World Health Organization (WHO) grading of astrocytic tumors is based on cellular features e.g., atypia, mitotic activity, endothelial proliferation, and necrosis, and predicts patient survival. Grade 2 tumors (median survival >5 yr) usually have cellular atypia, while grade 3 tumors (median survival ~3.5 yr) usually have both atypia and mitotic activity. However, the survival of grade 2 and 3 patients is heterogeneous, and there is no good marker to detect the atypical survivors. We hypothesized that DNA copy number would be a prognostic factor, and we used DNA microarrays to study 21 consensus grade 2 and 45 consensus grade 3 astrocytomas. Consensus was defined as diagnostic agreement among at least 4 of 6 neuropathologists. We performed Pearson’s correlation analyses to identify correlations among chromosomal aberrations, used cluster analysis to identify genetic groups, significance analysis of microarrays (SAM) to identify genetic differences between grades, and the Cox proportional hazards model to identify genetic markers associated with survival. Our analysis suggests that a) Copy number is correlated in grade 2 tumors between 1) chr 4 and chr’s 7, 9, 14, and 17; 2) chr 7, and chr’s 9, 14, and 17; and 3) chr’s 1 and X; and in grade 3 tumors between 1) chr’s 1 and 6; 2) chr’s 4 and 7/8; 3) chr’s 6 and 16/17; 4) chr’s 7 and 10; and 5) chr’s 9 and 15. b) Six of the poor survivors in grade 3 patients have chr.7 gain and/or chr.10 loss, a well known genetic marker for poor survivors in grade 4 patients. These cases formed a small subgroup within grade 3 patients by unsupervised cluster analysis. c) Losses of ~10Mb region on chr. 10q encompassing the PTEN gene and ~5Mb region on chr.13q22.2 were associated with poor survival. Kaplan Maier analysis suggests that PTEN loss segregates poor and better survivors more efficiently (0.0023) than tumor grade (p=0.0166). Furthermore, PTEN copy number loss was a far stronger independent prognostic marker for survival (p=0.0012) than age (p=0.0137) or tumor grade (p=0.0137) alone. This work was supported by NIH (NS42927 and CA85799), National Brain Tumor Foundation, and the Barrow Neurological Foundation.

98th AACR Annual Meeting-- Apr 14-18, 2007; Los Angeles, CA