How well do standard prognostic criteria predict oncotype DX (ODX) scores?

Abstract

576 Background: In node-negative, ER + breast cancer, gene expression profiling can identify level of risk and, in the case of ODX, may also identify pts with a higher chance of benefiting from adjuvant chemotherapy. Because the gene profile in ODX includes an assessment of ER, HER2, and proliferation, we hypothesized that clinicians using standardized criteria could discriminate risk (high versus low/intermediate) as specified by ODX. Methods: We identified Mayo Clinic patients with node-negative, ER + breast cancer, for whom ODX scores were available. Tumor slides were reviewed by an expert breast pathologist to confirm tumor size, histology, and tumor grade. Both ER and PR were quantitated; HER-2 was determined by IHC (FISH, if 2+). These clinical cases were presented to six academic oncologists, blinded to the ODX score, to predict ODX risk (low, intermediate, or high) and give their recommendation for chemotherapy (CTX) (yes/no). Afterwards, they were presented with the same cases with the actual ODX score, to give recommendations regarding CTX. Results: ODX scores in tumors from 31 patients were low in 18 pts, intermediate in 10 pts, and high in 3 pts. Concordance between predicted and actual ODX scores being low/intermediate vs high exceeded 87% for each oncologist. The most frequent discrepancies were actual low scores predicted as intermediate (31/80 discordant) and actual intermediate scores predicted as low (29/80 discordant). Overall agreement of predicted scores (high vs low/intermediate) among the oncologists was substantial (kappa=0.75, p<0.0001). CTX recommendations following provision of the ODX scores changed on average 18.2% (range 12.9%-25.8%) of the time, with slightly fewer changing for, versus against, a CTX recommendation. Conclusions: Our findings suggest that the proper evaluation and interpretation of traditional prognostic criteria will identify most node negative, ER + patients at high risk of recurrence (as predicted by ODX) but poorly discriminate low versus intermediate risk. The provision of ODX data changed the recommendation for CTX in approximately 20% of cases. No significant financial relationships to disclose.