Brief Summary

The project aims to evaluate factors underlying measurement and population variability in early cancer endpoint estimates, comparing real-world data to clinical trial data for melanoma and breast cancer. It seeks to establish the level of evidence supporting early cancer events, compare endpoint estimates using different definitions, assess how eligibility criteria stringency affects estimates, and evaluate heterogeneity across practices and clinical trials. The study utilizes electronic health record data from US community oncology clinics to generate insights on real-world endpoint variability and sources of heterogeneity, with the goal of informing future research.

Expanded Summary

Background and Objectives

This study focuses on understanding the variability in early cancer endpoint estimates between real-world data (RWD) and clinical trial data, specifically for melanoma and breast cancer. The research acknowledges that while clinical trials are often considered the gold standard for endpoint estimates, they may not fully represent real-world patient populations due to stringent eligibility criteria. The project aims to bridge this gap by analyzing data from electronic health records (EHRs) of US community oncology clinics.

The study has four primary objectives:

1. Establish the level of evidence underlying early cancer event determinations and its impact on time-to-event estimates.
2. Compare endpoint estimates derived from real-world definitions with those from clinical trial definitions.
3. Evaluate how endpoint estimates differ across levels of eligibility criteria stringency.
4. Assess heterogeneity in endpoint estimates across real-world practices and clinical trials.

Methodology

The study employs a retrospective observational cohort design, utilizing data from the iKM EHR system of The US Oncology Network and non-Network practices. The research focuses on patients diagnosed with early-stage (II and III) melanoma and breast cancer.

Key methodological approaches include:

• Additional abstraction of medical charts for a subset of patients to gather detailed information on the level of evidence supporting early cancer events.
• Comparison of time-to-event estimates across different sources of evidence (e.g., pathology reports, scan reports, progress notes).
• Development and application of "real-world stringent eligibility criteria" to mirror clinical trial populations within the real-world data.
• Statistical analyses including ANOVA, t-tests, multivariable linear regression, and mixed-effects models to evaluate differences in endpoint estimates and their variability.