Managing osteoarthritis of the knee

Editor—Tubach et al say that we misinterpret data by confusing group responses and individual responses. They present hypothetical calculations that supposedly show a number needed to treat of 5-6.6 for non-steroidal anti-inflammatory drugs (NSAIDs) in osteoarthritis of the knee.

Their calculations are based on a weighted mean difference of 10.1 mm on a visual analogue scale from our analysis. However, we showed that this group response value was inflated by patient selection bias in a subgroup of trials. When this subgroup of biased trials was removed from analysis in our paper, the unbiased weighted mean difference fell to 5.9 mm (3.8 to 7.9).

As a benchmark for response, Tubach et al have selected what the inventors termed the “minimally perceptible difference” at 9.7 mm on the visual analogue scale.¹ A more clinically relevant benchmark in knee osteoarthritis would be the “minimal clinically important difference,” which corresponds to 19.9 mm on the scale.² Using the unbiased weighted mean difference and the latter benchmark for treatment success, the resulting number needed to treat is in the 9-20 range.

Tubach et al make a point of the fact that the occasional patient may experience benefit from NSAID treatment. In our view, it seems more relevant to question if patients should be put at jeopardy for adverse effects from oral NSAIDs when only a few of them are likely to notice if their NSAID is replaced by a placebo.

We stand by our conclusion that there is a lack of evidence for clinically relevant effects from oral NSAIDs in knee osteoarthritis pain.