Fig. 2

HER2 copy number investigated with subpopulation treatment effect pattern plot (STEPP) and survival analysis. STEPP investigated the relationship between manually- (A) and digital image analysis (DIA) algorithm-scored (C) mean HER2 copy number and the probability of being recurrence-free at the median recurrence-free survival (RFS) (8.7 years). The lower left corner in A, C shows the cutoff levels suggested by STEPP analysis and identified by positional scanning (PS). These cutoffs divided the cohort into three subgroups, namely low, intermediate and high. The locally estimated scatterplot smoothing curve fit is shown with the blue line. Kaplan–Meier estimates and multivariable Cox regression model of manually- (B) and DIA algorithm-scored (D) HER2 copy number related to RFS. The study cohort was divided into three subgroups according to the suggested optimal first and second cutoffs for the manually scored HER2 copy number and the DIA algorithm scored HER2 copy number as shown in A, C. *The multivariable Cox regression model was adjusted for lymph node status, tumor size and stratified by estrogen receptor status. Adjusted hazard ratios (HR) with 95% confidence interval (CI) are presented