Predicting Oncotype DX score with clinicopathological variables.

Abstract

e11015 Background: Oncotype Dx (OD) is a commercially available gene-profiling test that has been shown to be effective in predicting risk of recurrence after therapy with CMF or Tamoxifen. The aim of our study is to develop a method that uses easily available clinical data to predict cases with favorable OD category (ODC). Methods: We identified 145 ER+/Her-2 neg cases that had information on histologic grade (HG), PR status and Oncotype Dx Category (ODC). Of these, 75 also had information on Ki-67. We evaluated HG, ER status, PR status, Nottingham score and tumor size for their correlation with ODC using multiple regression analysis. Results: The median tumor size was 12.0 mm (0.6-45). 33.7% were HG 1, 54% were grade 2, and 12% were grade 3. 88% were invasive ductal carcinomas, 9% were invasive lobular carcinomas, 3% were other types. The median OD score was 15 (0-57) with 86 (59%) ODC=1, 53 (37%) ODC=2 and 6 (4%) ODC=3; median Ki-67 was 11% (1-50). In a multivariate analysis excluding Ki-67, the only factors that were significantly correlated with ODC were PR status and HG (p= 0.004). PR status had the strongest overall association with ODC (P=0.008). Defining PR> 60% with HG=1 as a positive result we calculated the positive predictive value (PPV) or the capacity to predict cases with ODC=1, as 85% and negative predictive value (NPV) as 46%. None of the cases with PR> 60% and HG=1 had an ODC=3 (high risk). A patient with a favorable ODC=1 was considered as a positive case for purpose of calculating PPV, NPV, sensitivity and specificity. Sensitivity to detect positive cases was 26% and specificity was 93%. We explored a 2nd model that included Ki-67 and found that only PR >60% and Ki-67 >11% significantly correlated with ODC (p=0.004). PR status again had the strongest overall association with ODC (P=0.007) followed by Ki-67 (P=0.019). HG was not significant in the 2nd multivariate model because of its strong correlation with Ki-67. PPV for this 2nd model was 79% and NPV=50%. Conclusions: 1-Pts who have PR>60% with HG=1 are highly likely to have a favorable ODC=1, however the ODC of pts with either PR<60% or HG>1 is not predictable with either of our two models. 2- The use of Ki-67 doesn’t add much to the capacity of the model to predict ODC. 3- PR is the strongest predictive factor in the model.