Objective:To investigate the correlation between the trajectory of triglyceride-glucose index multiplied by body mass index (TyG×BMI) and the incidence of new-onset non-alcoholic fatty liver disease (NAFLD).Methods:It was a retrospective cohort study. A total of 2 304 subjects who underwent health examinations at the Health Management Center of the First Affiliated Hospital of Zhengzhou University from 2017 to 2019 were included as the study population. Based on the TyG×BMI values from the health examinations, a latent class modeling approach was used to determine four distinct TyG×BMI trajectory groups: low-stable, moderate-stable, high-stable, and extremely high-stable group. The incidence of NAFLD was followed-up during the 2020 and 2021 health examinations for each group. The differences in NAFLD incidence among different TyG×BMI trajectory groups were compared using the log-rank test, and the correlation between different TyG×BMI trajectories and the incidence of new-onset NAFLD was analyzed using Cox proportional hazards regression models.Results:The incidence of NAFLD increased with the elevation of TyG×BMI trajectories. The cumulative incidence rates of NAFLD for the low-stable, moderate-stable, high-stable, and extremely high-stable groups was 13.00%, 16.70%, 20.10% and 26.60%, respectively, with statistically significant differences ( χ2=35.155, P<0.01). Compared with the low-stable group, the high-stable and extremely high-stable groups had higher risks of NAFLD (HRs for high-stable group was 1.564, 1.428, 1.426, 1.289, respectively; HRs for extremely high-stable group was 2.121, 1.670, 1.659, 1.607, respectively; all P<0.05). After adjusting for various confounding factors such as gender, waist circumference, BMI, blood glucose, blood lipids and liver function in model 4, the risks of NAFLD for the high-stable and extremely high-stable groups were still 1.389 and 1.607 times higher than that in the low-stable group (95% CIs: 1.035-1.864, 1.207-2.140). Conclusion:The risk of NAFLD increases with the elevation of TyG×BMI trajectories, suggesting that TyG×BMI can serve as a predictive index for NAFLD.