P1.04-15 Smoking Status Is Not a Replacement Biomarker for Tumor Mutation Burden in Non-Small Lung Cancer

Abstract

Recent clinical studies suggest tumor mutation burden (TMB) as a promising therapeutic biomarker of anti-tumor immune checkpoint blockade (ICB). Given the causal link between cancer-causing mutations and tobacco smoking, patients with a significant smoking history may respond better to ICB. However, it is not clear if smoking history is an adequate surrogate biomarker for TMB. Here, we sought assess the clinical utility of smoking history in predicting tumor mutation burden. Publicly available smoking history and DNA somatic alteration data from NSCLC were downloaded from The Cancer Genome Atlas and a large dataset of lung adenocarcinoma tumors published by Imielinski, et al (Cell, 2012) Tumor mutation burden was calculated as the sum of all somatic mutations divided by the exome sequencing coverage. Smoking history was analyzed both as categorical (ever, never, former) and semi-continuous variables (pack years). Hypermutancy was defined as greater than or equal to 10 mutations per megabase. A total of 395 LUAD and 419 LUSC patients were included in this analysis. Smokers had significantly higher tumor mutation burdens than non-smokers; however, in both LUAD and LUSC, there were smokers with low TMB and non-smokers with high TMB. Smoking pack year history (SPY) was weakly positively correlated (Spearman ρ = 0.20, p = 2.5x10-4) in LUAD but uncorrelated (Spearman ρ = -0.026, p = 0.61) in LUSC. Non-smokers and patients without a recorded SPY were excluded from the SPY analysis. We calculated AUCs for predicting hypermutancy in tumors, using variable thresholds of SPY. In LUAD and LUSC, SPY had an AUC of 0.38 and 0.47 in predicting TMB, showing that SPY was not better than random prediction. We also sought to predict TMB from smoking as a binary variable. In LUSC, 8/18 (44%) non-smokers and 253/447 (57%) smokers were hypermutant. In LUAD, 9/61 (15%) non-smokers and 219/391 (56%) smokers were hypermutant. Additionally, we repeated this analysis on matched smoking history and TMB from an independent cohort of 162 LUAD tumors published by Imielinski, et al. Similarly, we found that 1/27 (4%) of nonsmokers and 66/135 (49%) of smokers were hypermutant. In this cohort, the AUC in predicting TMB with SPY was 0·21. In this study, we investigated the relationship between tobacco smoking and TMB. While the average lung cancer patient with a history of tobacco smoking has a higher TMB than the average never-smoker, there is not a clear relationship between the extent of exposure in pack years and TMB. In general, smoking is not an informative biomarker for TMB, however, non-smokers who develop LUAD are unlikely to have high TMB. This study highlights the value of next generation sequencing for TMB in predicting therapeutic response to ICB.