Asian Lung Cancer Absolute Risk Models for lung cancer mortality based on China Kadoorie Biobank

Abstract

AbstractBackgroundLung cancer is the leading cause of cancer mortality globally. Early detection through screening can markedly improve prognosis and prediction models can identify high-risk individuals for risk-based screening. However, most models have been developed in North American cohorts of smokers and much less is known about risk factors for never-smokers, which represent a growing proportion of lung cancers, particularly for Asian populations.MethodsBased on the China Kadoorie Biobank, a population-based prospective cohort study of 512,639 adults age 30-79 recruited between 2004-2008 with up to 12 years of follow-up, we built an Asian Lung Cancer Absolute Risk Model (ALARM) for lung cancer mortality using flexible parametric survival models, separately for ever- and never-smokers, accounting for competing risks of all-other-cause mortality. Model performance was evaluated in a 25% hold-out test set using the time-dependent area under the receiver operating characteristic curve (AUC) and by comparing the model-predicted and observed risks for model calibration.ResultsPredictors assessed in the never-smoker lung cancer mortality model were age, sex, household income, lung function, history of emphysema/bronchitis, family history of cancer, personal cancer history, BMI, passive smoking, and indoor air pollution. The ever-smoker model additionally assessed smoking status (former vs. current), duration, and intensity. The 5-year AUC based on the hold-out test set for the never and ever-smoker models were 0.77 (95% CI: 0.73-0.80) and 0.81 (95% CI: 0.79-0.84), respectively. The maximum 5-year risk for never and ever smokers were 2.6% and 12.7%, respectively.ConclusionsThis study is among the first to develop and test risk models specifically for Asian populations, separately for never (ALARM-NS) and ever-smokers (ALARM-ES). Our models identify Asian never- and ever-smokers at high-risk of death due to lung cancer with a high degree of accuracy and may identify those with risks exceeding common eligibility thresholds who would likely benefit from lung cancer screening.