E-poster Presentation 2014 World Cancer Congress

A novel literature-based approach to identify genetic and molecular predictors of survival in GBM: Analysis of 14678 patients using systematic review and meta-analytical tools (#1049)

Matthew Thuy 1 , Katharine Drummond 2
  1. Department of General Medicine, Ryde Hospital, Sydney, NSW, Australia
  2. Department of Neurosurgery, Royal Melbourne Hospital, Melbourne, VIC, Australia


Glioblastoma multiforme (GBM) has a poor prognosis despite maximal multimodal therapy. Biological markers relevant to prognosis could be potential treatment targets. A few hundred genetic and molecular factors have been implicated in the literature, however apart from two factors (IDH-1 and O6-MGMT), their clinical significance has been unknown.


To identify which genetic and molecular factors are associated with overall survival in adults with GBM using established systematic review and meta-analytical methods.


A systematic search of MEDLINE (1998-July 2010) was performed. Eligible papers studied the effect of any genetic or molecular marker on univariate overall survival in adult patients with histologically diagnosed GBM. Primary outcomes were median survival difference in months and univariate hazard ratios. Analyses included converting 126 Kaplan Meier curves and 27 raw data sets into primary outcomes. 74 random effects meta-analyses were performed on 39 unique genetic or molecular factors. Objective criteria were designed to classify factors into the categories of clearly prognostic, weakly prognostic, non-prognostic and promising.


Included were 304 publications and 174 studies involving 14678 unique patients from 33 countries. 422 genetic and molecular factors were identified, of which 52 had ≥2 studies. IDH-1 mutation (n = 1114) and O6-MGMT (n = 1232) were classified as clearly prognostic. High Ki-67/MIB-1 (n = 1099) and loss of heterozygosity (LOH) of chromosome 10/10q (n = 1300) were classified as weakly prognostic. Four factors were classified as non-prognostic (EGFR, p53, PTEN, CDKN2A) and 13 factors (BAX, 1p19q, EphA2, p-p70s6k, PI3K, Survivin, p-Akt, Cathepsin, hTERT, Rb, Ras, MAPK and PCNA) were classified as promising. Funnel plot analysis did not identify any publication bias.


This large study demonstrates a novel and statistically powerful literature and meta-analytical based methodology that with proper refinement by the oncology community could guide very effective and efficient basic sciences research.