P5-11-02: Prediction of Results of MammaPrint's 70 Gene Signatures by Conventional Histopathological and Biological Approaches in Patients with Breast Cancer.

Abstract

Abstract Background: Several studies have shown that MammaPrint's gene signatures allow prognostic stratification of breast cancers that is superior to the currently used clinico-pathological risk factors. MammaPrint has also been reported to predict chemotherapy response. The aim of this study was to find out whether conventional histopathological and biological approaches can be used to predict MammaPrint results. Materials and Methods: 96 invasive breast carcinomas were studied (of them, 34 cases with positive lymph nodes). For MammaPrint, fresh material from breast carcinomas was sent to Agendia (Amsterdam) in the provided RNA-laters. Conventional assays included: (i) nine histopathologic parameters combined according to a so called “Graz Risk Assessment” scheme (“GRA”, see below), (ii) immunohistochemistry for MIB-1 (Ki-67) as the “only parameter”, (iii) Nottingham Prognostic Index (NPI), (iv) uPA/PAI-1 (Elisa, cut-off for uPA: 3 ng/mg, cut-off for PAI-1:14 ng/mg), and (v) flow cytometry (either >6% or >10% for SPF). The “GRA” consisted of 4 “main” factors (high nuclear grade, MIB-1 >20%), Her2/neu pos. (immunohistochemistry or FISH), vascular invasion (either lymphatic or blood vessels) or lymph node metastasis) and 5 “minor” factors (ER neg., PR neg., p53 >10%, tumor necrosis and age > 40 y.o.). Tumors were classified as high risk when at least 4 parameters or 2 “main” factors were present. For MIB-1 scoring, at least 500 cancer cells were counted in each case. Different MIB-1 cut-offs (10%, 15%, 20%, 25%, and 30% or more) were tried. All analyses were performed independently and the results of each method were blinded. Overall accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of conventional assays were calculated. The “best” combination of the assays was evaluated by multivariate logistic regression. Results: Significant association with MammaPrint on univariate analysis showed “GRA” (p=8.2e-09), MIB-1 (p=1.1e-08 at 10% cut-off), NPI (good vs. moderate + poor prognosis, p=0.001) and flow cytometry (p= 0.012 for 10% cut-off and p=0.024 for 6% cutoff for SPF). uPA/PAI, uPA alone, or PAI alone did not correlate with MammaPrint (p-value: 0.22, 0.28, and 0.27, resp.). The overall accuracy, sensitivity, specificity, PPV, and NPV of “GRA” were 79%, 72%, 86%, 83%, and 77%, resp. MIB-1 >10% as “only parameter” showed overall accuracy of 78%, sensitivity of 87%, specificity of 70%, PPV of 73%, and NPV of 85%. With increasing MIB-1 cutoff, its predictive value dropped continuously. The accuracy of NPI and flow cytometry was poor. On multivariate analysis, only “GRA” and MIB-1 >10% were significant (p-value: 0.011 and 0.017, resp.). Conclusion: The overall accuracy of histopathologic and immunohistochemical assessment in predicting the results of MammaPrint is almost 80% when using “Graz Risk Assessment” (“GRA”), closely followed by MIB-1 (at 10% cut-off) with the overall accuracy of 78%. By adjusting cut-off values for either MIB-1 or “GRA” it might be possible to reliably stratify the vast majority of breast cancers into low and high risk groups without MammaPrint, whereas MammaPrint would be required only for a minority of “uncertain” tumors. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P5-11-02.