Abstract P3-10-09: Genomic Grade Index (GGI): Tumor Grading Performance and Prognostic Value Compared to Ki67 and Mitotic Index in Early Invasive Breast Cancer — A Reference Centre Experience

Background: Genomic grading represents a molecular approach to tumor grading. The GGI is a 97-gene assay which improves tumor grading by resolving histological grade (HG) 2 into high or low genomic grade. GGI carries added clinical value particularly in ER+/HER2-/N0 tumors where grade is a key decision factor for adjuvant treatment. The GGI algorithm has been validated for use in routine practice, including the definition of an appropriate cut-off and statistical confidence interval corresponding to a 3:1 odds-ratio. This study aimed at retrospectively testing the GGI classification performance and prognostic value in a series of 169 early breast cancers from a comprehensive cancer centre. Methods: Female breast cancers from the Institut Curie Database were selected on the following criteria: small size (pT1-2), node negative (pN0), 10-year follow-up data, availability of extracted RNA and frozen tissue. Genomic profiles were obtained using Affymetrix HGU133 Plus 2.0 gene chips. GGI was computed using Ipsogen MapQuant Dx®. Elston-Ellis (EE) grade, mitotic index (MI, mitosis/mm2), and Ki-67 IHC (continuous -% of positive cells-, or binary-median=20%-) were assessed retrospectively on the most representative tumor block in routine conditions by several pathologists. The GGI, pathological features, and proliferation markers were correlated. Survival analyses were performed using Kaplan-Meier, with comparisons using the logrank test and hazard ratios estimated using Cox proportional hazard model. Results: Profiles were obtained in 163 cases (96%).Tumor size ranged from 7 to 45 mm. 86% of tumors were ER+, and 94% were HER2-, 32.5% were EE-1, 43% EE-2 (76% mitotic score 1 and 24% mitotic score 2-3), 24.5% EE-3. Chemotherapy was given to 11 patients. Median follow-up was 12.9 yrs [0.5-15.2 yrs]. GGI classified 79% of all cases with a 95 % concordance with HG1 and HG3, and reclassified 69% of HG2 tumors (50% GGI-1 and 19% GGI-3). Concordance was 81% between Ki67 and GGI. In the ER+/HER2-/no chemotherapy subgroup (n=126), using GGI and Ki67 as continuous variables, GGI was the only significant factor in multivariate Cox regression including EE, age and size (GGI: HR=2.36 and p=0.005; Ki67: HR=1.02 and p=0.11). This was also true when GGI and Ki67 were analyzed as binary indexes (GGI: HR=5.23, p=0.02; Ki67: HR=2.44 and p=0.13). Relative risk (RR) were 4, 2.55 and 2.85 for GGI, Ki67 and MI resp. (p=0.007, p=0.052 and p=0.053). In the ER+/HG2 cohort (n=63, 10 metastasis events), RR were 3.27 and 1.78 for GGI and Ki67, none reaching statistical significance, certainly due to insufficient statistical power issue. Conclusion: In a cohort of pT1-2, pN0 early invasive breast cancers, GGI has a 95% concordance with EE grading (1 and 3) and reclassifies 69% of EE-2 tumors. In ER+/pN0/HER2- breast cancer, GGI, either continuous or binary, has a higher prognostic value than Ki67 or MI. In a reference comprehensive cancer centre setting, GGI should add clinical information in this particular breast cancer subgroup where adjuvant treatment decision remains a routine challenge. GGI and EE tumor grading: contingency table Citation Information: Cancer Res 2010;70(24 Suppl):Abstract nr P3-10-09.