Is oncotype DX recurrence score (RS) of prognostic value once HER2-positive and. low-ER expression patients are removed?

Oncotype DX has been criticized for not providing significantly more prognostic information than histopathologic analysis. Oncotype DX was validated in cohorts that included poor prognostic factors (HER2-positive, low-estrogen receptor [ER] expression), raising the question: if patients with known high recurrence rates are excluded, is the Recurrence Score (RS) still valid? Our purpose was to determine if RS can be predicted with readily available measures. One hundred and twenty samples from August 2006 to November 2010 that underwent Oncotype DX testing were analyzed. Data included RS, ER, progesterone receptor (PR), HER2, and Ki67 status by immunohistochemistry (IHC). IHC data were used to create two linear regression models to predict RS. SAS's JMP-7 was used for statistical analysis. When comparing Oncotype DX- and IHC-derived ER and PR values, there were 21 discordant samples. The linear regression model PRS-F created with IHC data (ER, PR, HER2, Ki67) from all samples (n = 120) had an adjusted R(2) = 0.60 indicating a good model for predicting RS. The PRS-R model was built without low-ER and HER2-positive samples (n = 110). It had an adjusted R(2) = 0.38 indicating poor prediction of RS. Oncotype DX data showed good concordance with IHC for ER- and PR-expression in this cohort. Low-ER samples had high RS. After removing low-ER and HER2-positives, calculating RS with PRS-R from remaining data showed poor predictive power for RS (adjusted R(2) = 0.38). This result questions whether RS is prognostic in this subgroup (who would most benefit from further clarification of recurrence risk) and independent of pathology, or is simply producing random RS values. Data bases available to Genomic Health can resolve this issue.