OBJECTIVES: Generating real-world evidence (RWE) after clinical trials is crucial for understanding the long-term effectiveness and safety of medical interventions. The integration of real-world data (RWD) strategies together with clinical trials has many advantages to accelerate this RWE generation. One effective approach is "trial tokenization," using privacy preserving record linkage to connect clinical trial data with de-identified RWD at the patient level for post-trial health outcome assessment. Here, we outline an objective process to evaluate RWD sources for this purpose for a type 2 diabetes (T2DM) example use case., which is critical to ensure reliable RWD selection to answering the research questions.

METHODS: Computable operational definitions (cODs) were modeled for key eligibility criteria and outcomes. An assessment plan and instrument were developed. Candidate RWD sources were identified and assessed on sample size and demographics, as well as availability and completeness of outcomes, specifically weight and cardiovascular outcomes. For example, as change in weight over time was an outcome of interest, we assessed the average number of weight measurements per patient over the prior 3 years.

RESULTS: 15 standards-based cODs were developed: 5 for sample size calculations, and 10 for outcomes assessments including variable-specific completeness assessments. Additionally, 39 study-relevant data variables were assessed for their general availability in candidate data sources. Data sources were identified & contacted via the Datavant ecosystem. 369k–2.5M patients meeting the T2DM cohort criteria were found across 6 data sources. On average, 6.1–9.2 weight measurements per patient were recorded in the prior 3 years with an average range of 88–116 days between measures. Cardiovascular outcomes were reliably documented via diagnosis codes.

CONCLUSIONS: Objective assessment using standards-based cODs ensures the reliability of data to support post-trial real-world evidence generation. This enables stakeholders like regulatory agencies to consider such data with confidence and supports the use of trial tokenization for future research.