Abstract Background Screening with computed tomography has been shown to reduce the rate of death from lung cancer, but at substantial cost in terms of morbidity associated with overdiagnosis. The performance of screening programs can be improved by restricting screening to those people who are at sufficiently high risk of lung cancer. We sought to develop a lung cancer risk prediction model incorporating a measure of lung function (forced expiratory volume in one second: FEV1) as well as other information that is routinely available to general practitioners. Methods We built the model using participants and data from the UK Biobank prospective cohort study, which recruited 500000 people aged 37 to 73 years between 2006-2010. Follow-up for cancer incidence and death was conducted via linkage to registries. Each participant was followed up for an average of 3 years and a maximum of 6 years. We investigated factors that are routinely available or can be easily ascertained by general practitioners: sex, variables related to smoking history and addiction to nicotine, personal medical history, and family history of lung cancer. We additionally investigated lung function, which was assessed via sprirometry as the forced expiratory volume in 1 second (FEV1, litres). Conditional on these predictors, we separately modeled the hazard of lung cancer and the hazard of death using flexible parametric survival models with age as the timescale, and combined the estimated hazards to predict the 2-year absolute risk of lung cancer. Internal validation of model discrimination was assessed by calculating the bootstrap optimism-corrected c-statistic. Results There were 738 incident lung cancer diagnoses and 3956 deaths from all causes among the UK Biobank population in the follow-up period. FEV1 at baseline was strongly and inversely associated with subsequent lung cancer risk: hazard ratios [95% confidence intervals] of 0.57 [0.43,0.77] per additional liter of FEV for never, 0.50 [0.40,0.63] for former, and 0.62 [0.48,0.80] for current smoking participants. Model discrimination was high (c-statistic 0.85) and the model is likely to discriminate well when applied to new data (bootstrap optimism-corrected c-statistic 0.83). Applying the current National Lung Screening Trial(NLST) inclusion criteria to the UK Biobank data yielded a specificity of 0.95 and sensitivity of 0.37. By applying a risk prediction model-based inclusion criteria where additional factors such as FEV1 were taken into account, we could improve the sensitivity to 0.47 for the same specificity. Conclusions Comprehensive risk prediction models based on standard risk factor information, along with FEV1, can clearly outperform currently used screening criteria in terms of specificity and sennsitivity in predicting future lung cancer diagnoses, and applying a model-based screening strategy has the potential to improve the performance of lung cancer screening programs. Citation Format: David C. Muller, Mattias Johansson, Paul Brennan. A model incorporating spirometry to predict absolute risk of lung cancer: The UK Biobank prospective cohort study. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 5581. doi:10.1158/1538-7445.AM2015-5581