Exploratory analysis of the TREAT EARLIER trial: is the efficacy of temporary methotrexate treatment dose-dependent?

The TREAT EARLIER trial¹ showed sustained improvement in MRI-detected subclinical joint inflammation and related joint symptoms after a 1-year methotrexate course in patients with both clinically suspect arthralgia and subclinical joint inflammation. Discussions are ongoing regarding whether pharmacological treatment has a place in individuals at risk of rheumatoid arthritis and if advancing the current first-line rheumatoid arthritis treatment (methotrexate) for a limited period to the clinically suspect arthralgia phase is beneficial. Crucial to this discussion is whether patients with clinically suspect arthralgia require the same dose of methotrexate as that which is currently prescribed to patients with rheumatoid arthritis.² Therefore, we studied the relationship between methotrexate dose and treatment efficacy in patients with clinically suspect arthralgia who used methotrexate in the TREAT EARLIER trial.¹ Despite being commonly used in rheumatoid arthritis, the evidence on optimal dose of methotrexate is scarce, as illustrated by the literature overview presented in the appendix (pp 9–10). Two double-blind randomised dose-finding trials were done in patients with rheumatoid arthritis more than three decades ago, which included 56 patients (46 in the study by Furst and colleagues³ and ten in the study by Seideman⁴). On the basis of these data and general clinical experience in patients with rheumatoid arthritis, we hypothesised that in clinically suspect arthralgia there is also an association between methotrexate dose and effectiveness in improving subclinical joint inflammation, joint pain, functional impairment, and work-related impairment.

The hypothesis was assessed in a post-hoc analysis of the TREAT EARLIER trial¹ in which all participating patients with clinically suspect arthralgia were analysed according to the treatment group to which they were originally assigned (treatment group or control group). Trial intervention consisted of a single intramuscular methylprednisolone injection or corresponding placebo injection, followed by a 52-week course of methotrexate or placebo tablets. After 4 weeks of a dose increase of methotrexate (as outlined in appendix pp 2–3), a target dose of 25 mg per week was pursued in all participants. If this target dose was not tolerated, the dose was lowered by the rheumatologist according to best practice. Participants recorded the number of tablets they took each week in their study diary. We evaluated the mean dose of methotrexate over 45 weeks (excluding the dose-increase and tapering periods). Following the first year with study medication, participants were followed up for another year without study medication with visits occurring every 4 months. Using linear mixed models, the mean treatment effects over 2 years of different mean doses of methotrexate on MRI-detected joint inflammation, pain, physical functioning (as measured with the Health Assessment Questionnaire [HAQ]) and presenteeism at work (Work Productivity and Activity Impairment questionnaire), were compared with placebo (appendix pp 2–8). Although the treatment did not prevent rheumatoid arthritis development in the trial, for completeness we also evaluated the different doses for this endpoint using Cox regression.

Of the 119 participants in the treatment group, 68 (57%) took a mean methotrexate dose of 20 mg or more per week, 36 (30%) took a mean dose of 1 mg, and 15 (13%) took less than 10 mg (baseline characteristics shown on appendix p 11). Mean dose per m² of body surface area was 2·3 mg (SD 1·7) in the less than 10 mg group, 8·0 mg (1·3) in the 1 mg group, and 12·5 mg (1·8) in the 20 mg or more group. Dose reductions occurred mostly due to raised liver panels, gastrointestinal intolerance, or other side-effects (mostly general discomfort; appendix pp 12–14).¹ A higher dose was associated with a larger improvement in MRI-detected joint inflammation (improvement of 0·07 in inflammation score per mg [95% CI 0·05–0·10]; figure; appendix pp 15–16). A similar association was observed for pain (0·39 improvement per mg [0·22–0·57]), HAQ score (0·005 per mg [0·002–0·008]) and presenteeism (0·45 per mg [0·24–0·66]; figure; appendix pp 15–16). The association was also observed within the subgroups of patients who were anti-citrullinated-protein antibody positive and had a high (>70%) risk of progression to rheumatoid arthritis, but these results have to be interpreted with caution because of the small patient population (appendix p 17). Sensitivity analyses using alternative categories of doses that patients took for 80% of the treatment period or adjusting for baseline differences between the dose groups, showed similar results (appendix pp 18–19). Within all dose categories, methotrexate did not prevent persistent clinical arthritis, which is in line with the original Article (appendix p 20).¹

Figure. Effect of different methotrexate doses in patients with clinically suspect arthralgia and subclinical inflammation.

(A) Mean improvement in MRI-detected inflammation (i), pain (ii), HAQ (iii), and presenteeism (iv) across each dose group over 2 years. The placebo group was the reference group. Higher values on the y axis represent greater improvements in each outcome. (B) Change in MRI-detected inflammation (i), pain (ii), HAQ (iii), and presenteeism (iv) over 2 years in patients receiving different mean doses of methotrexate. Time was a categorical variable in this linear mixed model. Mean dose per m² of body surface area was 2·3 mg (SD 1·7) in the less than 10 mg group, 8·0 mg (1·3) in the 10–20 mg group, and 12·5 mg (1·8) in the 20 mg or more group. MRI-detected inflammation was defined as the sum of scores of osteitis, synovitis, and tenosynovitis. The difference in improvement of MRI-detected inflammation between the 20 mg or more group and the 10–20 mg group was not statistically significant (0·4 points in MRI-detected inflammation [–0·6 to 1·5]; p=0·40). HAQ=Health Assessment Questionnaire. See appendix pp 15–16 for further information.

Here we showed, for the first time, the differences between methotrexate doses in improving MRI-detected joint inflammation and related symptoms and impairments in patients with clinically suspect arthralgia. When using MRI as a sensitive measure to assess inflammation and treatment efficacy, improvements were observed in participants who took more than 10 mg per week. For pain, physical functioning, and presenteeism, doses of 20 mg or more were particularly associated with improvements. This difference could be because MRI is a more sensitive measure than pain and physical function. However, MRI-detected joint inflammation is less clinically relevant if it is not associated with an improvement in disease burden. Considering cutoffs for clinical relevance of changes in pain and HAQ as used in rheumatoid arthritis, only doses of 20 mg or more per week were associated with clinically relevant improvements.^5,6 Thus, different doses could suffice in the at-risk stage of clinically suspect arthralgia, depending on the pursued treatment goal: varying from disease modification by lowering subclinical joint inflammation to sustained improvement of pain or function and work-related impairments.

In rheumatoid arthritis, guidelines recommend methotrexate as first-line therapy and advise to rapidly escalate it to 25 mg per week.⁷ Our literature overview (appendix pp 9–10) shows that the evidence on optimal dose of methotrexate is scarce, which has also been addressed previously.⁸ Although our findings are especially relevant for individuals with clinically suspect arthralgia and subclinical joint inflammation, results of this study might also strengthen the sparse evidence on dose-effectiveness in rheumatoid arthritis.

This study is not an official dose-finding study because participants were not randomly assigned to receive different methotrexate doses. Although we use data from a randomised trial, the original study design was not optimised to assess dose-dependent efficacy; therefore, our results do not have the same level of validity as a randomised trial. Relatively few participants deviated from the target dose towards lower doses, leading to smaller groups and less statistical power, especially in the lower dose categories. Reassuringly, all effect sizes found in the current study show a dose-dependent efficacy, and statistical significance was reached for the higher doses. In addition, sensitivity analyses using alternative dose categories showed consistent findings. Because participants were not randomly assigned to receive different doses, confounding factors influencing both adherence and efficacy cannot be ruled out with certainty. It is reassuring that the sensitivity analysis adjusting for possible confounders revealed similar results. Nevertheless, our findings need to be validated, preferably in a randomised trial comparing different methotrexate doses.

The side-effects reported by each patient defined the dose they took during the study. Consequently, these data are not suitable for studying relationships between doses and side-effects. However, side-effects in the trial were fully in line with those known from other indications (appendix p 13).^1,9

The previous trials on methotrexate efficacy^3,4 and a retrospective analysis² on dose-dependent efficacy of methotrexate used dose information from medical files and did not take adherence into account (appendix pp 9–10). Here, we have considered patient reported adherence. This more closely resembles clinical practice compared with previous studies, and can therefore be regarded as a stength. However, other methods, such as pill counting, would be even more optimal to evaluate adherence than the currently used study diaries.

In conclusion, higher methotrexate doses were associated with larger sustained improvements in subclinical joint inflammation, related symptoms and functional impairments in patients with clinically suspect arthralgia with subclinical joint inflammation. If temporary treatment in the clinically suspect arthralgia phase is to be implemented in practice, the optimal methotrexate dose depends on the pursued treatment goal and shared-decision making on the balance between effectiveness and tolerability. 10 mg per week appears to be the minimal dose to achieve a biological effect as measured by reduction of subclinical joint inflammation, while at least 20 mg per week is necessary to improve symptoms and functioning.