Safety of TNF-α inhibitors, IL-17 inhibitors, and JAK inhibitors in the treatment of axial spondyloarthritis: a protocol for a multi-method study integrating FAERS data mining, dose–response network meta-analysis, and drug-target Mendelian randomization

Abstract

Biological and targeted synthetic disease-modifying anti-rheumatic drugs (b/tsDMARDs), including tumor necrosis factor-α inhibitors (TNFi), interleukin-17 inhibitors (IL-17i), and Janus kinase inhibitors (JAKi), are standard treatments for axial spondyloarthritis (axSpA). However, comparative evidence on the safety profiles of these three drug classes, particularly regarding infection-related adverse events, remains limited. No previous study has simultaneously integrated real-world pharmacovigilance data, interventional trial evidence, and causal inference methods within a triangulation framework to systematically evaluate the safety of these agents in axSpA. This study will employ three complementary components. First, a disproportionality analysis of the FDA Adverse Event Reporting System (FAERS) database (Q1 2004–Q4 2025) will use four signal detection algorithms (ROR, PRR, MGPS, BCPNN), supplemented by time-to-onset analysis (Weibull distribution), geographic analysis, and a weighted composite scoring system. Only drug–event combinations where the target drug is coded as Primary Suspect (PS) will be included in the primary analysis. Second, a systematic review incorporating dose–response network meta-analysis (NMA) of double-blind randomized controlled trials will focus on upper respiratory tract infections (URTIs) and opportunistic infections (OIs), with OIs analyzed by subtype (tuberculosis, herpes zoster, candidiasis, PJP). An Emax dose–response model will serve as the primary analytical approach, with linear and conventional NMA models as fallbacks. Databases will include PubMed, Embase, Cochrane Library (CENTRAL), Web of Science, MEDLINE via Ovid, SCOPUS, and ProQuest, in addition to the ClinicalTrials.gov trial registry. The certainty of evidence will be assessed using the CINeMA framework. Third, multivariable Mendelian randomization (MVMR) and causal mediation analysis (FLOW-MR) will employ a sequential complementary framework: MVMR will estimate independent direct effects of each drug target pathway on infection outcomes while conditioning on AS, and FLOW-MR will decompose these effects along the temporal causal cascade (AS → drug target pathway → infection outcome), enabling mediation-based target validation. Signals identified for URTIs and OIs from both FAERS and NMA will be carried forward to the final triangulation stage. By triangulating evidence from real-world pharmacovigilance, randomized trials, and genetic causal inference within a structured qualitative framework with pre-specified convergence criteria, this multi-method protocol will provide a comprehensive and robust safety evaluation of b/tsDMARDs in axSpA. The integration of dose–response modeling and a sequential MVMR-to-FLOW-MR drug-target Mendelian randomization strategy represents a methodological advancement in the safety assessment of biologic therapies. Findings will inform clinical decision-making regarding the comparative safety of these agents. PROSPERO CRD420261417010.