.0 (95% CI 1.5 to 2.7), and the NNT was 3.1 (95% CI 2.7 to 3.8). Better quality studies had the same results. The effectiveness of individual drugs could not be analyzed because there were not enough studies to combine.
No significant benefit of oral over topical medication could be found in three studies of acute conditions and two of chronic conditions. Local adverse effects were found in 2.6% of subjects in studies of acute conditions (versus 3.0% for placebo) and 5.9% in studies of chronic conditions (versus 5.3% for placebo). Incidence of systemic adverse effects was about 1% for acute and chronic conditions and was the same for treatment and placebo groups.
Analysis of methodology
This study was generally well done. It addressed a focused clinical question and made use of explicit inclusion criteria. Failure to contact authors and search out abstracts might have meant important studies were missed. The authors’ own analysis suggested that some negative studies were absent from the list of those identified. Publication bias was, therefore, a problem, but likely no more of one than in other systematic overviews: especially for acute conditions, an implausibly large body of evidence would be needed to outweigh the 36 positive trials identified. Validity of the studies included was appraised, and findings did not appear to rest on poor-quality or small studies. Important steps in selecting studies for inclusion and abstracting results were done by more than one author and are probably reproducible.
The study’s major flaw reflects that of the trials included, none of which assessed outcomes beyond 30 days. Especially for chronic conditions, long-term efficacy of several months or even years would be important to patients, yet only short-term data are available. Even more problematic, adverse effects were also measured based on a short duration of treatment.
It is not clear from the study how often topical treatments should be applied, whether it matters which joints are involved, what effect concentration of the medication has on efficacy and adverse events, and how costly these medications are. It is also not clear how well blinding worked, especially if the active topical medications had a smell, appearance, or sensation on the skin that was different from placebo. Additional patient-oriented outcomes, such as function and quality of life, would be desirable. The study was sponsored by two manufacturers of topical nonsteroidal anti-inflammatory drugs (NSAIDs).