ILC cases in our pathology database with Oncotype DX testing were identified. Histologic tumor characteristics, immunohistochemical (IHC) of estrogen receptor (ER)/progesterone receptor (PgR) percent, HER2, E-cadherin expression, and Ki-67 levels were obtained for cases. Discriminant analysis was used to test the hypothesis that tumors classified as lower/higher risk based on Oncotype DX RS would differ significantly on a linear combination of variables.
From 2006 – 2014, 158 cases of ILC having Oncotype DX testing were identified; 90 low risk (RS < 18), 66 intermediate risk (RS 18 – 30) and 2 high risk (RS > 30). Discriminant analysis showed that PgR% followed by Ki-67 provided the greatest contribution to discern low versus elevated RS. A subset of 57 cases (∼36%) with predicted probabilities > 86% for either low or high RS yielded 96.5% correct classification, 92.3% sensitivity, and 97.7% specificity.
Our analytical model may be useful in predicting lower RR in patients with ILC. If validated, this provides a faster and less expensive alternative to Oncotype DX testing in certain patients with ILC.