Time-Varying Impact of Steatosis and Alcohol on Mortality: A Marginal Structural Model of the Canadian Longitudinal...

Subclassifications of steatotic liver disease (SLD) are intended to provide a clinically meaningful framework for management. However, steatosis, cardiometabolic comorbidities and alcohol consumption are known to change over time. We investigated the association of the dynamic classification of SLD subtypes and all-cause mortality.

We analysed data from the Canadian Longitudinal Study on Aging, a cohort of adults aged 45-85 years that is followed every 3 years or until death. Hepatic steatosis was identified using the NAFLD Ridge Score, and alcohol consumption was used to classify MASLD, MetALD and ALD. Marginal structural Cox proportional hazards models estimated associations between time-varying SLD subtypes and mortality. Models were adjusted for socio-demographic, lifestyle and cardiometabolic factors, and stratified by sex.

Among 16 560 participants (52.1% female; median age 58 [IQR 51-66] years), 1041 deaths occurred within a median of 7.8 years. Within the first 3 years, 35.4%, 23.7% and 22.6% of participants with ALD, MetALD and MASLD, respectively, transitioned to different SLD subtypes. Mortality risk increased with increasing levels of alcohol among people with and without steatosis, and ALD demonstrated the highest mortality risk (aHR 6.11; 95% CI 1.76-21.19). Sex-stratified analyses suggested higher mortality among females, with the highest mortality rate in females with ALD compared to males (11.37 vs. 5.60 per 1000 person-years).

SLD phenotypes are dynamic, with substantial transitions between disease subtypes over time. Reliance on static classifications may therefore misclassify risk. Longitudinal, time-varying approaches better capture disease evolution and may improve prognostic assessment and risk stratification.