The findings of the Dutch study [2] and the MIRA trial [1] are remarkably similar. However, 80 patients in the Dutch study were equally randomly assigned to receive either placebo or minocycline, whereas 110 and 109 patients in the MIRA study received placebo and minocycline, respectively. Both trials used 100 mg of oral minocycline twice a day; patients were treated for 48 weeks in the MIRA study but for only 26 weeks in the Dutch study (with 4 weeks of post-withdrawal follow-up). Patients in the Dutch study had had rheumatoid arthritis for an average of 12 to 14 years compared with 8 years for those in the MIRA study. Patients in the Dutch study were also more likely to have positive test results for rheumatoid factor (89% compared with 56%), joint erosions (95% compared with 57%), and higher baseline erythrocyte sedimentation rates. In both trials, previous therapy with nonsteroidal anti-inflammatory drugs and low-dose prednisone was continued. However, two thirds of the patients in the Dutch study also continued to receive disease-modifying antirheumatic drugs such as methotrexate, gold, and hydroxychloroquine, whereas in the MIRA trial, therapy with these drugs was discontinued 1 month before study entry.

These baseline differences suggest that patients in the MIRA study had somewhat less severe disease than did patients in the Dutch study. Clinical outcome measures improved in both trials, but measures also improved in the placebo groups; in addition, most measures did not significantly differ between groups. Nevertheless, in the Dutch study, changes in the Ritchie index score (for tender joints) and in the number of swollen joints significantly favored minocycline (P = 0.007 and P = 0.008, respectively). The proportion of participants in the MIRA study with 50% or greater improvement in swollen joints (P = 0.023) and tender joints (P = 0.021) also favored minocycline. Both studies used conservative intention-to-treat analyses. In the MIRA study, initial improvement in patients receiving placebo and in those receiving minocycline was similar, but during the last 24 weeks, patients in the minocycline group continued to improve, whereas those in the placebo group did not. In the Dutch study, the number of swollen joints and acute-phase reactants had significantly worsened 4 weeks after minocycline had been withdrawn at the end of the study, reinforcing the impression that the observed improvements were related to minocycline treatment.

Acute-phase reactants and rheumatoid factor titers improved with minocycline in both trials [1, 2]; however, these measures either did not change or worsened in the patients receiving placebo. Intergroup differences in the following measures were significant: C-reactive protein (P < 0.001), erythrocyte sedimentation rate (P < 0.001), IgM rheumatoid factor (P < 0.001), hemoglobin level (P < 0.001), hematocrit (P = 0.002), and platelet count (P < 0.001). In the Dutch study, the erythrocyte sedimentation rate had improved by week 4, whereas the Ritchie index score of tender joints did not improve significantly until week 12.

Adverse effects were modest except for two fractures (an elbow and a humerus) in patients in the Dutch study that were attributed to dizziness. Dizziness is a recognized adverse effect of minocycline that is not associated with other tetracyclines. Ten percent of patients in the Dutch study and 3% of those in the MIRA study dropped out because of dizziness. The number of opportunistic infections did not increase in either study.

Thus, two recent independent, randomized, placebo-controlled clinical trials with substantial sample sizes showed a modest but significant clinical benefit in patients who had had rheumatoid arthritis for an average of 8 to 14 years, as well as highly significant improvement in acute-phase reactants and IgM rheumatoid factor titers, with minimal adverse events. Separately, either trial could be considered an aberration; together they must be taken seriously. Indeed, together they could be submitted to the Food and Drug Administration as the two positive "pivotal" controlled clinical trials required for approval of a new drug application. Their combination of modest clinical efficacy with minimal toxicity compares favorably with currently used disease-modifying antirheumatic drugs, clearly justifying further evaluation of this therapy.

Neither study clarifies the mechanism of action of minocycline in rheumatoid arthritis [1, 2]. The improvement of acute-phase reactants and rheumatoid factor resembles that seen with treatment of a smoldering chronic infection, such as subacute bacterial endocarditis or chronic osteomyelitis. This suggests that a well-protected infection may be at least partially responsible for rheumatoid arthritis manifestations and that the treatment may suppress this infection. The absence of "cures" in the controlled studies, the occurrence of relapse after antibiotic withdrawal, and the apparent need for prolonged treatment suggest that either the postulated infection is difficult to eradicate or it is merely a cofactor and that once established, rheumatoid arthritis can continue without it. An intracellular infection that is partially protected from the administered antibiotic agents (for example, Mycoplasma infection) might fit the clinical observations. Alternatively, minocycline may produce clinical benefit in rheumatoid arthritis because of a slowly developing anti-inflammatory effect, similar to that seen with disease-modifying antirheumatic drugs. Circumstantial evidence supports each of these hypotheses, but neither the MIRA study nor the Dutch study addressed possible mechanisms of action.

Both Mycoplasma and Chlamydia have been considered as possible causes of arthritis. Mycoplasma hyorhinis causes an acute arthritis in Yorkshire swine that becomes chronic and destructive [8]. The organism was first cultured from synovial fluid and synovial tissue but later could not be cultured, even though joint inflammation continued and persistent microbial antigens could be shown [8]. Chlamydia organisms isolated from the knee of a patient with arthritis were reported to produce arthritis when injected into the knees of monkeys and of New Zealand white rabbits [9]. Moreover, Chlamydia trachomatis DNA has been detected in the joints of patients with reactive arthritis [10].

As discussed in the article by Tilley and colleagues [1], tetracyclines, particularly minocycline and doxycycline, are potent inhibitors of collagenases and gelatinases, two matrix metalloproteinases [11]. The role of metalloproteinases in the destructive inflammatory process of rheumatoid arthritis is not entirely clear but may be an important contributor [12]. Greenwald and colleagues [13] showed a significant reduction of the collagenolytic activity of synovial tissue obtained from six patients with rheumatoid arthritis who received minocycline before total joint arthroplasties compared with the collagenolytic activity of synovium obtained from the same patients during a previous arthroplasty when they had not received minocycline. Minocycline and doxycycline also inhibit phospholipase A2 in vitro [14], another mediator of inflammation [15]. Tetracycline has also been reported to decrease leukotaxis and to inhibit phagocytosis by neutrophils, lymphocyte proliferative responses, and anti-CD3-induced synovial T-cell proliferation in a dose-dependent manner [16, 17]. Some of these anti-inflammatory properties have been shown in a chemically modified tetracycline analog that has no antibacterial effects [16].

In summary, the results of the large, well-designed, and well-executed MIRA study reported by Tilley and colleagues [1], supported by the somewhat smaller but similar study by Kloppenburg and colleagues [2], provide substantial evidence of the beneficial clinical effect of minocycline in patients with rheumatoid arthritis. Several reasonable but unproved mechanisms of action are possible, any of which, if proved, could add a new dimension to the current understanding of the pathogenesis of rheumatoid arthritis and provide the basis for new approaches to its treatment. Additional clinical research is needed to differentiate the antibacterial and anti-inflammatory effects of minocycline treatment, perhaps by comparing minocycline with a nontetracycline antibiotic agent that lacks anti-inflammatory activity, and to determine whether doxycycline, which does not cause dizziness, is also efficacious in rheumatoid arthritis. In addition, studies of longer duration should be done in patients with recently developed rheumatoid arthritis to see whether prolonged early antibiotic treatment induces lasting remissions or decreases the progression of joint destruction.