A probabilistic framework reveals life course heterogeneity in global diet and healthy longevity associations

Conventional dietary adherence scores may limit cross-population comparability and allow strong components to offset marked shortfalls. We developed a probabilistic framework to place established dietary patterns on a common scale while retaining information on limiting components.

We constructed the Dietary Matching Index (DMI) for eight dietary patterns using age-specific and sex-specific dietary estimates from 185 countries from 1990 to 2018. Parameters were selected using outcome-blind sensitivity analyses. Construct validity was assessed using global dietary data and six harmonised individual-level dietary datasets. Spatial clustering was evaluated using global and local Moran's I. Associations with health-adjusted life expectancy (HALE) were estimated using geographically and temporally weighted regression models adjusted for sex and socioeconomic and ecological covariates.

DMI rankings were robust to calibration choices and Monte Carlo resampling. Across joint parameter settings, minimum Spearman correlations with baseline rankings ranged from 0.971 to 0.993, and Spearman correlations between the 250-draw and 500-draw estimates were 1.00 for all eight patterns. Component-level analyses reproduced the intended structure of the original dietary patterns, and individual-level gradients were consistent across six validation datasets. Dietary alignment varied substantially across countries and showed persistent geographic clustering. DMI-based models showed comparable or modestly better fit than conventional-score models. Associations with HALE were modest and varied by age and region. The largest positive pooled association was 0.17 additional HALE years for a one-standard-deviation higher Mediterranean-style DMI in ages 2-19 years (95% uncertainty interval 0.08 to 0.22). Between-country heterogeneity was greatest in midlife, with I² exceeding 50% for most patterns.

DMI provides a common probabilistic scale for comparing dietary patterns and identifying limiting components. Dietary alignment showed persistent geographic structure, while associations with HALE varied by age and region. These ecological findings may support dietary surveillance and prioritisation, but causal interpretation requires evidence from longitudinal or intervention studies.