PredictSelect - Identifying breast cancer patients who would benefit from EndoPredict Testing

Abstract

BACKGROUND AND PURPOSE: EndoPredict is a gene expression profiling (GEP) test that can risk stratify patients with ER-positive, HER2-negative breast cancer, by providing an EPClin score corresponding to the risk of distant recurrence within 10 years. However, like other GEP tests, its use is controversial due to high cost (currently 2980AUD) and reduced accessibility. Traditional clinical and pathological features can be used to predict chemotherapy benefit in most patients but there is a subset of patients that would benefit from GEP testing. The aim of this study was to create a clinical decision tool to identify those patients who would most benefit from EndoPredict testing. METHODS: This multi-institutional retrospective cohort study included all patients who had EndoPredict testing at DHM Pathology between 2017 and 2018. Data collected included all clinicopathological variables available on routine histopathology reports. Backwards stepwise linear regression modelled the molecular component of the EPClin score and an accompanying 95% prediction interval. Separate models for overall and hotspot Ki-67 were included as the latter is not always reported. The estimated molecular scores were included in the EPClin formula alongside observed lesion size and nodal status values to produce estimated EPClin scores. RESULTS: The model created based on this cohort of 369 patients included the variables lesion size, nodal status, mitotic rate, Ki-67 and PR percentage. The model using overall Ki-67 predicted 59 low risk patients (94.92% specificity), 6 high risk patients (100% sensitivity) and 304 uncertain cases. The model using hotspot Ki-67 predicted 50 low risk patients (92% specificity), 7 high risk patients (100% sensitivity) and 152 uncertain cases. The uncertain cases would be the subset that we would recommend to have EndoPredict testing. Using this model, patients can avoid EndoPredict testing in 18% of cases using the overall Ki-67 model and 27% using the hotspot Ki-67 model resulting in cost savings of 193,700AUD and 343,000 AUD respectively. CONCLUSIONS: In patients with ER-positive, HER2-negative breast cancer, this model can be used to select patients who can safely forego EndoPredict testing and those who should have further EndoPredict testing.