Evaluation of Postmastectomy Radiotherapy in Women with HER-2 positive Breast Cancer

Statistical Strategy: Performed by Hani Tamim (PhD in Epidemiology and Biostatistics) and Aurelie Mailhac (MS in Epidemiology and Biostatistics)1. Power CalculationThe planned study is a secondary data analysis, which entails a predetermined sample size. Based on the published numbers in the primary study on this cohort [11] the estimated numbers of participants with mastectomy in the PMRT(+) and PMRT(-) are of 1214 and 724, respectively. We estimated the LRR proportion in the unexposed group to be of 15%. With an expected 5% reduction in the exposed group (LRR proportion = 10%) and an alpha of 5%, the power calculation yielded 80%. Larger reductions will yield higher power.For the subgroup analyses (HER-2 positive, pT1-pT2 tumors, post-mastectomy, on Trastuzumab) with the estimated numbers of 364 participants in the exposed group and 217 in the unexposed group, an LRR proportion of 10% among the non-exposed, a reduction of 6% in the exposed (LRR proportion = 4%) and an alpha of 5%, the power calculation yielded 75.4%.2. Statistical AnalysisUnivariate analyses will be conducted to describe the cohort. Categorical variables will be presented as frequencies and percent, and continuous variables as means and standard deviations (SD). Bivariate analyses will be used to compare characteristics as well as outcomes in frequencies and percent between exposed PMRT(+) and unexposed PMRT(-) groups, using the independent t-test (continuous variables) or the chi-square test (categorical variables). Multivariate regression analyses will be used to evaluate the association between radiation and LRR while adjusting for potentially confounding variables. Moreover, results will be presented as odds ratios (OR) with their associated 95% confidence intervals (95% CI). Factors considered for inclusion in the multivariate analyses will depending on the statistical significance of the bivariate analyses, as well as the clinical significance of different characteristics. Moreover, Kaplan Meier curves will be constructed to describe the time to recurrence in both radiation groups. Furthermore, to adjust for the confounding variables, while considering time to recurrence as an outcome, Cox proportional hazards regression modeling will be used to estimate the Hazard Ratios (HR) with 95% CI. Statistical significance will be set at an alpha of 5% for a two-sided p-value. Analyses will be conducted using SAS 9.4.