Does Subdividing High Grade Ovarian Serous Carcinoma into Silverberg Grade 2 and 3 Have Biologic Validity?
Linda Feeley, Anna Marie Mulligan, Cindy Yao, Hanbert Chen, Blake Gilks, Paul Boutros, Blaise Clarke. St Michaels Hospital, Toronto, Canada; Ontario Institute for Cancer Research, Toronto, Canada; Uinversity of British Columbia, Vancouver, Canada; UHN, Toronto, Canada
Background: There are several grading systems for ovarian serous carcinoma (OSC), mostly ternary systems. The Silverberg grading system is based on a composite score of architectural pattern, cytologic atypia and mitotic index. A nominal binary system (high-grade/low-grade) has been proposed reflecting that these are distinct entities rather than a disease spectrum. All high grade OSC correspond to Silverberg grade 2 and 3. Grade categories should reflect therapeutic, prognostic or biologic difference. We have previously shown no prognostic difference between Silverberg grade 2 and 3 OSC and there are no therapeutic differences. It has been suggested that subclassifying high-grade OSC is not relevant, however, molecular studies comparing Silverberg grade 2 and 3 OSC are limited. We analyzed a publicly available microarray dataset of 285gynecological tumors. We compared Silverberg Grade 2 and Grade 3 OSC to determine if they are molecularly distinct entities.
Design: We included 140 arrays in our analysis: 49 Grade 2 and 91 Grade 3 OSC. Data analysis was performed in R (version 2.9.1) using BioConductor. Pre-processing employed the RMA algorithm, and quality-control with boxplots, principal components analysis and clustering indicated that all samples were of good quality. Non-expressed genes were filtered for and excluded based on Y chromosome signals. Two-sided, unpaired, unequal variance t-tests were used to compare grades, followed by an empirical Bayes moderation of the standard error. False-discovery rate adjustments were used to control for multiple testing.
Results: Differential gene expression was not identified between Grade 2 and Grade 3 tumors at either stringent (1% FDR) or relaxed (25% FDR) thresholds. As a control we compared Grade 1 and Grade 3 tumors, and hundreds of genes showed differential expression. We applied a naive p-value threshold of <0.001 to assess functional enrichment. Even at this level only 38 genes were found to be differentially expressed and there was no functional enrichment based on Gene Ontology enrichment analysis.
Conclusions: Based on the current data set, Silverberg Grade 2 and Grade 3 OSC do not differ molecularly (gene expression or pathway analysis), even in a very relaxed statistical analysis. These findings support the use of a two-tier classification system which is in keeping with clinical differences, molecular and morphologic evidence and therapeutic decisions and has been shown to be highly reproducible in practice.
Category: Gynecologic & Obstetrics
Monday, February 28, 2011 8:30 AM
Platform Session: Section D, Monday Morning