Genetic Evidence for Causal Relationships Between Maternal Obesity-Related Traits and Birth Weight

Abstract

Importance: Neonates born to overweight or obese women are larger and at higher risk of birth complications. Many maternal obesity-related traits are observationally associated with birth weight, but the causal nature of these associations is uncertain.

Objective: To test for genetic evidence of causal associations of maternal body mass index (BMI) and related traits with birth weight.

Design, setting, and participants: Mendelian randomization to test whether maternal BMI and obesity-related traits are potentially causally related to offspring birth weight. Data from 30,487 women in 18 studies were analyzed. Participants were of European ancestry from population- or community-based studies in Europe, North America, or Australia and were part of the Early Growth Genetics Consortium. Live, term, singleton offspring born between 1929 and 2013 were included.

Exposures: Genetic scores for BMI, fasting glucose level, type 2 diabetes, systolic blood pressure (SBP), triglyceride level, high-density lipoprotein cholesterol (HDL-C) level, vitamin D status, and adiponectin level.

Main outcome and measure: Offspring birth weight from 18 studies.

Results: Among the 30,487 newborns the mean birth weight in the various cohorts ranged from 3325 g to 3679 g. The maternal genetic score for BMI was associated with a 2-g (95% CI, 0 to 3 g) higher offspring birth weight per maternal BMI-raising allele (P = .008). The maternal genetic scores for fasting glucose and SBP were also associated with birth weight with effect sizes of 8 g (95% CI, 6 to 10 g) per glucose-raising allele (P = 7 × 10(-14)) and -4 g (95% CI, -6 to -2 g) per SBP-raising allele (P = 1×10(-5)), respectively. A 1-SD ( ≈ 4 points) genetically higher maternal BMI was associated with a 55-g higher offspring birth weight (95% CI, 17 to 93 g). A 1-SD ( ≈ 7.2 mg/dL) genetically higher maternal fasting glucose concentration was associated with 114-g higher offspring birth weight (95% CI, 80 to 147 g). However, a 1-SD ( ≈ 10 mm Hg) genetically higher maternal SBP was associated with a 208-g lower offspring birth weight (95% CI, -394 to -21 g). For BMI and fasting glucose, genetic associations were consistent with the observational associations, but for systolic blood pressure, the genetic and observational associations were in opposite directions.

Conclusions and relevance: In this mendelian randomization study, genetically elevated maternal BMI and blood glucose levels were potentially causally associated with higher offspring birth weight, whereas genetically elevated maternal SBP was potentially causally related to lower birth weight. If replicated, these findings may have implications for counseling and managing pregnancies to avoid adverse weight-related birth outcomes.

Box 1

The maternal obesity-related traits hypothesized to cause increased or decreased fetal growth, based on observational associations with birth weight: body mass index (BMI); fasting glucose; gestational or type 2 diabetes; triglycerides; HDL-cholesterol; systolic blood pressure; vitamin D status (as indicated by 25-hydroxyvitamin D, 25[OH]D level); adiponectin.

Figure 1

Principle of Mendelian randomization: If a maternal trait causally influences offspring birth weight, then a risk score of genetic variants associated with that trait will also be associated with birth weight. Since genotype is determined at conception, it should not be associated with factors that normally confound the association between maternal traits and birth weight (e.g. socio-economic status). Estimates of the genetic score-maternal phenotype association (w) and the genetic score-birth weight association (x) may be used to estimate the association between the maternal trait variation that is due to genetic score, and birth weight (y = x/w), which is expected to be free from confounding. If the estimated causal relationship, y, is different from the observational association between the measured maternal phenotype and birth weight, this would suggest that the observational association is confounded (assuming that the assumptions of the Mendelian randomization analyses are valid). The dashed line connecting maternal trait with fetal growth has no arrow, to indicate that the causal nature of the association is uncertain. It is important to adjust for possible direct effects of fetal genotype (z).

Figure 2

Comparison of the observational with the genetic change in birth weight (in grams) for a 1 standard deviation (SD) change in each maternal obesity-related trait. For 25[OH]D and adiponectin, we present the change in birth weight for a 10% change in maternal trait level because these variables were logged for analysis. The genetic change was estimated from Mendelian randomization analysis, in which a genetic score was used to estimate the possible causal relationship between the maternal trait and birth weight. The genetic estimate is presented twice: in the second case it was adjusted for fetal genotype using a subset of available studies. The error bars represent the 95% confidence intervals around the effect size estimates. For maternal pre-pregnancy BMI and fasting glucose, the 95% confidence intervals for both the observational and genetic approaches exclude the null, suggesting positive possible causal relationships between maternal BMI and fasting glucose and birth weight. For maternal SBP, the observational analysis suggested a weak positive association with birth weight, whereas the genetic analysis showed evidence of a negative possible causal relationship. Observational analyses suggested that higher maternal triglyceride levels, lower maternal adiponectin and lower maternal HDL-cholesterol levels were associated with higher birth weight, while lower maternal vitamin D status was associated with lower birth weight, but none of these were supported by the genetic analyses.