Correlation among pathology, genotype, and patient outcomes in glioblastoma.

Glioblastomas are histologically and genetically heterogeneous. We have investigated to what extent histologic features reflect the genetic profile and whether they are predictive of clinical outcome. Key histologic characteristics, including major cell types (small cell, nonsmall cell), other components such as oligodendroglial components, gemistocytes, multinucleated giant cells, as well as necrosis and microvascular proliferation, of 420 cases of glioblastoma within a population-based study (1) were reassessed and correlated with patients' clinical outcome and key genetic alterations. EGFR amplification and p16 homozygous deletion were significantly more frequent in small cell glioblastomas than in nonsmall cell glioblastomas (EGFR, 46% vs 26%, p = 0.0002; p16 39% vs 25%, p = 0.0167). Multivariate analyses with adjustment for age and gender showed that small cell glioblastomas had frequent EGFR amplification and p16 deletion but infrequent PTEN mutations. An oligodendroglial component was detected in 20% of glioblastomas; these patients were significantly younger (54.4 +/- 13.6 vs 59.2 +/- 13.8 years; p = 0.0049) and survived longer (10.3 +/- 8.3 vs 8.2 +/- 8.4 months; p = 0.0647). However, multivariate analyses with adjustment for age and gender did not show the presence of an oligodendroglial component to be predictive of longer survival. After adjustment for age and gender, LOH 1p was associated with longer survival (hazard ratio, 0.7; 95% confidence interval [CI], 0.5-1.0), whereas LOH 10q was associated with shorter survival (hazard ratio, 1.4; 95% CI, 1.0-1.8) of patients with glioblastoma. Glioblastomas containing >or=5% multinucleated giant cells showed more frequent TP53 mutation and infrequent EGFR amplification than those containing <5% multinucleated giant cells (TP53, 45% vs 24%, p = 0.0001; EGFR, 24% vs 42%, p = 0.0005). Vascular proliferation was observed in all glioblastomas, whereas large ischemic and/or pseudopalisading necrosis was observed in 366 of 420 (87%) cases. Glioblastomas with necrosis were associated with older age (59.2 +/- 13.3 vs 51.6 +/- 15.3 years; p = 0.0001) and shorter survival (7.9 +/- 6.8 vs 12.9 +/- 14.2 months; p = 0.0017). Multivariate analyses with adjustment for age and gender confirmed this observation (hazard ratio, 1.5; 95% CI, 1.1-2.0). Multivariate analysis with adjustment for age and gender showed that necrosis was significantly associated with wild-type TP53 and absence of an oligodendroglial component. These results suggest that some histologic features in glioblastomas are associated with specific genetic alterations and with clinical outcome.