Frailty and cardiovascular safety of JAK inhibitors versus TNF inhibitors in rheumatoid arthritis: a real-world comparative study of drug effects and patient profiles

Abstract

Objective: The aim of this study was to comparatively assess the risk of cardiovascular events (CVEs) in rheumatoid arthritis (RA) patients treated with Janus kinase inhibitors (JAKis) or tumor necrosis factor inhibitors (TNFis) and to explore the interactions with patient profiles [including age, baseline cardiovascular (CV) risk, and frailty, which is a state of decreased physiological reserve, assessed using a validated frailty index (FI) for healthcare administrative databases (AHDs)].

Methods: This retrospective study was based on AHDs of the Lombardy region, Italy (from 1st January 2020 to 31st December 2023). Cox regression models, both crude and adjusted, were applied to estimate the association between treatments and outcomes [CVEs, major adverse cardiovascular events (MACEs), and thromboembolic events (TEs)]. We tested the interaction between the drug treatment and the regulatory agency prescription rule changes [before or after 06th July 2021, the date of the first European Medicine Association (EMA) pronouncement on tofacitinib safety] or patient profiles.

Results: We identified 7,541 therapeutic courses in 5,563 patients: 2,343 started as TNFi users, 1,443 as JAKi users, and 1,777 started with other drugs (1,459 days of follow-up). The crude incidence rates (IRs) for new CVEs were 16.6 [95% confidence intervals (95% CI): 12.8-21.2] and 18.6 (95% CI: 14.2-23.9) per 1,000 person-years (PYs) for TNFi and JAKi users, respectively. Exposure to JAKis was not associated with a significantly increased risk of CVEs [adjusted hazard ratio (HR): 0.92; 95% CI: 0.64-1.32], MACEs (adjusted HR: 0.71; 95% CI: 0.37-1.33), or TEs (adjusted HR: 1.53; 95% CI: 0.65-3.65) compared to TNFis. Each 0.1-point increment of the FI significantly increased the HR for new CVEs (HR: 1.80; 95% CI: 1.48-2.19), MACEs (HR: 1.66; 95% CI: 1.10-2.51), and TEs (HR: 1.69; 95% CI: 1.03-2.78). When assessing the interaction between the period of drug delivery and the treatment with JAKis on the risk of new CVEs, no significant interaction was observed (p = 0.838), while the interaction was statistically significant for baseline CV risk (p = 0.007).

Conclusion: RA patients treated with JAKis in real-world settings have a risk of developing CVEs no higher than those of TNFi users, but potential signals remain for TEs, even if the sample was not sufficiently powered. Patient profiles, particularly the frailty, have a more substantial impact on the risk of CVEs than the specific disease-modifying anti-rheumatic drug (DMARD) choice.

Keywords: JAK inhibitors; TNF inhibitors; cardiovascular events; frailty; healthcare administrative databases; rheumatoid arthritis.

FIGURE 1

Hazard ratios (95% CI) resulting from the unadjusted [panel (A)] and adjusted [panel (B)] time-dependent Cox proportional hazard regression models on CCV events by JAKi treatment (TNFi as reference). Adjustments were made by age, e-RHD-FI, gender, treatment line, and assumption of NSAIDs, GCs, MTX, and CV drugs. Abbreviations: CCV, cardio-cerebrovascular; TNFis, tumor necrosis factor inhibitors; JAKis, Janus kinase inhibitors; HR, hazard ratio; CI, confidence interval; CVE, cardio-cerebrovascular event; MACE, major cardiovascular event; TE, thromboembolic event.

FIGURE 2

Hazard ratios (95% CI) resulting from the unadjusted [panel (A)] and adjusted [panel (B)] time-dependent Cox proportional hazard regression models on CV events by the subgroups of JAKi exposure (TNFi as reference). Adjustments were made by age, e-RHD-FI, gender, treatment line, and assumption of NSDAIDs, GCs, MTX, and CV drugs. Abbreviations: CV, cardio-cerebrovascular; TNFis, tumor necrosis factor inhibitors; JAKi, Janus kinase inhibitors; HR, hazard ratio; CI, confidence interval.

FIGURE 3

Forest plot of hazard ratios (95% CI) resulting from pairwise comparisons based on adjusted time-dependent Cox proportional hazard regression models on CV events by the interaction between JAKi exposure and variables of interest (TNFi as the reference). Adjustments were made by age, e-RHD-FI, gender, treatment line, and assumption of NSDAIDs, GCs, MTX, and CV drugs. P-values were derived from the interaction between JAKi exposure and variables of interest. Abbreviations: CVE, cardio-cerebrovascular event; TNFis, tumor necrosis factor inhibitors; JAKis, Janus kinase inhibitors; HR, hazard ratio; CI, confidence interval; e-RHD-FI, electronic-regional healthcare database frailty index; ORAL, oral rheumatoid arthritis trial; PRAC, Pharmacovigilance Risk Assessment Committee.

FIGURE 4

Cumulative incidence plots (and 95% CI bands) of CVE in patients on first-line TNFi or JAKi, according to specific characteristics: e-RHD-FI < 0.056 [panel (A)], e-RHD-FI ≥ 0.056 [panel (B)], not fulfilling ORAL inclusion criteria [panel (C)], and fulfilling ORAL inclusion criteria [Panel (D)]. Abbreviations: CV, cardio-cerebrovascular; TNFis, tumor necrosis factor inhibitors; JAKis, Janus kinase inhibitors; CI, confidence interval; e-RHD-FI, electronic-regional healthcare database frailty index.