The effect of patient and tumour genetics on survival from melanoma

The American Joint Committee on Cancer (AJCC) staging system based on tumour histopathological characteristics (Breslow thickness, ulceration and mitotic rate) is used in clinical practice to assess melanoma patients’ prognosis. Although a reasonable predictor of deaths from melanoma (area under the curve 0.68), AJCC staging does not always provide an accurate assessment of individual risk. Recent studies have shown that gene expression levels within the tumour are associated with survival from melanoma, and there is also preliminary evidence that a patient’s genotype may influence survival. Combining clinical data with gene expression data may improve prediction, but so far no study has analysed the combined effects of the three different types of factor on melanoma-specific survival (MSS). This study used patient and tumour characteristics, whole-genome gene expression levels, and genome-wide single nucleotide polymorphism (SNP) data to identify predictors of MSS in a training set of ~2000 patients from the Leeds Melanoma Cohort (LMC). The selected clinical and genomic predictors were combined to build melanoma prognostic models in a test set of 190 patients from LMC, using several approaches. In addition, heritability of survival from melanoma and of Breslow thickness (the most important predictor of MSS) was estimated using genome-wide SNP data. In the training set, five established clinical predictors (age, sex, tumour site, Breslow thickness and presence of ulceration) were associated with MSS; in addition penalized Cox regression identified 16 gene expression levels and 13 SNPs predictive of MSS. In the test set, the selected genomic predictors did not substantially improve on the predictive performance of the clinical factors. The 16 gene expression levels were also predictive, but were highly correlated with the clinical predictors, especially Breslow thickness, suggesting gene expression influences MSS through clinical predictors. The heritability analyses in this study provided some evidence that germline SNPs influence Breslow thickness