At the dawn of the era of biologic therapy, tumor necrosis factor-α inhibitors (TNF-αi) were the first agents that changed treatment paradigm and guidelines for patients with rheumatoid arthritis (RA) [1]. It has enabled patients with an active disease to achieve rapid and significant clinical improvement, maintain low disease activity (LDA), or enter clinical remission [2]. On the other hand, TNF-αi are well-known to cause adverse events, including bacterial infections, tuberculosis, congestive heart failure, demyelinating syndrome, autoimmune phenomena, and an increased risk of certain malignancies [3]. Increased prescriptions for TNF-αi also have raised medical cost for RA treatment [4].

Therefore, further interest towards developing strategies for tapering or discontinuing biologic therapy have risen, especially for patients who have sustained clinical remission or LDA [5]. Although many studies have investigated approaches to down-titrate or discontinue TNF-αi, it remains difficult to distinguish patients whose TNF-αi can be discontinued and predict who would experience a flare-up or relapse after such modification of their treatment regimen [6–16]. Previously published studies are mostly clinical trials focusing on down-titrating or discontinuing TNF-αi with follow-up periods of a year or less. They might not reflect real-world outcomes.

Among various TNF-αi, golimumab is a human anti-TNF-α monoclonal antibody purified from TNF-α-immunized transgenic mice with altered expression of human IgG [17]. Pivotal trials of golimumab in RA have been conducted in methotrexate (MTX)-naïve patients (GO-BEFORE) and in patients with inadequate responses to MTX (GO-FORWARD) [18,19]. Both trials were conducted on Korean patients followed by a 5-year extension study in which golimumab was discontinued for all participants at the time of study completion.

The aim of the present study was to investigate 2-year outcomes of patients with RA whose golimumab treatment was discontinued after completing the 5-year GO-BEFORE and GO-FORWARD extension study. Possible risk factors leading to subsequent intensive treatment after golimumab discontinuation were also identified.

Even with recent advances in treatment recommendations and strategies for RA, discontinuing or tapering TNF-αi is a challenging decision to make. A consensus is needed among rheumatologists as to when this should be done. Our study aimed to analyze clinical outcomes in patients who discontinued golimumab after a 5-year extension open-label trial. In our study, 38.3% of patients who achieved LDA or remission (43.5% of study subjects were in DAS28 remission) maintained or tapered cDMARDs during 2 years after discontinuing golimumab without stepping-up cDMARD use or restarting biologic therapy. To the best of our knowledge, this is the first report to present data of a 2-year MB in RA patients after discontinuing a single biologic agent used for 5 years.

Outcomes after biologic discontinuation shown in previous clinical studies are relatively heterogeneous. Tanaka et al. [9] reported that 55% of patients with RA who discontinued infliximab after attaining LDA (DAS28 < 3.2) maintained stable disease activity at a 1-year follow-up while 45% of patients experienced disease flare-up. A study investigating patients who achieved DAS28 remission after using adalimumab showed that 48% of subjects were able to maintain DAS28 remission for 1 year after discontinuing adalimumab [7]. Of note, Saleem et al. [20] reported that only 15% of patients who underwent delayed (median symptom duration of 9 months). TNF-αi treatment were able to attain sustained DAS28 remission after its discontinuation and that a shorter duration of untreated symptoms was the main predictor of successful cessation. Harigai et al. [21] demonstrated that 18.2% of long-standing RA patients who discontinued adalimumab after attaining LDA maintained similar disease activity for 1 year without resuming biologic therapy. Furthermore, Brocq et al. [22] reported that, among patients with long-standing RA who had achieved DAS28 remission for at least 6 months, 75% experienced disease relapse within 12 months after discontinuing TNF-αi. Results of previous studies utilizing clinical disease activity index (CDAI) were dissimilar. Smolen et al. [13] showed that only 17.6% of established RA patients who stopped certolizumab pegol maintained CDAI remission until week 52 and suggested that it would be difficult for patients to withdraw from certolizumab pegol even if they had achieved remission. According to data from the Corrona registry, 42% of RA patients who exhibited disease activity of LDA or lower (CDAI≤10) successfully discontinued their first TNF-αi and experienced an MB for 24 months [6]. Despite the fact that baseline disease activity and cDMARD/corticosteroid use in these studies were different from those in our study, results of the Corrona data are similar to what we demonstrated.

In the multivariable logistic regression analysis, SJC and disease duration at the time of golimumab discontinuation were significant factors associated with restarting biologic therapy after golimumab discontinuation. For every 1-point increase in SJC, the chance of resuming biologics increased 8.34-fold. Previous studies have shown that higher level of disease activity at the time of tapering biologics and presence of RF or anti-citrullinated peptide antidbody may reduce the success of dose-reduction regimen [16,23]. In a study of Tanaka et al. [24], patients achieving sustained deep or stringent remission with biologics had a higher probability of remaining in remission or LDA after dose reduction or withdrawal of etanercept. Interestingly, corticosteroids were used more frequently in the MB group than in the LB group at the time of golimumab discontinuation, though the mean dose of prednisolone was below 5 mg/day. This suggests that concomitant use of low-dose corticosteroids may have aided in maintaining LDA or a stable disease state and a longer biologic-free period, although European Alliance of Associations for Rheumatology (EULAR) guidelines recommend to start with tapering steroids, then biologics, followed by cDMARDs [25].

Given that two-thirds of patients lose their good disease state after discontinuation of golimumab in this study, gradual tapering biological disease-modifying antirheumatic drug rather than discontinuation may be needed to maintain a long-term clinical response in patients with RA in remission or LDA. On the other hand, recent studies showed non-inferiority of tapering or stopping cDMARD versus maintaining cDMARD therapy in stable patients who treated with cDMARD plus biologic therapy [26,27]. In a study comparing cDMARD tapering and TNF-αi tapering, cDMARD tapering tended to be more effective by a 10% difference in flare rate [28], underscoring the need for further research to assess the DMARD tapering strategy in patients with RA.

Our study was the first to determine patient outcomes after golimumab discontinuation for a relatively long period of time (2 years). In addition, our patients had been treated with a single TNF-αi at regular intervals for an extensive period. However, our study has several limitations. First, the study population was relatively small. Of the 91 patients who participated in the GO-BEFORE and GO-FORWARD trials, 31 were excluded from this study due to lack of 2-year follow-up data. Therefore, the MB group may have been underestimated in evaluating long term clinical outcomes after discontinuation of golimumab. In addition, the list of medications and disease activity at baseline were heterogeneous. Furthermore, most patients showed mild to moderate disease activity at the time of golimumab discontinuation. Most previous studies investigated patients who had experienced LDA or remission for at least 6 months prior to TNF-αi discontinuation. However, our higher baseline disease activity was not associated with an increased incidence of restarting biologic therapy compared to other studies. This in part might be due to the national health insurance reimbursement policy that requires a period of cDMARD combination therapy before restarting biologic therapy. Second, as this was a retrospective analysis, there were some missing data on disease activity during the 2-year follow-up period. Therefore, we used indirect measures of disease flare-up or relapse by assessing stepped-up/added-on/switched cDMARDs or restarting biologic therapy which reflected uncontrolled disease activity. Third, the baseline parameter was the time point of discontinuation of golimumab. Initial and earlier responses to golimumab (i.e., American College of Rheumatology response at 52 weeks, duration of remission, and timing of clinical response) were not considered. However, these indices were too remote to directly affect clinical outcomes after discontinuation.

In summary, controlling disease activity of patients with RA is a long-term, unremitting process that is in part affected by underlying clinical characteristics at key time points of treatment. Beyond guidelines on when to start TNF-αi, post-discontinuation treatment strategies are needed to better maintain disease control and quality of life of patients with RA.