Modeling prognosis for patients with malignant astrocytic gliomas: quantifying the expression of multiple genetic markers and clinical variables.

The disparate lengths of survival among patients with malignant astrocytic gliomas (anaplastic astrocytomas [AAs] and glioblastoma multiforme [GBM]) cannot be adequately accounted for by clinical variables (patient age, histology, and recurrent status). Using real-time quantitative reverse transcription-polymerase chain reaction, we quantified the expression of four genes that were putative prognostic markers (CDK4, IGFBP2, MMP2, and RPS9) in a set of 43 AAs, 41 GBMs, and seven adjacent normal brain tissues. We previously explicated the expression and prognostic value of PAX6, PTEN, VEGF, and EGFR in these glioma tissues and established a comprehensive prognostic model (Zhou et al., 2003). This study attempts to improve that model by including four additional genetic markers, which exhibited a differential expression (P < 0.001) among tumor grades and between tumor and normal tissues. By including eight log-scaled gene expression variables, three clinical variables, and interaction terms among the eight genes, we established a prognostic model that accounted for two thirds of the variation (R2) in survival for this set of patients. To improve the R2 of the model without compromising its clinical utility, our data demonstrated that incorporating genes from different pathways markedly strengthens the model. Spearman rank correlation analysis of gene expression demonstrated a statistically significant positive correlation (P < 0.01) between the expression of IGFBP2-MMP2 and IGFBP2-VEGF in GBMs, but not in AAs. This finding suggests that the expression of IGFBP2 is associated with pathways activated specifically in GBMs that result in enhancing invasiveness and angiogenesis.