Estimating the Effects of Nordic Diets on the Risk of Major Adverse Liver Outcomes: a Target Trial Emulation across Two Cohorts in Sweden

BackgroundRandomized trials have shown that a healthy Nordic diet (HND) improves liver steatosis, but there is limited evidence on the effects of Nordic dietary patterns on the risk of major adverse liver outcomes (MALO). We specified a hypothetical target trial protocol to estimate the effects of adhering to a HND or the Nordic Nutrition Recommendations 2023 (NNR23) on the 24-year risk of MALO in a middle-aged to elderly Swedish population. MethodsTwo pooled population-based cohorts including n=64,406 men and women (Cohort of Swedish Men (COSM) and the Swedish Mammography Cohort (SMC)) with repeated measurements on diet and confounders in 1997, 2008/2009 and 2019 were used to emulate population-adapted versions of the diets. Under the assumptions of no unmeasured confounding, selection bias or measurement error, the parametric g-formula was used to estimate 24-year risks of MALO from each hypothetical intervention. Secondary analyses included comparing the HND and NNR23 with a low-adherence group; reducing alcohol as an additional hypothetical intervention; and assessing risk of all-cause mortality. ResultsThe estimated 24-year risk of MALO in the HND was 0.53% (95% CI: 0.38, 0.73), in the NNR23 diet 0.70% (95% CI: 0.57, 0.90) and in no intervention 0.64% (95% CI: 0.56, 0.77). Estimated risk differences (RDs) of MALO for HND versus no intervention and NNR23 versus no intervention were -0.11% (95% CI: -0.27, 0.07) and 0.06% (95% CI: -0.04, 0.15), respectively. Compared to NNR23, the estimated RD for the HND was -0.17% (95% CI: -0.38, 0.05). Meaningful risk reductions following the HND were estimated when compared to a low-adherence diet group (-1.50% (95% CI: -9.53, -0.05)), when including reducing alcohol, and for all-cause mortality (-2.67% (95% CI: -3.51, -1.85) versus no intervention; -1.68% (95% CI: - 2.75, -0.62) versus NNR23)). ConclusionWe estimated no clear risk reductions from a population-adapted HND or a NNR23 diet on the 24-year risk of MALO when compared to each other or no intervention. However, when either compared to a low-adherence group or when including reducing alcohol as a hypothetical intervention or when specifying all-cause mortality as the outcome, we estimated meaningful risk differences following the HND.