Discussion

In this high fish-eating cohort from the SCDS, total fish consumption throughout pregnancy was not associated with repeated measures of child weight at birth, 20 months, 7 years or 13 years. There were, however, several associations between the n-3 and n-6 PUFA and child weight. The lowest quartile of maternal AA concentrations (<0·148 mg/ml) was associated with a lower risk of LGA. Furthermore, higher maternal total n-6 PUFA concentrations were associated with an increased risk of overweight/obesity at 7 years and 13 years. In cord blood, DHA concentrations <0·071 mg/ml were associated with a significantly higher risk of LGA within the child. None of the observed associations between the cord n-3 or n-6 PUFA and child weight were evident at 7 or 13 years of age. Overall, these findings appear to suggest differing influences of lower cord n-3 and higher maternal n-6 PUFA status on child adiposity at birth until adolescence.

In the current study, there was no clear evidence of any associations between maternal total fish consumption and child weight at birth or throughout childhood, despite the uniquely high fish intakes within this cohort. Observational studies investigating the associations between maternal fish consumption and measures of child growth are few and have been overall inconclusive. Findings from this study appear to be consistent with a number of other mother–child cohorts, albeit lower fish consumers, where no associations were found^{(27,Reference Guldner, Monfort and Rouget39,Reference Mohanty, Thompson and Burbacher40)} . In contrast, one larger Norwegian Mother and Child Cohort Study with a much lower average fish intake of 36 g per d reported positive associations between maternal seafood consumption and birth weight^{(Reference Brantsaeter, Birgisdottir and Meltzer41)}. Furthermore, one pooled analysis of fifteen European and USA birth cohorts reported higher fish intakes (>3 times/week) to be associated with an increased risk of rapid infant growth and child overweight/obesity at 4 and 6 years which may be owing to contaminants found within fish^{(Reference Stratakis, Roumeliotaki and Oken36)}. We found that the consumption of >57·14 g/d lean fish was associated with an increased risk of overweight/obesity at 13 years; however, this association was not observed at any other time point, nor was it dose-responsive, and thus may be a chance finding. Studies which have reported a positive association between maternal fish intakes and child weight have noted that these associations appear to be dependent on the type of fish, with nutrients and contaminants within fish varying depending on the species and source which may also account for some variation between populations in addition to the differences in consumption patterns^{(Reference Stratakis, Roumeliotaki and Oken36,Reference Willer, Newton and Malcorps42)} . Nevertheless, fish is an important source of protein and nutrients required for growth and development both in pregnancy and childhood^{(Reference Mendivil22)}, and uncertainty surrounding the benefits and/or risks of fish consumption during pregnancy on child body composition may lead to transitions towards more westernized-style diets which have been associated with metabolic complications including CVD^{(Reference Clemente-Suarewz, Beltran-Velasco and Redondo-Florez43)}. This transition is particularly relevant amongst the Global South, with developing nations such as the Seychelles undergoing a period of rapid dietary transition resulting in lower fish intakes^{(Reference Cardoso, Bovet, Viswanathan and Luke44)}. Given the limited and inconsistent data, alongside the public health implications, future longitudinal analyses using repeat measurements throughout childhood and which utilise direct measures of body composition are warranted to elucidate the relationships between maternal fish consumption during pregnancy and child adiposity.

Previous research has emphasised the importance of n-3 PUFA, primarily obtained in the diet from fish consumption, during pregnancy in relation to fetal growth, brain structure and neurodevelopment^{(Reference Makrides45,Reference Olsen, Halldorsson and Thorne-Lyman46)} . There was no evidence of any associations between the maternal n-3 PUFA and child weight at birth, 20 months, 7 years or 13 years, which confirms results from our previous analysis from the SCDS, where we did not detect any associations between maternal n-3 PUFA and birth outcomes^{(Reference Van Wijngaarden, Harrington and Kobrosly34,Reference Yeates, Zavez and Thurston35)} . The existing literature with regard to the associations between maternal n-3 PUFA status and child adiposity remains controversial and has primarily focused on early childhood which makes comparison to the current study challenging. Findings from the much larger Generation R Study noted higher maternal total n-3 PUFA concentrations within the second trimester were associated with a lower childhood body fat percentage and a lower android:gynoid fat mass ratio at 6 years^{(Reference Vidakovic, Gishti and Voortman47)}. Others have reported no associations between maternal n-3 PUFA concentrations and child fat mass at 6 years^{(Reference Moon, Harvey and Robinson48)}. It is noteworthy that the two latter studies measured PUFA concentrations in plasma phospholipids, whereas in the current study maternal PUFA were measured in serum which makes comparisons more difficult. Moreover, 13 years may be a period of rapid pubertal growth where changes in the hormonal milieu result in adiposity accretion^{(Reference Solorzano and McCartney49)}. In this cohort, maternal fish consumption was not correlated with maternal PUFA status, which is also likely reflective of other factors influencing biomarker status. Aside from the differences in n-3 PUFA intakes, genetic variations can also influence PUFA metabolism and may also explain some of the inconsistencies across studies^{(Reference Yeates, Love and Engström50)}. Future studies should consider the various genotypes which regulate PUFA metabolism when investigating the associations between maternal n-3 PUFA and child adiposity.

Within this cohort, maternal AA concentrations <0·148 mg/ml were associated with a lower risk of LGA at birth when compared with higher maternal AA concentrations. Additionally, the maternal n-6 PUFA were consistently associated with an increased risk of overweight/obesity throughout childhood. The current results are broadly consistent with the limited previous human studies^{(Reference Vidakovic, Gishti and Voortman47,Reference Moon, Harvey and Robinson48,Reference De Vries, Gielen and Rizopoulos51)} . The Southampton Women’s Survey reported positive associations between maternal plasma n-6 PUFA concentrations and child fat mass at 4 and 6 years^{(Reference Moon, Harvey and Robinson48)}. Similar findings were also reported in the Generation R Study where higher maternal plasma n-6 PUFA were found to be associated with a higher childhood total body fat percentage and abdominal preperitoneal fat mass area^{(Reference Vidakovic, Gishti and Voortman47)}. Taken together, these findings, along with the current study, appear to suggest that higher prenatal exposures to the n-6 PUFA may influence adipogenesis development in utero and may be associated with an increase in child adiposity until early adolescence.

There are several existing mechanisms which may help to explain these observations. The n-6 PUFA are known to elicit a more pro-inflammatory immune response, leading to the production of pro-inflammatory lipid mediators known as eicosanoids^{(Reference Calder52)}. These eicosanoids are highly biologically active, exerting effects on many cell types via cell membrane G-protein-coupled receptors^{(Reference Tang, Hou, Schwartz and Haeggstrom53)}. For example, the n-6 PUFA are known to act as a precursor for prostacyclin (PG I₂) which has been shown to bind to PPARγ which regulates adipogenesis via the proliferation of preadipocytes into mature adipocytes^{(Reference Majed and Khalil54,Reference Rahaman55)} . Moreover, there is also some evidence from murine models which suggests differing effects of fatty acids in the metabolic programming of the fetus in utero through epigenetic modification such as DNA methylation. Such epigenetic changes may influence growth in later life by influencing appetite control and energy balance^{(Reference Kabaran and Besler56)} and, thereby, may offer some plausible biological basis as to why associations were not apparent until later in childhood, albeit further work is needed to confirm this. We also noted an increased risk of overweight/obesity at 20 months and 7 years with a higher maternal n6:n3 ratio, which is generally associated with a greater pro-inflammatory immune response. The observed associations between the n-6 PUFA and child adiposity during childhood likely reflect the influence of the n-6 PUFA on the inflammatory milieu throughout pregnancy, and future work may need to measure eicosanoid production to improve mechanistic understanding in this area.

Typically, LA is abundant within the diet and found within several sources, including vegetable oils, nuts and seeds^{(Reference Orsavova, Misurcova and Ambrozova57)}. Endogenously, LA can undergo the elongation and desaturation pathway to the more biologically active n-6 PUFA, although AA is also found in high amounts in red meat from grain-fed animals^{(Reference Mariamenatu and Abdu58)}. Westernised-style diets are high in processed foods, red meat and eggs, which are typically higher in n-6 PUFA^{(Reference Liput, Lepczyński and Ogłuszka59)}. Previous research from our group has shown that this style of dietary pattern is most typically followed within this cohort of pregnant Seychellois women, reflecting the transition from the traditional Seychellois diet to a more westernized-style diet within the Republic of Seychelles^{(Reference Conway, Mulhern and McSorley30)}. The westernised dietary pattern has also been shown to be associated with an enhanced pro-inflammatory immune response and may be involved in promoting intracellular inflammatory pathways in adipocytes^{(Reference Monaco-Brown and Lawrence60)}. These findings have significant public health implications in light of the transition towards westernized-style diets throughout pregnancy and thus higher intakes of n-6 PUFA, particularly within the Global South. Such dietary changes may have important consequences for adiposity accretion and subsequently obesity development in the child in later life.

Although maternal n-3 PUFA were not found to be associated with child weight, cord DHA concentrations <0·071 mg/ml were associated with an increased risk of LGA. Much of the existing evidence has focused on maternal n-3 PUFA exposure earlier in pregnancy and the risk of preterm birth and gestational duration^{(Reference Baker, Calder and Kermack61)}, whereas fewer have examined cord PUFA status at delivery and its potential influence on fetal growth at term. One study of 847 mother–child pairs reported a negative association between cord erythrocyte DHA concentrations and birth weight^{(Reference Cinelli, Fabrizi and Ravà62)}, whilst another, smaller observational study of 253 mother–child pairs also found a negative association between cord DHA concentrations and birth weight^{(Reference Kilari, Mehendale and Dangat63)}. Whilst preterm births were excluded in the current analyses, and therefore the results may not be generalisable to pregnancies ending preterm or to outcomes related to gestational duration, the findings of the current study suggest that later-stage n-3 PUFA exposure may be important in relation to birth size. The n-3 PUFA are essential in mediating the immune response, acting as precursors for eicosanoids and docosanoids which are considered to be more anti-inflammatory^{(Reference Calder64)}. The n-3 PUFA inhibit nuclear transcription factor kappa B, promoting a more anti-inflammatory immune response, which in turn may modulate adipokine secretion and thus lower adipogenesis and promote a more metabolically healthy phenotype^{(Reference Albracht-Schlulte, Kalupahana and Ramalingam65)}. Within this cohort, only thirty-eight (3·2 %) of children were considered LGA which is considerably lower than many other birth cohorts within Africa^{(Reference Sridhar, Ferrara and Ehrlich66,Reference Younes, Samara and Salama67)} but is reflective of the Seychellois population which has an incidence of 4·4 %. The traditional Seychellois diet is rich in oily fish and fish dishes, which has been previously shown to be associated with a higher n-3 PUFA status^{(Reference Conway, Mulhern and McSorley30)}. Given both the rising incidence of LGA babies in many populations and the long-term metabolic complications that are often associated with LGA^{(Reference Hong and Lee68)}, the unique dietary patterns within the Seychelles paired with the low incidence of LGA within this population may provide promising evidence of potential dietary interventions in helping to lower the risk of LGA within the child through increasing n-3 PUFA intakes throughout pregnancy.

It is important to recognise the differing associations between the maternal and cord PUFA observed within this study. For example, in this cohort, maternal n-6 PUFA were consistently associated with an increased risk of overweight/obesity at 7 and 13 years, whilst few associations were found with the cord PUFA. These differing associations are interesting, considering the developing fetus relies entirely on the maternal supply of PUFA owing to insufficient desaturase activity within the placenta or fetus to synthesise PUFA^{(Reference Hanebutt, Demmelmair and Schiessl69)}. The circulating concentrations of PUFA fluctuate throughout gestation with the placenta controlling the uptake and transfer of maternal PUFA to the fetus^{(Reference Duttaroy70)}. Previous evidence has suggested that placental uptake and transfer differ, with the transfer favouring DHA and AA, particularly during the last trimester of pregnancy. The differing associations apparent within the current study may be owing to differences in genetics influencing PUFA metabolism, in addition to the altered fatty acid metabolism throughout pregnancy; the maternal PUFA status reflects exposure throughout the entire pregnancy, whilst the cord PUFA status may more accurately depict exposure throughout the last trimester. The placenta has been shown to play an active role in fetal programming^{(Reference Kwon and Kim72)}, and given that PUFA transfer across the placenta differs throughout gestation, it may point towards specific periods during pregnancy where the developing fetus may be particularly sensitive to the effects of the n-3 and n-6 PUFA. In future, a focused examination using repeat measurements of PUFA status throughout pregnancy may help to elucidate the differing associations between the maternal and cord PUFA and child adiposity.

The current study has several strengths. First, the use of robust biological measures of both cord and maternal PUFA concentrations and the assessment of dietary intakes of fish provide a robust depiction of exposures throughout pregnancy. The use of numerous time points also helped to determine if associations persisted throughout childhood. The large cohort size and the high participant retainment throughout the follow-up appointments are also particular strengths of this research. Finally, this cohort was well characterised at numerous time points, allowing the comprehensive adjustment of confounders, including fish intakes throughout childhood. There are, however, several limitations. Like all observational epidemiological studies, no cause-and-effect relationships can be established and can only identify associations, which might be non-causal. Although models were adjusted for a range of key maternal and child characteristics, we cannot exclude the possibility of residual confounding from unmeasured factors. Variables such as gestational weight gain, paternal BMI, maternal physical activity, broader dietary patterns or other environmental exposures may have influenced child body weight and should be considered when interpreting these findings. These variables were not available as the present study was a secondary analysis of an existing cohort that was not originally designed to address these specific outcomes. The practical challenges of conducting data collection in a developing country also limited the feasibility of capturing all potential confounders. Maternal fish intakes were assessed using an FUQ administered at 28 weeks’ gestation which may not accurately capture fish consumption across the entire pregnancy. Moreover, the FUQ cannot capture other dietary factors which may have influenced results. Additionally, maternal PUFA status was measured at one time point only, and as mentioned, future research may wish to use repeated measures to determine any sensitive periods as to when the developing fetus may be differentially susceptible to the effects of the n-3 and n-6 PUFA. We were also unable to assess PUFA status throughout childhood. It is worth noting that in the present study, fatty acids were reported as absolute concentrations rather than as percentages of total fatty acids which may limit comparability to other studies. During pregnancy, however, absolute concentrations may more closely reflect maternal–fetal exposure, as placental transfer depends on circulating levels. In contrast, percentage values can be affected by changes in other lipid fractions, which fluctuate substantially in pregnancy^{(Reference Brenna, Plourde and Stark73)}. Finally, the methods used were to enable the detection of subtle associations between exposures and measures of child weight, and thus, the clinical significance of these findings is unclear.