We agree with Dr. Patmas that minocycline's mechanism of action in producing benefit in patients with rheumatoid arthritis is unknown; however, we do not believe that any of the patients enrolled in our trial were infected with Borrelia burgdorferi. To be eligible for the study of ceftriaxone treatment of chronic inflammatory arthritis cited by Dr. Patmas, patients had to have antibodies to B. burgdorferi in titers of 1:64 or greater on two occasions before entry. We actively excluded patients with a history or clinical evidence of Lyme disease, and all patients tested negative for antibodies to B. burgdorferi on serum samples collected at the baseline evaluation and assayed by the Centers for Disease Control and Prevention.

We did not do the analysis suggested by Dr. Meehan, which would exclude protocol violations, because these violations appeared to be related to the assigned treatments. While receiving study medication, more patients in the placebo group required increases in background medications because of worsening arthritis; in contrast, more patients in the minocycline group reduced their use of these agents. By eliminating such patients from the analysis, we might not detect true treatment effects. We did do analyses restricted to patients who had completed the full 48-week protocol and found a significant benefit of minocycline over placebo in our primary outcome measures.

Drs. Canvin and Madhok correctly point out that we did not include data on a direct measurement of pain and note that a composite score may be a more valid measure of disease activity. The protocol for the trial was modeled after rheumatoid arthritis clinical trials done by the Cooperative Systematic Studies of Rheumatic Diseases group, which had not included pain assessments. Joint tenderness was assessed and is a measure of pain. We added a patient's assessment of pain using a visual analog scale after the first Outcome Measurements in Rheumatoid Arthritis Clinical Trials meeting in 1992, during which the international rheumatology community agreed that such a measure should be included in all future clinical trials on rheumatoid arthritis [1]. Unfortunately, this was after our trial had been designed and implemented. Subsequently, the American College of Rheumatology published its recommended core criteria set of outcome measures for rheumatoid arthritis clinical trials [2] and a method for assessing improvement [3]. We evaluated this criteria set using data from our trial but, because our patients did not assess their pain at baseline, we modified the method to require 20% improvement in tender and swollen joint counts and three of the four remaining outcome measures (instead of three of five). Using this approach, we found that 44% of the patients receiving minocycline would be classified as improved compared with 26% of the placebo recipients [4]. We support the view that future rheumatoid arthritis clinical trials should include a patient's assessment of pain to make full use of the method for assessing improvement proposed by the American College of Rheumatology.