Abstract P3-10-08: Combining Genomic Grade and NPI Refines Risk Group Classification

Abstract

Abstract Background: The Nottingham Prognostic Index (NPI) is a clinically validated prognostic model which predicts survival for patients with early primary breast cancer. The NPI combines information on tumor size, lymph node stage and tumor grade and stratifies patients into risk groups useful for treatment decisions. Classically, NPI defines 5 risk categories: Excellent, Good, Moderate I, Moderate II and Poor Prognostic Group (EPG, GPG, MPGI, MPGII and PPG). However, the value of NPI to identify EPG patients has recently being challenged. The Genomic Grade Index (GGI) is a 97-gene test which has been shown to improve tumor grading by resolving the intermediate Histological Grade (HG) 2 tumors into high or low Genomic Grades. In this study, we investigated the potential of GGI to refine the NPI classification, and especially to discriminate patients with the lowest risk. Material and methods:Breast cancer microarray profiles from node negative, untreated patients with the required clinical and pathological information were retrieved from 3 public databases. GGI was calculated using the Ipsogen MapQuant Dx® protocol which defines two classes, GG1 and GG3 (results falling in the CI are reported as Undetermined, “UD”). Standard NPI was calculated based on HG according to the usual formula: [0.2 x tumor size (cm) + HG(1-3) + LN stage (I-III)]. NPI scores were also calculated using GGI (“G-NPI”): the HG was replaced by the GG in the equation, and UD patients were given a value of 2 for tumor grade. Patients were classified into prognostic groups (cut-offs 2.4, 3.4 and 5.4) for both NPI (EPG, GPG, MPG and PPG) and G-NPI (G-EPG, G-GPG, G-MPG and G-PPG). NPI and G-NPI risk classifications were compared and prognostic performances were analyzed using Kaplan Meier survival curves (10-yr MFS). Results:Unique microarray expression profiles with adequate MapQuant Dx® quality criteria and relevant clinical information were identified for 472 patients. HG was distributed as follows: 18 % HG1, 52 % HG2 and 30% HG3 with size ranging from 0.1 to 6 cm (mean = 2.1 cm). HG2 cases (n=246) were reclassified into 47% GG1 and 28% GG3, with 25% remaining UD. There were no patients classified in the PPG, using the standard NPI or the G-NPI. Comparison of risk group classification showed that G-NPI globally shifted patients from low to better prognostic groups: 28% of patients were classified in the G-EPG as opposed to only 12% in the EPG [Table 1]. Within the HG2 subgroup, none of the patients were classified in the EPG by standard NPI, while 75 patients were classified in this group by G-NPI. In the entire cohort, EPG and G-EPG 10-yrs MFS were comparable (82 and 88% resp.). In the HG2 subgroup, 10-yrs MFS was 92% for the G-EPG, 87 % for the G-GPG and 57% for the G-MPG. Conclusion: Using GGI to calculate NPI scores provides additional information compared to standard NPI. This is particularly true in the HG2 group where G-NPI is able to identify patients matching the EPG definition with a 92% 10-yr MFS. Combining genomic information with NPI may facilitate adjuvant therapy decision making. This approach deserves further validation. Prognostic Group Classification according to classic NPI and G-NPI Citation Information: Cancer Res 2010;70(24 Suppl):Abstract nr P3-10-08.