Blood samples undergo polymorphism analysis of common single-nucleotide polymorphisms and haplotypes to examine genetic variation, gene-gene interactions, and the population structure.

Correlative studies

Determined on all SNPs by chi-square tests.

Logistic regression will be used to estimate odds ratios and 95% confidence intervals for the association between each SNP and OS under co-dominant, dominant and recessive genetic models. Stratified analyses will be conducted to examine sex, tumor subtype and outcome differences.

Assessed using a multiplicative model. Haplotypes will be constructed using both Bayesian and expectation-maximization algorithms. Differences between cases and controls will be evaluated with HaploStats which uses haplotype posterior probabilities as weights to update the regression coefficients in an iterative manner.

Kaplan-Meier survival curves will be used to determine outcome relative to genotype.

Annotation and filtering of each whole-exome variant locus will be performed using a custom software pipeline. Variants in >= 2 OS cases will be validated, and then subsequently replicated in additional OS cases (samples previously received for the GWAS from international collaborators). Variants will also be evaluated for presence in known biologically plausible pathways and genes.