A combination of plasma phospholipid fatty acids and its association with incidence of type 2 diabetes: The EPIC-InterAct case-cohort study

Background Combinations of multiple fatty acids may influence cardiometabolic risk more than single fatty acids. The association of a combination of fatty acids with incident type 2 diabetes (T2D) has not been evaluated. Methods and findings We measured plasma phospholipid fatty acids by gas chromatography in 27,296 adults, including 12,132 incident cases of T2D, over the follow-up period between baseline (1991– 1998) and 31 December 2007 in 8 European countries in EPIC-InterAct, a nested casecohort study. The first principal component derived by principal component analysis of 27 individual fatty acids (mole percentage) was the main exposure (subsequently called the fatty acid pattern score [FA-pattern score]). The FA-pattern score was partly characterised by high concentrations of linoleic acid, stearic acid, odd-chain fatty acids, and very-longchain saturated fatty acids and low concentrations of γ-linolenic acid, palmitic acid, and long-chain monounsaturated fatty acids, and it explained 16.1% of the overall variability of the 27 fatty acids. Based on country-specific Prentice-weighted Cox regression and random-effects meta-analysis, the FA-pattern score was associated with lower incident T2D. Comparing the top to the bottom fifth of the score, the hazard ratio of incident T2D was 0.23 (95% CI 0.19–0.29) adjusted for potential confounders and 0.37 (95% CI 0.27–0.50) further adjusted for metabolic risk factors. The association changed little after adjustment for individual fatty acids or fatty acid subclasses. In cross-sectional analyses relating the FA-pattern score to metabolic, genetic, and dietary factors, the FA-pattern score was inversely associated with adiposity, triglycerides, liver enzymes, C-reactive protein, a genetic score representing insulin resistance, and dietary intakes of soft drinks and alcohol and was positively associated with high-density-lipoprotein cholesterol and intakes of polyunsaturated fat, dietary fibre, and coffee (p < 0.05 each). Limitations include potential measurement error in the fatty acids and other model covariates and possible residual confounding. Conclusions A combination of individual fatty acids, characterised by high concentrations of linoleic acid, oddchain fatty acids, and very long-chain fatty acids, was associated with lower incidence of T2D. The specific fatty acid pattern may be influenced by metabolic, genetic, and dietary factors.