Characteristics of patients and treatments
We identified 178 incidences of DMARD re-employment in 163 patients. The median number of DMARDs between original and re-employed course was 1 (range 1–7). There were no patients who had stopped the original courses because of pregnancy or remission. In the 15 patients with two documented incidences of re-employment of the same DMARD (ten MTX courses and five courses of other DMARDs), we only used the first observed event for all analyses to preserve independency of observations. Of the remaining 163 events, 86 included methotrexate and 77 included other DMARDs (Table ). The group of 'other DMARDs' included mainly sulfasalazine (n = 27) and chloroquine (n = 25). Rheumatoid factor was present in 59% of all patients in the source data set, and was somewhat more frequent in patients in whom methotrexate or other DMARDs were re-employed (71% and 65%, respectively). By their very nature, the re-employed courses were instituted in late disease. However, the original methotrexate courses were also not applied in early but rather in established disease, mostly after previous DMARD failures. Among the 163 patients, the original courses had been started between 1985 and 2001 and terminated between 1986 and 2002; re-employed courses started between early 1990 and 2002 and were terminated between 1990 and 2004. The mean time interval between the end of original and the start of re-employed courses was 2.1 ± 1.8 years for methotrexate and 2.9 ± 3.1 years for other DMARDs. Concomitant low-dose glucocorticoid treatment was similar among original and re-employed courses (42% and 49%, respectively; P = 0.32).
Characteristics of patients and treatments
Baseline acute phase reactants were higher for original methotrexate courses than for re-employed courses of methotrexate, but this difference did not reach statistical significance (CRP, P = 0.42; ESR, P = 0.07). For other DMARDs, baseline CRP and ESR were also similar in both courses (CRP, P = 0.75; ESR, P = 0.86; Table ).
Reasons for termination of methotrexate courses
We analyzed all methotrexate pairs for which the re-employed course had already been terminated or was still ongoing after at least one year of therapy (n = 79). The maximum weekly dose was significantly lower with original courses (12.1 ± 4.8 mg/week, median 10.0 mg/week) than in re-employed courses (16.1 ± 6.0 mg/week, median 15.0 mg/week; P = 0.0001). The frequency of parenteral methotrexate application was under 10% among both original and re-employed courses.
Among these 79 patients, methotrexate was terminated for ineffectiveness in 51 patients in at least one of the two courses, whereas 19 patients (25.0%) had no evidence of inefficacy in either course. Twenty-three therapies (29.1%) were not ineffective upon re-employment but the original course had been ineffective, while for only nine patients (11.4%) was it vice versa (Table ). This shift toward a lower frequency of ineffective therapies during re-employment was significant (P = 0.02, by McNemar test).
Treatment terminations for ineffectiveness in original and re-employed methotrexate therapies
Among the 51 patients with inefficacy of the original methotrexate course, the drug was effectively re-employed in 23 (45.1%) patients (Table ). Although in the original courses in these 51 patients mean doses of methotrexate reached a maximum of 13.3 ± 5.1 mg/week (median 10 mg/week; first and third quartiles 7.5 and 15 mg/week, respectively) before the decision to terminate therapy because of inefficacy was made (Table ), methotrexate was increased to up to 18.1 ± 5.6 mg/week (median 15 mg/week; first and third quartiles 10 and 20 mg/week) in re-employed courses before termination for ineffectiveness (P < 0.001, by Wilcoxon test; Table ). Figure indicates that discontinuation rates for ineffectiveness of re-employed courses were lower if the methotrexate dose of the original courses had been low: if patients had originally been treated with 10 mg/week or less, then re-employment was ineffective in only about one-third of patients; if they had been treated with a dose greater than 17.5 mg/week in the original course, then the frequency of ineffectiveness upon re-employment was 75%. The effect of the original methotrexate dose on effectiveness of re-employed courses was statistically significant (P = 0.02, using Logistic regression). The same effect was observed in subgroups of patients whose re-employed MTX dose was higher than the original dose, and for those whose dose was lower or equal upon re-employment (Figure ).
Figure 1 Ineffectiveness of re-employed courses in relation to dose of original course. Bars show the association of treatment termination of re-employed courses for ineffectiveness with the dose of the original methotrexate (MTX) course (total numbers shown in (more ...)
We analyzed the same 79 pairs of methotrexate courses with respect to treatment termination for adverse events (Table ). In 16 (20.3%) of these courses adverse events limited treatment continuation originally but not during re-employment, whereas in 10 (12.7%) it was vice versa. There was no statistical evidence that the frequency of adverse events triggering termination of methotrexate was different between original and re-employed courses (P = 0.33, by McNemar test), although the methotrexate dose was higher in the 79 re-employed courses (see above). Better tolerance to higher methotrexate doses was probably due to the higher rates of folate substitution in re-employed courses compared with the original courses (35.4% and 24.1%, respectively; P = 0.03, by Wilcoxon test). Folate substitution has regularly been instituted since 1995 in original and since 1997 in re-employed courses. When compared with all other subgroups, patients with adverse events in both courses (n = 8) had the lowest mean methotrexate doses (Table ), suggesting that in a significant proportion of patients (about 25%) methotrexate is not tolerated even at very low doses. These patients are therefore possibly more likely to re-encounter adverse effects if methotrexate is re-employed.
Treatment terminations for adverse events in original and re-employed methotrexate therapies
In 5.1% of the original methotrexate courses and 1.3% of the re-employed courses reasons of discontinuation were unknown. Among re-employed courses ongoing at more than 12 months from initiation (which were not excluded in the above analyses; n = 22), reasons for termination of the original course of methotrexate was classifiable as due to ineffectiveness in 68.2% (n = 15) and as due to adverse events in 31.8% (n = 7). Among all original courses, treatment termination after 12 months occurred in 36 patients (42%). When the original course was terminated for inefficacy within the first 12 months of treatment (n = 22), inefficacy recurred in 59.1% (n = 13) of the re-employed course.
Reasons for termination of other DMARDs
For the group of pairs of all other DMARDs (n = 69), there was no significant difference between original and re-employed courses in the frequency of termination due to ineffectiveness (62.3% concordant; P = 0.56, by McNemar test) or adverse events (76.8% concordant; P = 0.45).
Among these non-methotrexate DMARDs, sulfasalazine was the most frequently used drug (n = 23). Four courses were excluded because they were currently ongoing for less than 12 months. The maximum stable dose of sulfasalazine was 2.0 ± 0.6 g/day in original courses and 2.1 ± 0.5 g/day in re-employed courses (P = 0.30, by Wilcoxon test). Among patients with ineffective original sulfasalazine therapies (n = 16), doses of re-employed SSZ courses were on average similar, specifically 2.1 ± 0.6 g/day before termination for ineffectiveness (n = 11) and 2.2 ± 0.5 g/day before termination of re-employed sulfasalazine for other reasons (n = 12).
Improvement in acute phase response
We compared the relative changes in CRP and ESR during the first year of treatment between original and re-employed courses with methotrexate. The improvements in both CRP and ESR throughout the first year were not significantly different between original and re-employed courses (P = 0.80 and P = 0.43, respectively; Wilcoxon test comparing area under the curves; Table ). Also, we did not find significant differences in CRP and ESR levels when we compared the subgroups of patients with ineffectiveness in the original and effectiveness in the re-employed course (CRP, P = 0.69; ESR, P = 0.18) or those experiencing ineffectiveness in both courses (CRP, P = 0.58; ESR, P = 0.89). In the heterogeneous group of other DMARDs, the re-employed courses were also not associated with significant changes in CRP or ESR (data not shown).
Changes in CRP and ESR with treatment
Retention of treatment
Kaplan-Meier estimates were used to analyze cumulative drug retention rates while accounting for the fact that some re-employed courses were not yet terminated at the time of evaluation ('censoring'). The overall retention rate was numerically slightly longer for re-employed methotrexate (24.3 ± 2.7 months) than for original methotrexate (21.7 ± 2.8 months), but this did not reach statistical significance (P = 0.31, by Log-rank test; Figure ). However, retention because of therapeutic effectiveness was significantly higher in the re-employed methotrexate group (45.6 ± 5.9 months) than in the original methotrexate group (30.3 ± 4.1 months; P = 0.01; Figure ), and retention because of safety was slightly higher with re-employed methotrexate courses (87.3 ± 8.6 months) than with original methotrexate courses (70.8 ± 6.2 months), although the latter finding was not statistically significant (P = 0.23; Figure ).
Figure 2 Cumulative retention rates of methotrexate therapies. Kaplan-Meier analysis of retention rates of methotrexate. The different panels show (a) the overall drug retention, (b) retention of drug effectiveness, and (c) retention of drug safety for original (more ...)
Retention rates for therapies might partly be a function of differences in methotrexate dose and the choice of alternatives at the time when a change in therapy is considered [34
]. In fact, 54.7% (n
= 30) of patients were given either biological agents or leflunomide immediately after the re-employed course, whereas this was the case for only 11.3% of original courses (P
< 0.001). We therefore assessed the independent effect of re-employment on methotrexate retention rates adjusting for methotrexate dose and the year of DMARD prescription (dichotomized at 1999, when the first new DMARDs became available). There was an independent significant benefit of re-employment (using the original therapy as the referent), with a hazard ratio (HR) of 0.64 (95% confidence interval [CI] 0.42–0.97; P
= 0.04) (Table ). There was a trend toward better methotrexate retention rates with increasing methotrexate dose (HR per dose increment of methotrexate: 0.97, 95% CI 0.93–1.00; P
= 0.06). As hypothesized, methotrexate courses prescribed in 1999 or later tended to exhibit a greater risk for termination compared with methotrexate courses prescribed in 1998 or earlier (HR 1.63, 95% CI 0.97–2.73; P
= 0.06) (Table ). Figure illustrates the predicted survival functions of original and re-employed courses of methotrexate in the absence of dose differences and independently of the time period in which it had been prescribed.
Cox regression models on DMARD retention rates
Figure 3 Adjusted cumulative retention rates of MTX therapies. Based on a Cox proportional hazards regression model, cumulative retention rates for original therapies (grey lines; oMTX) and re-employed therapies (black lines; rMTX) are shown adjusted for the use (more ...)