This cohort provides an insight into the “real world” response to MTX within an unselected inception cohort of patients with IP starting MTX as their first DMARD. Although the total numbers starting MTX as their first DMARD (with clinical data within 3 months of drug start) was relatively small this analysis provides additional evidence that MTX is effective and generally well tolerated in an unselected IP cohort, with 75% of patients remaining on monotherapy at 1 year and 61% at 2 years. Within this cohort, standard clinical and laboratory variables collected at first visit were poorly predictive of treatment outcome. Interestingly, the most strongly associated predictor of MTX inefficacy at both 1 and 2 years was possession of the SE, whereas other variables, such as CRP, DAS28 and HAQ, did not appear to be associated with inefficacy. The majority of patients stopping for adverse events did so within the first 12 months, within this group higher HAQ score, higher TJC and lower CRP were associated with stopping treatment for adverse events. However, for neither outcome were these variables, even when used in combination, sufficiently predictive of outcome to be of clinical use.
Studies suggest that high levels of inflammation (as measured by erythrocyte sedimentation rate, CRP or joint counts) predict a poor prognosis in terms of radiological damage.13
Our results suggest that disease activity does not influence treatment response in either direction. It could be hypothesised that the patients with the least active disease will do better because of their better underlying natural history or conversely those who have the most active disease have the greatest potential for showing a response.
Increased disability, as measured by the HAQ score, was predictive for stopping treatment for adverse events, although not for inefficacy. The explanation for this finding is unclear. In other studies,7
higher levels of disability (as measured by the Steinbrocker functional class) were associated with a poorer response to treatment, which may be explained on the basis that chronic severe disease is less responsive to treatment, The association in the current study seen with adverse events may be explained by a high HAQ score being a marker for other psychosocial factors that may increase adverse events such as depression or altered illness beliefs. Previous studies have shown that the HAQ score, although accepted as one of the gold standard measures of disability in RA, is also associated with pain,14
depression and disease activity.15
In a meta-analysis, Anderson et al7
found that female patients had a poorer response to treatment. Although within the NOAR cohort, female patients were more likely to stop treatment for adverse events, the difference was not statistically significant. Other studies9
suggest that male patients are more likely to respond to MTX, although whether this relates to gender differences in MTX clearance or other factors remains unclear.
Interestingly, the strongest factor demonstrated was the association between being positive for the HLA DRB1 SE and stopping MTX for inefficacy. Small numbers limited the ability to be able to detect a dose effect, or the relative importance of different SE alleles. No association was observed between stopping MTX for adverse events and SE status. Other studies support a link between SE and MTX efficacy; O’Dell et al16
found that patients who were SE positive were less likely to respond to MTX monotherapy compared with combination treatment with MTX, sulfasalazine and hydroxychloroquine, whereas patients who were SE negative did equally well regardless of treatment allocation. Our results are in contrast to those of Criswell et al17
who found that patients who were SE positive were more
likely to respond to MTX. Further studies are required to reconcile these apparently contradictory results.
The strength of this study is that we were able to recruit an unselected cohort of patients with IP and so the results are generalisable to other unselected inflammatory arthritis cohorts. The treatment of early IP is difficult because most of the evidence base relates to patients who satisfy the American College of Rheumatology (ACR) criteria. Yet it is increasingly recognised that treatment should be started as early as possible before the ACR criteria have been satisfied, with the PROMPT (Probable rheumatoid arthritis: Methotrexate versus Placebo Treatment) study showing benefit of early introduction of MTX in patients with undifferentiated IP in terms of slowing the progression to “ACR criteria” RA.18
Thus it is a strength of this study that it adds to the evidence base of the efficacy of treatment in patients with undifferentiated IP. There are, however, a number of limitations. First, the measure of disease response was not based on a formal protocol but reflected physician opinion. Hence the decision to stop treatment because of inefficacy was not standardised between patients. Further, as the NOAR follow-up was based on anniversary since entry, we do not have the detailed clinical data on disease activity and drug dose; for example, at the time of stopping treatment to obtain objective data to support the decision to stop treatment. This also meant that a number of patients were excluded from the analysis either because they had already started on MTX, or did not start within 3 months of their clinical assessment. Furthermore, there are additional, unmeasured predictors that may influence treatment response in this cohort, and hence improve the fit of the prediction models. We did not have radiographic data available on all patients and so these could not be included in the analysis. As this was a primary care based cohort of early disease, with short symptom duration, it was assumed that the prevalence of erosive disease at baseline would be low and, therefore, x
-rays were not routinely taken on all patients at baseline.
Despite its limitations, this study highlights that clinical and laboratory factors per se, even in combination, are relatively poor at predicting treatment response to MTX at least as judged by physician decision to stop treatment. This suggests that, first, there is no simple decision rule that could guide which patients should and should not be commenced on MTX. Secondly, given that response to treatment is unlikely to be a truly random event, there must be a number of other biomarkers, for example, not measured in this study that may influence outcome of MTX treatment. There have been several studies suggesting that polymorphisms in some of the pathways involved in the metabolism of MTX may be useful in predicting drug outcome,19
with a recent pharmacogenetic model being used to predict drug outcome in a subset of patients.20
To date, the clinical utility of such a model has still to be fully established. Further attempts to refine our ability to predict treatment response may therefore need to incorporate both additional genetic biomarkers and psychosocial factors, such as compliance, patient education and illness beliefs, which may influence treatment response.
In summary, routinely gathered clinical and laboratory factors are poor at predicting outcome of treatment with MTX in patients with newly diagnosed inflammatory arthritis. Nevertheless, given the central importance of MTX in the management of RA further work is required to optimise our use of this treatment.