Linking an early access program (EAP) to the National Health Data System (NHDS) in France and assessing its performances and potential biases.

Abstract

11137 Background: Early access programs (EAP) are a unique opportunity to document first use of innovative treatments in real-world settings. The French HTA body requires mandatory patients’ clinical data collection for EAP and since July 2021, has recommended anticipating linkage EAP to the National Health Data System (NHDS). Prior to drug initiation, the collected data is almost complete contrary to follow-up data. To the best of our knowledge, no other study has proposed a methodology to link EAP to the French Cancer cohort (FCC), an extract of the NHDS of over 8 million persons diagnosed or at high risk of cancer, and to assess its assets and downsides. Methods: This cohort study has reused data of informed French EAP patients treated with durvalumab from October 1st, 2017 to December 31st, 2018 (n=457) in locally advanced unresectable NSCLC and of patients who received durvalumab at the time of the EAP in the FCC (n=666). The linkage methodology consists in harmonizing available data across sources, defining matching, control and study variables, testing several indirect deterministic linkage algorithms and assessing their performances. Consistency and differences between data sources were described for the best performing algorithm. Results: Proposed linkage algorithms required matches on a combination of data among sex, birth year and month, treating hospital ID and department. Performance started at 80.9% (370/457) patients from the EAP cohort linked to eligible patients from FCC up to 85.3% (390/457) for the best performing algorithm. The consistency of linkage variables between the two sources was substantial (Cohen’s kappa > 0.8) or excellent (intraclass correlation > 0.9). It was the same for control variables. Comparisons between linked patients and unlinked patients from the EAP cohort showed no statistically significant difference on matching and control variables, except for few cases on history of chemotherapy and history of radiation therapy prior to durvalumab. Similarly, no statistically significant difference was found on matching and control variables between linked patients and unlinked patients from FCC, except for history of radiation therapy and treating hospital department. Median of time difference between foreseen date of first durvalumab administration in EAP versus the effective date in FCC was -1 (IQR = [-9 ; 0]) day. Lack of reported diagnosis related to prior radiation therapy administration in out-of-hospital settings refrained from using them as matching or control variables. Conclusions: The study illustrates the feasibility of matching EAP patients to the French NDHS with indirect deterministic linkage. Additional EAP linkages will be required to ensure the validation and reproducibility of such methodology. Further works in progress will document the adequacy of French NHDS data to address effectiveness.