Surveys show consistently that pain is not treated well. Improvement depends on knowing which treatments are the most effective. We used systematic review to compare the relative efficacy of two common analgesics, ibuprofen and diclofenac, in post-operative pain. Studies were identified by searching MEDLINE (1966 to Dec 1996), EMBASE (1980 to Jan 1997), the Cochrane Library (Aug 1996), Biological Abstracts (Jan 1985 to Dec 1996) and the Oxford Pain Relief Database (1950 to 1994). We sought randomised, controlled, single-dose comparisons of oral ibuprofen or diclofenac against placebo. Summed pain relief or pain intensity difference over 4-6 h was extracted and converted into dichotomous information yielding the number of patients with at least 50% pain relief. This was then used to calculate the relative benefit and the number-needed-to-treat for one patient to achieve at least 50% pain relief. Thirty-four reports compared ibuprofen and placebo (3591 patients), six compared diclofenac with placebo (840 patients), and there were two direct comparisons of diclofenac 50 mg and ibuprofen 400 mg (130 patients). In post-operative pain, the numbers-needed-to-treat for ibuprofen 200 mg were 3.3 (95% confidence interval 2.8-4.0) compared with placebo, for ibuprofen 400 mg 2.7 (2.5-3.0), for ibuprofen 600 mg 2.4 (1.9-3.3), for diclofenac 50 mg 2.3 (2.0-2.7) and for diclofenac 100 mg 1.8 (1.5-2.1). Direct comparisons of diclofenac 50 mg with ibuprofen 400 mg showed no significant difference between the two. Both drugs worked well. Choosing between them is an issue of dose, safety and cost. Copyright 1998 European Federation of Chapters of the International Association for the Study of Pain.