A composite single-nucleotide polymorphism prediction signature for extranodal natural killer/T-cell lymphoma

Abstract

Current prognostic scoring systems based on clinicopathologic variables are inadequate in predicting the survival and treatment response of extranodal natural killer/T-cell lymphoma (ENKTL) patients undergoing nonanthracyline-based treatment. We aimed to construct a classifier based on single-nucleotide polymorphisms (SNPs) for improving predictive accuracy and guiding clinical decision making. Data from 722 patients with ENKTL from international centers were analyzed. A 7-SNP-based classifier was constructed using LASSO Cox regression in the training cohort (n = 336) and further validated in the internal testing cohort (n = 144) and in 2 external validation cohorts (n = 142 and n = 100). The 7-SNP-based classifier showed good prognostic predictive efficacy in the training cohort and the 3 validation cohorts. Patients with high- and low-risk scores calculated by the classifier exhibited significantly different progression-free survival (PFS) and overall survival (OS) (all P < .001). The 7-SNP-based classifier was further proved to be an independent prognostic factor by multivariate analysis, and its predictive accuracy was significantly better than clinicopathological risk variables. Application of the 7-SNP-based classifier was not affected by sample types. Notably, chemotherapy combined with radiotherapy significantly improved PFS and OS vs radiotherapy alone in high-risk Ann Arbor stage I patients, whereas there was no statistical difference between the 2 therapeutic modalities among low-risk patients. A nomogram was constructed comprising the classifier and clinicopathological variables; it showed remarkably better predictive accuracy than either variable alone. The 7-SNP-based classifier is a complement to existing risk-stratification systems in ENKTL, which could have significant implications for clinical decision making for patients with ENKTL.