Background Observational studies may provide context for adverse events observed in clinical trial programs, particularly for infrequent events or potential long-term risks. For such purposes, however, published observational data often have limitations.
Objectives To develop methods for providing improved safety contextualization for an RA drug development program, using a coordinated and consistent approach across multiple registries.
Methods We included 4 existing RA registries: Swedish Rheumatology Registry, Sweden; Consortium of Rheumatology Researchers of North America, USA; Norfolk Arthritis Register, UK; Institute of Rheumatology Rheumatoid Arthritis cohort, Japan (Table). A new registry (CORRONA International) for Eastern Europe, Latin America and India, is also being established to improve geographical “matching” to the global clinical trial program. We identified differences and similarities across registries, and investigated how selecting restricted subcohorts might improve comparability with a typical clinical trials population. To address the potential for confounding, we identified the most important risk predictors across registries for outcomes of interest (mortality, CVD, infection, malignancy). A coordinated approach with patient-level analyses at each registry followed by a central analysis of standardized data was implemented.
Results Despite different data collection procedures and formats, the collaborative approach enabled a more consistent definition of variables for key baseline characteristics and outcomes. Selection of subcohorts based on therapy switching or joint count criteria (Table) improved baseline comparability with trial patients for some disease activity measures, e.g. DAS28, HAQ (Table; DAS28 ~5.8, HAQ ~1.6 in trial patients), but less for other characteristics, e.g. age. This, however, did decrease the sample size and statistical power (Table). For most outcomes, age was by far the most important risk predictor, emphasizing the importance of age/sex standardization. Beyond age and sex, HAQ was most consistently a strong risk predictor across registries for many outcomes incl. death; joint counts were not (Table). The prospective approach ensured optimally current and relevant data, and the distributed analysis safeguarded the confidentiality of each registry’s individual data.
Conclusions In using observational data to provide context for safety observations from clinical trials, a forward-looking coordinated approach across several data sources can provide improved comparability and consistency, and gives better support for sensitivity and exploratory analyses and interpretation of data, than using published data alone.
Acknowledgements This study was funded by AstraZeneca
Disclosure of Interest F. Nyberg Employee of: AstraZeneca, J. Askling Grant/research support from: AstraZeneca, Consultant for: AstraZeneca, J. Greenberg Shareholder of: CORRONA, Inc., Consultant for: AstraZeneca, CORRONA, Novartis, Pfizer, K. Michaud Grant/research support from: ACR Rheumatology Research Foundation; NDB receives funds from AstraZeneca, Employee of: National Data Bank for Rheumatic Diseases (NDB), H. Yamanaka Grant/research support from: IORRA cohort is supported by various grants from a large number of pharmaceutical companies including AstraZeneca, Consultant for: AstraZeneca, D. Symmons Grant/research support from: AstraZeneca, Consultant for: AstraZeneca, M. Ho Employee of: AstraZeneca