Comparative safety and effectiveness of first-line conventional synthetic disease-modifying antirheumatic drugs in adults with newly diagnosed rheumatoid arthritis: a real-world, population-based cohort study using the United Kingdom clinical practice research datalink aurum

Kadi, Tahaneiy 2025. Comparative safety and effectiveness of first-line conventional synthetic disease-modifying antirheumatic drugs in adults with newly diagnosed rheumatoid arthritis: a real-world, population-based cohort study using the United Kingdom clinical practice research datalink aurum. PhD Thesis, Cardiff University. Item availability restricted.

	PDF - Accepted Post-Print Version Restricted to Repository staff only until 14 May 2027 due to copyright restrictions. Download (9MB)
	PDF (Cardiff University Electronic Publication Form) - Supplemental Material Restricted to Repository staff only Download (125kB)

Abstract

To evaluate real-world patterns of first-line conventional synthetic DMARD (csDMARD) initiation, treatment failure, and comparative effectiveness among adults with newly diagnosed rheumatoid arthritis (RA) in the UK, and to assess whether baseline characteristics support personalised treatment selection. Methods A retrospective population-based cohort study was conducted using UK Clinical Practice Research Datalink (CPRD) AURUM data linked to hospital records (2012–2022). Adults (≥18 years) with incident RA initiating methotrexate (MTX), hydroxychloroquine (HCQ), sulfasalazine (SSZ), or leflunomide (LEF) were included (n = 53,991). Descriptive analyses and multinomial logistic regression identified predictors of csDMARD initiation. Treatment failure (switch, add-on, or discontinuation) was analysed using Kaplan–Meier methods and inverse probability-weighted time-dependent Cox regression with practice-level clustering, informed by a directed acyclic graph. Sensitivity analyses included RA-related hospitalisation and negative control outcomes. Random Forest and XGBoost models were used to predict treatment failure from baseline variables. Results MTX was the most frequently initiated csDMARD (39.7%), followed by HCQ (31.6%), SSZ (24.0%), and LEF (4.6%). Age, sex, ethnicity, cardiometabolic disease, and corticosteroid use were the strongest determinants of treatment choice. LEF showed the greatest treatment persistence, followed by SSZ and HCQ, while MTX had the highest failure rates. In adjusted time-dependent Cox models, LEF, SSZ, and HCQ were each associated with lower hazards of treatment failure than MTX. Cardiovascular disease, chronic kidney disease, obesity, glucocorticoid exposure, alcohol use, and socioeconomic deprivation predicted failure. Machine-learning models demonstrated modest discrimination (AUC ~0.60–0.62). ii Conclusion First-line csDMARDs selection and treatment durability in early RA are strongly shaped by clinical and sociodemographic factors. LEF demonstrated particularly favourable persistence compared with MTX. Integrating causal inference with exploratory machine learning supports more individualised prescribing while highlighting the limits of baseline EHR data for prediction.

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Schools > Pharmacy
Subjects:	Q Science > Q Science (General)
Date of First Compliant Deposit:	14 May 2026
Last Modified:	14 May 2026 15:40
URI:	https://orca.cardiff.ac.uk/id/eprint/187002

Actions (repository staff only)

Edit Item