Which advanced treatment should be used following the failure of a first-line anti-TNF in patients with rheumatoid arthritis? 15 years of evidence from the Quebec registry RHUMADATA

Abstract

Background

Since 2000, advanced therapies (AT) have revolutionized the treatment of moderate to severe RA. Randomized control trials as well as observational studies together with medication availability often determine second-line choices after the failure of first TNF inhibitors (TNFi). This led to the observation that specific sequences provide better long-term effectiveness. We investigated which alternative medication offers the best long-term sustainability following the first TNFi failure in RA.

Methods

Data were extracted from RHUMADATA from January2007. Patients were followed until treatment discontinuation, loss to follow-up or 25 November 2022. Kaplan–Meier and Cox regression models were used to compare discontinuation between groups. Missing data were imputed, and propensity scores were computed to reduce potential attribution bias. Complete, unadjusted and propensity score-adjusted imputed data analyses were produced.

Results

Six hundred eleven patients [320 treated with a TNFi and 291 treated with molecules having another mechanism of action (OMA)] were included. The mean age at diagnosis was 44.5 and 43.9 years, respectively. The median retention was 2.84 and 4.48 years for TNFi and OMAs groups. Using multivariable analysis, the discontinuation rate of the OMA group was significantly lower than TNFi (adjHR: 0.65; 95% CI: 0.44–0.94). This remained true for the PS-adjusted MI Cox models. In a stratified analysis, rituximab (adjHR: 0.39; 95% CI: 0.18–0.84) had better retention than TNFi after adjusting for patient characteristics.

Conclusion

Switching to an OMA, especially rituximab, in patients with failure to a first TNFi appears to be the best strategy as a second line of therapy.

Introduction

The advent of biologic therapies, particularly TNF inhibitors (TNFi), has revolutionized the treatment landscape for patients with various autoimmune diseases, including RA. Even in recent years, the first agent used is usually an agent targeting TNF. However, many patients experience a lack of response, loss of response or intolerance to TNFi treatment, leading to what is known as ‘TNFi failure’. approximately 60% of patients with RA may show nonresponse, insufficient response, loss of response or adverse events related to a TNFi agent [1].

The decision-making process for subsequent treatment becomes crucial in achieving successful long-term disease management for patients who fail TNFi. Most advanced therapies (AT) were shown in the short term to have some degree of efficacy compared with placebo after first or multiple TNFi failures [2–5].

The response to a second agent is usually not adequate or long-term [6]. As most of these observations were short-term (<2 years) [6], real-world data give a unique opportunity to evaluate the effectiveness of second-line AT during longer periods. They may inform clinicians of the best choice of therapy after failing a first TNFi agent. The introduction of AT with other mechanisms of action (OMA), such as interleukin-6 and Janus kinase (JAK) inhibitors, anti-B cell agents (rituximab) and agents targeting the co-stimulatory signal of T cells such as abatacept have provided alternative therapeutic options. Research suggests that specific sequences provide better long-term effectiveness. Several observational studies have published data on the effectiveness of TNFi and OMA in RA patients who had previously received a TNFi. Using data from the Swedish National Registry, Chatzidionysiou et al. showed that up to 40% of patients with first TNFi failure achieved low disease activity or remission by switching to a second biologic [7]. Favalli et al. showed that in patients with first TNFi failure, switching to an OMA is a better treatment strategy than a second TNFi agent [8]. However, a recent study did not show significant outcome differences between groups [9, 10].

To the best of our knowledge, little data compares long-term OMA and TNFi retention as a second therapy choice in patients with prior experience with TNFi. Our previous study compared abatacept vs TNFi and showed that abatacept had longer retention than TNFi as second-line biologic therapy [11]. Few other studies have also investigated the retention of second bDMARDs and their effectiveness [10, 12–16]. Therefore, in this study, we proposed examining the retention of second biologic (TNFi or OMA) in patients who fail (primary or secondary) the initial TNFi.

Methods

Data source

The RHUMADATA clinical database and registry monitors the clinical care of all the patients with inflammatory articular diseases seen at the Institut de Rhumatologie de Montréal (IRM), the Centre de l’Ostéoporose et de Rhumatologie de Québec (CORQ) and the Clinique de Santé Jacques-Cartier (CSJC), some of the largest rheumatologic clinics in the province of Québec, Canada. RHUMADATA has been collecting real-world observational data since 1998. The database currently includes >6000 patients with inflammatory disease (RA, axial and peripheral spondyloarthropathies including ankylosing spondylitis, psoriatic arthritis, reactive arthritis and Inflammatory bowel disease associated arthritis). Data collected at all visits includes demographics, disease history, laboratory values [RF, ACPA at least once, CRP and ESR], all disease activity scores [DAS- CRP and ESR, clinical disease activity index (CDAI) and simplified disease activity index (SDAI)], patient report outcomes (PROs) including HAQ disability index (HAQ-DI), morning stiffness, patient global evaluation of disease activity using a visual analogue scale from 0 to 10 (VAS), patient evaluation of pain(VAS), patient evaluation of fatigue (VAS) and physician global evaluation of disease activity. Comorbidities, including but not limited to cardiovascular disease, diabetes mellitus, hypertension, cancer and infections, are also collected. Medication usage for disease control is entered into the database {[start and termination dates and the reason for termination are classified as ineffectiveness, adverse events (AEs), infectious events and other reasons]}.

This study was conducted following the ethical principles of the Declaration of Helsinki. Investigators obtained central or local institutional review board approval for conducting noninterventional research involving human patients. All investigators also signed an agreement to access their electronic medical records for data review by RHUMADATA outcome assessors, who also signed a confidentiality agreement. All these procedures are regulated by standard operating procedures approved by the institutional review board services. All registry patients provided written informed consent (RHUMADATA Institutional Review Board Services #IRB00005290).

Study population and data collection

Data for AT prescribed since January 2007 was extracted from RHUMADATA. 2007 was the starting point because abatacept and rituximab were introduced to the Canadian market then. Patients who failed a TNFi as their first AT were included and followed until treatment discontinuation, loss to follow-up or 25 November 2022, whichever came first. We applied no time restriction on time to switch from the first TNFi failure to the second AT.

Statistical analysis

Descriptive statistics, including mean and s.d. for continuous variables and counts and proportions for categorical variables, were generated for all baseline (defined as the start of medication) characteristics. Comparisons between patients on TNFi vs OMA were conducted using the independent-samples t test for continuous variables and the χ² or the Fisher’s Exact test for categorical variables. Time to discontinuation due to (1) any reason, (2) lack/loss of response or adverse events (AEs) were assessed using unadjusted Kaplan–Meier survival analysis and Cox proportional hazards regression (HR) analysis for TNFi vs different OMA users. The level of statistical significance is set to 5.0%.

Missing data were imputed using a sequential imputation method known as fully conditional specification (FCS) or chained equations. FCS uses an iterative variable-by-variable approach to impute multivariate missing data under arbitrary patterns of missingness [17]. Consequently, missing values are addressed separately for each incomplete variable. To apply this method, it is necessary to specify an imputation model for each incomplete variable. The discriminant function method was used to impute binary variables (gender and variables taking yes/no values), and continuous variables were imputed using regression models. All variables used in the imputation process are listed in Table 1. A total of 20 imputed datasets were used in this analysis.

We used the propensity scores (PS) matching method to address confounding by indication. Propensity scores were calculated from the imputed data based on variables that were notably different between treatment groups (variables with a P-value of 0.1 or less when compared between the TNFi and OMAs groups at treatment initiation). PS were matched based on the difference in the propensity score logit between groups. Standardized differences between groups of variables listed in Table 1 are presented in Supplementary Table S1, available at Rheumatology online for pre- and post-PS adjustments. Complete data, unadjusted and propensity score-adjusted imputed data analyses were produced.

Results

Baseline characteristics of patients at the time of initiating the second biologic by choice of the second biologic

In total, 611 patients were included in the analysis. Three hundred twenty patients (52.4%) started a second TNFi, and 291 (47.6%) used OMA (Supplementary Fig. S1, available at Rheumatology online).

Table 1 describes the baseline characteristics of patients at the time of the second treatment initiation by type of AT. The mean age at diagnosis was 44.5 (14.4) and 43.9 (14.8) years in the TNFi and OMA groups. Women made up 72.8% and 81.1% of these groups, respectively. Compared with TNFi, patients with different OMA had numerically shorter disease duration (12.9 vs 14.1 years, P = 0.1741), positive RF (69.8% vs 65.4%, P = 0.2887), higher patient global (5.3 vs 4.0, P < 0.0001), pain (5.8 vs 4.3, P < 0.0001) and fatigue (5.5 vs 4.1, P < 0.0001), longer duration of morning stiffness (124.6 min vs 71.7 min, P = 0.0451), HAQ-DI (1.4 vs 1.1, P < 0.0001) and disease activity measured by swollen (8.0 vs 4.9, P < 0.0001), tender joint count (7.2 vs 4.3, P < 0.0001), physician global assessment (4.6 vs 2.9, P < 0.0001). In the OMA group, inflammatory markers, including ESR (26.9 vs 21.6, P = 0.0136) and CRP (15.4 vs 8.9, P = 0.0005) were also higher. There was no statistically significant difference between the two groups for the Charlson’s Comorbidity Index (CCI) (0.52 vs 0.68, P = 0.0631). However, individuals reported comorbidities were numerically higher in the OMA group. Reasons for initial TNFi treatment cessation, duration and time to second treatment for two treatment group are also presented in Supplementary Table S2, available at Rheumatology online.

Drug retention of the second advanced treatment

Among patients with first TNFi failure, 247 (40.4%) were censored and remained on treatment (while on their second biologic) at the end of follow-up; 144 (45.0%) were in the TNFi group and 103 (35.4%) in the OMAs group. Discontinuation due to ineffectiveness was the most common reason in the two treatment groups (50.0%). Discontinuation due to AEs was the second most reason (17.6% in TNFi and 13.3% in OMAs) (Supplementary Table S3, available at Rheumatology online).

The median retention was 4.48 years (95% CI: 3.43–5.39) and 2.84 years (95%CI: 1.83–3.56) in OMA and TNFi treatment groups, respectively (Table 2).

The unadjusted retention rate of second advanced treatment at 6 months was 84.8% (95% CI: 80.1%–88.5%) and 79.6% (95%CI: 74.8%–83.6%) and declined to 76.8% (95% CI: 71.5%–81.3%) and 67.5% (95% CI: 61.8%–72.5%) after 12 months for OMA and TNFi, respectively (Table 2). The retention rate at 24 months after treatment initiation was 63.9% (95% CI: 58.0%–69.2%) and 53.7% (95% CI: 47.3%–53.9%) for OMA and TNFi groups, respectively (Table 2). At 60 months after treatment initiation, the retention reduced to 45.5% (95%CI: 39.5%–51.4%) in OMA and 36.4% (95%CI: 30.0%–42.8%) in the TNFi group. The retention rates at different time points were also higher in the OMA group than TNFi for discontinuation due to ineffectiveness or AEs (Table 2).

The Kaplan–Meier survival curve showed that retention of the second biologic OMA was higher (logrank = 0.0134) than TNFi retention (Fig. 1).

Using univariate Cox regression, the discontinuation of the OMA group was significantly lower than TNFi (HRs: 0.791; 95%CI: 0.634–0.987). After adjustment for baseline sociodemographic (age, gender) and clinical characteristics (disease duration, RF, ACPA, CDAI, HAQ-DI and CCI) of discontinuation of OMA remained significantly lower than TNFi (adjHR: 0.645: 95% CI: 0.441–0.945) (Table 3). This remained true for the unadjusted and PS-adjusted MI Cox models (MI adjHR: 0.665: 95% CI: 0.528–0.837 and MI-PS adjHR: 0.694: 95% CI: 0.545–0.885).

A Kaplan–Meier for discontinuation due to ineffectiveness or AEs also showed that OMA had significantly longer retention than TNFi (Supplementary Fig. S2, available at Rheumatology online). However, after adjusting for baseline covariates in the multivariate Cox regression model, there was no significant difference in the risk of discontinuation between treatment groups (Supplementary Table S4, available at Rheumatology online). Still, the unadjusted and PS-adjusted MI Cox models showed significant differences in time to discontinuation due to ineffectiveness or AEs (MI adjHR: 0.645: 95% CI: 0.487–0.854, and MI-PS adjHR: 0.680: 95% CI: 0.506–0.914).

Comparison of discontinuation between TNFi and individual OMAs agents

In a stratified analysis, rituximab (logrank = 0.0023) had better retention (due to any reason) than TNFi (Fig. 2).

Table 4 also shows a head-to-head comparison of discontinuation due to any reason between TNFi and individual OMA agents. Rituximab had higher retention than the TNFi group [adjHR of 0.515 (95% CI: 0.346–0.766)], abatacept [adjHR of 0.619 (95% CI: 0.404–0.949)] and JAKi [adjHR of 0.430 (95% CI: 0.259–0.715)] in unadjusted proportional hazard model. After adjusting for baseline variables (gender, age at diagnosis, disease duration, RF, ACPA, CDAI, HAQ-DI and CCI), Rituximab had higher retention than the TNFi group [adjHR of 0.392 (95% CI: 0.182–0.843)].

Rituximab also had a lower discontinuation due to ineffectiveness or AEs than TNFi, other individual OMAs or JAKi (Supplementary Fig. S3, available at Rheumatology online).

Discussion

Our results showed that although the core discontinuation frequency was higher for agents with a different mechanism of action, such as abatacept, interleukin-6 inhibitor or rituximab, they had higher mean survival than TNFi agents.

The significantly higher disease activity (swollen joint count, tender joint counts, physician global assessment, ESR, CRP) and lower physical function (HAQ-DI, patient global assessment, pain and fatigue) at baseline seen in our study’s OMA group were also demonstrated in Frisell et al. study [17]. This observation could be influenced by the fact that many patients discontinued their first-line anti-TNFi because of primary failure and that RA is a heterogeneous disease when we look at the underlying inflammatory processes involved. Indeed, the proportion of discontinuation due to nonresponse for the second line of therapy was numerically higher in TNFi (36.9%) than in OMA (23.4%).

The retention findings from other studies are consistent with our results. Bonafede et al. (2016) showed that, compared with TNFi users (n = 5020), non-TNFi switchers (n = 1925) were significantly more likely to be persistent on therapy at 12 months (61.8% vs 58.2%; P < 0.001) [10] which are comparable to our 76.5% and 67.5%.

Rotar et al. (2015) using Slovenian registry data, showed that after the first-line TNFi failure, a second-line TNFi is more likely to fail earlier than non-TNFi [12].

Wie et al. (2017) using retrospective data from 613 RA patients with prior failure to TNFi, showed that compared with new OMA switchers, TNFi cyclers were 51% more likely to be nonpersistent (adjusted hazard ratio, 1.511; 95% CI 1.196, 1.908) [13]. Conversely, Park et al. (2022) using a small population of RA with prior use of TNFi (n = 209), showed that the non-TNFi group had a lower likelihood of discontinuing their treatment than the second TNFi group [HR = 0.326, 95% CI: 0.170–0.626, P = 0.001].

In contrast to our results, Curtis et al. in a recent retrospective comparative study, showed that achieving LDA at 12 months based on CDAI and DAS28-CRP was not significantly different between non-TNFi and TNFi groups in patients with prior exposure to TNFi [9]. However, they did not present findings 6 months after treatment, which clinicians consider a strategic treatment time point. Moreover, there may be better approaches than comparing our retention findings with their effectiveness results.

Migliore et al. in a network meta-analysis of RCTs and observational studies, also showed that those with a prior failure to first TNFi switched to an OMA had better effectiveness and lower discontinuation than cycling to a different TNFi [16]. Using the Truven Health MarketScan Research database, Matusevich et al. also showed a significantly longer time to discontinuation for non-TNFi than for TNFi (median 605 days compared with 489 days) after initial TNFi discontinuation [15]. To improve treatment response in RA patients, it has been suggested that rather than TNFi cycling, we switch to a non-TNFi agent after initial TNFi failure [18].

Strengths of our study include the use of multicentre data, the inclusion of 95% of patients treated at our centers without exclusion, making RHUMADATA not only a real patient registry but a real-life one, the excellent collaboration of patients to complete questionnaires either online or at the clinics, similar access to advanced therapy independent of your socio-economic status controlling for disease severity, comorbidities and demographics by adjusting our models for these potential confounders. RHUMADATA database has an excellent long follow-up duration and includes patients with over 10 years of follow-up. Access and coverage of AT in Canada, especially in the province of Quebec, is usually rapid after approval of the Health Canada agency. It allows RHUMADATA to include newer AT more rapidly than in other countries.

There are several limitations of this study. First, given its observational nature, patient and physician variables cannot be reliably accounted for. Additionally, there may be systematic differences in the practice patterns of physicians participating in RHUMADATA within the registry compared with other RA registries. Finally, unmeasured variables and residual bias are some of the most common limitations of observational studies.

In conclusion, our study demonstrated that patients with failure to a first TNFi have higher retention if they switch to a different OMA agent than another TNFi. Rituximab is the best choice at this moment. Janus kinase should be included in the future.

Acknowledgements

The authors would like to thank all the participating providers and patients in the RHUMADATA Registry who contributed data to this study.

Supplementary material

Supplementary material is available at Rheumatology online.

Data availability

The data that support the findings of this study are available from the corresponding author, (Dr Denis Choquette), upon reasonable request.

Contribution statement

Denis Choquette (Collection of data, Conception and Design, Revised the work critically and approved), Boulos Haraoui (Collection of data, Revised the work critically and approved), Mohammad Movahedi (Conception and Design, Revised the work critically and approved, drafted the manuscript), Louis Bessette (Collection of data), Loïc Choquette Sauvageau (Collection of data, Conception and Design, Revised the work critically and approved, data analysis), Isabelle Ferdinand (Collection of data), Maxine Joly-Chevrier (Collection of data), Ariel Masetto (Collection of data), Frédéric Massicotte (Collection of data), Valérie Nadon (Collection of data), Jean-Pierre Pelletier (Collection of data), Jean-Pierre Raynauld (Collection of data), Diane Sauvageau (Collection of data), Édith Villeneuve (Collection of data) and Louis Coupal (Collection of data)

Funding

RHUMADATA is supported by unrestricted grants from Abbvie Canada, Amgen Canada, Eli Lilly Canada, Fresenius Kabi Canada, Jamp Canada, Novartis Canada, Pfizer Canada, Sandoz Canada, Sanofi Canada and Teva Pharmaceuticals.

Disclosure statement: D.C. has received honoria from RHUMADATA as scientific director. He also served as an advisor and has given talks on different topics for the companies funding RHUMADATA (AbbVie, Amgen, Eli Lilly, Fresenius Kabi, Novartis, Pfizer, Sandoz, Tevapharm).

B.H. has received grants from AbbVie, Amgen, Pfizer and UCB. He is also a consultant at AbbVie, Amgen Eli Lilly, Pfizer, Sandoz and UCB and a speaker at Pfizer.

M.M. has a faculty position (status) at the University of Toronto and is a staff scientist at the University of Health Network. He received an honorarium for this work and served as a medical writer.

L.B. has received grants from AbbVie, Amgen, BMS, Celgene, Janssen, Novartis, Pfizer, Roche, Sandoz, Sanofi-Genzyme and UCB; He is also a consultant at AbbVie, Amgen, BMS, Celgene, Fresenius Kabi, Janssen, Novartis, Pfizer, Sandoz, Sanofi-Genzyme, UCB.

L.C.S. has no disclosure; he is an employee of Rhumadata and graduate student in pharmaceutical sicences at l'Université de Montréal.

I.F. is a consultant for AbbVie, Eli Lilly, Fresenius Kabi, Janssen, Novartis, Pfizer and UCB and a speaker at Pfizer.

M.J.C. has no disclosure. She is MD candidate at l'Université de Montréal.

A.M. has received grants from Novartis and is a consultant at AbbVie, Janssen, Novartis, Pfizer, Sanofi-Genzyme and UCB and a Speaker for AbbVie, Janssen, Novartis.

F.M. is a consultant in AbbVie, Eli Lilly, Janssen, Pfizer and a speaker at Janssen.

V.N. is a consultant in AbbVie, Eli Lilly, Janssen, Pfizer, Roche and Sanofi-Genzyme.

J.P.P. has received grants from TRB Chemedica SA. He is also a consultant at TRB Chemedica SA.

J.P.R. is a consultant for AbbVie, ArthroLab Inc., Janssen, Pfizer and Sanofi-Genzyme. He is also a speaker for AbbVie, Amgen, BMS, Eli Lilly, Janssen, Novartis, Pfizer, Sandoz and Sanofi-Genzyme.

D.S. has no disclosure; she is an employee of Rhumadata.

E.V. is a consultant at AbbVie, Amgen, Novartis and Pfizer and a speaker at AbbVie,

B.M.S., Novartis and Pfizer.

L.C. has received honoraria from Rhumadata.

References