Serum metabolite signature of the mediterranean diet adherence and cognitive function: Evidence from two cohorts.

BACKGROUND AND AIMS: The Mediterranean diet (MD) has been consistently associated with favorable cognitive outcomes. However, reliance on self-reported dietary assessment limits accurate characterization of diet-health associations. Circulating metabolites may capture the endogenous metabolic imprint of habitual dietary patterns and provide complementary biological information beyond questionnaire-based measures. We aimed to identify a serum metabolite signature associated with MD adherence and to examine its cross-cohort reproducibility and associations with cognitive performance and trajectories. METHODS: We analyzed 595 serum metabolites measured by mass spectrometry in 337 adults aged ≥65 years (cognitively normal or with mild cognitive impairment [MCI]) from the Hellenic Longitudinal Investigation of Aging and Diet (HELIAD). MD adherence was estimated using a validated food-frequency questionnaire (FFQ). A two-stage penalized regression approach was used to derive a metabolite signature of MD adherence. External validation (applying the HELIAD-derived metabolites and coefficients) and recalibration (re-estimating coefficients) were performed in 148 adults aged ≥40 years from the Aiginition Longitudinal Biomarker Investigation of Neurodegeneration (ALBION) (cognitively normal or MCI), where MD adherence was assessed using four repeated 24-h dietary recalls. Associations with cognitive performance (global and five domain-specific scores) were examined cross-sectionally and prospectively (median follow-up: 2.8 years in HELIAD; 1.2 years in ALBION). RESULTS: A 23-metabolite signature enriched in lysophospholipids, plasmalogen and ether-linked phospholipids, long-chain polyunsaturated phosphatidylcholines, and steroid hormone sulfates explained 28.7% of the variance in the MedDietScore in HELIAD (r = 0.54). Direct cross-cohort transferability was limited in ALBION (R2 = 0.02), but predictive performance improved following cohort-specific recalibration (R2 = 0.21). The metabolite signature was positively associated with memory performance cross-sectionally and with subsequent language change in HELIAD. Calibration analyses suggested attenuation of FFQ-based associations, with larger effect estimates after correction but wider confidence intervals. CONCLUSION: A reproducible serum metabolite signature reflects diet-related metabolic variation associated with MD adherence and is linked to cognitive performance in older adults. Rather than functioning as a direct biomarker of adherence, this profile captures the endogenous metabolic imprint of habitual dietary patterns and may complement self-reported dietary measures in nutritional epidemiology.