Date of Award

May 2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Biomedical Engineering

First Advisor

Enrique Alvarez Vazquez

Abstract

Demographic underrepresentation in clinical trials continues to limit the external validity, equity, and generalizability of biomedical evidence. While regulatory reforms have increased the availability of aggregate demographic reporting, most existing approaches assess representation retrospectively, after recruitment has begun, offering limited support for design-stage decision-making. This dissertation presents a digital twin–based framework for prospective assessment of demographic diversity in clinical trials using trial-level metadata. The work integrates three core contributions. First, an open, reproducible, no-code extraction tool was developed to enable bulk retrieval and harmonization of baseline demographic data (race, ethnicity, sex/gender, and age) from ClinicalTrials.gov using National Clinical Trial identifiers. The tool standardizes heterogeneous reporting formats to U.S. Office of Management and Budget and FDA conventions, reducing technical barriers to equity audits and portfolio-level analyses. Applied to over 2,400 U.S. cardiovascular trials registered after implementation of the Final Rule (42 CFR Part 11), the tool reveals high reporting completeness for sex and age, substantially lower availability for ethnicity, and persistent representation gaps across demographic groups. Second, a multi-output digital twin model was developed to predict trial-level demographic enrollment proportions from structured and narrative protocol features prior to re- cruitment. Model outputs are summarized using a decomposable Clinical Diversity Representation (CDR) score based on an ICER-style, disease-burden–adjusted formulation, providing an interpretable, fixed-scale measure of demographic alignment across race, sex, and age domains. Finally, these components were operational- ized within an interactive application-layer digital twin that supports deterministic inference and human-in-the-loop “what-if” exploration of trial design choices. A rule-based neuro-fuzzy policy layer translates quantitative representation shortfalls into qualitative design guidance without prescribing or optimizing trial designs. To- gether, this work demonstrates how digital twin methods can support governance- aware, design-stage decision support for improving demographic diversity in clinical trials.

Download

Share

COinS