Predicting oncotype DX recurrence scores using locally available immunohistochemical markers: experience in a...

Oncotype DX testing is a reliable widely used gene assay to determine whether chemotherapy is of additional value in oestrogen receptor (ER) positive Human Epidermal Growth Factor receptor 2 (HER2) negative, node negative breast cancer, but the high cost of the test can be a barrier for optimal therapy guidance for a substantial proportion of eligible patients around the world. We aimed to determine whether the commonly available immunohistochemical markers Ki67 and progesterone receptor (PR) can predict Oncotype DX Recurrence Score (RS) scores in a district general hospital setting.

The Oncotype DX RS scores from 58 tumours were regressed against corrected Ki67 values in a simple regression model, and against ER-derived and PR-derived indices and corrected Ki67 values in a multiple model. Model portability was assessed using leave-one-out cross-validation (LOOCV).

All terms in both regression models were significantly associated with RS scores at the 5% significance level (p<0.001 for all parameters). The multiple model was a better fit to the data (adjusted R²=0.784), and performed better under LOOCV (root mean square error=7.26), suggesting good predictive capability and model portability.

Locally available, cheaper alternatives to multigene assays to determine therapy in ER positive HER2 negative patients is of benefit both from patient management and financial perspectives. A model has been derived with high capability to predict RS scores accurately from linear combinations of predictive biomarkers in a district general hospital setting, which should show good properties when applied to other samples.