We thank Dr. Bastick for his suggestion that adjustment for baseline severity of OA (K/L score 0 or 1) might lead to different results. In response to an earlier reviewer, we included a sensitivity analysis restricted to participants with a K/L score of 0 only (Supplementary Table 2 in our original article), which replicated our results.
We have since repeated our analysis and adjusted for baseline K/L grade as suggested, and we again found associations comparable to those reported in our article, although the confidence intervals were much wider for these estimates and consequently far less precise. Additionally, we found no evidence to support effect modification by K/L grade. We did not find support for the idea that baseline K/L grade confounds our results.