The relative effectiveness of six nonsteroidal anti-inflammatory agents was studied in 33 patients with rheumatoid arthritis and 32 patients with ankylosing spondylitis in a double-blind, randomized, prospective study employing a six-way multiple crossover design with six-week trials of each agent. In ankylosing spondylitis, naproxen, indomethacin, and fenoprofen calcium were the most effective agents. In rheumatoid arthritis, relatively little mean difference between drugs was found. Most of this difference could be attributed to compliance factors, which favored drugs that required only a small number of pills daily. Despite the small differences in effect, patients had strong preferences. More than 85% of patients were still taking their preferred medication after a mean follow-up period of one year.