To the Editor:

We believe that the strength of a prospective followup study is in the fact that relationships between potential determinants and the onset or progression of a disease can be investigated. The generated hypothesis within such a study design, i.e., that the long-term use of diclofenac is associated with the occurrence of progression of OA, should be confirmed in other prospective studies, and preferably in a randomized clinical trial.

We agree with Dr. Ding that the reported associations could be confounded by pain severity. However, a confounding factor must be 1) a risk factor for the outcome (progression of OA), 2) associated with the exposure under study (long-term use of diclofenac), and 3) not an intermediate factor in the relationship between the outcome and the exposure (Rothman KJ, Greenland S, editors. Precision and validity studies. In: Modern epidemiology. 2nd ed. Boston: Little, Brown and Co.; 1998. p. 123–5). As stated by Dr. Ding and also reported by our group (Reijman M, Hazes JM, Pols HA, Bernsen RM, Koes BW, Bierma-Zeinstra SM. X-ray findings strongly predict progression of osteoarthritis of the hip: a prospective cohort study. BMJ 2005;330:1183), baseline joint pain is a risk factor for progression of OA. The question is whether there is a relationship between the duration of use of diclofenac and baseline joint pain. In the subjects with hip OA, we found a minor difference in the duration of use of diclofenac between those with and those without baseline pain. However, in those with knee OA, no such difference was noted.

When we adjusted the analyses for baseline pain, the risk estimate for the hip was similar but reached only borderline significance (the odds ratio [OR] changed from 2.4 to 2.3, with a 95% confidence interval [95% CI] of 0.9–6.0), and for the knee it remained unchanged (OR 3.2, 95% CI 1.0–9.9). Among subjects without joint pain at baseline and followup, we found an OR of 1.8 (95% CI 0.5–7.3) for the hip, and for the knee we found no reliable estimate due to the small sample size.

Taking into account the magnitude of change of the reported risk estimate, there is no basis to assume that baseline joint pain is an important confounder for the relationship between diclofenac and progression of OA. It is possible that subjects with joint pain more frequently use NSAIDs for a longer period of time and that the relationship between baseline joint pain and progression of OA could be explained by the use of these NSAIDs. We found, as reported above, that persons with hip OA used diclofenac more often when they also had baseline hip pain, compared with those without baseline hip pain. When we adjusted the relationship between baseline hip pain and progression of hip OA for the use of diclofenac, the risk estimate for hip pain decreased significantly (OR from 3.4 to 2.2, 95% CI 1.1–4.1). Therefore, it appears that the relationship between hip pain and progression of OA is partly due to the use of diclofenac.

M. Reijman MSc*, S. M. A. Bierma-Zeinstra PhD*, H. A. P. Pols MD, PhD*, B. W. Koes PhD*, B. H. C. Stricker MD, PhD*, J. M. W. Hazes MD, PhD*, * Erasmus MC, Rotterdam, The Netherlands.