Prediction of Oncotype DX and TAILORx risk categories using histopathological and immunohistochemical markers by classification and regression tree (CART) analysis.

Oncotype DX is an RT-PCR assay used to predict which patients with ER-positive node-negative (NN) disease will benefit from chemotherapy. Each patient is stratified into a risk category based on a recurrence score (RS) and the TAILORx trial is determining the benefit of chemotherapy for patients with mid-range RSs. We tested if Oncotype DX and TAILORx risk categories could be predicted by standard pathological features and protein markers corresponding to 10 genes in the assay (ER, PR, Ki67, HER2, BCL2, CD68, Aurora A kinase, survivin, cyclin B1 and BAG1) on 52 patients who enrolled on TAILORx. Immunohistochemistry for the protein markers was performed on whole tissue sections. Classification and regression tree (CART) analysis correctly classified 69% of cases into Oncotype DX risk categories based on the expression of PR, survivin and nuclear pleomorphism. All tumours with PR staining (Allred score ≥ 2) and marked nuclear pleomorphism were in the high-risk category. No case with PR <2, low survivin (≤ 15.5%) and nuclear pleomorphism <3 was high-risk. Similarly, 77% of cases were correctly classified into TAILORx categories based on nuclear pleomorphism, survivin, BAG1 and cyclin B1. Ki67 was the only variable that predicted the absolute RS with a cut-off for positivity of 15% (p = 0.003). In conclusion, CART revealed key predictors including proliferation markers, PR and nuclear pleomorphism that correctly classified over two thirds of ER-positive NN cancers into Oncotype DX and TAILORx risk categories. These variables could be used as an alternative to the RT-PCR assay to reduce the number of patients requiring Oncotype DX testing.