(SLB-058) Data Source Selection for Clinical Trial Diversity Planning: Rare Data in Rare Disease

Background: There is increasing awareness of the need to enroll participants in clinical trials that represent the real world population for which treatment is intended. To inform clinical trial planning, characterization of the racial/ethnic profile of the indicated patient population is of utility. Particularly for rare conditions, however, identifying data for the establishment of diverse patient cohorts poses challenges.

Objectives: To characterize the racial and/or ethnic attributes of US patient populations sourced from disease-specific registries relative to those found in the TriNetX database across two different rare disease conditions and assess their generalizability.

Methods: Categorical data on race and/or ethnicity were collated for US patients enrolled in two rare disease registries, as well as those identified with the corresponding conditions in the TriNetX electronic health record database among US patients. Race and/or ethnicity distribution for the patient cohorts obtained from the registries and TriNetX were compared for each indication, in addition to being compared to the total US population from the 2020 Decennial Census.

Results: TriNetX and the rare disease registries categorized race/ethnicity consistent with FDA guidance. The overall racial/ethnic distribution of patients in the TriNetX database closely mirrored the US census; a similar assessment could not be performed for the registry population. Data on ethnicity was collected only in one registry and was >99% available but missing for 19.5-25.6% of the TriNetX cohorts. Non-Hispanic patients were overrepresented within the registry (90.2%) and underrepresented in the TriNetX cohorts (63.4-71.1%), perhaps due to missing data, as compared to the US Census (81.3%). White patients (73.7-77.4%) were also overrepresented in the registry cohorts (100% data availability) compared to the US Census (61.6%). TriNetX yielded cohorts that closely approximated the racial distribution of the US population with the difference in proportion per race ranging from 0.4-4.0% for the 83.5-89.1% of patients with available data.

Conclusions: For rare diseases, real world data sources leveraged to inform clinical trial diversity planning may have limitations with respect to available data and representativeness. While TriNetX yielded a racial patient distribution more consistent with the US Census, it had a higher proportion of missing data as compared to patient registries. When establishing trial targets, available epidemiological data should consider such limitations and be complemented by available data on the pathophysiology and clinical management of the disease in different racial and ethnic populations.