Introduction
Mean body mass index (BMI) has increased worldwide, and obesity reached epidemic proportions Reference Risk and Collaboration1 . In Pelotas, a southern Brazilian city, the prevalence of maternal prepregnancy overweight increased from 22.1% in 1982 to 47.0% in 2015 Reference Horta, Barros and Lima2 . In the short term, maternal overweight during pregnancy is associated with a higher risk of preeclampsia, gestational diabetes, spontaneous preterm labor, macrosomia, congenital anomalies, stillbirth, unsuccessful breastfeeding, and even maternal death Reference Gesink Law, Maclehose and Longnecker3-Reference Turcksin, Bel, Galjaard and Devlieger5 . Furthermore, it has also been suggested that maternal overweight in the pregnancy may program the development of cardiometabolic risk factors in the offspring. Maternal prepregnancy BMI would be positively associated with higher offspring BMI, fat mass percentage, and mean levels of metabolic cardiovascular risk factors such as blood pressure, glucose, and cholesterol Reference Lemas, Brinton and Shapiro6-Reference Brandt, Moß and Lennerz18 .
Concerning the mechanisms for the association of maternal prepregnancy BMI with offspring cardiovascular risk, maternal overweight/obesity during pregnancy would induce changes in fetal metabolism that would have an effect on offspring body composition and the development of metabolic cardiovascular risk factors Reference Lawlor, Smith and O’Callaghan19 . Overweight or obese women would have higher glucose, lipids, and fatty acids levels, which may lead to fetal overfeeding and consequently to changes in energy metabolism and the fetal endocrine system, resulting in differences in appetite control, risk of obesity, and cardiometabolic outcomes Reference Lawlor20,Reference Farrar, Simmonds and Bryant21 . Furthermore, offspring anthropometric measurements or body composition could be another mechanism for this association. As previously mentioned, offspring of overweight or obese mothers have higher BMI Reference Daraki, Georgiou and Papavasiliou8,Reference Rath, Marsh and Newnham13,Reference Stamnes Køpp, Dahl-Jørgensen and Stigum14,Reference Dias, Matijasevich and Barros16 , which is positively associated with blood pressure, glucose, and changes in lipid profile Reference Malta, Bernal and Vieira Neto22–Reference Nyberg, David Batty and Pentti24 . Concerning the control for confounding, most of the previously published studies adjusted their estimates for possible mediators, such as gestational age, birth weight, offspring smoking status, offspring schooling, offspring physical activity, age, and maturation stage Reference Daraki, Georgiou and Papavasiliou8,Reference Tan, Roberts and Catov9,Reference Stamnes Køpp, Dahl-Jørgensen and Stigum14,Reference Gademan, Vermeulen and Oostvogels15,Reference Hochner, Friedlander and Calderon-Margalit17,Reference Brandt, Moß and Lennerz18 . By adjusting the estimates to possible mediators, a causal pathway is blocked, the total effect underestimated, and a collider bias is introduced Reference Shrier and Platt25 . Some studies may be susceptible to residual confounding because they have not adjusted their estimates for known confounding factors, such as socioeconomic status Reference Rath, Marsh and Newnham13,Reference Morea, Miu, Morea and Cornean26 . It is important to distinguish between confounder and mediator. A confounder is a variable that has a direct effect on the exposure and the outcome, whereas a mediator is caused by the exposure and is associated with the outcome, i.e. the mediator is in the causal pathway between exposure and outcome Reference Discacciati, Bellavia, Lee, Mazumdar and Valeri27,Reference VanderWeele28 .
To our knowledge, only one study assessed the possible mediators for the association of maternal prepregnancy BMI with offspring cardiometabolic risk factors. Daraki et al evaluated the role of gestational weight gain, birth weight, breastfeeding duration, and TV watching as possible mediators Reference Daraki, Georgiou and Papavasiliou8 . The possible mediators were included in the multiple linear regression model and adjustment for these variables did not change substantively the regression coefficients Reference Daraki, Georgiou and Papavasiliou8 . This approach has been criticized because it does not adjust for confounders of the mediator – outcome association and does not consider a possible interaction between exposure and the mediator Reference Discacciati, Bellavia, Lee, Mazumdar and Valeri27 .
The present study was aimed at evaluating the association of maternal prepregnancy BMI with offspring blood pressure, random blood glucose, and non-HDL cholesterol in adolescence and adulthood in three Brazilian birth cohorts. We also estimated the indirect effect of offspring BMI in the association of maternal prepregnancy nutritional status with metabolic cardiovascular risk factors using statistical methods that adjust for mediator-outcome confounders.
Methods
Study design and participants
In 1982, 1993, and 2004, all maternity hospitals in Pelotas, a southern Brazilian city, were daily visited and all births identified. Those live births whose mothers lived in the urban area of the city were examined, and their mothers were interviewed soon after delivery (1982, n = 5914, 1993, n = 5249, and 2004, n = 4231). These subjects have been followed up for several times at different ages. In the present study, we used data from the last follow-up of each cohort, which was carried out at mean ages of 30.2 (SD: 0.3), 22.6 (SD: 0.3), and 10.9 (SD: 0.3) years of age for the 1982, 1993, and 2004 cohorts, respectively. In these visits, subjects were invited to visit the research clinic to be interviewed and examined.
Maternal prepregnancy BMI
In the three cohorts, information on maternal prepregnancy weight was gathered from prenatal cards or when absent by self-report in the perinatal visit. With respect to height, in 1982 and 1993, the mothers were measured by the hospital staff and this data were retrieved from the hospital records, whereas in 2004, the mothers were measured at home in the 3-month visit by the research team. In all cohorts, height was evaluated using a locally made portable stadiometer with a precision of 1 mm. As suggested by the World Health Organization (WHO), those mothers with a BMI < 18.5 kg/m2 were considered as underweight, normal weight was defined by a BMI between 18.5 and 24.9 kg/m², overweight by a BMI ≥ 25.0 and ≤ 29.9 kg/m² and a BMI ≥ 30.0 kg/m² defined the presence of obesity 29 .
Metabolic cardiovascular risk factors
In the three cohorts, blood pressure was measured using an automatic device, model HEM-705CPINT – Omron, on the left arm, with a cuff appropriate for arm circumference. Two measurements were taken with the subject seated, one at the beginning of anthropometry and other at the end, and the mean of the two readings (in mmHg) was used in the analyses.
Blood cholesterol and glucose were evaluated only for the 1982 and 1993 cohorts, using plasma or serum that was collected during the visit and was stored at −80ºC. Total and HDL cholesterol were measured using an enzymatic colorimetric method, with a chemical analyzer BS 380 Mindray. Non-HDL cholesterol (mg/dl) was calculated as the difference between total and HDL cholesterol. Random blood glucose (mmol/l) was measured from fingertip blood, using a portable glucose meter (Accu-Check Advantage – Roche). Information on the time of the last meal and the blood collection were recorded so that the time elapsed since the last meal could be calculated.
Confounders and mediators
The following variables, measured in the perinatal study, were considered as possible confounders: maternal schooling in complete years (0–4, 5–8, 9–11, ≥ 12), family income at birth in tercile, maternal age in years (<20, 20–25, 26–30, >30), parity (1, 2, ≥3), and maternal smoking during pregnancy (No/Yes).
Offspring BMI at the last visit was considered as a possible mediator. Cohort members were weighed to the nearest 0.1 kg using a scale coupled to the BodPod (COSMED, Chicago, USA) with a maximum of 150 kg, and height was measured with a portable stadiometer (SECA 240; SECA, Birmingham, UK). BMI was calculated by dividing the weight by the squared height in meters (kg/m²).
In the mediation analysis, offspring schooling in complete years (0–4, 5–8, 9–11, ≥ 12), family income at the last visit, physical activity, and diet were considered as post-confounders. Physical activity was measured using GENEActiv accelerometer (ActivInsights, Kimbolton, UK) for the 1982 cohort, and ActiGraph, wGT3X-BT, wGT3X, and ActiSlee models (ActiGraph, USA) for the 1993 and 2004 cohorts. Accelerometers were worn for 7 consecutive days and we evaluated the time spent in moderated and vigorous physical activity. Diet was assessed with a food frequency questionnaire, and the Block score was used to estimate the intake of fiber and fat.
Statistical analysis
Analyses were carried out using Stata version 14.0. Analysis of variance (ANOVA) was used to compare means and multiple linear regression to adjust for possible confounders (family income at birth, maternal schooling, maternal age, parity, maternal smoking during pregnancy, and cohort membership).
Figure 1 illustrates the potential pathways between maternal prepregnancy BMI with offspring non-HDL cholesterol, systolic and diastolic blood pressure. Mediation analysis was carried out using G-computation formula to decompose the total effect into natural direct and indirect effects of maternal prepregnancy BMI on offspring metabolic cardiovascular risk factors. Standard errors for mediation analyses were calculated using bootstrapping with 10,000 simulations. Separate models were fitted for each offspring outcome. The data from the three cohorts were pooled, analyzed, and all models were adjusted for base confounders (family income at birth, maternal schooling, maternal age, parity, maternal smoking during pregnancy, and cohort membership), and post-confounders (offspring schooling, family income at last visit, physical activity, and diet). A term for exposure–mediator interaction was included in all models. Statistical significance was set at heterogeneity and trend p-values < 0.05, and all tests were two-tailed. The beta coefficients were converted into outcome units (mmHg, mg/dl ,and mmol/l) to facilitate understanding and interpretation of results.
Fig. 1. Directed acyclic graph showing potential pathways between maternal prepregnancy BMI with offspring non-HDL cholesterol, systolic and diastolic blood pressure.
This study was carried out according to guidelines established in the Declaration of Helsinki and all procedures involving human subjects were approved by the Research Ethics Committee of the School of Medicine, Federal University of Pelotas.
Results
Table 1 shows that the prevalence of maternal prepregnancy obesity increased from 4.3% in 1982 to 6.2% in 2004, maternal schooling increased across the three cohorts, whereas the prevalence of maternal smoking in the pregnancy slightly decreased from 35.6% in 1982 to 28.6% in 2004. Mean birth weight was similar in the three cohorts, but the prevalence of preterm birth increased from 6.2% in 1982 to 13.3% in 2004. The mean systolic blood pressure was 121.1 mmHg (SD.: 13.8) for the 1982 cohort, 123.4 mmHg (SD: 13.7) for the 1993 cohort, and 112.5 mmHg (SD: 11.1) for the 2004 cohort. Mean non-HDL cholesterol was 132.4 mg/dl (SD: 36.7) for the 1982 and 110.4 mg/dl (SD: 32.9) for the 1993.
Table 1. Characteristics of the studied population of three Pelotas birth cohorts, Rio Grande do Sul State, Brazil
Table 2 shows that after controlling for confounding, systolic blood pressure was associated with maternal prepregnancy BMI. Offspring of overweight or obese mothers had a higher mean systolic blood pressure than those from normal-weight mothers, with mean differences of 1.25 mmHg (95% CI: 0.45; 2.05) and 2.13 mmHg (95% CI: 0.66; 3.59), respectively. For diastolic blood pressure, the magnitude of the association increased after adjusting for confounding, and a positive association was already observed in the crude analyses. Mean random blood glucose was slightly higher among offspring of obese mothers, but the confidence interval included the reference. Therefore, the observed difference may have been due to the random variation. Non-HDL cholesterol among offspring of overweight and obese mothers was 2.70 mg/dl (95% CI 0.25; 5.15) and 4.22 mg/dl (95% CI − 0.58; 9.02) higher, respectively, than that among those from normal-weight mothers.
Table 2. Association between maternal prepregnancy nutritional status and offspring cardiometabolic risk factors, a pooled analysis on three Pelotas birth cohorts, Rio Grande do Sul State, Brazil
* Adjusted: family income at birth, maternal schooling, maternal age, parity, maternal smoking during pregnancy, and cohort membership.
a Data from three cohorts.
b Data from 1982 and 1993 cohort.
˦ Heterogeneity p-value.
ǂ Trend p-value.
Table 3 shows that offspring BMI captured the effect of maternal prepregnancy nutritional status on non-HDL cholesterol, and on systolic and diastolic blood pressure. The natural indirect effect of offspring BMI was 1.91 mmHg (95% CI 1.63; 2.19) for mean systolic blood pressure, 1.21 mmHg (95% CI 1.02; 1.40) for mean diastolic blood pressure, and 3.14 mg/dl (95% CI 2.55; 3.72) for mean non-HDL cholesterol. The controlled direct effect was similar between the different levels of the mediator, suggesting that there was no interaction.
Table 3. Total effect, natural direct effect, and natural indirect effect (via offspring body mass index) of maternal prepregnancy body mass index on offspring systolic and diastolic blood pressure, a pooled analysis for the three Pelotas birth cohort studies (n = 8827)
* Offspring body mass index.
Base confounders (family income at birth, maternal schooling, maternal age, parity, maternal smoking during pregnancy, and cohort membership).
Post-confounders (offspring schooling, family income at last visit, offspring physical activity, and offspring diet).
Discussion
In the present study, maternal prepregnancy BMI was positively associated with offspring systolic and diastolic blood pressure and non-HDL cholesterol, whereas we did not observe an association with blood glucose. The mediation analysis showed that offspring BMI captured the total effect of maternal prepregnancy BMI.
Concerning the mediation analysis, as expected, offspring BMI was a mediator of the association between maternal prepregnancy BMI and offspring outcomes. Since BMI is positively associated with blood pressure and lipids Reference Miranda, Carrillo-Larco and Ferreccio30 , and it has also been reported that maternal prepregnancy nutritional status is associated with offspring BMI Reference Daraki, Georgiou and Papavasiliou8,Reference Rath, Marsh and Newnham13,Reference Stamnes Køpp, Dahl-Jørgensen and Stigum14,Reference Dias, Matijasevich and Barros16,Reference Jacota, Forhan, Saldanha-Gomes, Charles and Heude31 . We tested whether these associations were modified by breastfeeding and offspring BMI; however, we did not observe any indication of interaction by these variables. In a previously published study, we evaluated the association of maternal prepregnancy BMI with offspring BMI in adolescence and early adulthood. Offspring BMI was positively associated with maternal prepregnancy BMI, even after controlling for confounders Reference Dias, Matijasevich and Barros16 . Mean BMI of offspring of obese mothers was 0.65 kg/m² (95% CI: 0.48; 0.81) higher than that among those of normal-weight mothers in 1982 Pelotas birth cohort (mean age at anthropometric assessment: 30.2 years), the mean difference for 1993 Pelotas birth cohort, at a mean age of 22.6 years, was 0.65 kg/m² (95% CI: 0.45; 0.85), and for 2004 Pelotas birth cohort (mean age at anthropometric assessment: 10.9) was 0.83 kg/m² (95% CI: 0.64; 1.01). And, we also observed that offspring diet and physical activity did not modify this association Reference Dias, Matijasevich and Barros16 .
We failed to observe an association of blood glucose with maternal prepregnancy BMI. Wander et al also did not observe an association between maternal prepregnancy BMI with offspring blood glucose at a mean age of 32 years Reference Wander, Hochner and Sitlani11 . Similar finding was reported by Tan et al among 12-year-old subjects, when comparing the mean blood glucose concentration of offspring whose mothers were overweight or obese with offspring whose mothers were normal weight, but the estimates were not adjusted for confounding factors Reference Tan, Roberts and Catov9 . On the other hand, Westberg et al showed that maternal BMI was negatively associated with 2-hour glucose concentration among male offspring, whereas no association was observed among women Reference Westberg, Kautiainen and Salonen7 . In that cohort, participants were aged 62 and the analyses were adjusted for offspring age, BMI, smoking, and leisure time physical activity Reference Westberg, Kautiainen and Salonen7 . As these variables are possible mediators, the adjustment may have introduced a collider bias, and we cannot predict in which direction the measures of association were affected Reference Shrier and Platt25 . Concerning cholesterol, similarly to other studies Reference Tan, Roberts and Catov9,Reference Wander, Hochner and Sitlani11,Reference Santos Ferreira, Williams and Kangas12,Reference Hochner, Friedlander and Calderon-Margalit17 , we observed a worse lipid profile among offspring of obese or overweight mothers.
In relation to study strengths, the association of maternal prepregnancy nutritional status with offspring cardiometabolic risk factors was evaluated using information from three birth cohorts, carried out in a southern Brazilian city. And information on maternal anthropometry and confounders were collected just after delivery using standardized methods, minimizing measurement error and reducing the probability of residual confounding and nondifferential misclassification.
As limitations, we can point out that blood samples were not collected during the fasting period. In relation to cholesterol measures, it has been reported that nonfasting blood collection is better for estimating individual cardiovascular risk Reference Langsted and Nordestgaard32,Reference Nordestgaard33 . On the other hand, for blood glucose, the analyses should be carried out with fasting blood. To minimize this limitation, we corrected the estimates for the time since the last meal. Maternal prepregnancy BMI was based on self-reported data for height and prepregnancy maternal weight. However, a validation study based on data from the Brazilian National Health Survey observed a high agreement between self-reported and measured weight, height, and BMI Reference Nordestgaard33 . Furthermore, information on maternal prepregnancy BMI was collected just after delivery, reducing the likelihood of recall bias. Regarding a possible collinearity between the confounding factors, the correlation was not very high among the confounders, decreasing the likelihood of collinearity. Furthermore, even if these variables were collinear, it is unlikely that the magnitude of the association between maternal BMI and offspring cardiovascular risk factors would be biased.
In conclusion, our results showed that offspring BMI explained the association of maternal prepregnancy nutritional status with offspring blood pressure and non-HDL cholesterol levels, reinforcing the importance of implementing interventions aimed at reducing the individual’s BMI.
Acknowledgments
This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – Brasil (CAPES) –- Finance Code 001. This article is based on data from the three Pelotas Birth Cohorts studies (1982, 1993, and 2004) conducted by Postgraduate Program in Epidemiology at Federal University of Pelotas with the collaboration of the Brazilian Public Health Association (ABRASCO).
Financial support
This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – Brasil (CAPES) – Finance Code 001. This article is based on data from the three Pelotas Birth Cohorts studies (1982, 1993, and 2004) conducted by Postgraduate Program in Epidemiology at Federal University of Pelotas with the collaboration of the Brazilian Public Health Association (ABRASCO). From 2004 to 2013, the Wellcome Trust supported the 1982 and 1993 birth cohort studies and for the 2004 birth cohort study from 2009 to 2013. The International Development Research Center, World Health Organization, Overseas Development Administration, European Union, National Support Program for Centers of Excellence (PRONEX), the Brazilian National Research Council (CNPq), and the Brazilian Ministry of Health supported previous phases of these studies. The European Union, National Support Program for Centers of Excellence (PRONEX), the Brazilian National Research Council (CNPq), and the Brazilian Ministry of Health and Children’s Pastorate supported previous phases of the study. AM, ANBM, FCB, FCW, HG, ISS, and BH are supported by the CNPq. MSD received a scholarship for a PhD degree from the Coordination of Improvement of Higher Level Personnel (CAPES).
Author’s contributions
MSD and BLH participated in all stages from conception, written part, data analysis, discussion of results. AM, ANBM, FCB, FCW, HG, and ISS revised the final version.
Conflicts of interest
The authors declare that they have no conflict of interest.
Ethical standards
This study was carried out according to guidelines established in the Declaration of Helsinki and all procedures involving human subjects were approved by the Research Ethics Committee of the College of Medicine, Federal University of Pelotas.
