Detection of novel drug-adverse drug reaction signals in rheumatoid arthritis and ankylosing spondylitis: analysis of Korean real-world biologics registry data

**Background:** Biological disease-modifying antirheumatic drugs (DMARDs) and targeted therapies are effective for rheumatoid arthritis (RA) and ankylosing spondylitis (AS) but can cause unexpected adverse drug reactions (ADRs) due to immune system complexity. Pharmacovigilance using real-world data is essential, yet registry data have been underutilized for signal detection. This study aimed to discover novel ADR signals from the Korean nationwide KOBIO registry, with external validation using FAERS and a single-center common data model (CDM).

**Methods:** Data were retrieved from KOBIO (December 2012 to June 2020) for RA (2279 initial patients, 3972 follow-ups) and AS (1940 initial patients, 4969 follow-ups) who continued the same biologic DMARD or targeted therapy. ADRs reported as 'related' or 'indeterminate' were included. Relative risks (RR) were calculated from 2x2 contingency tables for each drug-ADR pair; RR>1 was considered significant. Significant pairs were checked against the latest FDA labels (2019–2021) to exclude known ADRs. Remaining pairs were externally validated using FAERS data (2012–2018) via disproportionality analysis (PRR, ROR, BCPNN with FDR<0.05) and KYUH-CDM laboratory data (2012–2019) using the MetaLAB algorithm (RR>1, p<0.05). Final candidate signals were the intersection of KOBIO and FAERS significant pairs.

**Key Results:** For RA, 51 significant drug-ADR pairs (RR>1) were identified from 8 drugs; after excluding 16 known ADRs, 35 pairs remained. For AS, 36 significant pairs from 4 drugs; after excluding 10 known ADRs, 26 pairs remained. External validation with FAERS identified 13 of 35 RA pairs and 6 of 26 AS pairs as significant (any of BCPNN, PRR, ROR with FDR<0.05). KYUH-CDM validation confirmed 2 pairs: tocilizumab-anemia and tofacitinib-thrombocytosis. The final 14 novel signals included: adalimumab (white matter lesion, benign prostatic hyperplasia); golimumab (iron deficiency anemia, spinal compression fracture, fibromyalgia, Sjogren's syndrome, endometriosis, conjunctivitis, rhinitis allergic); abatacept (cataract, asthma, interstitial lung disease); tocilizumab (osteoporosis); and tofacitinib (thrombocytosis). Tofacitinib-thrombocytosis was the only signal significant across all three datasets.

**Clinical Implications:** These 14 novel signals represent potential new ADRs not previously reported in FDA labels, warranting further evaluation. The tofacitinib-thrombocytosis signal is biologically plausible given JAK2 inhibition, and GWAS associations with JAK2 mutations support this. Clinicians should be vigilant for these ADRs when prescribing these agents. The study demonstrates the value of registry data for pharmacovigilance and the importance of multimodal signal detection. Limitations include potential underreporting, inability to integrate TNF inhibitor data across diseases, exclusion of drug switches, and single-center CDM validation. Future studies should incorporate multi-hospital CDM data and assess causal relationships.