Patients were obtained from the Leiden Early Arthritis Clinic (EAC), which has been previously described . In short, the Leiden EAC is an inception cohort, including all patients presenting with recent-onset arthritis with a symptom duration ≤ 2 years. Research visits took place at baseline, after 4 months and 8 months, and annually thereafter. During these visits, joint counts were performed, disease activity scores calculated , laboratory measurements performed, and questionnaires filled out. Visits to the treating rheumatologists were more frequent, as often as it was found necessary.
Biomarker measurements were performed in consecutive RA patients included in the Leiden EAC between 2010 and 2015, representing the period in which early methotrexate (MTX) and treat-to-target strategies were common. RA was stringently defined by a clinical diagnosis of RA by an experienced rheumatologist plus fulfillment of the 1987 and/or 2010 criteria [12, 13]. Patients diagnosed with conditions other than RA (e.g., reactive arthritis/psoriatic arthritis/inflammatory osteoarthritis), or who had a high suspicion on these diagnoses, were excluded. Biomarker levels were measured in 312 RA patients. RA patients who did not use DMARDs during follow-up (n = 13), who had a delayed DMARD start (> 4 months after diagnosis) (n = 16), or who concomitantly participated in a clinical trial (n = 17), and therefore not routinely, were excluded. Consequently, 266 RA patients were selected as the primary study population, of whom 135 were ACPA-positive and 131 were ACPA-negative RA patients (Fig. S1). The baseline characteristics of the study population and RA patients excluded from the study population did not remarkably differ (S2).
Second, for the replication of findings, ACPA-negative RA patients, consecutively included in the Leiden EAC between 2007 and 2010, were studied (Fig. S1). The same criteria for the classification of RA and patient selection were applied as described above.
Treatment strategies in the Leiden EAC are described elsewhere . In brief, all included RA patients were promptly treated with conventional synthetic DMARDs (csDMARDs) after diagnosis; methotrexate (MTX) was the first choice. Subsequently, DAS-steered treatment adjustments were made. When the initial treatment failed, another csDMARD was initiated or added. A biological DMARD (bDMARD) was allowed when RA patients failed ≥ 2 csDMARDs. When low disease activity (DAS44 < 2.4) was sustained, and clinical synovitis was absent, treatment could be tapered and eventually discontinued. Guidelines were to taper DMARDs in case of DAS44 < 2.4 in subsequent visits to the rheumatologist; however, decisions on DMARD cessation were taken in shared decision-making between rheumatologists and patients.
Sustained DMARD-free remission (SDFR) was defined as the absence of clinical synovitis (swollen joints at physical examination) after cessation of DMARD treatment (including systemic/intra-articular corticosteroids) that persisted for the entire follow-up thereafter, and this follow-up should be ≥ 1 year. RA patients experiencing a late flare (defined as reoccurrence of clinical synovitis) after SDFR development were also included in the non-SDFR group. These stringent definitions were chosen to ensure the sustainability of DMARD-free remission. Medical files were studied on the occurrence of SDFR until August 2020.
Serum samples were collected at disease presentation (before DMARD initiation, including corticosteroids) and after 12 and 24 months (stored at − 80 °C). Biomarker levels were determined using three separate multiplex, sandwich immunoassays (as previously described) . Measurements were performed blinded to clinical data and outcome.
A previous biomarker study found that baseline levels of three out of twelve studied biomarkers were associated with SDFR development: MMP-3, SAA, and CRP . We hypothesized that the levels of these three biomarkers might subsequently change differently over time in RA patients achieving SDFR compared to those who do not (hypothesis 1, Fig. 1). Therefore, these biomarkers were the main subject of this study. The other 9 biomarkers from this previous study, matrix metalloproteinase-1 (MMP-1), tumor necrosis factor receptor superfamily member-1A (TNF-R1), interleukin-6 (IL-6), leptin, resistin, human cartilage glycoprotein-39 (YKL-40), epidermal growth factor (EGF), vascular endothelial growth factor-A (VEGF-A), and vascular cell adhesion molecule-1 (VCAM-1), were also evaluated and considered as “negative controls” to preclude that changes reflected regression-to-the-mean. Thus, for these biomarkers, we did not presume to find changes related to achieving SDFR.
The statistical methods are extensively described in supplementary S3. In short, to test our first hypothesis (Fig. 1), the course of the individual biomarkers was compared between the SDFR group and non-SDFR group using linear mixed models (LMM), stratified for ACPA status. Since biomarker levels were measured at three specific time points (0/12/24 months), differences in biomarker course between the SDFR groups were compared for two time periods: 0–12 months/12–24 months. Biomarker levels were log-transformed because of non-normal distribution. Because logarithmic results are measured on a multiplicative scale, differences in biomarker course between the SDFR group and non-SDFR group, in the two specific time periods, were expressed as ratios.
Of the included 266 RA patients, follow-up biomarker measurements were not available in 42 RA patients. Missing biomarker measurements were not imputed or excluded since LMM analysis can handle missing data, assuming missingness is at random. Baseline characteristics of these 42 RA patients did not remarkably differ from patients who did have follow-up biomarker data (S4).
As it was previously observed that biomarker levels at diagnosis were related to SDFR-development , correlations between baseline biomarker levels and decline in levels in the first 12 months were plotted (testing hypothesis 2, Fig. 1) and expressed using Spearman’s rho. Additionally, considering the previous findings on the relation between early DAS remission (DAS4 months < 1.6) and SDFR , median biomarker change within the first 12 months was compared between RA patients with and without early DAS remission using the Mann-Whitney U test (testing hypothesis 3, Fig. 1).
Several sub-analyses were carried out. First, RA patients achieving SDFR within 3 years of follow-up were excluded from the analyses. From all RA patients achieving SDFR, these were the RA patients who reached SDFR after the shortest disease duration. If overtreatment (treatment of patients who might otherwise have had spontaneous resolution) would be an issue, these excluded patients would presumably be part of this group. Second, to explore whether the effects differed per initial DMARD, analyses were repeated in the subgroup of RA patients initially treated with MTX, as this was the most frequently prescribed initial DMARD. Finally, ACPA-negative rheumatoid factor (RF)-positive RA patients were excluded to restrict analyses to autoantibody-negative RA patients.
STATA (V16) was used. p-values < 0.05 were considered statistically significant.