Cardiovascular risk in rheumatoid arthritis patients treated with targeted synthetic and biological disease‐modifying antirheumatic drugs: A multi‐centre cohort study

This study aimed to compare the cardiovascular safety of interleukin‐6 inhibitors (IL‐6i) and Janus Kinase inhibitors (JAKi) to tumour necrosis factor inhibitors (TNFi).


Introduction
Biological and targeted synthetic diseasemodifying antirheumatic drugs (b/tsDMARDs) which target the specific inflammatory pathway can slow rheumatoid arthritis (RA) progression rapidly. Therefore, a standard RA management method for inadequate responses to conventional synthetic DMARDs (csDMARDs) is optimal [1,2]. Chronic inflammation from RA would alter body composition, induce insulin sensitivity, change lipid profile and further mediate premature atherosclerosis [3][4][5][6]. It has been established that RA is associated with an increased risk of cardiovascular events (CVE), and half of the excess mortality among RA patients is attributed to CVE [6][7][8][9].
Although the exact mechanisms remain unanswered, ample literature suggests that certain bDMARDs -for example the tumour necrosis factor alphainhibitor (TNFi) -have a beneficial CVE effect compared to csDMARDs [9,[10][11][12][13] through the efficient reduction of systemic inflammation. However, the effects of non-TNFi bDMARDs on CVE risk are still debatable [14]. Some studies have found that anti interleukin-6 inhibitors (IL-6i) may increase CVE risk through elevated circulating low-density lipoprotein (LDL) levels [15], but the increased outcome of CVE has yet to be identified [12,16,17]. Janus Kinase inhibitors (JAKis) are targeted synthetic DMARDs (tsDMARDs) [18]. Compared to bDMARDs, tsDMARDs can increase patient compliance with lower rates of immunogenicity and can be given orally [19], but they are not free from the risk of side effects. In some aspects, IL-6i shares the same pathway with JAKi (the JAK-STAT) in reducing RA inflammation [20,21]. Inhibiting such a pathway may block the gene network that regulates the physiological response of inner organs, such as cardiovascular systems [14]. Modulatory effects of IL-6i and JAKi on the cardiovascular system are not conclusive; although systemic inflammation is reduced, they may increase the traditional CVE risk factors -such as altering lipid profiles and increasing atherosclerosis risk [22][23][24].
To our knowledge, current post-marketing studies comparing the cardiovascular safety between b/tsDMARDs in real-world settings are limited [12,13,[25][26][27]. This study aims to examine the CVE risk of JAKi or IL-6i in comparison to TNFi using multiple population-based electronic health and claims databases from Hong Kong, Taiwan and Korea.

Data sources
The study utilized three population-based electronic health records and claims databases in Hong Kong, Taiwan and Korea. The Hong Kong Clinical Data Analysis and Reporting System (CDARS) is a territory-wide database that covers all public hospitals' and ambulatory clinics' medical records [28]. CDARS routinely centralizes electronic medical records, including demographics, date of registered death and cause, date of hospital admission and discharge, prescriptions and diagnoses for research and audit purposes. CDARS' coding accuracy and record quality have been demonstrated through numerous high-quality epidemiology and health service studies [29,30]. Taiwan's data are obtained from the National Health Insurance Research Database (NHIRD), which exemplifies a population-level data source for generating real-world evidence to support clinical decisions and healthcare policymaking. By the end of year 2014, NHIRD covered 99.9% of the Taiwanese population. Over 2700 peer-reviewed studies had been published using NHIRD data by 2018 [31]. Patients who used b/tsDMARDs would be 100% eligible for the RA diagnosis because the use of b/tsDMARDs would have been reviewed and approved by a second rheumatologist. In addition, the diagnosis of cardiovascular-related disease had been validated in previous studies [32][33][34]. Korea's data were obtained from the National Health Insurance (NHI) database. NHI covers all citizens in South Korea and contains comprehensive information pertaining to healthcare services, such as treatments, pharmaceuticals, procedures and diagnoses. The positive predictive values of diagnostic codes for major clinical outcomes in the database were reported as 71.5%-92.0% [35]. All patient data are anonymized to protect confidentiality and identity.

Participants and study design
The study population in this multi-centre retrospective cohort study were patients with RA who had their first diagnosis after 1 January 2010 and were prescribed b/tsDMARDs for the first time. We excluded patients with other major autoimmune diseases (systemic lupus erythematosus, psoriasis, ankylosing spondylitis, multiple sclerosis Fig. 1 Study schema of cohort study design. Superscripted "a" occurrence of outcome, treatment discontinuation (defined as prescription gap more than 6 months), treatment switch to another b/tsDMARDs in different modes of action, death or study end (31 December, 2020). The study period in Hong Kong database is used for illustration. and inflammatory bowel disease). b/tsDMARDs assessed in this study include TNFi (etanercept, infliximab, adalimumab, certolizumab pegol and golimumab), IL-6i (tocilizumab, sarilumab) and JAKi (tofacitinib, baricitinib and upadacitinib). This study obtained ethics approval from the Institutional Review Board via The University of Hong Kong/Hospital Authority Hong Kong Western Cluster (UW21-338), National Cheng Kung University Hospital in Taiwan (B-ER-107-378) and Sungkyunkwan University in Korea (SKKU2021-06-006). Patient information was de-identified with no patient consent required.
The index date of the study is the dispensing date of the first b/tsDMARDs. We followed patients from the index date to the occurrence of outcome, or any of the earliest censoring events -including treatment discontinuation (defined as a prescription gap of more than 6 months), treatment switch to another b/tsDMARDs with different modes of action, death or study end. Figure 1 illustrates the study design. The study reporting is in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline.

Study outcomes and covariates
The outcome of interest in this study is the composite of the earliest hospitalized CVE (including acute coronary heart diseases, stroke, heart failure, venous thromboembolism events and systemic embolism) after the first b/tsDMARD treatment [12,36]. Validities of recorded diagnoses were verified in previous cardiovascular studies [32,34,35,37]. Subjects' age, sex, disease duration (defined as years since RA diagnosis to the first treatment of b/tsDMARDs); stage of RA (early RA [disease duration less than or equal to 2 years] and established RA [disease duration longer than 2 years]), medical histories, the Charlson Comorbidity Index (CCI) and recent 1year medication usage (csDMARDs, non-steroidal anti-inflammatory drugs [NSAIDs], corticosteroid and cardiovascular drugs) before index date were treated as baseline covariates.
Diseases of outcomes and covariates were identified using the International Classification of Diseases, Ninth (ICD-9-CM) and Tenth Revision Clinical Modification (ICD-10-CM, Table S1). Drugs were selected by molecular name, WHO ATC code or drug ingredient code (Table S2).

Statistical analysis
We reported the descriptive statistics of baseline characteristics in each dataset and pooled results [38]. We used means (standard deviations [std]) or medians (interquartile ranges [IQR]) to report continuous variables and used number (percentage) to report dichotomous variables. We used chi-square or Fisher's test to compare different groups and used t-tests to compare mean values. We estimated the crude incidence rates of CVE with 95% confidence interval (CI) for each group (TNFi, IL-6i and JAKi) using the Poisson distribution. We used a generalized linear regression model (Poisson distribution) considering the followup time to estimate the incidence rate ratio (IRR) of IL-6i and JAKi with TNFi as reference. We further adjusted IRRs with age, sex, disease duration and CCI.
We conducted data analysis via R (version 4.1.4, R Foundation for Statistical Computing, Vienna, Austria) and SAS (version 9.4, SAS Institute, Cary, NC, USA). Pooled analyses and the visualization of crude and adjusted IRRs were conducted through Review Manager (RevMan) (Computer programme, Version 5.4. The Cochrane Collaboration, 2020). Statistical analysis was conducted and cross-checked by XT, CYS, HLJ and YL.

Participants and baseline characteristics
Between the corresponding data periods available (1 January 2010 and 31 December 2020 in Hong Kong and Korea; 1 January 2010 and 31 December 2018 in Taiwan), 1054, 3828 and 3807 biological naïve patients with RA who started b/tsDMARDs were identified from Hong Kong, Taiwan and Korea, respectively (Fig. S1). Baseline characteristics are shown in Table 1 After adjustment for age, sex, RA disease duration and CCI, we observed that compared to TNFi, IL-6i and JAKi had no increased risk of CVE in either centre (Fig. 2). Pooled adjusted IRR (95% CI)

Discussion
Our study compared the risk of CVEs among b/tsDMARDs with different modes of action using electronic databases from three Asian regions covering 8689 newly diagnosed RA patients who were first treated with TNFi, IL-6i or JAKi. This study found no significant CVE risk differences among IL-6i, JAKi and TNFi. The findings were replicated  Abbreviations: CI, confidence interval; CVE, cardiovascular events; IL-6i, interleukin-6 inhibitors; IQR, interquartile range; JAKi, Janus Kinase inhibitors; TNFi, tumour necrosis factor inhibitors.
across Hong Kong, Taiwan and Korea. Independent of b/tsDMARDs treatment, the risk of CVE may be associated with patient-related factors, such as age, sex, disease duration and comorbidities. Our multi-centre, population-based observational study found that RA patients receiving TNFi, IL-6i or JAKi presented a similar subsequent cardiovascular risk. However, given the limited sample size, follow-up time and the potential unmeasured residual confounding, cardiovascular safety monitoring among RA patients receiving t/bsDMARDs should continue and be evaluated on an individual basis.
Paradoxical effects of IL-6 inhibitors on the risk of cardiovascular disease have been identified in previous studies. On the one hand, IL-6 as an inflammation cytokine is at a high level among RA patients and is associated with accelerating atherosclerosis and increased risk of cardiovascular diseases [39,40]. IL-6 receptor inhibitors would significantly reduce systemic inflammation levels, which alleviates RA disease activity and reduces the risk of CVEs. On the contrary, IL-6i used for RA treatment is known to increase LDL cholesterol, which is an established risk factor of CVEs, but this may reflect a normalization of lipid profiles that were lowered during the acute phase response in RA [39,[41][42][43]. In our study, both individual settings and pooled analysis showed that RA patients who received IL-6i as the first non-csDMARDs did not have a higher risk of CVE compared to those with TNFi, at least during the 2-year observation period. Our study findings imply that the irregular lipid profiles induced by IL-6i may not result in a higher overall risk of CVD in the short term. These results should be carefully evaluated with larger sample sizes and broader population settings [12,16,17].
There has also been uncertainty about the potential CVE risk associated with JAKi due to their complex pharmacological mechanism. JAKi are small molecules targeting the intracellular transduction pathways -essential signals in downstreaming inflammatory molecules with a wide spectrum of cytokines [44,45]. A growing body of literature indicates that JAKi potentially increase thrombotic risk and adversely affect several cardiovascular risk factors, such as serum lipid profiles and platelet count [46][47][48][49]. One recent randomized, open-label trial conducted by Ytterberg et al. showed that tofacitinib has a higher risk of major CVEs than TNFi [50]. However, observational evidence has suggested that JAKi-based therapy has no association with CVE [51,52], rather they could positively modify the risk of cardiovascular disease by reducing systemic inflammatory cytokines [53][54][55]. In our study's pooled observational data from three populations, we found no increased risk of CVE associated with JAKi compared to TNFi. The numerically higher but nonsignificant adjusted IRRs in Hong Kong may be due to the small number of participants receiving JAKi (N = 294) and short follow-up period (median: 0.91 years).
It is possible that patient-related risk factors may also play an important role in CVE risk independent of b/tsDMARD use. A previous review found that well-controlled RA is associated with a decreased risk of CVE [8]. One prospective nested case-control study found that inflammation control among patients with RA was related to the decreased risk of myocardial infarction, independent of the b/tsDMARDs used [56]. Treatment duration of b/tsDMARDs might also contribute to CVE risk. One longitudinal observational conducted by Delcoigne et al. found the elevated risk of acute coronary syndrome only when follow-up duration reached up to 5 years but found no difference within the short-to-medium term [57]. In our study, the follow-up time for the three comparison groups was in general less than 2 years. Therefore, it is possible that CVE risk in the longer term may be associated with IL-6i and JAKi, although such effects are not observed in our analysis. Additionally, it is possible that other independent CVE risk factors, such as age, systemic inflammation level and comorbidities, may have differential impacts on CVE risk.
There have been limited post-marketing safety studies of b/tsDMARD treatments associated with CVE in patients with RA. This study has used real-world observational data from three Asian regions and compared the risk of CVE among b/tsDMARDs. However, several limitations exist, and our findings have to be interpreted cautiously due to the large range of the CIs. First, our databases did not include direct disease activity indicators such as C-reactive protein or disease activity score-28 that might reflect the risk of subsequent development of CVE. Therefore, we used recent 1-year prescriptions of csDMARDs, NSAIDs and corticosteroids, and the disease duration as a proxy of disease severity. Second, our databases could not provide information on lifestyles and lab tests with high quality. Consequently, we did not report the characteristics of some traditional CVE risk factors, such as smoking habits, blood pressure and lipid profiles. However, we assumed the diagnosis of comorbidities, such as hypertension and hyper lipidaemia, could represent for the lab tests. There was no significant difference in the percentage of patients with hypertension between the three treatment groups. Moreover, despite the higher percentage of patients with hyper lipidaemia in the IL-6i and JAKi groups (37.3% and 34.6%, respectively) than that in the TNFi group (33.1%) in the combined population, the risk of CVEs was not higher for IL-6i and JAKi compared to TNKi. Third, although we pooled the results from three databases, the number of CVEs was still too small, which may impact the statistical power to conduct a sub-analysis on individual CVEs or a more rigorous adjustment on confounders. Future studies should consider the addition of more centres with larger sample sizes and longer observation.

Conclusion
This study suggests there are no differences as to the risk of CVE among RA patients receiving IL-6i or JAKi compared to those receiving TNFi in the populations from Hong Kong, Taiwan and Korea during approximately 2 years of follow-up. These findings suggest that clinicians, from a short-term aspect, should consider patient-related factors, such as disease activity and cardiovascular comorbidities rather than the specific b/tsDMARDs when aiming to prevent CVE. This study presents a foundation for future longitudinal studies, and it should be replicated in additional countries and with longer follow-up times to support our findings.