P338: Applying Target Trial Emulation to Natural History Studies for Regulatory Decision Making in Biomarker-Defined Oncology
Poster Presenter
Priyanka Bobbili
Vice President
Analysis Group, Inc. United States
Objectives
To demonstrate how a target trial emulation (TTE) framework can be applied to construct a regulatory-grade natural history study in biomarker-defined oncology populations to support contextualization of single-arm trials and inform clinical development strategy.
Method
Using curated U.S. oncology EHR data, a retrospective natural history study was designed under a TTE framework to emulate key design elements of the single-arm clinical trial in previously untreated stage II/III dMMR/MSI-H rectal cancer.
Results
The TTE framework was operationalized across five core components: eligibility criteria, time zero (index date), treatment strategy, outcome definitions, and analytic approach.
First, eligibility criteria were explicitly aligned to those of the target trial, including newly diagnosed, previously untreated stage II/III dMMR/MSI-H rectal cancer. Biomarker status was ascertained using structured laboratory data and scalable AI-aided extraction of unstructured pathology reports, ensuring mutation-defined population identification consistent with a trial-like cohort. Second, time zero was defined to mirror the estimand of interest. For response-based endpoints, follow-up began at the first disease assessment following completion of standard-of-care therapy; for event-free survival endpoints, follow-up began at initiation of standard-of-care treatment. This minimized immortal time bias and aligned real-world and trial estimands. Third, treatment strategies were defined to reflect real-world standard of care, including chemoradiotherapy, total neoadjuvant therapy, surgery, and non-operative management. Treatment sequencing was curated through structured data and clinician-verified chart abstraction to reduce misclassification. Fourth, endpoints were operationalized to parallel clinical trial definitions. Disease assessments were derived from radiology and endoscopy reports using predefined abstraction rules, with sensitivity analyses to address missing or variably timed assessments. Fifth, analytic populations were constructed to support both a broad natural history cohort and a trial-emulation subgroup applying additional eligibility constraints. Sensitivity analyses evaluated data completeness, surveillance frequency, and potential misclassification. Collectively, this approach illustrates how TTE principles enhance internal validity, transparency, and regulatory interpretability of natural history studies derived from oncology EHR data.
Conclusion
Target trial emulation provides a structured, estimand-driven framework for designing natural history studies capable of supporting regulatory decision-making. Rigorously designed real-world contextualization may be critical in biomarker-defined populations such as stage II/III dMMR/MSI-H rectal cancer, where low prevalence and limited use of immuno-oncology therapies in routine practice make randomized comparator arms difficult to implement, and single-arm trials are necessary. By explicitly aligning eligibility criteria, time zero, treatment definitions, outcome operationalization, and analytic strategy with a hypothetical target trial, TTE reduces bias, improves interpretability, and facilitates transparent comparison between real-world control and interventional treatment evidence. The approach also requires feasibility assessment of fit-for-use real-world data sources and proactive and transparent identification of data limitations, conducts sensitivity analyses to assess robustness of results, evaluates intercurrent events, and clarifies assumptions required for regulatory use.
Beyond contextualizing single-arm trials, TTE-based natural history studies can inform endpoint selection, feasibility assessment, sample size planning, biomarker testing patterns, and expected standard-of-care outcomes during early clinical development. When implemented using curated oncology EHR data with technology-enabled abstraction and quality controls, such studies can achieve a level of methodological rigor aligned with regulatory expectations. As precision oncology increasingly focuses on molecularly defined subpopulations, TTE offers a scalable and reproducible strategy to generate fit-for-purpose natural history evidence, strengthening the bridge between real-world data and regulatory-grade decision making.
