OBJECTIVES: KADor is a retrospective national observational study combining a cross-sectional registry and a cohort design. That registry aimed to estimate the size of the French target population: HER2+ eBC patients with residual disease (non-pCR) at surgery after neoadjuvant chemotherapy and trastuzumab-based therapy.

METHODS: The population size was estimated using a 5-steps method: First, a Poisson model was developed on active sites while considering their location, their level and type of activity to predict the number of HER2+ eBC patients treated with trastuzumab-based neoadjuvant. A sensitivity analysis with a negative binomial model was conducted to control for data overdispersion. Then, a 5-fold-validation assessed the predictive quality of the models. Thirdly, the model was applied to all French sites that were identified through the national hospitalization database (PMSI) for extrapolation. Subsequently, the proportion of non-pCR patients (as locally defined) was estimated from the registry and applied to the size estimation. Finally, a bootstrap method with 1000 random samples was used to construct a 95% confidence interval (95%CI).

RESULTS: Out of 460 French sites, 48 included at least one patient in the registry and were included in the model. The latter were evenly geographically distributed across France and all types of sites (academic, community, private) were represented. The potential number of new patients over one year in France with HER2+ eBC treated with neoadjuvant trastuzumab followed by surgery estimated with the Poisson model was 2015 patients (95%CI: 1516-2871). Among them, non-pCR patients were estimated at 1152 (95%CI: 867-1642). Results from the negative binomial model were consistent with the Poisson model.

CONCLUSIONS: The estimation relies on the combination of data from a dedicated registry with the French exhaustive hospitalization database (PMSI). This study uses real world data in a predictive model to estimate the size of the French HER2+ eBC patients with non-pCR.