Does persistence to methotrexate treatment in early rheumatoid arthritis have a familial component?

In rheumatoid arthritis (RA), early and effective treatment reduces inflammation, prevents joint damage, and improves quality of life [1]. For the typical newly diagnosed patient, most treatment guidelines currently suggest methotrexate (MTX) in DMARD monotherapy as first-line treatment [2, 3]. However, this approach does not take the heterogeneity of RA and treatment responsiveness into account; at 3 months, only ~30% of patients initiating MTX in DMARD monotherapy have achieved a good response [4, 5], and only two-thirds of patients remain on this regimen at 1 year [6]. As treatment guidelines suggest a treat-to-target approach, those unable to reach the target require treatment with other DMARDs, or combination therapy [2, 3]. Thus, early-stage identification of patients with a low chance of remaining on initial treatment (here: MTX in DMARD monotherapy) is of utmost importance as these individuals should be offered alternative therapies up-front.

Previous studies have identified several factors, genetic as well as clinical, that associate with primary response to MTX, although they have generally either been weak predictors or predictors that call for replication [5, 7‐10]. Most of these studies have used disease activity scores and their components, which often suffer from missing data, as treatment outcome. Few studies have studied treatment persistence as the outcome, which, in a treat-to-target paradigm should serve as a proxy for favorable tolerance and sustained treatment response. Using persistence as the outcome measure, a previous study from our group found that family history of RA, in itself, was unable to predict persistence to MTX in DMARD monotherapy at 3 and 6 months, though that study did not take family history of MTX treatment response into account [11].

Genetic factors contributing to MTX treatment response have been extensively studied with multiple genetic variants found to be associated with MTX efficacy [8, 9]. Studies have further indicated that models including genetic factors predict MTX treatment response better than those relying solely on clinical factors [12, 13]. This would imply the existence of a non-negligible genetic component and, per extension, a familial component to MTX treatment response. Whereas the existence of a familial component in the etiology of RA has been well documented [14‐16], evidence of a familial component within aspects of the clinical presentation of the disease, such as disease severity and treatment response, has so far been limited [11, 17, 18].

Evidence of a familial aggregation would potentially allow for early identification of patients who are less likely to respond to and remain well on MTX through assessment of their family history of this response. Furthermore, in clinical practice, questions on the information contained within a family history of a certain treatment outcome are not infrequent, yet the evidence to back up the response to such questions is scarce.

In this study, we aimed to assess whether a familial component contributes to the persistence to MTX used as DMARD monotherapy in early RA. We aimed to do this by testing whether treatment persistence aggregates within families of first-degree relatives concordant for RA and treatment with MTX and to quantify its magnitude by estimating the corresponding heritability.

We identified 347 unique individuals with at least one first-degree relative concordant for early RA and treatment with MTX in DMARD monotherapy. Together, these 347 individuals constituted 354 distinct pairs of first-degree relatives, making up the study cohort for the main analysis (Fig. 1). The majority of relative pairs were siblings (44%), followed by equal shares of offspring (28%) and parents (18% mothers and 10% fathers). Mean age at treatment start was 58 years (IQR: 47–69), and 246 (71%) of all individuals were female.

Among the 347 unique individuals, 231 (67%) were still on MTX in DMARD monotherapy at 1 year, and 175 (50%) at 3 years. Earlier age at treatment start and a greater fraction of female patients were observed among the non-persistent individuals, but the groups had similar proportions with seropositive RA and a similar number of identified first-degree relatives (Table 1). Among the 354 pairs of first-degree relatives, in 158 (45%), both individuals were persistent at 1 year, with 106 (30%) pairs wherein both individuals were persistent at 3 years.

Familial persistence was not associated with persistence at 1 year (RR=1.02, 95% CI 0.87–1.20) but a significant association with persistence was observed at 3 years (RR=1.41, 95% CI 1.14–1.74). Narrow-sense heritability was estimated to be 0.08 (95% CI 0*–0.43) for persistence at 1 year, and 0.58 (95% CI 0.27–0.89) at 3 years.

We here present the first investigation of the familial aggregation and heritability of persistence to MTX in DMARD monotherapy in early RA. We did not observe any statistically significant familial aggregation nor heritability during the first year of treatment. However, at 3 years, both estimates were significant, indicating that a familial component, including a possible genetic element, might be present.

We chose to perform a family-based study as this takes the full genetic component of individuals into account (as an extension to the genetic component captured by the markers included in a genome-wide association study). Using persistence as a treatment outcome allows us to circumvent the problem of dependence on the availability of clinical visits with disease activity measures, while still enabling analyses of both short- and long-term outcomes. Previous studies have mostly focused on primary treatment response after 3 or 6 months. Here, we assessed persistence at 1 and 3 years; at 1 year, we capture also those with a slow(-er) response to the drug (compared to assessment at 3 months) and after 3 years we should retain only a minimal number of false-positive responders.

Previous studies have implicated a broad set of genetic variants associated with MTX response [8, 9], as well as adverse events [36], though most variants found through candidate gene approaches have been only weakly associated, or call for replication [37]. Studies have also indicated that models including genetic data predict MTX treatment response better than those relying solely on clinical factors [12, 13]. While the presence of associated genetic variants is an indicator of an existing genetic component, their individual influence on the phenotype is often minor for complex traits. This is commonly taken as a sign of a polygenic trait, meaning that an analysis of genetic variants on an aggregated level, such as through the heritability, may be more informative of the magnitude of the genetic component underlying the phenotype.

When estimating heritability with tetrachoric correlations, an assumption is made that the only similarity between first-degree relatives is due to genetics, thus assuming that potential environmental influence is minor and negligible. Although this assumption is (most likely) false, the magnitude of the individual contributions of genetic and environmental components require further modeling, something the sample size in our study did not allow. Nevertheless, heritability estimates observed with tetrachoric correlations can be regarded as an upper bound of the additive genetic contribution. Setting the precision of the estimates aside, in our sensitivity cohort including only siblings, we observed point estimates that in general were larger than their counterparts for all first-degree relatives. As siblings share an environment during their upbringing, this indicates that not only genetics, but also a shared environmental familial component could have an impact on persistence to treatment.

The present study extends our previous findings in which we investigated familial aggregation of EULAR response to treatment with MTX in monotherapy and tumor necrosis factor inhibitors (TNFi) [11]. Due to small sample sizes, that study could not investigate familial aggregation within MTX-treated individuals, but found a significant familial component related to TNFi discontinuation after 12 months. In general, in RA, larger heritability estimates have been observed for treatment response with TNFi than what is observed in this study for MTX [38‐40], something that could be indicative of different treatment response phenotypes for different DMARDs. Studies have also found that different RA treatment outcomes vary highly in heritability estimates and that they in general correlate poorly within outcomes assessed at a single time-point [41], making singular point estimates difficult to compare across treatment type, period and outcome.

Our study has certain limitations. First, as persistence status was assessed based on drugs from the prescribed drugs register, some individuals might be incorrectly classified as they may have chosen to not follow the treatment protocol prescribed by their rheumatologist. However, due to the regular follow-up visits for early RA patients, such discontinuation should generally be captured by the treating rheumatologist leading to only a small number of such patients and a negligible bias. Secondly, as our tetrachoric correlation estimates of the heritabilities could not account for patients with multiple first-degree relatives, the standard errors will be biased leading towards more narrow confidence intervals than expected, had dependencies between observations been accounted for. Despite this, the number of individuals with multiple first-degree relatives was low and their influence on the estimates should be minor. Third, the sample size was a limitation to our study which lead to reduced precision within our estimates. This primarily affected the possibility of analysis of sub-cohorts, leading to sample sizes that were unlikely to produce meaningful inference, such as for analysis of the reason for MTX non-persistence. It should, however, be noted, as previously mentioned, that the cohort used in the study still contains virtually all individuals in Sweden fulfilling the study inclusion criteria. Obtaining a larger sample size using the same unbiased cohort selection is therefore difficult.

Our study has multiple strengths. Our outcome measure of persistence had very low missingness, compared with many outcome studies using composite disease activity measures such as DAS28. This was evidenced in our study by the major decrease in sample size between the primary analysis of persistence and the exploratory analysis of treatment response in the same cohort. Furthermore, the extensive follow-up enabled by the persistence outcome allows for additional capturing of patients with a slower response to MTX, who might have been missed when considering a 6-month follow-up. An analysis that requires complete case data might also suffer from potential inclusion bias, as patients may be assessed differently based on the severity of their disease, something that is avoided through the use of treatment persistence. Moreover, the use of register data from Swedish nationwide population-based registers with near-complete coverage allowed us to estimate valid and precise familial aggregation and heritability as our register-based approach covers virtually every first-degree relative pair in Sweden where both individuals were concordant for early RA and first-line treatment with MTX in DMARD monotherapy.

In contrast to short-term persistence, long-term persistence to treatment with MTX, here in the form of persistence at 3 years, aggregates within families, and the heritability of this phenotype is rather substantial. These findings imply the existence of a familial component within the long-term persistence to MTX. Whether any such familial component is reflective of characteristics of the underlying RA disease or determinants for MTX response in itself (that is not mediated via primary response) will require further investigation.