Statistics from Altmetric.com
Significance of this study
What is already known about this subject?
Methotrexate (MTX) is an effective medication to induce and maintain remission in Crohn’s disease (CD) when given parenterally.
Literature from rheumatology suggest that the oral route is inferior to the subcutaneous/intramuscular IM route but not to an extent to preclude using MTX orally.
The effectiveness of the oral route has never been compared with the parenteral route in CD.
What are the new findings?
Our study did not find consistent superiority of delivering MTX via the subcutaneous versus the oral routes.
However, the oral route was associated with longer time to remission, lesser improvement in linear growth, and lower numerical rates of sustained steroid-free remission.
Therefore, the subcutaneous route is probably somewhat more effective than the oral route in paediatric CD.
How might it impact on clinical practice in the foreseeable future?
Our study suggests that MTX should be first commenced subcutaneously.
Since the inferiority of the oral route does not seem to be substantial, it is not unreasonable to consider switching children in complete remission to the oral route, while closely following inflammatory markers and growth.
Methotrexate (MTX) has proven to be effective in inducing1 and maintaining2 ,3 clinical remission in adult Crohn’s disease (CD). Initially, MTX was used in patients with IBD who did not respond to azathioprine or 6-mercaptopurine.4 However, because of data that suggest an increased risk of lymphoma with thiopurine use, MTX is now being increasingly used in children as a first-line immunomodulator. We and others have shown that the 1-year steroid-free remission rate in children treated with MTX following failure of or intolerance to thiopurines is ∼40%.4–9 Maintenance of remission with MTX may also result in improved height velocity,4 which has not been consistently shown with thiopurines.10 However, these studies and the aforementioned clinical trials in adults primarily included patients treated with parenteral MTX, which may deter patients, especially children, from accepting its use.
The comparison of oral versus parenteral route of administration of MTX has never been the primary aim of any study in IBD. Moreover, simple comparisons of these routes of administration in retrospective studies are flawed by confounding-by-indication bias. For instance, the parenteral route may be preferred in younger patients, those with more active disease, those requiring higher dosing or those whose adherence to therapy is questionable. This was evident in our previous multicentre paediatric study, in which patients treated orally (PO) had lower baseline Paediatric Crohn's Disease Activity Index (PCDAI) values than those treated with subcutaneous injections.4
In this multicentre retrospective cohort study, we aim to use a robust statistical method of propensity score (PS), to compare effectiveness and toxicity of PO versus subcutaneously (SC) administered MTX in paediatric CD. The PS method is a powerful statistical tool to control for confounding-by-indication bias in retrospective longitudinal cohorts, but requires large data sets. We therefore established the largest cohort to date of children with CD treated with MTX.
Study design and eligibility
This is a multicentre, retrospective cohort study of children with CD diagnosed using established criteria11 with at least 1 year follow-up in paediatric IBD centres in North America, Europe and Israel. Some of the data were used in previous publications, unrelated to the aim of this manuscript.4 ,8 The choice of oral or subcutaneous administration and dosing were dictated by the physicians’ discretion and centre preference. Prior treatment with thiopurines was allowed, but all patients were naïve to biological therapy. Children may have received MTX either PO (‘PO group’) or SC (‘SC group’) continuously or may have been switched from subcutaneous to oral administration during the 1st year (‘SC/PO group’), as has been common practice in one of the participating centres. Switching from oral to SC was considered treatment failure and not a separate treatment group.
The end point of primary interest was sustained steroid-free remission (SSFR) defined as PCDAI ≤10 and (where applicable) no fistula discharge at 6 months and 12 months without stopping MTX or requiring treatment escalation (eg, adding anti-TNF or need for surgery). Secondary end points included linear growth, need for treatment escalation, adverse events, time to remission with MTX and corticosteroid use.
Linear growth was assessed by its most sensitive measure, height velocity, during the year after starting MTX therapy. Anthropometric measurements are expressed as SD z-scores using age-matched and gender-matched reference standards.12 ,13 Height velocity data were standardised using the JMP-derived polynomial calculations based on centile data.12 ,14 Computations were performed according to bone age when growth or pubertal delay was present. Treatment escalation included addition of a new steroid course, biologicals or surgery within the 1st year. In the PO group, the definition was extended to also include the need to switch to the subcutaneous route. Recorded adverse events included nausea (categorised as none/mild and moderate/severe) and significant elevation of liver enzymes (at least twice the upper normal limit and requiring dose change or MTX cessation). Corticosteroids use during the year after MTX initiation was measured as cumulative daily dose of prednisone. Steroid dependency was defined stringently as at least 24 cumulative weeks of treatment during the year.
Values of growth and PCDAI were carried forward in those requiring to stop MTX or to escalate therapy to anti-TNF or surgery.
The following data were extracted from the medical records: demographic and phenotypical data, medical and surgical treatments in the year before and the year after MTX initiation, dosing (mg per body surface area), route of administration of MTX and reason for changing route, serial anthropometric data including bone age, serial liver enzymes determinations and complete blood counts, PCDAI at the time of initiating MTX and at 6 months and 12 months thereafter, global assessment of the response to MTX therapy and time to remission after MTX.
The PS method attempts to balance the treatment groups so as to reduce bias of treatment selection, thereby creating a quasi-randomised experiment. The PS is the conditional probability of receiving the treatment given the observed covariates (particularly pretreatment variables which may serve as confounders15–26). Two PSs were developed using the SAS LOGISTIC procedure; one conditioning the likelihood of allocation to one of the three treatment groups (PO, SC and SC/PO groups) and the other focusing only on the two more homogenous PO versus SC groups. A total of 23 pretreatment variables were included in both models (see online supplementary tables S1 and S2). Univariate analyses were first applied for each variable using PROC general linear model (GLM) to compare the groups for continuous variables, and logistic regression for binary variables. Next, the treatment groups were compared using the PS-weighted observations ((1/propensity)×(proportion of observations from the respective treatment)) demonstrating that the treatment groups were similar across all pretreatment variables (see online supplementary tables S1 and S2). Three variables were considered fundamental and thus included in all final analyses despite being balanced in the PS- adjusted database: disease duration, pre MTX disease activity (measured by the PCDAI) and the initial MTX dose. This uses the ‘doubly robust’ estimation in which some pretreatment variables used in the propensity are also included in the final weighted model to ensure comprehensive adjustment.27
In addition to using the PS method as a weighting factor in the models, the PS calculated for the two groups (ie, PO vs SC groups) was also used for individual matching.17 Empirical and theoretical research has demonstrated that matching on the PS eliminates a greater proportion of baseline differences between the treatment groups than does stratification.15 The matching analysis was repeated using different callipers,17 starting with a quarter of a SD of the logit of the PS, according to Rosenbaum and Rubin28 using the MatchIt R package.29 The larger the matching calliper, the more matches found, thereby increasing sample size at the expense of homogeneity. All pretreatment variables were tested for similarity between the treatment groups to ensure successful matching.
Post hoc power calculation of the PS-weighted primary model of SSFR was conducted by sample simulations with the same structure as the data (sample size, values of covariates and PSs) but with varying ORs.
Data are presented as means (±SD), or medians (IQR), as appropriate for the distribution normality. Paired categorical data were compared using the McNemar test and continuous data were compared using the paired Student's t test and the Sign test. Time-to-remission was analysed using Kaplan-Meier survival estimates, and strata were compared using log rank and the Fleming (0, 1) tests. All comparisons were made using two-sided significance level of p<0.05. Statistical analyses were performed using SAS V.9.2, ‘R’ software and SPSS V.15.0.
A total of 226 children were included, of whom 38 (17%) were commenced on oral MTX from the outset (ie, PO group), 98 (43%) started with subcutaneous and switched to oral during the 1st year (ie, PO/SC group), and 90 (40%) were treated with subcutaneous only (ie, the SC group). In total, 136 children (60%) were treated with oral MTX during at least part of the 1st year. In comparison to the SC or PO/SC groups, the PO cohort at baseline had lower disease activity, shorter disease duration, lower rate of prior thiopurine treatment and received a lower MTX dosage (table 1). These differences were successfully addressed below using the matching and PS-weighted methods. A total of 117 (52% of the entire cohort, 86% of those treated with oral at least part of the year) received oral MTX within the first 4 months of treatment (73 were switched from subcutaneous to oral exactly at 4 months as per the treatment protocol practiced in Edinburgh which provided 76 children to the total cohort; 62 of whom (82%) after attaining complete remission). Eighteen of the 136 children (13%) who were treated PO were switched to the subcutaneous route during the 1st year, due to lack of effect, of whom 15 had available long-term outcome data; 7 (47%) achieved complete remission after the switch, and 2 (13%) improved without remission (ie, non-response was noted in 40%). MTX dose was altered in 119 (53%) children of the entire cohort, within a median period of 100 (IQR 48–193) days from initiating MTX.
Unadjusted, 76 children of the entire cohort (34%) had SSFR at 1 year (see figure 1A for unadjusted subgroup rates). Treatment escalation or new steroid course was required during the 1st year in 91 (40%) children. Height velocity in the total cohort showed mild mean catch up growth (z=0.19±3), but was significantly higher in the 76 children who qualified as having SSFR (z=1.76±3). A height velocity z-score >0 indicate supernormal growth rate and this was apparent in 92 (41%) children (50 (66%) of the 76 children with SSFR). Fifty (22%) children were steroid dependent during the year post MTX. Any elevation of liver enzymes at least once during the year was noted in 89 children (39%), of whom 40 (18% of the entire cohort) required dose reduction (n=30, 13%) or treatment cessation (n=10, 5%) due to elevation at least twice the upper limit (ie, significant elevated liver enzymes (ELEs)). Nausea was reported in 101 (45%) children, of whom, 49 (22%) reported severe symptoms.
In an unadjusted analysis, there were no significant differences in the various outcomes between the PO and the SC groups (figure 1A). The SC/PO group seemed to have a higher SSFR rate and a higher adverse event rate but it should be emphasised that this group was recruited primarily from one site with likely different clinical practice. The outcomes were therefore compared using the PS method. Of the 23 pretreatment variables 11 and 6 were initially significantly different across the treatment groups, in the three-arm and two-arm comparisons, respectively (see online supplementary tables S1 and S2). In the three-arm comparisons, none remained significant after adjustment, indicating that the PS modelling managed to balance well the treatment groups (see online supplementary table S1).
In the two-arm comparisons, only disease duration remained significant (see online supplementary table S2), and thus was included in all final models.
Sustained steroid-free remission
Using the propensity-weighted ‘doubly robust’ logistic model, no statistically significant differences were found between the PO and the SC groups in the SSFR rate (OR=1.72 (95% CI 0.5 to 5.8) in favour of the SC group; p=0.39). It must be emphasised that the model was underpowered to show statistical significance at OR of 1.7 (power of 80% would have been achieved with an OR of ∼5). However, the SC/PO group was superior to the SC group (OR=0.29 (0.15 to 0.58); p=0.0004) and the PO group (OR=5.8 (1.7 to 19.6); p=0.004).
Using the propensity-weighted linear regression model, the PO group showed significantly lower height velocity in the year after commencing MTX compared with the SC group (p=0.006) and the SC/PO group (p=0.0004).
Need for treatment escalation
The propensity-weighted model showed no differences among the three groups in the need to escalate therapy to biologicals, surgery, new steroid course or switch to parenteral MTX route if treated PO (SC vs PO, OR=0.51 (95% CI 0.2 to 1.6, p=0.24); SC vs SC/PO 1.2 (0.6 to 2.3, p=0.58); PO vs SC/PO 0.4 (0.1 to 1.3, p=0.13)).
There were no differences in the rate of significant ELE between the SC and the PO groups (OR=0.61 (95% CI 0.1 to 3.7; p=0.59)). However, the SC/PO group had significantly higher rates as compared with the SC group (OR=0.14 (0.05 to 0.4; p=0.0003)) and marginally compared with the PO group (OR=4.4 (0.9 to 23; p=0.08)).
Despite the slightly higher numerical rate of severe nausea in the SC/PO group, the propensity-weighted model showed no statistically significant differences between the groups (SC vs PO, OR=0.9 (95% CI 0.3 to 3, p=0.85); SC vs SC/PO 0.7 (0.3 to 1.4, p=0.28); PO vs SC/PO (1.3 (0.4 to 4.3, p=0.65)).
In a propensity-weighted ordinal regression, cumulative steroid dose during the year after commencing MTX (divided into five centiles), was not different among the three groups (SC vs PO, OR=0.66 (95% CI 0.3 to 1.7, p=0.37); SC vs SC/PO (1.1 (0.64 to 1.9, p=0.73)); PO vs SC/PO (0.6 (0.3 to 1.5, p=0.27)). There was also no significant difference between the PO and the SC groups in time to remission when considering the entire year after MTX (log rank test p=0.23). However, when using the Fleming (0, 1) test that gives more weight on early events, time to remission was shorter in the SC group (p=0.036) alluding to slower effect of the oral route (figure 2).
Matching only the PO and the SC groups
Since the PO group had shorter disease duration and lower baseline disease activity (table 1) and since these variables are known predictors to influence treatment outcome in CD, we first compared the PO group with a random subset of the SC group, matched for disease duration and disease activity (median disease duration of the SC group 0.8 (0.54–1.45) years and mean PCDAI 23±14 points, similar to the PO group as tabulated in table 1). In this subgroup-matched analysis, SC administration had a very slightly better outcome which did not reach statistical significance (figure 1B).
Next, matching was performed using the PS, based on all 23 pretreatment variables described above. A total of 23 of the 38 children in the PO group were individually matched to 23 children from the SC group (totaling 46 children in this matched analysis) using a matching calliper of 0.75. All 23 pretreatment variables were balanced between the matched groups (figure 3). In this highly selective and homogenous subcohort, no differences were found in the SSFR rate (McNemar's χ2 test, p=1.0), need for treatment escalation (p=0.34) and height velocity (Wilcoxon signed test, p=0.56) between the PO and the SC groups. Repeating this analysis with 17 matched pairs (using a conservative matching calliper of 0.25) and repeating the analysis with all 38 children in the PO group liberally matched by the nearest neighbour with 38 children from the SC group, yielded similar insignificant results (data not shown).
This study is the first to systematically address the question whether MTX can be administered PO in CD. It is also the largest cohort ever assembled for MTX in paediatric CD, including 226 children. We have used robust statistical methods across several outcomes and subgroups which managed to successfully balance 23 possible confounders, approximating random allocation. For most outcomes and analyses we could not find a superiority of the SC group over the PO group, including SSFR rate, the need for treatment escalation, steroid use and adverse events. However, in the primary outcome of SSFR, the SC group was superior to the PO group with an OR of 1.7 but the model was underpowered to show statistical significance of this modest effect size. Moreover, height velocity was lower in the PO group in the PS-weighted analysis and it may be assumed from the survival analyses that the time to remission may be slower with the oral route. In some of the outcomes, the SC/PO group seemed unexpectedly superior to the PO and the SC groups with higher incidence of adverse events. It is plausible to assume that this stems from some unknown (and therefore uncontrolled) differences in local practice rather than real difference in treatment effect, since almost all patients in that group came from one centre. To account for this we performed the subgroup-matched analysis, using a separately calculated PS for the homogenous PO vs SC groups.
A 6 month double-blind randomised controlled trial of 375 patients with rheumatoid arthritis, showed that the subcutaneous route is more effective than the oral route in all measured outcomes. However, similar to our results, the effect size was only moderate, inferring that the oral route is also effective.30 In that study, switching from the oral to the subcutaneous route resulted in further 30% response rate versus approximately 50% of the 18 children in our study. Similarly, retrospective reports of patients with rheumatoid arthritis reported improved outcomes after switching from oral to subcutaneous MTX, regardless of whether the oral route failed due to adverse events (mainly nausea) or ineffectiveness.31 ,32 Rozin et al33 reported a high relapse rate of patients with rheumatoid arthritis after switching from the parenteral to the oral route.
Evidence-based guidelines in rheumatoid arthritis recommend starting with the oral route and switching to the parenteral route only in case of insufficient response. Two small uncontrolled studies in adult CD demonstrated some efficacy of oral MTX in maintaining remission.34 ,35 In another trial, 27 patients with CD treated with intravenous MTX for 3 months followed by oral MTX for 3 months showed similar remission rates as with azathioprine.36 In contrast, two other small IBD trials found similar remission rates between PO administered MTX and placebo.37 ,38 Two uncontrolled studies that allowed oral or parenteral administration of MTX (without randomisation) showed no difference in relapse rate but in the paediatric report, steroid sparing and growth improvement was noted only in the SC group, quite similar to the findings of our study.4 ,39
Reduced effectiveness of the oral route may be explained by the lower bioavailability of the oral versus the parenteral formulation40–42 with higher variability in the former,40 ,42–44 especially in high doses. In contrast, Stephens et al45 found similar MTX interpatient variability and similar bioavailability in 11 paediatric patients with IBD treated with oral and subcutaneous formulation. A more recent study in CD showed somewhat lower bioavailability of the oral route (86% bioavailability of the subcutaneous route).46
Of particular interest is the fact the rate of ELEs and nausea was similar between the groups, contrary to the notion that PO administered MTX is associated with more nausea.31
This study has several strengths. It is the first to systematically explore whether the route of MTX administration matters in CD using robust statistical methods to control for known biases. It is by far the largest cohort ever assembled for children with CD treated with MTX and it is a multicentre, multinational study. However, the retrospective nature of the study and our inability to exclude the possibility that unknown cofounding variables were missed are weaknesses. The PO group included milder patients as reflected by several basic parameters (table 1) and it is possible that despite our extensive efforts, the PS strategy failed to account for all of the confounding effect. It must also be acknowledged that the children treated PO were registered mainly from two centres, while the SC group was balanced across all sites. Another limitation of our study is that the impact of treatment on endoscopic disease activity (ie, ‘mucosal healing’) was not assessed.
In concluding our practice recommendation, we cannot ignore the consistent literature from rheumatology indicating superiority of the subcutaneous route, nor the inferior bioavailability demonstrated in almost all studies of oral administration, the longer time to remission identified in this study, the lesser improvement in linear growth, the OR of 1.7 in favour of the PO group in achieving SSFR (underpowered to reach significance) and the fact that half of patients with active disease on oral MTX in this study responded to the subcutaneous route. Taken together, we can cautiously conclude that the subcutaneous route is somewhat more effective than the oral route in paediatric CD, but still the oral route may be effective in many children and the difference in effect size is not large. It is therefore reasonable to consider switching to the oral route following initial period of subcutaneous MTX injections leading to complete remission. If switching, patients should be followed closely to ensure continued clinical and biological remission, and in particular, maintenance of normal height velocity.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.