A methodological review identified several options for utilizing registries for randomized controlled trials.

Abstract

Objective: Registry-based Randomized Controlled Trials (RRCTs) can provide internally valid results in a real-world context at relatively low effort and cost. However, the main characteristics, the extent to which the registry is utilized (e.g., proportion of data from registry), and registry-related limitations are not well characterized. This methodological review of RRCTs aims to analyze the trial design features, investigate potential usage options, and identify possible limitations of using registry data for RCTs.

Study design and setting: A systematic search in PubMed for ongoing and published RRCTs was conducted up to 2023/28/02. Studies that reported at least one outcome derived from a registry were included. Study selection was independently performed by two reviewers. All data were extracted into a standardized table, and descriptive statistics were generated.

Results: We included 162 RRCTs (41 protocols and 121 studies). Most RRCTs were multicenter trials (n=127; 78.4%) comprising a large number of participants (median=1,787; range=41-683,927) and a long follow-up period (median=60 months; range=1-367 months) with a minimal loss to follow-up. The inclusion criteria of participants were mostly broadly defined. Types of interventions ranged from surgical procedures to behavioral interventions and almost half of the interventions (46.9%) had a preventive purpose. The main registry outcome was mostly a clinical endpoint (40.1%) or a composite endpoint of major clinical events (30.9%) that was objectively measurable. We found different degrees of registry utilization, ranging from the exclusive use of long-term monitoring of previously published data to the more comprehensive registry utilization for patient recruitment, endpoint collection, and long-term follow-up. Limitations related to the use of registry data comprised potential coding errors or incomplete data (e.g., due to under-recording of mild cases). In addition, technical challenges must be considered (e.g., failed linkages or time-delayed data entry).

Conclusions: A broad spectrum of potential usage options and usage extent of registry data exist. Our analysis suggests that in many cases, the potential of using registry data and thus their benefits were not fully utilized. In addition, the study illustrates that there is not a single, unified methodology for designing RRCTs but that registries can support RCTs in various ways. Therefore, future RRCTs should specify for what purposes and to what extent registries were utilized. Moreover, a clear definition and taxonomy of RRCTs appears necessary for facilitating future dialogue and research on RRCTs.