1. Ten patients with bilateral knee joint effusions were treated topically with a gel containing 1 g diclofenac/100 g (80 mg three times daily). They were randomized to receive diclofenac gel to one knee and a placebo gel preparation to the other knee. 2. Diclofenac was assayed in synovial fluid and blood plasma by GC/ECD as the pentafluorobenzyl-ester derivative. 3. Total concentrations of diclofenac in synovial fluid (day 4) were significantly higher in the diclofenac gel treated knee than in the contralateral placebo treated knee (25.5 +/- 3.6 ng ml-1 vs 21.6 +/- 2 ng ml-1; P less than 0.05). These concentrations were lower than total plasma drug concentrations (40.6 +/- 4.7 ng ml-1, n = 10, P less than 0.01). Unbound concentrations of diclofenac in synovial fluid from either the diclofenac gel treated or the placebo treated knee were not significantly different from each other or from plasma free concentrations (115 +/- 16 and 99 +/- 12 vs 108 +/- 19 pg ml-1). 4. Clinical parameters showed improvement of joint mobility and a small reduction of swelling (circumference) in both knees with time. However, the differences between knees were not significant. 5. We conclude that direct transport of diclofenac from the skin into the ipsilateral knee joint after cutaneous application is minimal. Distribution seems to be predominantly via the blood. Whether the observed improvements of clinical parameters were due to drug effects or to the spontaneous course of the underlying disease cannot be distinguished.