This study found a complete lack of association between EPA and DHA intake during pregnancy and risk of type 1 diabetes; these data were consistent between the DNBC and the MoBa, with relatively narrow 95% CIs.

Our results are consistent with data from a birth cohort with high genetic risk for type 1 diabetes from Finland [9]. Our results therefore confirm and extend this lack of association in larger, general population cohorts. Although not investigating fatty acids directly, another cohort study including children at high genetic risk for type 1 diabetes found no association between maternal fish intake and risk of type 1 diabetes in offspring [12].

The major strengths of our study include the very large, population-based cohorts from two different countries and prospectively collected exposure data during pregnancy, and the complete follow-up of type 1 diabetes diagnosis using nationwide registries. Limitations include the lack of biomarkers of exposure and the low numbers of non-European participants. Below we discuss the strengths and limitations in more detail.

Our cohorts were population based and assessed dietary intake, including dietary supplements, during pregnancy. Biomarker data on EPA and DHA intake were not available in our study; however, the ability of the FFQ to quantify n-3 fatty acid intake has been validated against biomarkers in both cohorts [13, 14]. In addition, a nested case–control prospective study from Norway found no association between offspring type 1 diabetes and EPA and DHA levels in the phospholipid fraction of maternal serum collected in late pregnancy [15]. While biomarkers are not influenced by recall and self-report, they are not without problems. Storage and handling of samples may lead to oxidation. Furthermore, biomarkers may be influenced by fasting or recent meals, as well as genetic and other factors regulating metabolism of fatty acids, depending on the type of specimen and assay methods used [16]. Common variants in the fatty acid desaturase (FADS1/FADS2/FADS3) gene cluster are associated with a lower efficiency of conversion of the dietary precursor n-3 fatty acid alpha-linolenic acid to EPA and DHA and lower blood levels of EPA and DHA [1]. In the asthma prevention trial cited in the introduction, both lower baseline blood levels of EPA and DHA and a genetic variant associated with lower blood levels of these fatty acids were associated with a stronger relative effect of fish oil supplementation on the prevention of asthma [8]. On the other hand, studies of other disease outcomes in adults have not found consistent interactions between intake of EPA or DHA and genetic variants in the FADS gene cluster with regard to cardiometabolic disease outcomes [1]. Future studies of maternal n-3 fatty acid intake in pregnancy in relation to childhood type 1 diabetes could consider including genetic variants in the FADS gene cluster influencing conversion of alpha-linolenic acid to EPA and DHA.

We cannot exclude the possibility that higher intakes of EPA and DHA than those observed in our studies may show an association with type 1 diabetes. However, analysis of quintiles of the sum of EPA and DHA intake did not suggest any threshold effect. We adjusted for intake of vitamin D in pregnancy in robustness analyses, even though previous evidence suggests no association with type 1 diabetes [17]. Unmeasured confounding, for instance from toxicants in fatty fish, may have influenced our results; however, there is currently no strong evidence supporting an association between such toxicants and type 1 diabetes [2] and we do not believe that toxicants have confounded our results substantially. Given the largely Scandinavian origin of our participants, we believe that our results are generalisable to other European-origin populations, but not necessarily to populations with large proportions of people of other ancestries. Finally, our results do not exclude a potential effect of EPA and DHA intake in children rather than in their pregnant mothers.

In conclusion, the hypothesis that a higher maternal n-3 fatty acid intake during pregnancy reduces the risk of type 1 diabetes in offspring was not supported by this study. In a setting where primary prevention trials are extremely expensive and time-consuming [18], we believe that our results, together with the evidence discussed above, clearly indicate that a trial of EPA and DHA during pregnancy to prevent type 1 diabetes in offspring should not be prioritised.