Researchers publishing in the Journal of the National Cancer Institute have proposed a novel real-world data framework for validating surrogate endpoints in oncology, applying it to a large retrospective cohort of patients with metastatic breast cancer (mBC). The study evaluated four intermediate clinical endpoints (ICEs), real-world progression-free survival (rwPFS), time to next line of therapy (TTNT), time to last treatment administration (TTLA), and time to end of first-line therapy (TTEFL), as potential surrogates for overall survival (OS).
Among the endpoints assessed, rwPFS emerged as the strongest surrogate for OS, satisfying both required surrogacy conditions: a high individual-level correlation with OS and a strong correlation of group-level treatment effects. The authors note that no validated framework for evaluating real-world surrogate endpoints has previously existed across most cancer types, positioning this analysis as a potential methodological blueprint for future studies.
Key finding
In a retrospective analysis of 19,164 patients with mBC drawn from the Flatiron Health database, rwPFS was the only intermediate endpoint to satisfy both individual-level and group-level surrogacy conditions for OS, though the retrospective, observational nature of the data and its derivation from a single nationwide US database may limit generalizability.
Study Design
| Study | Retrospective cohort analysis using the nationwide US Flatiron Health electronic health record–derived database |
|---|---|
| Population | 19,164 patients with mBC |
| Primary objective | Validate a novel framework for assessing intermediate clinical endpoints as surrogates for OS using real-world data |
| Endpoints evaluated | rwPFS, TTNT, TTLA of first-line therapy, and TTEFL |
| Surrogacy conditions | Condition 1: individual-level correlation between each intermediate endpoint and OS; Condition 2: correlation of group-level effects between each intermediate endpoint and OS |
The study used Kendall’s tau to assess individual-level correlations and R² to quantify group-level effect correlations across prespecified subgroups. Surrogacy thresholds required Kendall’s tau ≥ 0.5 and R² ≥ 0.7 to be met for both conditions. Median OS across the full study population was 36.3 months (95% CI, 35.6–37.2).
Results and Safety
For condition 1 (individual-level correlation), three outcomes, rwPFS, TTNT, and TTLA, met the prespecified threshold, each demonstrating a Kendall’s tau ≥ 0.5 and an R² ≥ 0.7 between 36-month OS rates and 12-month rates for each endpoint across subgroups of interest. TTEFL was not reported to have met this threshold.
For condition 2 (group-level effect correlation), only rwPFS demonstrated strong surrogacy, achieving R² ≥ 0.7 across each subgroup comparison. The other three outcomes, TTNT, TTLA, and TTEFL, did not meet this second condition. The authors conclude that rwPFS was the sole endpoint to satisfy both conditions required for validated surrogacy.
Safety data were not assessed, as this was a retrospective observational study focused on endpoint methodology rather than treatment evaluation.
Clinical Context
The absence of a validated surrogate endpoint framework for real-world data has represented a longstanding methodological gap in oncology research. Overall survival remains the gold standard for regulatory and clinical decision-making, but it requires lengthy follow-up and large patient populations to power adequately. If rwPFS can serve as a reliable OS surrogate in mBC, it could support more efficient real-world evidence generation and potentially inform regulatory and payer decisions where randomized trial data are limited or unavailable.
The authors position this analysis as a replicable blueprint for other tumor types, though key questions remain. The framework was developed and applied within a single US database, and its performance across different health systems, geographies, and cancer settings has yet to be established. Whether rwPFS surrogacy holds across specific mBC subtypes or treatment classes was not reported in the abstract, and further prospective validation efforts would help determine the broader applicability of these findings.
Sources
- Labaki C, et al. JNCI: Journal of the National Cancer Institute. “Novel framework for the validation of intermediate clinical endpoints in oncology real-world data.” Original article. 2026; djag390. View source
