A systematic evaluation of seven different scores representing the EAT–Lancet reference diet and mortality, stroke, and greenhouse gas emissions in three cohorts

Different approaches have been used for translation of the EAT-Lancet reference diet into dietary scores that can be used to assess health and environmental impact. Our aim was to compare the different EAT-Lancet diet scores, and to estimate their associations with all-cause mortality, stroke incidence, and greenhouse gas emissions. We did a systematic review (PROSPERO, CRD42021286597) to identify different scores representing adherence to the EAT-Lancet reference diet. We then qualitatively compared the diet adherence scores, including their ability to group individuals according the EAT-Lancet reference diet recommendations, and quantitatively assessed the associations of the diet scores with health and environmental outcome data in three diverse cohorts: the Danish Diet, Cancer and Health Cohort (DCH; n=52 452), the Swedish Malmö Diet and Cancer Cohort (MDC; n=20 973), and the Mexican Teachers' Cohort (MTC; n=30 151). The DCH and MTC used food frequency questionnaires and the MDC used a modified diet history method to assess dietary intake, which we used to compute EAT-Lancet diet scores and evaluate the associations of scores with hazard of all-cause mortality and stroke. In the MDC, dietary greenhouse gas emission values were summarised for every participant, which we used to predict greenhouse gas emissions associated with varying diet adherence scores on each scoring system. In our review, seven diet scores were identified (Knuppel et al, 2019; Trijsburg et al, 2020; Cacau et al, 2021; Hanley-Cook et al, 2021; Kesse-Guyot et al, 2021; Stubbendorff et al, 2022; and Colizzi et al, 2023). Two of the seven scores (Stubbendorff and Colizzi) were among the most consistent in grouping participants according to the EAT-Lancet reference diet recommendations across cohorts, and higher scores (greater diet adherence) were associated with decreased risk of mortality (in the DCH and MDC), decreased risk of incident stroke (in the DCH and MDC for the Stubbendorff score; and in the DCH for the Colizzi score), and decreased predicted greenhouse gas emissions in the MDC. We conclude that the seven different scores representing the EAT-Lancet reference diet had differences in construction, interpretation, and relation to disease and climate-related outcomes. Two scores generally performed well in our evaluation. Future studies should carefully consider which diet score to use and preferably use multiple scores to assess the robustness of estimations, given that public health and environmental policy rely on these estimates.