Estimating years of life lost and gained from dietary risks using an accelerated failure time model

Abstract Traditional mortality rates often obscure the impact of lifestyle on premature death. To guide more effective public health strategies, there is a pressing need to quantify how specific exposures, particularly diet, influence years of life lost or gained. We used data from 58,929 U.S. individuals across the 30 years of the National Health and Nutrition Examination Survey (NHANES 1988-2018), linked to mortality outcomes via the National Death Index. We included 26 dietary variables, serum cotinine (as a smoking indicator), and demographic factors. To capture non-linear effects, dietary variables were squared. We developed a penalized Weibull accelerated failure time (AFT) model, adjusting for age at study entry. Acceleration factors (AF) were estimated across the intake distribution to translate diet exposure into years of life lost or gained. Predictive accuracy was assessed on a test set using a rank-based R² for observed vs predicted age at death. Higher serum cotinine (AF = 0.989; CI = [0.988, 0.991]) and cholesterol (AF = 0.992; CI = [0.987, 0.998]) levels were associated with reduced lifespan, up to 1.54 and 1.20 years lost, respectively. In contrast, a greater intake of omega-3 eicosapentaenoic acid (EPA PUFA 20:5; AF = 1.005; CI = [1.003, 1.007]) and a higher poverty-income ratio (AF = 1.021; CI = [1.018, 1.023]) were linked to gains of up to 0.77 and 3.20 years, respectively. Fiber (AF = 1.034; CI = [1.022, 1.046]) and palmitic acid (SFA 16:0; AF = 1.051; CI = [1.008, 1.095]) showed non-linear effects: moderate intake of fiber was associated with a 0.83-year gain, while very low or high intakes of both nutrients led to significant lifespan reductions. Survival modeling provides a powerful tool for quantifying life-years lost or gained from diet using large-scale population data. These findings can inform precision nutrition strategies and support resource allocation toward dietary interventions with the greatest potential impact. Key messages • Dietary exposures can shorten or extend life expectancy by several years. • Survival modeling supports data-driven nutrition policy and public health planning.