Abstract P4-09-01: A multi-study correlative analysis of change in Ki67 in the neoadjuvant setting and disease free survival improvement in the adjuvant setting

Abstract

Abstract Background: Based on results from the ATAC and IMPACT trials, Dowsett et al.1 hypothesized that “short-term changes in proliferation in the neoadjuvant setting may be able to predict outcome during adjuvant use of the same treatments.” Recent neoadjuvant studies2,3,4 of novel agents for HR+ breast cancer used Ki67, a biomarker for cell proliferation, as an endpoint for evaluating biological activity. However, the potential of this endpoint for predicting the adjuvant efficacy of therapeutic agents remains highly uncertain, motivating continued study based on clinical data. After curating relevant data from multiple trials (in both the neoadjuvant and adjuvant settings), a statistical model was constructed to evaluate the current evidence regarding this hypothesis. Methods: Data were collated from randomized trials of systemic therapies in post-menopausal women with HR+ breast cancer reporting Ki67 reduction in the neoadjuvant setting (7 studies) or disease-free survival (DFS) in the adjuvant setting (5 studies). A Bayesian hierarchical joint network meta-analysis model was constructed and fit to the data to evaluate: (1) if there is an association between study-level neoadjuvant Ki67 reduction for a given regimen and corresponding adjuvant DFS hazard ratio, and (2) if the relationship does exist, what DFS effect does it predict for various regimens for which only neoadjuvant data are currently available. Results: The results of the statistical model indicate that effect-size based on neoadjuvant Ki67 correlates with adjuvant (DFS) hazard ratio. Using neoadjuvant Ki67 data for several therapies for which adjuvant data are presently unavailable, we demonstrated how forecasts (and associated confidence intervals) of adjuvant effect-size may be produced using the model. Conclusions: While randomized trials yielding paired data from both clinical settings are sparse, this assessment using a statistically rigorous approach sheds additional light on the hypothesis of a correlation between neoadjuvant Ki67 changes and adjuvant DFS improvements. This analysis suggests the potential utility of Ki67 as a surrogate endpoint to screen/prioritize experimental regimens for development in the adjuvant setting.