Factors affecting the growth of children till the age of three years with overweight whose mothers have diabetes mellitus: A population-based cohort study

Background The prevalence of diabetes mellitus (DM) during pregnancy and childhood obesity is increasing worldwide. Factors affecting the growth of children with overweight whose mothers had DM are complicated and inconclusive. Few longitudinal studies have focused on the growth of infants with macrosomia born to mothers with DM and the factors influencing their overweight. This study explored risk factors for childhood overweight/obesity (OWOB) among children of mothers with DM. Perinatal, maternal socio-demographic, infant care, and maternal body weight characteristics as well as child growth until age 3 years were analyzed using a longitudinal design. Methods In total, 24,200 pairs of mothers and their children from the Taiwan Birth Cohort Study were included. Combined Taiwan Children Growth Curve report classifications were analyzed for infant growth at birth and at 6, 12, 18, 24, and 36 months old (m/o). A multiple logistic regression analysis with different model settings was used to assess factors affecting the growth of high birth weight children of mothers with diabetic mellitus (HODM). Results Children in the HODM group had a higher average body weight than did those in the non-DM group at different age stages. Relative to the non-DM group, weight gain in the HODM group was slower before 18 m/o but faster from 18 to 36 m/o, particularly after 24 m/o. Maternal DM was a major risk factor for childhood OWOB (odds ratio [OR]: 3.25–3.95). After adjustment for related confounders, the OR was 2.19–3.17. Maternal overweight or obesity and higher gestational weight gain were greater risk factors for childhood OWOB at 3 years old after adjusted maternal DM and other selected confounders (OR: 1.45 and 1.23, respectively). Breastfeeding until 6 m/o was a protective factor against childhood OWOB (OR: 0.95). The HODM and non-DM groups did not differ significantly in perinatal, maternal socio-demographic, or infant care characteristics. Conclusions Maternal DM is a major factor of childhood OWOB. Maternal body weight before and after pregnancy affects childhood OWOB, and this effect increases with the child’s age.


Background
The prevalence of diabetes mellitus (DM) is increasing rapidly worldwide and is the most common complication of pregnancy, affecting up to 10% of expectant mothers [1]. In Taiwan, the prevalence of diagnosed DM in pregnancy increased from 2.0% in 1996-1998 [2] to 3.5% in 2001-2008 [3]. A Taiwanese study demonstrated that morbidity is higher among children whose mothers had DM than among children of mothers without DM, and overweight newborns had an increased risk of hospitalization and repeat admission [2].
Infants exposed to high glucose levels prenatally have an increased risk of long-term adverse outcomes, including childhood overweight and obesity (OWOB). Many risk factors contribute to childhood overweight. Maternal obesity is a known major risk factor for childhood obesity. Many studies have demonstrated that mothers with higher body mass index (BMI) are more likely to have overweight infants [4]. Although high maternal BMI is associated with a substantially increased risk of gestational DM (GDM) [5], the causal associations of childhood overweight with maternal DM and obesity remain unclear. In addition to genetic and DM factors of child obesity, the effects of family lifestyle [6], breastfeeding, parents' education level, and even whether mothers who are able to perceive their children's body weight status accurately have been discussed [7,8].
To our knowledge, no studies examining the association between maternal DM and childhood OWOB have considered all potential key confounders, including maternal obesity, gestational weight gain, maternal and infant lifestyle, and care factors. Women with preexisting DM have higher risks of preterm labor, congenital anomalies in their children, adverse perinatal complications, and slowed infant growth in the future. Studies that have not excluded infants with low birth weight born to mothers with DM, especially infants born to women with GDM and type 2 DM, have methodological limitations in their analyses of childhood obesity [9]. A study recommended that infants of women with DM be targeted specifically for obesity prevention [10]. By controlling for maternal BMI, our study compared high birth weight of diabetic mothers (HODM) with children whose mothers did not have DM to explore risk factors for subsequent childhood OWOB and growth. The results may provide information for mothers with the most severe DM to prevent childhood obesity in their children.

Data source and study population
Data used in this study were collected from the Taiwan Birth Cohort Study (TBCS). TBCS is a national population-based cohort study in Taiwan to investigate   the health determinants of this 2005-2006 birth cohort  at their newborn, infant, child, adolescent, and adult  stages, from the life-course Taiwan birth registration database for 2005, 24,200 pairs of mothers and their children were sampled in the study from the TBCS through a multistage stratified systematic sampling method [11]. This dataset was compiled from the Millennium Cohort Study (MCS) [12] and the National Children's Study (NCS) [13]. The mothers provided written consent to participate, and this study was approved by the Medical Ethics Committee and Data Protection Board in Taiwan. A total of 21,248 motherchild pairs (87.8%) completed the first wave survey when child's age is 6 months old. A total of 20,645 motherchild pairs were enrolled into our study. Of the 20,645 pairs who complete the first wave of study, 19,597 (92.22%, 18 months) and 19,344 (91.03%, 36 months) completed the second wave and the third wave, respectively. Our study excluded multiple pregnancies (561 cases) and mothers who did not know their diagnosed DM during pregnancy at the time of interview (43 cases). Children with major or minor congenital fetal anomalies (998 cases) were not calculated into multiple logistic regression analysis models.

Variable definitions
Children were classified into four groups: children born to mothers without DM (non-DM), Offspring of mother with Diabetic Mellitus (ODM) with high birth weight (>4000 g, HODM), ODM with appropriate birth weight (2500-3999 g, AODM), and ODM with low birth weight (<2500 g, LODM). OWOB was defined as body weight above the 85th percentile at 6, 18, 24, or 36 m/o (separated by sex) according to Taiwan Children Growth Curve report classifications [14].
Mothers were asked questions regarding awareness of diagnosed DM by doctors during pregnancy by trained interviewers (yes or no), maternal BMIs before pregnancy and 6 months after delivery (calculated based on the data reported from the first wave survey), nationality, educational level, area of residence (urban or rural), family income, whether they breastfed or staple fed their children at 6 m/o, and main daytime caregiver (mother or others). Maternal age was defined as mother's age at delivery (older or younger than 35 years). APGAR scores were classified as more or less than 7. Maternal body weight gain during pregnancy was divided into three categories: <10, 10-14, and >14 kg. Prepregnancy and postpartum BMI were divided into three categories: < 18.5, 18.5-25, and >25 kg/m 2 .

Statistical analyses
We classified the variables of influential factors into four major categories, namely perinatal condition, maternal socio-demographic characteristics, infant care factors, and maternal body weight characteristics. Categorical variables were analyzed using the chi-square test among the non-DM, HODM, AODM, and LODM groups. Fisher's exact test was conducted if necessary ( Table 1).
Analysis of variance (ANOVA) in combination with the Tukey honestly significant difference post hoc test was conducted on normally distributed variables to compare children's body weight at different time points (birth, 6, 18, 24, and 36 m/o) and perform a sensitive assessment of deviations in growth ( Table 2). Figure 1 plots the body weights and changes in growth of all children born in the non-DM and HODM groups from birth to 36 m/o as a result of a Cochran-Armitage test for trend. Separate models for the HODM and non-DM groups were developed to detect influential factors of perinatal conditions, maternal socio-demographic characteristics, infant care, and maternal body weight characteristics. Multiple logistic regression analyses were used to estimate odds ratios (ORs) and 95% confidence intervals for the associations of influential variables between the HODM and non-DM groups at different ages, including maternal BMI, gestational weight gain, breastfeeding, and perinatal conditions ( Table 3). The non-DM group was used as the reference group, and P < 0.05 indicated statistical significance. Statistical analyses were performed using SAS 9.4. (SAS Institute Inc., Carrie, North Carolina, USA).

Results
A total of 20,645 (85.3%) pairs of mothers and children were enrolled into our study. Among the pairs, 446 (2.16%) children were born to mothers with diagnosed DM during pregnancy. Regardless of birth weight, all selected influential categories, including perinatal conditions, maternal socio-demographic characteristics, and maternal body weight characteristics, revealed significant differences between the non-DM group and ODM groups, except infant care factors. Compared with mothers without DM, mothers with DM were more likely to have advanced maternal age and live in urban areas. DM was less likely in immigrant women and those with higher education levels (P < 0.05). Differences in family income, main caregiver, and breastfeeding versus staple-food feeding at 6 m/o were non-significant. More adverse birth outcomes, such as preterm labor, lower APGAR scores 1 and 5 min after birth, and increased prevalence of birth defects, were present in infants in the ODM groups. Maternal BMIs before and after pregnancy were higher in the DM groups than in the non-DM group. In particular, mothers in the HODM group had obviously higher prevalence of high prepregnancy BMI (46.2% vs. 9.7%) and postpartum BMI (100.0% vs. 63.8%), (P < 0.001). Differences in maternal gestational weight gain among the four groups were nonsignificant (P = 0.157) ( Table 1). Figure 2 plots the growth of all children in the HODM and non-DM groups. The average birth weight in these two groups was 4294.5 and 3122.7 g, respectively. From birth to 18 m/o, child growth in both the groups HODM had similar velocity and slope in terms of body weight, and average body weight was higher in the non-DM group. However, the velocity of growth in body weight was faster in the HODM group between 18 and 36 m/o. Average body weight at 36 m/o was 16,830.8 and 14, 939.4 g in the HODM and non-DM groups, respectively. Growth trends in both the groups were significantly different (P < 0.05). Figure 3 presents the percentages of childhood OWOB in the four groups. Children in the HODM group had higher rates of child OWOB at each age stage, especially at 36 m/o.
The disparities in infant body weight and body weight gain between the HODM and non-DM groups at birth, 6, 18, 24, and 36 m/o are shown in Table 2. Children in the HODM group had a higher average body weight than did the non-DM group at different age stages. Body weight gain among infants differed significantly between the two groups. Body weight gain was slower in the HODM group than in the non-DM group before 18 m/o but faster than the non-DM group, from 18 to 36 m/o showed latter catch-up in HODM group, particularly after 24 m/o. Table 3 presents the results of the stepwise model analysis by using a multiple logistic regression. In the null model, the incidence of OWOB in children in the HODM group was greater than that in children in the non-DM group from birth to 36 m/o (OR: 3.95-3.74). Model A, which adjusted for maternal sociodemographic and perinatal conditions, yielded similar results. When infant care factors in model B were considered, breastfeeding to 6 m/o reduced childhood OWOB among children in the HODM group by approximately 4%, however, mothers as the main daytime caregiver or early staple-food feeding did not affect childhood OWOB (P > 0.05). Model C adjusted for the factors included in model A and maternal body weight factors to analyze the effects of maternal body weight on childhood OWOB. Compared with the non-DM group, mothers in the HODM group were more likely to have OWOB children when their gestational weight gain was >14 kg (OR: 1.12-1.25). Obesity (BMI > 25 kg/m 2 ) before and after pregnancy also increased the risk of childhood

Discussion
GDM prevalence in Asia has been reported to range from 0.7 to 51.0% [15]. A meta-analysis yielded a pooled GDM prevalence rate of 11.5% [16]. This vast disparity in prevalence rates may be due to differences in diagnostic criteria [17], screening strategies [18], and ethnicity and population characteristics [19]. Our study included 446 mothers with DM who accounted for approximately 2.2% of all included participants. The TBCS study defines whether a mother has DM during pregnancy, in addition to the self-recognition from the interviewees, must be judged by trained interviewers (mainly public health nurses) to determine whether there is "diagnosed DM by doctors during pregnancy", and refer to the     check-up records on the maternal health booklet issued by Taiwan's National Health Insurance (NHI). The coverage rate of NHI is >99%. Pregnant women can receive at least 10 prenatal health examinations by obstetricians, and barriers in access to prenatal care are lower in Taiwan compared with those in other countries. This discrepancy in prevalence rates might be related to the design of the TBCS, which was connected as a selfreport questionnaire by mothers administered in interviews. Although the urine glucose level is a routine parameter tested during prenatal examinations, serum glucose levels and oral glucose tolerance tests are not routine items designated by Taiwan's NHI for prenatal examinations, mothers with DM may not be aware of their DM status. Such issue may cause some GDM mothers not being diagnosed and not included in DM group of this study. However, it is reported that nearly 95% of pregnant women received free prenatal care by the NHI in Taiwan. Doctors are required to screen for maternal DM by NHI guidelines, so we believe it has consistent validity. Our study using a dataset from the TBCS is the first to use a multistage stratified systematic sampling design, and the content of the questionnaire covered a wide range of information. TBSC has been widely applied in research to reveal insights into the health profiles of children growing up in Taiwan. Advanced maternal age, family history of DM, ethnicity, overweight or obesity, and smoking are welldocumented risk factors for GDM [20]. In addition to these risk factors, data increasingly indicate that diet and lifestyle factors both before and during pregnancy are associated with GDM [21]. A study in Taiwan indicated that advanced maternal age, a family history of DM, a less than high school education, high prepregnancy BMI, and an indigenous ethnicity were risk factors for GDM [22]. In our study, mothers with DM were older, lived in urban areas, were less likely to be immigrants, and had higher education levels (p < 0.05). Their BMIs before pregnancy and postpartum at 6 months were higher than those of the non-DM group. In particular, mothers in the HODM group had higher prepregnancy BMIs. The disparity in maternal age may explain why mothers with DM had higher education levels; moreover, immigrant and indigenous mothers in Taiwan are relatively younger.
Because of the potential confounders including genetic factors, mother and infant lifestyle, and maternal obesity factors are difficult to control, the relationships between maternal DM and risk of childhood OWOB are complex. The Taiwan's NHI has a high coverage rate and people face few barriers to medical assessments. We focused on the HODM group because mothers in the LODM group tended to have pregestational DM that increased the incidence of congenital anomalies and preterm labor. Our results indicate that maternal obesity before and after pregnancy increase the risk of childhood OWOB, and this risk increases as children age. These findings suggest that genetics may play a central role. A previous study identified many genetic polymorphisms in genome-wide association studies of adult BMI and the genetic susceptibility to childhood obesity; the association was partially explained by appetitive traits in infancy followed by an early childhood increase in BMI [23].
Studies of the effects of gestational DM on childhood overweight and obesity have yielded inconclusive results. A meta-analysis reported inconsistent evidence of an association of GDM with childhood overweight and obesity because of methodological limitations in the included studies [9]. However, our study adjusted for prepregnancy obesity, infant care, and maternal sociodemographic factors. Maternal DM was an independent determinant of childhood OWOB, but the associations were attenuated after adjustment for prepregnancy and postpartum maternal BMI. Previous studies have not focused on infants with macrosomia, which may have confounded their results. Because the infants of mothers with DM, particularly type 1 DM, have a higher risk of congenital anomalies, premature birth, and low APGAR scores, their growth is expected to be slower than that of infants of mothers without DM or even mothers with GDM. We conducted a stepwise logistic regression by using five models to eliminate these confounders.
Because of mothers with overweight or obesity have a higher prevalence of DM, our study could not identify the prevalence of DM in ODM children. Mothers with obesity may have a genetic predisposition to have children who become overweight; in addition, maternal lifestyle, diet, and maternal cognition to infant body weight are factors of childhood overweight [24]. In our study, main daytime caregiver and early staple-food feeding were not influential factors in Taiwanese cohort. However, high calorie intake in infancy and large appetite from 1 year were reported to be related to a higher incidence of subsequent childhood obesity [23]. Children who were breastfed until 12 m/o were reported to be 2.7 times less likely to develop childhood obesity [24,25]. We demonstrated that breastfeeding until 6 m/o is a protective factor against childhood OWOB (OR: 0.96). Although the benefits of breastfeeding appear to be a key part of the positive health outcomes associated with the parent-child relationship, the effect of the dose is not informed in this study. The odds ratio of HODM group to the non-DM group for childhood OWOB is range from 3.25 to 3.95 at different age stages. Regardless of whether neonatal care factors were adjusted in Model A or Model B, the odds of HODM in the probability of children being OWOB has similar findings. The effects to the odds of HODM between the former two are much lower than those of Model C and Model D. Therefore, we infer that HODM is an independent factor affecting childhood OWOB, and the impact of maternal overweight or obesity is much greater than that of perinatal conditions and acquired care. Control of DM and restricted weight gain during pregnancy are keys to preventing childhood obesity, especially in mothers with DM.
The strength of our study is its focus on a particular infant population for comparison with children whose mothers did not have DM and its exclusion of infants with low birth weight and congenital anomalies to eliminate the effect of congenital or genetic factors that could confound the statistical results. We also used a stepwise logistic regression model to detect confounders and their interactions between DM and diabetes-related perinatal complications. In addition, we analyzed prepregnancy and postpartum BMI to compare the effects of maternal BMI and gestational body weight gain with childhood OWOB. Moreover, this cohort study can offer information on the determinants of childhood OWOB at different ages.
Our research has some limitations. The design of the TBCS questionnaires did not differentiate between pregestational DM and GDM, neither lack of medical records about the severity of hyperglycemia and related medical intervention. This limitation makes genetic factors and the effects of DM control difficult to identify. Another limitation is the high rate of cesarean deliveries in Taiwan; some infants with macrosomia can be delivered earlier if their body weight is estimated to be overweight during prenatal examinations. This phenomenon may have reduced the number of infants in the HODM group. The maternal height and weight were selfreported instead of measurements might latent errors. This cohort study followed children to 36 m/o. Future studies may prolong the follow-up period to analyze the long-term relationships between mothers with DM and childhood obesity.

Conclusion and suggestions
In this population-based cohort study, maternal DM and BMI of mothers more than 25 kg/m 2 prepregnancy and postpartum were the significant factors of childhood OWOB, and the maternal BMI effects increased as children aged. Prolonged breastfeeding to the age of 6 months and weight gain control during pregnancy are protection factors. DM control during pregnancy, restricted maternal gestational body weight gain, maternal BMI control, and prolonged breastfeeding are strategies for mothers with DM to prevent childhood overweight.