Association of pre-pregnancy body mass index with offspring metabolic profile: Analyses of 3 European prospective birth cohorts

Background A high proportion of women start pregnancy overweight or obese. According to the developmental overnutrition hypothesis, this could lead offspring to have metabolic disruption throughout their lives and thus perpetuate the obesity epidemic across generations. Concerns about this hypothesis are influencing antenatal care. However, it is unknown whether maternal pregnancy adiposity is associated with long-term risk of adverse metabolic profiles in offspring, and if so, whether this association is causal, via intrauterine mechanisms, or explained by shared familial (genetic, lifestyle, socioeconomic) characteristics. We aimed to determine if associations between maternal body mass index (BMI) and offspring systemic cardio-metabolic profile are causal, via intrauterine mechanisms, or due to shared familial factors. Methods and findings We used 1- and 2-stage individual participant data (IPD) meta-analysis, and a negative-control (paternal BMI) to examine the association between maternal pre-pregnancy BMI and offspring serum metabolome from 3 European birth cohorts (offspring age at blood collection: 16, 17, and 31 years). Circulating metabolic traits were quantified by high-throughput nuclear magnetic resonance metabolomics. Results from 1-stage IPD meta-analysis (N = 5327 to 5377 mother-father-offspring trios) showed that increasing maternal and paternal BMI was associated with an adverse cardio-metabolic profile in offspring. We observed strong positive associations with very-low-density lipoprotein (VLDL)-lipoproteins, VLDL-cholesterol (C), VLDL-triglycerides, VLDL-diameter, branched/aromatic amino acids, glycoprotein acetyls, and triglycerides, and strong negative associations with high-density lipoprotein (HDL), HDL-diameter, HDL-C, HDL2-C, and HDL3-C (all P < 0.003). Slightly stronger magnitudes of associations were present for maternal compared with paternal BMI across these associations; however, there was no strong statistical evidence for heterogeneity between them (all bootstrap P > 0.003, equivalent to P > 0.05 after accounting for multiple testing). Results were similar in each individual cohort, and in the 2-stage analysis. Offspring BMI showed similar patterns of cross-sectional association with metabolic profile as for parental pre-pregnancy BMI associations but with greater magnitudes. Adjustment of parental BMI–offspring metabolic traits associations for offspring BMI suggested the parental associations were largely due to the association of parental BMI with offspring BMI. Limitations of this study are that inferences cannot be drawn about the role of circulating maternal fetal fuels (i.e., glucose, lipids, fatty acids, and amino acids) on later offspring metabolic profile. In addition, BMI may not reflect potential effects of maternal pregnancy fat distribution. Conclusion Our findings suggest that maternal BMI–offspring metabolome associations are likely to be largely due to shared genetic or familial lifestyle confounding rather than to intrauterine mechanisms.


Northern Finland Birth Cohort 1966 and 1986 (NFBC66 and NFBC86)
The Northern Finland Birth Cohort studies (http://www.oulu.fi/nfbc) are two longitudinal birth cohorts established to study factors affecting preterm birth and consequent morbidity in the two northernmost provinces of Finland, Oulu and Lapland. The NFBC66 includes 12,058 live births (12,231 children) covering 96% of all eligible births in this region during January-December 1966. Two decades later, a second cohort of 9,432 births (9,479 children) was obtained (NFBC86) which covered 99% of all the deliveries taking place in the target regions during July 1985-June 1986. In both cohorts, mothers and children have been followed-up since mothers enrolled at their first antenatal clinic visit (10-16th week). For NFBC86, the 16-year follow-up data collection (2001)(2002) included clinical examination and serum collection for 6621 adolescents (71% of the original cohort) [3,4]. In NFBC86: parental height, weight, occupation, smoking status, offspring sex and maternal parity were collected using questionnaires given to all mothers at their first antenatal clinic visit. Level of education was 2/3 obtained from questionnaires in 2001-02. Parental and offspring age was derived from their date of birth and date of assessments. Parental education was categorized into 8 categories from no occupational education to University degree, and occupation into 6 categories from entrepreneur to no-occupation. In NFBC66: maternal height, weight, occupation, smoking status, parity, child sex were reported by mothers at the first antenatal clinic visit (16 th week of gestation), or in questionnaires administered between the 24 th and 28 th week of gestation. Offspring age at serum collection was derived from their date of birth and date of attendance at the 1997-1998 follow-up clinic. Maternal age in pregnancy was derived from year of birth and the date of pregnancy questionnaire completion. Education was categorized into 9 categories from none or circulating school to beyond matriculation exam and occupation into 5 categories ranging from I (highest social class) to V (no-occupation). Information on paternal BMI was not collected in NFBC66.
Informed written consent was obtained from all participants. The research protocols were approved by the Ethics Committee of Northern Ostrobotnia Hospital District, Finland.
NMR-based metabolomics was measured on 5500 adolescents, of which 95% were overnight fasting serum. In NFBC66, the 1997 follow-up also included clinical examinations and serum sampling for 6007 participants aged 31 years (52% of target sample) [5]. This subsample was representative of the original cohort. In total, 5714 participants had NMR-based metabolite profiles measured, of which 96% were performed on overnight fasting serum [6]. In both cohorts, serum samples were stored at -80 °C until thawing NMR profiling in 2012.

Variable harmonization across cohorts
For the one-stage individual participant analysis (IPD) meta-analysis, education and head of house hold social class occupation categories were harmonized (see S1 Table). The former was collapse into basic or none, secondary, higher and other, and the latter into I (highest social class) to IV (lowest social class).

Metabolic NMR profiling
Each lipoprotein measurement is characterized by three elements: size (e.g. extremely large, very large, large, medium, small, very small), density (e.g. very low density lipoprotein (VLDL), Intermediate density lipoprotein (IDL), low density lipoprotein (LDL), high density lipoprotein (HDL)) and property (e.g. particle concentration, total lipids, triglycerides, phospholipids, total cholesterol, cholesterol esters, free cholesterol). The 14 lipoprotein subclass sizes were defined as follows: VLDL is subdivided into six subclasses, the largest being extremely large VLDL with particle diameters from 75 nm upwards and a possible contribution of chylomicrons, and five other VLDL subclasses (average particle diameters of 64.0 nm, 53.6 nm, 44.5 nm, 36.8 nm, and 31.3 nm); IDL (28.6 nm), three LDL subclasses (25.5 nm, 23.0 nm, and 18.7 nm), and four HDL subclasses (14.3 nm, 12.1 nm, 10.9 nm, and 8.7 nm) [7]. The mean sizes for VLDL, LDL and HDL particles were calculated by weighting the corresponding subclass diameters with their particle concentrations. The lipoproteins traits obtained are the concentration of the lipoprotein size-density-property combination in the total serum sample. For example, 0.5 mmol/l of very large VLDL cholesterol means 0.5 mmol of cholesterol embedded in very large VLDL particles per litre of serum. Remnant cholesterol was defined as VLDL-cholesterol + IDL-cholesterol, which is equivalent to total-cholesterol -HDLcholesterol -LDL-cholesterol [7]. For fatty acids (FA), only the cis configuration was quantified since the trans fatty acids are below the platform's detection limit [7]. FA were modelled as individual (absolute) measures and also as ratios (expressed as a %) to total FAs.

Statistical analyses, multiple testing correction
Principal component analysis (PCA) was performed separately on each individual cohort. In each cohort, all offspring who had data on the metabolic traits were used and PCA was performed on the z-scored metabolic data. This method assumes that the independence of the principal components (PCs) is equivalent to the degree of freedom of the original metabolic dataset, and that retaining a number of PCs that is enough to explain at least 95% of the variance will only result in a small chance of a type 1 error. Since the number of samples available varies across cohorts, the number of PCs needed to explain 95% of the variation in the metabolic traits also varies. The PCA results are as follows: ALSPAC: 2440 observations, 17 PCs; NFBC66: 4874 observations, 14 PCs; NFBC86: 2937 observations; 15 PCs. The highest number (17 components) was observed in the ALSPAC cohort and it was used as a conservative estimate of the number of independent tests been performed. Therefore, the threshold of p-value < 0.05 becomes p-value < 0.003 (i.e. α/17 where α=0.05), when multiple testing is considered, for assessing associations with the 153 metabolic traits.