To assess the potential benefits of methotrexate in patients with systemic lupus erythematosus (SLE).
To assess the potential benefits of methotrexate in patients with systemic lupus erythematosus (SLE).
A 12-month, double-blind, placebo-controlled trial of methotrexate with folic acid was conducted. Intent-to-treat analyses were performed with mixed linear models and α = 0.04 (96% confidence interval [96% CI]) to account for interim analysis of longitudinal data to assess the treatment effects on lupus disease activity and daily steroid dose across monthly measurements, and to test if the treatment effects depended on selected participant characteristics.
Of 215 participants screened, 94 were excluded, 35 declined, and 86 were randomized (methotrexate = 41, placebo = 45). The groups were balanced for demographic and disease characteristics. Antimalarial use was more frequent in the placebo group, which was adjusted for in multivariable analyses. Sixty participants (27 methotrexate, 33 placebo) completed the study and 26 terminated early. Among participants who had the same baseline prednisone dose, those taking methotrexate received, on average, 1.33 mg/day less prednisone during the trial period (96% CI 0.06, 2.72 mg/day; a 22% reduction of their average-during-trial daily dose) compared with those in the placebo group. For the primary measure of disease activity (revised Systemic Lupus Activity Measure), methotrexate use was also associated with a marginally significant reduction in the mean during-trial score of 0.86 units (96% CI 0.01, 1.71; P = 0.039). A significant interaction between treatment and baseline damage was found (P = 0.001).
Methotrexate conferred a significant advantage in participants with moderately active lupus by lowering daily prednisone dose and slightly decreasing lupus disease activity. As a therapeutic option in moderate SLE, methotrexate can be considered to be steroid sparing.
Methotrexate, an antifolate drug that leads to adenosine release and other antiinflammatory and immunosuppressive effects, is the most prescribed drug for rheumatoid arthritis (1). Results of prospective case series (2–8) and 1 randomized controlled trial (9) suggest that methotrexate may have disease-remitting and steroid-sparing effects in systemic lupus erythematosus (SLE). Our objective was to determine the efficacy and safety of methotrexate with folic acid in moderately active SLE in a 12-month, double-blind, randomized, parallel-group, placebo-controlled trial. Our primary hypothesis was that methotrexate would decrease disease activity and would have a steroid-sparing effect.
Participants from 14 centers of the Canadian Network for Improved Outcomes in SLE were recruited, and consented to participate, if they were 18 years of age or older, fulfilled the 1982 American College of Rheumatology classification criteria for SLE (10), and had at least moderately active disease at baseline, as defined by a revised Systemic Lupus Activity Measure (SLAM-R) (11) score of ≥8 (12), despite stable conventional therapy. Patients whose lupus manifestations required use of azathioprine or cyclophosphamide were excluded. The complete list of inclusion and exclusion criteria is shown in Table 1. See Appendix A for a list of members of the Canadian Network for Improved Outcomes in SLE who contributed participants to the study. Each center had research ethics board review approval to conduct the study.
|Men and women with a definite diagnosis of systemic lupus erythematosus|
|Provincial legal age of adult consent|
|Practice of a proven effective form of birth control|
|Have active lupus as defined by a total SLAM-R score ≥8|
|Have limited damage as defined by a SLICC-DI/ACR score ≤15|
|Must be on stable dose of NSAIDs, prednisone, or antimalarial drugs for minimum of 4 weeks prior to study enrollment|
|Must agree to sign an informed consent form|
|Previous hypersensitivity to methotrexate or folic acid|
|Total SLAM-R score <8|
|Total SLICC-DI/ACR score >15|
|Inability to comply with instructions or a history of medical noncompliance|
|Intraarticular or intramuscular steroids in the previous 4 weeks|
|Clinically significant acute or chronic liver disease (with the exception of autoimmune liver disease)|
|Chronic alcohol use|
|Insulin-requiring diabetes mellitus combined with morbid obesity|
|Significant renal failure|
|Interstitial lung disease|
|Prior use of methotrexate|
|Concomitant use of sulfa drugs or cytotoxic medications|
|NSAID use in the presence of renal failure or other contraindications|
|Other research trial|
|Human immunodeficiency virus infection|
|Recently diagnosed malignancy|
|Vitamin B12 deficiency|
|Taking azathioprine or cyclophosphamide in the previous 4 weeks|
|Lupus manifestation that requires use of azathioprine or cyclophosphamide such as membranoproliferative lupus nephritis, vasculitis, or severe neuropsychiatric lupus|
Participants were assigned to either a methotrexate or placebo group using a stratified blocked randomization implemented through a customized Fortran program. The 4 strata were defined by baseline age (≤50 versus >50 years) and lupus damage, as measured by the Systemic Lupus International Collaborating Clinics Damage Index (SDI; with cutoff 4 versus >4) (13, 14). In each stratum, blocked randomization with a random block size of 2, 4, or 6 was used to minimize imbalance between the size of the 2 groups while reducing the risk of unblinding. The pharmacist at the central coordinating center was the only person aware of the participant treatment allocation.
Participants received a standard escalating dose of either methotrexate (starting dosage 7.5 mg/week) or placebo. Both groups were administered 2.5 mg/day of folic acid, except on the day of the trial medication. The dosage of methotrexate or placebo could be increased or decreased at monthly visits by 2.5 mg/week, according to response or side effects, up to a maximum of 20 mg/week. For participants receiving prednisone, a suggested tapering steroid schedule involved decreasing the dosage by 5 mg/day every 2 weeks for dosages ≥30 mg/day, and by 2.5 mg/day every 2 weeks for lower dosages. However, the decision whether or not to taper and to follow the suggested tapering guidelines was left to the discretion of the clinician.
At study entry, we collected information on age, sex, disease duration, SLAM-R and Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) scores (11, 15), lupus damage on the SDI, and the physical component score (PCS) and mental component score (MCS) of the Short Form 36 (SF-36) (16).
Participants were followed monthly for 1 year after randomization. At each monthly visit, an independent blinded assessor scored the SLAM-R and SLEDAI while the clinical investigator recorded the adverse events. To keep the independent assessor blinded, a research assistant recorded the laboratory values of the SLAM-R and SLEDAI, and did not make attribution as to whether the laboratory results were due to disease activity or treatment. The SDI was scored at 6 and 12 months' followup, and the SF-36 was collected monthly. Current daily prednisone dose was documented at baseline and at every monthly visit. Baseline and monthly blood and urine tests were performed to screen for toxicity, and to complete the disease activity measures.
Adverse events were recorded monthly and classified as treatment related or not. Individual events were collapsed by a research assistant into the following categories: cardiovascular, neurologic, otorhinolaryngologic, endocrine/metabolic, gastrointestinal, genitourinary, hematologic, infectious, mucocutaneous, musculoskeletal, psychological, renal, and respiratory. Adverse events that resulted in change of treatment were identified. Both groups were compared for the frequency of all and specific events, and of any adverse events leading to treatment modification.
The trial was designed to detect the difference of 3 SLAM-R units between mean changes (posttreatment minus baseline) in the 2 trial arms, suggested by a poll of lupus experts as a minimum clinically relevant treatment benefit. Based on our earlier prospective study (17), the within-group SD of the SLAM-R change scores was assumed to be 4.24. In the final analyses, we used the corrected significance level of α = 0.04, chosen to account for the interim analysis, while keeping the overall Type I error rate at 0.05 (18). A 1-tailed t-test was used for sample size estimation, because we hypothesized a priori that methotrexate would be more effective than placebo. Under these assumptions, 75 participants were required to complete the trial in order to obtain the estimated power, in the final analysis, of 90% for a 1-tailed independent-groups t-test to detect the mean difference of 3 SLAM-R points (19). Assuming a 10% attrition rate, 84 subjects needed to be recruited for 75 to complete the trial.
Interim analysis was planned after 46 participants completed the study, which corresponded to the estimated power of 78% of a 1-tailed t-test at P ≤ 0.01 to detect a 4-point between-group difference in the mean SLAM-R change score (18).
Descriptive statistics were used to compare the baseline distributions of clinical and sociodemographic characteristics in the 2 trial arms. Main analyses of primary and secondary outcomes were carried out on an intent-to-treat basis. Primary analyses tested the average followup difference, across monthly measurements and between the 2 groups, in the mean value of the 2 main outcomes: SLE disease activity, measured by the SLAM-R, and daily prednisone dose. Because of its positively skewed distribution, daily prednisone dose was log transformed.
To account for missing data and for dependence of repeated outcome measures on the same subject, all analyses relied on the multivariable mixed linear model (19). In particular, in the analyses of longitudinal studies with missing data, mixed model yields likelihood-based estimates of treatment effect, which have been shown to avoid bias and underestimation of the variance induced by the last observation carried forward approach or other simple imputation methods (20). In mixed model analyses, the 12 postbaseline monthly measurements of a given outcome corresponded to the repeated measures of the dependent variable. The heterogeneous autoregressive order 1 structure of the covariance matrix between repeated measures was used to represent the pattern of within-subject correlations. This structure assumes that the repeated within-subject observations closer together in time are more correlated than those observations that are further apart (21). Each multivariable mixed model analysis tested the effect of the treatment (methotrexate versus placebo) on a particular outcome, while adjusting for the following a priori selected potential confounders: baseline value of the respective outcome, time of measurement (in months from the baseline), sex, age, baseline value of the SDI, and use of antimalarial drugs at baseline. To account for the interim analysis, the adjusted treatment effect was tested using the corrected P < 0.04 for the 2-tailed mixed model–based F test, and was reported with 96% confidence intervals (96% CIs).
In additional post hoc analyses, we assessed if the treatment effect varied depending on the patients' characteristics. Thus, we tested the 2-way interactions between the treatment and each of the aforementioned potential confounders, while adjusting for all variables included in the multivariable mixed model. To correct for multiple testing, while taking into account the lack of independence between tests of different interactions, we tested each interaction at the 0.01 2-tailed significance level. If an interaction with a specific covariate met the P ≤ 0.01 significance criterion, we stratified the mixed model analyses of a given outcome accordingly, and reported separate effects of treatment for each of the covariate strata. Finally, because, for example, the longitudinal pattern of changes in SLE activity may vary depending on the initial activity level, for each outcome we tested the interaction between the corresponding baseline value and time (20) and kept it in the final model if it met the P ≤ 0.01 criterion.
Secondary outcomes (SLEDAI and the SF-36 PCS and MCS) were analyzed using the same approach. The overall proportions of participants with adverse reactions requiring treatment discontinuation were compared using 2-tailed Fisher's exact test, at a 0.05 significance level.
A Drug and Safety Monitoring Board (DSMB) consisted of 4 members: 2 expert lupus rheumatologists with 1 as Chair, an ethicist, and a biostatistician (Appendix A). They reviewed adverse events regularly. The DSMB biostatistician conducted an a priori planned interim analysis of the primary outcome. Based on the results, the DSMB recommended continuing the trial.
A total of 215 persons with SLE were screened. Of these, 35 (16%) refused to participate and 94 (44%) did not meet inclusion criteria. A total of 86 participants were randomized, 41 to methotrexate and 45 to placebo (Figure 1). The 2 groups were similar on almost all baseline characteristics (Table 2), except for a higher proportion of antimalarial drug use in the placebo group (91% versus 66%), and this difference was adjusted for in all multivariable analyses. No patients were receiving immunosuppressive drugs at entry into the study. In each group, at the baseline visit, 22 participants were taking corticosteroids, but the mean initial dose was slightly higher in the methotrexate group (Table 2).
|Variable||Methotrexate (n = 41)||Placebo (n = 45)|
|Age at baseline, years||40.2 (34.9–48.5)||40.2 (34.0–48.2)|
|Duration of SLE at baseline, years||5.7 (1.6–14.7)||4.5 (1.7–8.8)|
|Number of ACR criteria||6 (5–7)||6 (5–7)|
|Sex (female)||37 (90.2)||41 (91.1)|
|Education (university and above)||6 (14.6)||10 (22.2)|
|White||32 (78.1)||34 (75.6)|
|SLAM-R (total)||11.0 (10.0–14.0)||12.0 (10.0–14.0)|
|SLAM-R organ systems†|
|Constitutional||40 (97.6)||42 (93.3)|
|Integument||39 (95.2)||35 (77.8)|
|Eye||1 (2.4)||1 (2.2)|
|Reticuloendothelial||9 (22.0)||14 (31.1)|
|Pulmonary||21 (51.2)||18 (40.0)|
|Cardiovascular||32 (78.1)||29 (64.4)|
|Gastrointestinal||10 (24.4)||8 (17.8)|
|Neuromotor||30 (73.2)||28 (62.2)|
|Musculoskeletal||38 (92.7)||43 (95.6)|
|Other (ad hoc)||9 (23.1)||2 (4.4)|
|ESR||23 (56.1)||23 (53.5)|
|Hematologic||25 (61.0)||35 (77.8)|
|Renal||20 (48.8)||19 (42.2)|
|SLEDAI (total)||10.0 (6.0–14.0)||10.0 (6.0–14.0)|
|Standardized PCS||31.3 (26.4–39.9)||32.9 (28.0–36.6)|
|Standardized MCS||43.5 (34.9–52.2)||38.7 (31.8–49.3)|
|SLICC-DI (total)||1.0 (0.0–2.0)||1.0 (0.0–2.0)|
|SLICC-DI ≥1||21 (51.2)||27 (60.0)|
|Dose of oral steroids (mg/day)||5.0 (0.0–10.0)||0.0 (0.0–10.0)|
|Mean ± SD||7.1 ± 9.7||4.9 ± 6.0|
|Proportion on steroids||22 (54)||22 (49)|
|Antimalarial medication||27 (65.9)||41 (91.1)|
Sixty participants (70%; 27 methotrexate, 33 placebo) completed the blinded phase of the study at 12 months. Twenty-six (30%) terminated early (14 methotrexate, 12 placebo) due to lupus flares (7 methotrexate, 10 placebo), nausea (4 methotrexate, 0 placebo), neutropenia (1 methotrexate, 0 placebo), and loss to followup (2 methotrexate, 2 placebo) (Figure 1). A clinical investigator recommended discontinuation of the treatment in 2 participants, 1 (placebo) with generalized ulcerations and severe dermatitis, and 1 (methotrexate) with severe nausea and vomiting postdose.
Due to early termination and a failure of individual participants to show up for some prescheduled monthly data, the outcome measurements were missing for ∼10% (46 of the 492) monthly visits in the 41 subjects taking methotrexate, and for 5% (27 of the 540) visits in the 45 subjects taking placebo. The distribution of the number of completed postrandomization visits was similar in the methotrexate group (median 12, interquartile range [IQR] 7–12) and placebo group (median 11, IQR 9–12).
Over the duration of the trial, 1,019 adverse events were recorded, with a mean ± SD of 8.7 ± 8.9 in 38 (93%) of 41 individuals in the methotrexate group and 7.2 ± 7.1 in 38 (87%) of 45 individuals in the placebo group (P = 0.48 by Fisher's exact test on the proportion of participants with at least 1 adverse event). There was no difference between the 2 treatment groups in the frequency of adverse events possibly attributable to treatment, with 78% of subjects in the methotrexate group and 73% of subjects in the placebo group reporting at least 1 such event (P = 0.63). Table 3 shows the adverse events attributed to treatment, overall and by organ systems. The methotrexate group had a higher risk of gastrointestinal (56.1% versus 33.3%; P = 0.05) and psychological (essentially mood disorder) adverse events (9.8% versus 0%; P = 0.05).
|Type of event||Methotrexate (n = 41)||Placebo (n = 45)||P†|
|All categories||32 (78.1)||33 (73.3)||0.63|
|Cardiovascular||0 (0.0)||0 (0.0)||NA|
|Central/peripheral nervous system||3 (7.3)||2 (4.4)||0.67|
|Ear, eye, nose, throat, larynx||1 (2.4)||0 (0.0)||0.48|
|Endocrine/metabolic||1 (2.4)||0 (0.0)||0.48|
|Gastrointestinal||23 (56.1)||15 (33.3)||0.05|
|Genitourinary||1 (2.4)||0 (0.0)||0.48|
|Hematologic||11 (26.8)||10 (22.2)||0.80|
|Infection||2 (4.9)||1 (2.2)||0.60|
|Mucocutaneous||13 (31.7)||21 (46.7)||0.19|
|Musculoskeletal||1 (2.4)||1 (2.2)||1.00|
|Psychological||4 (9.8)||0 (0.0)||0.05|
|Renal||0 (0.0)||0 (0.0)||NA|
|Respiratory||0 (0.0)||0 (0.0)||NA|
Twenty-five clinically important adverse events required modification of treatment in 16 participants and included mucocutaneous, gastrointestinal, and hematologic events. Of these, 9 persons (5 methotrexate, 4 placebo) had their drug withheld at least once during the study. Seven (17.1%) persons taking methotrexate had to decrease their dose due to gastrointestinal events compared with only 1 (2.2%) person in the placebo group (P = 0.02).
In the multivariable mixed models analyses of primary and secondary outcomes, there were no statistically significant changes over time in the mean of all trial participants' log-transformed corticosteroids dose (P = 0.56) or SLEDAI score (P = 0.14). In contrast, the mean SLAM-R score of SLE disease activity decreased significantly during the 12 months of followup (P = 0.006), while the mean scores of SF-36 physical and mental components showed significant increases over time (P = 0.0009 and P = 0.03, respectively). In contrast, there was no significant (P ≤ 0.01) interaction between treatment and time for any of the primary or secondary outcomes (data not shown), which suggested that the estimated differences between the 2 trial arms remained stable during the 12-month followup. Accordingly, for each outcome, Table 4 shows the adjusted mean during-trial difference, across all 12 monthly postbaseline measurements, between the mean values of a given outcome in the methotrexate group versus the placebo group, together with 96% CIs and a 2-tailed P value.
|Outcome||Mean during-trial difference†||96% CI‡||P (2-tailed)|
|Log (prednisone dose; mean relative difference in daily dose, in %)§||−22.3||−36.2, −5.4||0.010¶|
|SF-36 PCS||1.77||−0.31, 3.85||0.085|
|SF-36 MCS||2.78||0.13, 5.44||0.034¶|
The first row of Table 4 shows that methotrexate treatment was associated with a statistically significant reduction in the (log-transformed) corticosteroids dose (P = 0.01). Specifically, patients in the methotrexate group reduced their average-during-trial daily dose by ∼22% [1 − exp(−0.25) = 0.223] (96% CI 5.4, 36.2% reduction) compared with those in the placebo group who had the same initial dose and the same values of all potential confounders. To facilitate the interpretation of this steroid-sparing effect of methotrexate, we performed additional analyses of the untransformed dose, which showed an average decrease of ∼1.33 mg (96% CI −0.06, 2.72 mg reduction) in the mean value of the average-during-trial daily corticosteroids dose for patients taking methotrexate relative to those receiving placebo. Whereas the mean difference in daily dose was small, over 12 months of the trial an important decrease in the cumulative exposure to steroids was observed. Additional analyses revealed very significant interactions between baseline dose and time (P = 0.0002 and P < 0.0001 for the log-transformed and untransformed dose, respectively), with the corticosteroids dose increasing over time for patients with low initial dose but decreasing for those with baseline dose >3 mg (data not shown). When these interactions, which are likely due at least in part to regression to the mean, were added to the multivariable mixed models, the steroid-sparing effects of methotrexate remained statistically significant (P = 0.007 and P = 0.013 for the log-transformed and untransformed dose, respectively).
Table 5 provides more detailed, descriptive information on prednisone use, which was limited to the changes in its use or dose between the baseline and the 12-month visits. Among patients who were taking prednisone at baseline, only 2 (9.1%) of 22 in the methotrexate group and only 1 (4.6%) of 22 in the placebo group did not take prednisone at the last visit. However, only 1 (5.3%) of 19 initially prednisone-free methotrexate subjects took prednisone at 12 months. In contrast, 6 (26.1%) of 23 patients initially not on prednisone in the placebo group took this drug at 12 months. Moreover, as shown in Table 5, methotrexate patients were more likely to decrease their prednisone dose over the 12 months, while in the placebo group more patients increased their dose.
|Dose||Methotrexate (n = 41)||Placebo (n = 45)|
|No prednisone at baseline or 12 months||18 (44)||17 (37)|
|Prednisone at baseline but not at 12 months||2 (5)||1 (2)|
|No prednisone at baseline but taken at 12 months||1 (2)||6 (13)|
|Dose decreased from baseline to 12 months||10 (24)||6 (13)|
|Same dose||5 (12)||7 (16)|
|Dose increased||5 (12)||8 (18)|
Furthermore, methotrexate use was associated with a marginally statistically significant reduction in SLE disease activity, as measured by the SLAM-R score, the second primary outcome (Table 4). On average, across the 12 monthly assessments, the mean SLAM-R score in the methotrexate group was 0.86 points lower (96% CI 0.02, 1.71; P = 0.039) than the corresponding mean score among participants randomized to placebo who had the same baseline SLE disease activity.
Among the secondary outcomes, the small reduction in the mean SLEDAI score of disease activity in the methotrexate group relative to placebo was not statistically significant (mean-during-trial difference −1.04 points; 96% CI –2.59, 0.52; P = 0.174). The mean increase of 1.77 points (96% CI −0.31, 3.85) in the physical component of the SF-36 measure of quality of life associated with the methotrexate treatment was also not significant (P = 0.09).
In contrast, Table 4 shows that the adjusted effect of methotrexate on the mental component of quality of life did reach our corrected criterion for statistical significance. During the trial, patients taking methotrexate had, on average, MCS scores 2.78 points higher (96% CI 0.13, 5.44; P = 0.034) than the patients in the placebo group who had the same baseline MCS score.
Post hoc analyses identified only 1 statistically significant interaction involving the treatment: the decrease in the mean-over-time SLAM-R score associated with methotrexate use became less important as the baseline SDI score increased (P = 0.001 for the treatment-by-SDI interaction in the multivariable mixed model). In the subgroup analysis restricted to 38 participants with no damage at baseline (SDI score = 0), the mean-during-trial SLAM-R score in the 20 participants taking methotrexate was 1.41 points lower (mean adjusted difference −1.41; 96% CI −2.42, −0.39; P = 0.008) than in the 18 participants assigned to placebo. In contrast, among participants with baseline SDI scores >0, a small decrease in the mean SLAM-R score in the methotrexate versus the placebo group was statistically nonsignificant (mean-during-trial difference of −0.65; 96% CI −1.94, 0.64; P = 0.306).
In our study, methotrexate conferred a significant advantage in reducing time-average prednisone use, with a marginally significant decrease in lupus disease activity in participants with moderately active lupus. Methotrexate was generally well tolerated, except for some excess in mild to moderate gastrointestinal and psychological toxicity. In fact, methotrexate showed a marginal advantage over placebo in decreasing lupus activity measured by SLAM-R but not by SLEDAI. This may be explained in part by the higher responsiveness of SLAM-R over SLEDAI to detect change in lupus activity over time (22). During the trial, participants in the methotrexate group also had significantly better scores on the mental health component (MCS) of the SF-36 quality of life scale compared with participants randomized to placebo, who had the same baseline MCS scores.
Table 5 demonstrates that more participants in the placebo group (13%) than in the methotrexate group (2%) had to start prednisone during the study. Similarly, more participants in the methotrexate group (24%) than the placebo group (13%) were able to taper their dose of steroids. Among participants who had the same baseline prednisone dose, those on methotrexate received, on average, 1.33 mg/day less during the 12 months, and benefited from a 22% reduction of their average-during-trial daily dose compared with those in the placebo group who started on the same dose. This modest reduction in the daily dose translates into an important difference in the cumulative 12-month dose. Thus, as a therapeutic option in SLE, methotrexate can be considered to be steroid sparing. This is an important finding, as steroid toxicity is common and often leads to significant comorbidity such as Cushingoid features, weight gain, mood disorders, glucose intolerance, osteoporosis, and increased susceptibility to infection and cardiovascular disease (23).
Our results differ somewhat from those of previous studies (2–7, 24) with smaller sample sizes, shorter followups, and, with one exception, uncontrolled design. The only controlled study of methotrexate in SLE, by Carneiro and Sato, was a 6-month, randomized, prospective, double-blind controlled trial of 41 participants with moderate baseline lupus activity. In that study, a greater proportion of participants on methotrexate than on placebo were able to decrease their prednisone dose and reduced their posttreatment SLEDAI scores, relative to baseline (9). Those with arthritis and rash responded best to methotrexate. However, these results were limited by the small number of participants, the possibility of unblinding of the assessor through concurrent assessment of outcome and side effects, and the absence of a predefined primary outcome.
Our study has some limitations. Recruitment for a randomized clinical trial in lupus is challenging due to the complexity of disease manifestations and the fluctuating pattern of activity. We achieved our sample size requirements but we had to screen as many as 215 participants to randomize 86 (40%). This might have affected the generalizability of our results. Although steroid tapering guidelines were suggested, it was ethically impossible to impose such guidelines, and some physicians chose to modify the steroid dose according to their own clinical experience rather than according to our guidelines. We recorded this carefully in our data and used this information in our mixed linear model analysis of longitudinal data. As in all longitudinal studies, some of the participants discontinued the trial, mostly due to lupus flare (17 of 26 terminations), and/or failed to show up for some of the scheduled monthly visits. However, the missing data likely had only a minor impact on our results, because the distribution of the number of completed visits was similar in the 2 groups, and the mixed model approach accounts for missing-at-random data in longitudinal trials (20).
Of interest, using the SLAM-R, the primary measure of disease activity, we found a significant interaction between presence of damage at baseline and treatment. In post hoc subgroup analyses, for participants with no damage at baseline, there was a higher reduction in disease activity in the methotrexate group. This may suggest that the absence of damage may identify a subgroup of participants more likely to respond to methotrexate. Damage is a function of cumulative severity of disease activity since the diagnosis of lupus (13, 14).
We have demonstrated that use of methotrexate in patients with moderately active lupus allows the reduction of steroid dose without increasing lupus activity, especially in patients without damage.
Dr. Fortin had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Fortin, Lacaille, Smith.
Acquisition of data. Fortin, Ferland, Lacaille, Smith, Zummer.
Analysis and interpretation of data. Fortin, Abrahamowicz, Lacaille.
Manuscript preparation. Fortin, Abrahamowicz, Ferland, Lacaille, Smith.
Statistical analysis. Abrahamowicz.
Coinvestigators for the Canadian Network for Improved Outcomes in SLE (in addition to the authors) are as follows: Ann E. Clarke, MD, MS, Roxane Du Berger, MS, Sasha Bernatsky, MD, MS, PhD (McGill University Health Centre, Montreal, Quebec); John M. Esdaile, MD, MPH, Howard B. Stein, MD (The Arthritis Research Centre of Canada, Vancouver, British Columbia); Jean-Pierre Mathieu, MD, Line Duchesne, MD, Suzanne Mercille, MD, Pierre Dagenais, MD, PhD (Hôpital Maisonneuve-Rosemont, Montreal, Quebec); Janet E. Pope, MD, MPH (St. Joseph's Hospital, London, Ontario); Steven Edworthy, MD, Susan Barr, MD, Garry Morris, MD (Calgary Health Sciences Centre, Calgary, Alberta); Michael Starr, MD (St. Mary's Hospital, Montreal, Quebec); Vivian Bykerk, MD (Mount Sinai Hospital, Toronto, Ontario); Janice Canvin, MD, Hani S. El-Gabalawy, MD, Christine Peschken, MD, MS (Winnipeg Health Sciences Centre, Winnipeg, Manitoba); Alfred Cividino, MD (MacMaster-Cherooke Hospital, Hamilton, Ontario); Jean-Luc Senécal, MD, Jean-Pierre Raynauld, MD, Eric Rich, MD (Hôpital Notre-Dame, Montreal, Quebec); C. Kirk Osterland, MD, Carol A. Yeadon, MD, André Beaulieu, MD, Simon Carette, MD (Centre Hospitalier de L'Université de Laval, Quebec City, Quebec); Gilles Boire, MD, MSc (Centre Hospitalier de L'Univerisité de Sherbrooke, Sherbrooke, Quebec).
Research assistants: Rhondda Morrison, RN, MSN, Elaine S. Clark, Karen Rangno, RN (The Arthritis Centre of Canada, Vancouver, British Columbia); Paula Dale, RN (Ottawa General Hospital, Ottawa, Ontario); Deborah L. Fenlon, BScN, Wendy Curran, Lynda Bere, RN, LHSC (Victoria Campus, London, Ontario); Rosario Talavera, MD, Beverly Green, BN, BScN, Elisia Teixeira, RN, BN (Calgary Health Sciences Centre, Calgary, Alberta); Jackie O'Farrell, RN (Credit Valley Hospital, Mississauga, Ontario); Sharyn Wood, RN, Ann Huggard, RN (Winnipeg Health Sciences Centre, Winnipeg, Manitoba); Joy Zahavich, RN (MacMaster-Cherooke Hospital, Hamilton, Ontario); Mariette Prave, RN (Centre Hospitalier de L'Université de Laval, Quebec City, Quebec).
Members of the Drug and Safety Monitoring Committee: Matthew H. Liang, MD, MPH (Brigham and Women's Hospital, Boston, Massachusetts); Susan Manzi, MD, MPH (University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania); Norbert Gilmore, MD, PhD (McGill Centre for Medicine, Ethics and Law, Montreal, Quebec); Richard J. Cook, PhD (University of Waterloo, Ontario).