Key points are not available for this paper at this time.
Abstract Background: Molecular profiling is powerful to match patients’ tumor profile with targeted therapies in oncology. While our understanding of treatments and biomarkers that enhance overall response rates is growing, the level of consistency at which treatment response is assured in biomarker-enriched (BE) cohorts (e. g. immune-checkpoint inhibitors in PD-L1-positive tumors) remains poorly understood. In essence, an optimal biomarker would not only result in high mean survival but also precise, closely clustered response of all patients around this mean. Objectives: The primary objective of this study is to develop a method to measure the variability of survival outcomes from published Kaplan-Meier (KM) curves in trials. The secondary objective is to assess whether novel treatments, when given in the presence of predictive biomarkers, improve not only mean survival, but also its precision. This is clinically relevant as higher precision guarantees patients a treatment outcome closer to the mean, thereby providing safeguards against non-response. Methods: We conducted a criteria-based search in PubMed and EMBASE to identify all phase-II and -III drug trials on breast cancer and non-small cell lung cancer (NSCLC) published between 2018-2022. We developed an algorithm to derive pseudo-individual patient data (p-IPD) from KM curves. We measured variability of survival outcomes in treatment and control groups using ratios of coefficients of variation (CVRs) derived from log-normal models and restricted mean survival time. This novel computational approach enables comparing the consistency of treatment responses in BE subgroups versus total, i. e. intention-to-treat (ITT), populations in each trial. Results: From screening 780 publication records (NSCLC: n = 405; breast cancer: n = 375), we identified 67 biomarker-stratified drug trials (NSCLC: n = 41; breast cancer: n = 26). We successfully constructed p-IPD using KM curves from ITT and BE cohorts for the first 18 trials (NSCLC: n = 10; breast cancer: n = 8) from a total of 11, 373 patients (NSCLC: n = 7, 547; breast cancer: n = 3, 826). The preliminary meta-analysis for overall survival on these first 18 trials shows that BE treatment groups respond more precisely, with a variability reduction from 0. 93 95%-CI: 0. 91, 0. 96 in the ITT group to 0. 86 0. 83, 0. 91 in the BE subgroup. This variability reduction varies across treatment, biomarker, and tumor types. Finally, we observe different variability estimates in trials (n = 6) that are stratified by subgroups of different PD-L1 expression. Conclusion: This is the first study that presents a viable approach to analyze the variability of biomarker-treatment pairs in oncology at large scale. We find a reduction of treatment variability in breast and lung cancer trials in BE populations. Our work offers an orthogonal approach to measure biomarker utility with the potential to inform biomarker discovery and trial design. Citation Format: Maximilian Schuessler, Elizaveta Skarga, Pascal Geldsetzer, Ying Lu, Maike Hohberg. Drivers of precision in oncology trials: A landscape analysis of biomarkers and treatments abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts) ; 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84 (6Suppl): Abstract nr 2387.
Schuessler et al. (Fri,) studied this question.