Fig. 4
ROC curve with optimal cut-off point. a PBI1‘no benefit’ (n = 32, AUC = 0.61). b PBI1 ‘benefit’ (n = 86, AUC = 0.63). c PBI2 no benefit’ (n = 32, AUC = 0.62). d PBI2 ‘benefit’ (n = 86, AUC = 0.63). ROC = receiver operating characteristic, PBI = Patient Benefit Index, AUC = Area under de curveThe anchor-based MIC distribution is displayed in Fig. 5. As visualised in Fig. 5, the SDC is larger than the MIC, especially for PBI2. There is much overlap between the curves, leading to much misclassification