Thirty-eight patients with active, definite, or classical rheumatoid arthritis were tested in a double-blind, 3-week-per-arm, multiple-crossover, randomized, block-design comparison of 100, 300, 600, and 800 mg/day carprofen given b.i.d. A linear dose-response relationship was demonstrated for six of nine efficacy measures (p less than 0.052). A plasma concentration to therapeutic response relationship was shown just before or 1 to 2 hours after a dose (p less than 0.05) for seven efficacy parameters for the patients with at least three serum carprofen concentrations. By nonparametric analysis, with the patients divided into three equal groups, the percent of responders rose from 38.1% to 50% to 59.1%. Sixty-nine percent of patients responded when carprofen concentrations were greater than 10 micrograms/ml, whereas only 9% responded when they were below 1.9 micrograms/ml. Although only seven patients had limiting side effects, there was a tendency toward a dose-toxicity relationship through 600 mg daily carprofen.