PCN216 ADJUVANT CHEMOTHERAPY GUIDED BY A 21-GENE EXPRESSION ASSAY IN HORMONE-RECEPTOR POSITIVE EARLY-STAGE BREAST CANCER: USE OF SIMULATION MODELING TO EXTEND AND TRANSLATE CLINICAL TRIAL RESULTS INTO PRACTICE

Abstract

Translation of clinical trial findings in to practice requires information on variation of trial results in real-world settings and outcomes beyond trial follow-up. We developed a simulation model to extend the "Trial Assigning IndividuaLized Options for Treatment” (TAILORx) and examine long-term outcomes and variation of trial findings for clinically-relevant subgroups of women. We conducted an empirical Bayesian analysis with a Monte Carlo sampling method for probabilistic sensitivity analysis using data independent of TAILORx. Women eligible were diagnosed with HER2-, hormone receptor positive (HR+), node–negative breast cancer and recurrence scores (RS) between 11-25. The trial was also simulated for alternative RS categories including 11-30, 18-30, and 26-30. Women were followed up to 20-years to assess long-term outcomes. Input parameters were derived from SEER, Oxford Overview and published data. Hazard ratios (HRs) of no-chemo vs chemo were estimated using Cox models. To assess predictive validity of the trial simulations, we compared estimates for each trial endpoint (invasive disease-free interval (iDFI) and distant recurrence-free interval interval (DRFI)) with TAILORx published results. The HRs for no-chemo vs. chemo for 9 and 20-year iDFI and DRFI were not statistically significant in the 11-25 and 18-30 RS groups. Majority (72%) of the simulations showed that omission of chemotherapy resulted in a statistically significant decrease in the 9-year iDFI rate among women with RS 26-30. The results varied by age. On average, simulated results were similar to the published TAILORx findings at 9-years in the RS 11-25 group. The model projections show that long-term recurrence rates remain low with omission of chemotherapy among women diagnosed with node-negative, HR+ breast cancer and RS 11-25 and 18-30. Chemotherapy may show some benefit in reducing recurrence risk in RS 26-30 category. Simulation modeling can be used to extend trial data by synthesizing information from historical trials.