Minocycline in Rheumatoid Arthritis: A 48-Week, Double-Blind, Placebo-Controlled Trial

Methods

Protocol Development

The protocol for the Minocycline in Rheumatoid Arthritis (MIRA) Trial was developed by the investigators from the clinical centers, the coordinating center, and the National Institute of Arthritis and Musculoskeletal and Skin Diseases on the basis of protocols for clinical trials of therapies for rheumatoid arthritis that were previously done by the Cooperative Systematic Studies of Rheumatic Diseases (CSSRD) program [23, 24]. Inclusion criteria for those CSSRD trials were two of the following: nine or more tender joints capable of response, an erythrocyte sedimentation rate of 28 mm/h or greater, and morning stiffness lasting 45 minutes or longer. To simplify recruitment of patients, we modified these criteria to require the presence of nine or more tender joints capable of response with no minimum specified for either the erythrocyte sedimentation rate or for morning stiffness. The study protocol was reviewed and approved by institutional review boards at each participating clinical center.

Patients

To be eligible for the minocycline trial, patients had to meet the 1987 American Rheumatism Association (now the American College of Rheumatology) classification criteria for rheumatoid arthritis [25]. Patients were required to have active disease, defined as nine or more tender joints and six or more swollen joints capable of assessable improvement. Patients had to be 18 years of age or older and to have experienced onset of rheumatoid arthritis at 16 years of age or older. Patients receiving NSAIDs or low-dose corticosteroids (prednisone equivalent of a daily dose of 10 mg or less) were required to be receiving stable concurrent medication for 4 weeks before entering the study; these medications were to be kept stable throughout the study. One change from equivalent doses of one NSAID to another was permitted during the trial. Phenylbutazone, oxyphenbutazone, or parenteral steroids (including intra-articular corticosteroids) and disease-modifying antirheumatic drugs were not allowed for 4 weeks before the study; the use of these drugs during the trial was a protocol violation. Women were required to be postmenopausal, surgically sterile, or practicing a reliable method of contraception and to have a negative pregnancy test result.

Before randomization, eligible patients for the minocycline trial could have received therapeutic trials of hydroxychloroquine and, at most, one other disease-modifying antirheumatic drug (oral or parenteral gold salts, D-penicillamine, azathioprine, methotrexate, or sulfasalazine) unless therapy was discontinued for toxicity. Patients could have tried any number of disease-modifying antirheumatic drugs if discontinued for toxicity, provided that no drug was administered for more than 3 months or, for parenteral gold salts, was not given in a cumulative dose of more than 500 mg. These exclusion criteria were designed to avoid recruiting patients with a history of repeated failures of therapeutic trials of disease-modifying antirheumatic drugs but did allow entry of those who had discontinued therapy with these drugs because of toxicity before prespecified dosages or durations were reached.

Patients were not eligible at screening if they had known hypersensitivity to tetracycline; required long-term tetracycline therapy for another disorder; required ongoing therapy with antacids containing aluminum, calcium, or magnesium; or had received anticoagulant therapy. We excluded patients with chronic illnesses that would limit successful participation in the trial and patients classified as American Rheumatism Association-Steinbrocker functional class IV [26].

Three patients who met eligibility criteria at their screening visit were found to have too few affected joints at their baseline (randomization) visit and were inadvertently randomized. We included these patients in the analysis. All three patients were in the placebo group and were not considered responders to treatment. Removing these patients did not change the results of the trial.

Study Design

The MIRA trial was a six-center, double-blind, randomized, placebo-controlled trial comparing minocycline and placebo in treating rheumatoid arthritis. Patients were randomly assigned within the clinical center (that is, the clinical center was a stratifying factor), and randomization was blocked using randomly chosen block sizes between 4 and 6 to minimize the possibility that staff would guess the next treatment assignment. For 48 weeks, patients received two 50-mg capsules of minocycline hydrochloride or a visually similar placebo taken with water twice daily on an empty stomach. To minimize interference with the absorption of minocycline, patients were instructed not to ingest food or iron within 2 hours of receiving study medication. Investigators were allowed to adjust the dosage (without breaking the treatment code) if patients experienced side effects.

Measurements

The following measurements were obtained at the baseline and randomization visits and every 12 weeks thereafter.

Evaluator Assessments

Sixty diarthrodial joints were examined for the presence or absence of tenderness, and 58 were examined for swelling (hips excluded). Joint counts for tenderness and for swelling were the sum of the number of affected joints [27]. These joints were also evaluated for the severity of both tenderness and swelling on a 4-point scale. For tenderness, 0 = none, 1 = mild (positive response on questioning), 2 = moderate (spontaneous response elicited on examination), and 3 = severe (withdrawal by patient on examination); for swelling, 0 = none, 1 = mild (detectable synovial thickening without loss of bony contours), 2 = moderate (loss of distinctness of bony contours), and 3 = severe (bulging synovial proliferation with cystic characteristics). The sum of these determinations for all joints constituted a collective scores for either swelling or tenderness [23]. To foster agreement, evaluators attended a training session at the coordinating center before the trial. During the trial, all joint assessments for each patient were done by the same evaluator if possible. (During the trial, 94% of patients receiving minocycline and 92% of those receiving placebo had their joints evaluated by the same examiner at every visit.)

Grip strength was measured bilaterally using a mercury strain sphygmomanometer [28]. An evaluator assessed disease activity on the day of the examination using a scale of 1 to 5: absent, mild, moderate, severe, or very severe activity [29]. For functional class, patients were assessed according to the American Rheumatism Association-Steinbrocker classification [26] (at baseline and 48 weeks only).

Patient Assessments

Patients were asked about the duration of morning stiffness they experienced on the day before each study visit [23]. Function was assessed on eight activities of daily living with the Modified Health Assessment Questionnaire (MHAQ) [30]. Patients used a scale of 0 to 3: activity without any difficulty, with some difficulty, with much difficulty, and unable to do; thus, the higher the score, the greater the incapacitation. Patients were also asked to give an overall assessment of their disease activity for the preceding day on a scale of 1 to 5: absent, mild, moderate, severe, and very severe [29].

Ancillary Assessments

Laboratory determinations included urinalysis, complete blood count with differential, and a chemistry profile at the screening visit and every 4 weeks through week 12 and then every 6 weeks through week 48 of their treatment. Erythrocyte sedimentation rate (Westergren sedimentation rate) was measured at baseline and every 12 weeks. These tests were done by laboratories at each of the clinical centers.

Blood was drawn at baseline and every 12 weeks thereafter for centrally done analyses and serum banking. Serum samples were shipped frozen to the coordinating center. Immunoglobulin M rheumatoid factor was assayed by SmithKline Beecham Clinical Laboratories (Farmington Hills, Michigan) using nephelometry [31]. Seropositivity was defined as an IgM rheumatoid factor level of 30 U/dL or greater. For the few samples that were too lipemic to be tested reliably by nephelometry, IgM rheumatoid factor was measured by the latex fixation agglutination method [32]. The National Center for Infectious Diseases of the Centers for Disease Control and Prevention assayed serum samples for the presence of antibodies to Borrelia burgdorferi using a whole-cell sonicate and flagella-based enzyme-linked immunosorbent assay [33-35].

A posteroanterior radiograph of both hands and wrists was taken at baseline. Erosions were identified using the average score of two independent readers.

Adverse Drug Reactions

The patients were screened for adverse drug reactions at every visit. Patients were to stop receiving study medication for significant leukopenia, thrombocytopenia, anemia, hepatotoxicity, renal toxicity, or other serious adverse drug reactions. Data on adverse events were reviewed monthly at the coordinating center and every 6 months by the independent Data Safety Monitoring Review Committee.

Compliance

Patients were queried at each visit about all medications taken, including background medications. Trial medication capsules were counted at each visit to monitor compliance. Blood samples for minocycline assays were drawn at baseline and every 12 weeks. Pharmaco, Inc. (Richmond, Virginia) analyzed serum samples for minocycline using high-performance liquid chromatography with ultraviolet detection. This method provides a minimum quantifiable level for minocycline of 0.02 µg/mL [36].

Statistical Analysis

We used two prespecified primary outcome measures: improvement in joint swelling and improvement in joint tenderness. Definitions of the primary outcomes were taken from protocols for the CSSRD rheumatoid arthritis treatment trials [23, 24, 37]. Patients were considered to show improvement in joint swelling (or joint tenderness) if they had a 50% or greater improvement in the number of swollen (tender) joints at 48 weeks compared with baseline. Improvement in an individual joint was defined as a change in swelling (tenderness) from severe, moderate, or mild to no swelling (tenderness). Worsening was defined as involvement of a previously uninvolved joint or a change in swelling (tenderness) to severe in a joint previously rated as moderate or mild or none. For example, if a patient with 10 moderately tender joints capable of responding at baseline had 8 joints improve and 3 worsen during the trial (a net of 5 joints or 50% improvement), the patient would be classified as showing an improvement in tenderness. If 4 joints improved in the same patient and 2 worsened (20% improvement), this patient would not meet the definition of improvement.

The study was designed to detect an absolute difference between the minocycline and placebo groups of 30% using a critical level of 0.05 (two-sided test). In calculating the sample size, we assumed that the joints of 20% of the patients receiving placebo would improve. We assumed a dropout rate of 24% and considered this rate in calculating the sample size [38]. A sample size of 200 (100 per group) would provide greater than 90% power to detect such a difference. We did not plan an adjustment for multiple comparisons for the two primary outcome measures because these were prespecified.

One interim analysis of the primary outcome data was done using a Lan-DeMets statistical approach [39]. Because of this interim analysis, the final P value for the primary outcomes should be interpreted against a critical level of 0.0498 instead of the planned 0.05. The clinical center investigators and all patients remained blinded for the duration of the study and were unaware of the results of the interim analysis.

The primary outcomes of the treatment and placebo groups were compared using a Mantel-Haenszel [40] test with the clinical center included because it was a stratifying factor for randomization. (We used a critical value of 0.05 as a criterion for identifying differences between treatment groups at baseline that should be included in the model to test for treatment differences at follow-up. No baseline variables met the criterion.) In the analysis, we ignored the blocking. A statistical analysis that ignores blocking is less powerful, making detection of treatment differences more difficult [38]. We used an intention-to-treat analysis [38]; that is, all patients were analyzed in the group to which they were assigned. All patients were followed to the completion of the trial, even if they had discontinued therapy with the study medication and received other treatments. For patients who did not have a visit at 48 weeks, we imputed a final visit value using the worst of the previous visits (including the randomization visit) as a proxy for the 48-week visit.

We examined 10 other outcomes and used the Mantel-Haenszel approach for binary variables, with the clinical center as the stratifying variable. For continuous or ordinal outcomes, we used analysis of covariance on the ranks, adjusting for clinical center (five dummy variables) and testing for a treatment effect. To adjust for multiple comparisons, we interpreted P values against a critical level of 0.005 (Bonferroni correction for 10 tests) [41]. Our study was not designed to have high power to detect differences in these 10 other outcomes.

Baseline comparisons between the minocycline and placebo groups were made using t-tests for continuous variables, chi-square tests for nominal variables, and the Wilcoxon rank-sum [42] or the robust rank-order statistical test [43] for ordinal variables. We used an exact test for 2-by-K tables to compare reasons for withdrawal from study medication between the treatment groups.

Results

Patients

Of the 219 patients entered in the study, 109 were randomly assigned to receive minocycline and 110 were assigned to receive placebo. The demographic and baseline clinical features for patients in the two treatment groups are shown in Tables 1 and 2. No patient had positive serologic results for B. burgdorferi.

Almost one half of the patients had received disease-modifying antirheumatic drug therapy before entering the study, including oral and parenteral gold salts (23%), hydroxychloroquine (22%), oral and parenteral methotrexate (12%), sulfasalazine (6%), D-penicillamine (2%), and azathioprine (1%); one third were receiving low doses of oral corticosteroids.

Drug Compliance and Withdrawal

At the 48-week visit, 79% of the minocycline group and 78% of the placebo group continued to receive the study medication. Of the 219 patients, 94% had a visit at 48 weeks (7 patients in the minocycline group and 7 patients in the placebo group missed this visit). Of the 47 patients who had previously discontinued therapy with the study drug, 36 had a visit at 48 weeks. For those receiving medication, the mean percentage of medication taken by capsule count was 84% ± 26% for the minocycline group and 85% ± 25% for the placebo group. Minocycline assays confirmed the capsule count data.

The primary reasons for discontinuation of therapy with the study medication were inefficacy in the placebo group and toxicity in the minocycline group (Table 3). We observed an association between treatment group and the primary reason for stopping medication (P = 0.002).

Concurrent Antirheumatic Drugs

At baseline, both groups were similar in the concurrent use of antirheumatic drugs. Twenty-nine percent of the minocycline group and 32% of the placebo group were receiving oral corticosteroids. As shown in Table 4, there were more therapeutic interventions in the placebo group than in the minocycline group during the trial; some of these interventions were clearly protocol violations (for example, 17% of patients in the placebo group and 6% in the minocycline group received an intra-articular corticosteroid injection, and 14% of patients in the placebo group and 7% in the minocycline group had more than one NSAID substitution). Other interventions occurred after therapy with the study drug had been discontinued.

Primary Outcomes

Table 5 shows the intention-to-treat data for the primary outcomes. For joint swelling, 54% of patients receiving minocycline showed improvement compared with 39% of patients receiving placebo (P = 0.023). We observed similar differences (56% and 41% of patients, respectively) for joint tenderness (P = 0.021).

Other Outcomes

At the week 48 evaluation, the median joint counts for tenderness and swelling decreased from the baseline counts for patients in both the minocycline and placebo groups. The median decreases for tenderness were 14 for the minocycline group and 8.5 for the placebo group; for swelling, the decreases were 9 and 6, respectively (P < 0.001). No patients met American Rheumatism Association criteria for remission [44]. Figure 1 shows the temporal change in the proportion of patients with meaningful improvement by intention-to-treat analysis. For both groups, improvement began by week 12. Improvement in joint swelling and in joint tenderness continued through week 48 for patients receiving minocycline but reached a plateau at week 24 for patients receiving placebo.

Table 6 shows the data for some other outcome variables at randomization and at 48 weeks. The Table also heart rate alterans shows differences between the two assessments and the corresponding summary statistics for these differences. With two exceptions, both groups on average showed some improvement from baseline to 48 weeks. The exceptions were in the placebo group, in which hematocrit decreased and IgM rheumatoid factor levels increased. We detected treatment group differences for erythrocyte sedimentation rate, hematocrit, platelet count, and IgM rheumatoid factor, with the minocycline group showing greater improvement for each measurement (all P values < 0.001).

Fifty-four percent of patients in the minocycline group and 39% in the placebo group had erythrocyte sedimentation rates within the normal range at 48 weeks. Eighty-seven percent and 77% of patients, respectively, had hematocrits within the normal range, and 98% and 96% had platelet counts within the normal range. Of the 60 patients in the minocycline group who tested positive for rheumatoid factor at baseline, 13 tested negative at 48 weeks compared with only 1 of 63 patients in the placebo group. However, 3 of the 48 patients receiving minocycline who had negative test results became positive, and 2 of 47 patients in the placebo group became positive.

In comparisons of the treatment groups for the remaining outcomes—grip strength, patient and physician global assessment, MHAQ, and hours of morning stiffness— P values ranged from 0.04 to 0.76, all greater than the critical value of 0.005 (Bonferroni adjustment for multiple comparisons).

Adverse Drug Reactions

One death occurred—a suicide by a patient receiving placebo who discontinued therapy with study medication early in the trial. Table 7 shows the reported adverse effects that lasted 1 day or more. Dizziness was expected to be an adverse effect of minocycline [19, 20, 45], yet we observed only slight differences between the two treatment groups. Thirty-nine percent of the minocycline group and 34% of the placebo group reported dizziness lasting more than 1 day at least once during the trial. Six percent of the minocycline group but no patients in the placebo group stopped medication because of possible toxicity. Elevated liver function test results were reported for two patients in the minocycline group and three patients in the placebo group. Two patients receiving minocycline had total bilirubin levels greater than 25.65 µmol/L (1.5 mg/dL) during the trial. The values in one of these patients subsequently returned to normal. The other patient, a suspected alcohol abuser, had persistently elevated liver enzymes, stopped receiving medication, and moved away from the clinical center areas. One patient receiving minocycline and one patient receiving placebo had elevated creatinine levels during the study (range, 132.6 to 203.3 µmol/L [1.5 to 2.3 mg/dL]). Two patients receiving minocycline developed raised blue spots on their tongues, and another reported “graying” teeth. Fewer patients in the minocycline group than in the placebo group had urinary tract infections (5% compared with 12%; P = 0.05) and upper respiratory tract infections (11% compared with 16%; P = 0.25). More patients in the minocycline group had vaginitis than did those in the placebo group (9% compared with 3%; P = 0.04).

Discussion

The MIRA trial showed that minocycline is both effective and safe for treating patients with mild to moderately active rheumatoid arthritis. Benefit became evident after 12 weeks of therapy, and the proportion of patients treated with minocycline showing improvement continued to increase through week 48 of the study. Along with clinical improvement in joint swelling and tenderness, objective laboratory features of active inflammation such as hematocrit, erythrocyte sedimentation rate, platelet count, and IgM rheumatoid factor level showed favorable changes. No serious toxicity occurred. This experience complements reports of other studies of minocycline for treating rheumatoid arthritis [19, 20, 22]. These data contrast with the negative results of a small trial of rheumatoid arthritis in which 250 mg of tetracycline daily was used [5]. However, minocycline is more potent and more readily absorbed and has a better tissue distribution than tetracycline hydrochloride [45].

The proportion of patients in the placebo group who improved was twice that anticipated in the study design. Several factors may have contributed to this increased response and may have minimized the differences in outcomes between groups at 48 weeks in an intention-to-treat analysis. Clinic coordinators maintained frequent contact with patients and reinforced study compliance. We provided NSAIDs to all patients as an incentive to maintain optimal and stable background therapy for the duration of the trial. In addition, patients receiving placebo had more protocol-violating interventions during the trial, such as injections of intra-articular corticosteroids and increases in prednisone or NSAID therapy. Finally, more patients receiving placebo than those treated with minocycline who discontinued therapy with the study medication began receiving a disease-modifying antirheumatic drug before their week 48 study visit. However, all available study patients were evaluated at 48 weeks, and their data were analyzed according to their original treatment group assignments.

In our study, adverse drug reactions clearly attributable to minocycline were both infrequent and mild. Patient-reported problems previously believed to be specifically related to the administration of tetracycline and its derivatives, such as vestibular dysfunction [19], occurred with similar frequency in the two treatment groups. Only 3 of the 109 patients receiving minocycline reported dizziness as a reason for stopping medication. This profile contrasts with the experience of others [20, 22], who reported treatment-limiting gastrointestinal and vestibular symptoms in their patients with rheumatoid arthritis who received minocycline. Consistent with the clear dose response for vestibular symptoms shown in patients receiving preventive minocycline therapy for meningitis [46], a possible reason for the low frequency of adverse effects in our minocycline trial is that investigators were allowed to adjust the dosage of the study medication.

Patients in the minocycline trial can be compared with patients in similar clinical trials of therapies for rheumatoid arthritis [22-24, 47-52]. Disease duration was shorter than in the study of minocycline by Kloppenburg and colleagues [22] but longer than in several studies of other antirheumatic agents [24, 48-51]. Our patients had a lower erythrocyte sedimentation rate at baseline [22, 47, 52], and fewer patients were positive for IgM rheumatoid factor when compared with patients in other studies [48, 49, 51]. Sixty-eight percent of our patients had erosive disease compared with 80% in a CSSRD trial [52].

We observed significant differences (as defined in the Methods section) favoring minocycline over placebo in the primary outcome measures, in improvement in joint swelling and tenderness, and in the laboratory measurements. For some outcomes, such as the patient and physician global assessments, we observed differences in favor of minocycline that did not meet our criteria for statistical significance. These results were similar to the observations of Kloppenburg and colleagues [22] and the investigators of the CSSRD sulfasalazine study [50].

We cannot completely explain the failure to find larger differences in the global assessments and MHAQ. We had power of 0.83 and 0.77 to detect effects of the size observed in Weinblatt and colleagues' trial [47] comparing methotrexate with auranofin for the physician and patient global assessments, respectively, using a critical level of 0.005. In Weinblatt and colleagues' trial [47], the methotrexate group showed larger mean differences in the physician and patient global assessments than we observed in our minocycline group. However, the methotrexate differences would not have been considered significant had there been an adjustment for multiple comparisons. In our trial, pain appeared to be well controlled by NSAIDs in both groups, and multitrait analyses [53] suggest that the two global outcomes represent another assessment of the patient's level of pain. By controlling pain, it is also possible that level of activity as measured by the MHAQ would also improve. Thus, the background effects of NSAIDs may have made it difficult to find differences in our 48-week trial. Given the effect of minocycline on disease activity as shown by the differences in laboratory measurements and in the primary outcomes, it is possible that over a longer period of observation, larger differences in physician or patient global assessments could have emerged.

Whether the antirheumatic activity of minocycline is mediated by its antimicrobial, anti-inflammatory, or immunomodulatory properties remains to be determined. Although the efficacy of minocycline became evident at 12 weeks, improvement in patients treated with minocycline continued through the end of the trial; this time lag by itself does not rule out the possibility that the antirheumatic properties of this compound are caused by its antimicrobial activity. In addition, patients receiving minocycline experienced beneficial changes in laboratory measurements such as hematocrit, erythrocyte sedimentation rate, platelet count, and IgM rheumatoid factor level. Although the magnitude of the median differences in Table 6 is small for some laboratory measurements, the change in median values in the minocycline group resulted in more patients in the minocycline group having values within the normal ranges at 48 weeks than patients in the placebo group. The consistency of these laboratory results suggests that other well-documented nonantibiotic properties of these compounds, as shown both in vitro and in experimental animals [6-18], could explain the antirheumatic effect of minocycline.

In conclusion, we showed that minocycline is a safe, effective treatment for patients with mild to moderate rheumatoid arthritis. The observed adverse effects were mild to infrequent. On average, all outcome measures in the minocycline group improved from baseline to 48 weeks. For patients in the placebo group, hematocrit and IgM rheumatoid factor level on average worsened. The proportion of patients in the minocycline group whose joint counts improved by 50% or more continued to increase through week 48 of the trial, whereas the corresponding proportion of patients receiving placebo appeared to reach a plateau by week 24. The long-term effects of minocycline, its relative potency and safety in relation to disease-modifying antirheumatic drugs, and its mechanisms of action remain to be determined.

Appendix

The following are members of the MIRA Trial Group: The Beth Israel Hospital, Boston, Massachusetts: D. Trentham, R. Dynesius-Trentham, K. Sewell, F. Kantrowitz, G. Craven, M. Finell; State University of New York Health Sciences Center, Brooklyn, New York: D. Kaplan, R. Neuner, K. Orloff, C. Mignone, I. Caicedo, I. Kansariwala, S. Kolaninski, E. Ginzler, I. Antoniadis, R. Joks, S. Bakels; The University of Alabama at Birmingham, Birmingham, Alabama: G. Alarcon, W. Blackburn, Jr., S. Baum, I. Mikhail, R. Ayers, N. Abraham, M. Sehn; The University of Utah Medical Center, Salt Lake City, Utah: D. Clegg, T. McAllister, C. Jackson, M. Lundberg, P. Knibbe, J. Williams, G. Cannon, D. Stromquist, J. Ward, J. Huan; Henry Ford Hospital, Detroit, Michigan: J. Leisen, C. Sherlitz, M. Lubetsky, B. Segal, J. Sigler, L. Lasichak; The University of Vermont College of Medicine, Burlington, Vermont: S. Cooper, M. Merchant, L. Buckley, E. Leib, K. Cartularo, R. Budd, M. Davitt, K. Clark; Case Western Reserve University, The Henry Ford Health Sciences Center, Detroit, Michigan: B. Tilley, S. Fowler, H. Duncan, G. Bluhm, S. Fagan, S. Kardasz, M. Rusinova, N. Maddy, H. Boldarini, D. Whitted; National Institute of Arthritis and Musculoskeletal and Skin Diseases, Bethesda, Maryland: L. Shulman, S. Heyse, S. Pillemer, M. Tuttleman, S. Shores, E. Webster-Cissel; Tifton Medical Center, Tifton, Georgia: J. Sharp. Data and Safety Monitoring Review Committee Members included the following: The Medical Center at Princeton, New Jersey: R. Pinals; The New York Hospital-Cornell Medical Center New York, New York: M. Reidenberg; Fred Hutchinson Cancer Research Center, Seattle, Washington: S. Green; Howard University, Washington, D.C.: W. Greaves; National Eye Institute, Bethesda, Maryland: D. Everett; Georgetown University Medical Center: Washington, D.C.: C. Taylor.