(231) Lessons learned for data harmonisation and integration from secondary use of randomised controlled trial (RCT) and real-world data in the context of a study to evaluate external comparator arm study results versus RCT results

Publication Number: 564

Background: A study using secondary data was performed to advance knowledge around using single-arm trials (SAT) with external comparators arms (ECA) for cancer drug development.

Objectives: To share lessons learned regarding data harmonisation and integration in the context of this study on using real-world data (RWD) to generate an ECA for completed RCTs.

Methods: Data from the experimental arm of an RCT was used as if it was a SAT and analysed with an ECA from RWD. To identify suitable secondary data for this study, selection criteria were defined for both RCT and RWD. For RCTs access to individual patient data was required and overall survival results had to be available and significant. RCT data for two indications were obtained via Yale University Open Data Access and Project Data Sphere. RWD for the same indications were utilised via the Guardian Research Network, a network of regional community health systems in the United States with experienced cancer research programmes and organised patient electronic medical record data. RWD were harmonised with RCT data.

Results: First, feasibility considerations played a prominent role in establishing sets of data for this study. Access to RCT data is still limited, sometimes only aggregated data are shared, and there is a delay between trial completion and data becoming available.
Second, in working with RCT data sometimes results as reported in publications on the study were difficult to reproduce. Preferably, together with datasets from an RCT, access is also provided to the protocol, SAP and study report(s).
Third, harmonising RWD with RCT data took a large effort. Not all prioritized eligibility criteria were available as structured variables in the RWD or required verification which also made it difficult to assess expected sample size. Understanding the raw data and constructing derived variables was a crucial and intensive step for some covariates. Further, there remained barriers in using the RWD in terms of capturing all relevant variables in a sufficiently structured way and completeness of data. The impact of these limitations was different depending on whether variables were part of eligibility criteria, treatments compared, important covariates or outcomes of interest.

Conclusions: An important learning from this study is regarding the extensive effort required to complete data harmonisation and integration. The study insights and lessons learned are useful when planning ECA studies and may also stimulate increased access to data for this type of research. To optimize planning of a regular ECA study it is important to have opportunity for extensive feasibility assessment and to consider different types of RWD sources and opportunities for data enhancement.