Adverse pregnancy outcomes on the risk of overweight offspring: a population-based retrospective study in Xiamen, China

The growth trajectory of Chinese preschoolers still remains unclear. Our objective was to determine whether there was an association between adverse pregnancy outcomes and overweight offspring. We analyzed population-based retrospective cohort data from the Medical Birth Registry of Xiamen, which comprised 33,157 children examined from 1 to 6 years of age. Longitudinal analyses were used to evaluate the growth trajectories of offspring body mass index (BMI). Multivariate logistic regression was used to assess the effects of two adverse pregnancy outcomes, gestational diabetes mellitus (GDM) and being large-for-gestational age (LGA), on childhood overweight. Offspring of mothers with GDM and LGA has a higher annual BMI z-score from 1 to 6 years of age (all P < 0.05). But, a higher annual BMI z-score was only observed in children aged 1–5 years in models 1–3. Overall BMI z-score of offspring aged 1–6 who were born to mothers with GDM and LGA were also higher in models 1–3 (all P < 0.05). Additionally, offspring of mothers with GDM and LGA had a higher risk for overweight in model 1, from 1 to 6 years of age (odds ratio (OR), 1.814; 95% confidence interval (CI), 1.657–1.985; P < 0.0001). However, this association was attenuated after adjusting for maternal pre-pregnancy BMI (OR, 1.270; 95% CI, 0.961–1.679; P = 0.0930). Offspring of mothers with GDM and LGA had a higher BMI z-score and increased risk for overweight. Indeed, intrauterine exposure to maternal GDM and LGA could bias offspring to overweight, whereas maternal pre-pregnancy BMI may play a key role in offspring overweight for children born to mothers with GDM and LGA.


Effects of m-GDM-
LGA on offspring BMI z-score. All pregnant women were divided into three groups according to maternal GDM and gestational age status, as shown in Table 2: m-nonGDM-AGA, m-GDM-AGA, and m-GDM-LGA. In models 1, 2, and 3, offspring who were born to m-GDM-LGA pregnancy were more likely to have higher BMI z-score from 1 to 5 years of age, compared with offspring who were born to m-nonGDM-AGA pregnancy or m-GDM-AGA pregnancy (all P < 0.05). However, no significant differences were observed for offspring BMI z-score at 6 years of age, even after adjusting for pre-pregnancy BMI and other covariates (all P > 0.05).
Our mixed model showed that offspring exposed to m-GDM-LGA pregnancy had the highest BMI Z-score trajectory compared with those exposed to m-GDM-AGA and m-nonGDM-AGA pregnancy in an unadjusted model (Fig. 2). Variable Available n Exposed to m-nonGDM-AGA (n = 26379) Exposed to m-GDM-AGA (n = 5179) Exposed to m-GDM-LGA (n = 1599) P value

Effects of m-GDM-
LGA on offspring BMI z-score and overweight. Offspring who were born to m-GDM-LGA pregnancy were more likely to have higher BMI z-score (models 1, 2, and 3) and become overweight (models 1 and 2) in early childhood (Table 3). After adjusting for model covariates, offspring who were born to m-GDM-LGA pregnancy had 1.814-fold higher odds of becoming overweight from 1 to 6 years of age (OR, 1.814; 1.657-1.985; P < 0.0001). 1.366-fold higher odds after adjusting for model 2 (model 1 plus maternal gestational weight gain) covariates (P = 0.0273), and 1.270-fold higher odds after adjusting for model 3 (model 2 plus pre-pregnancy BMI) covariates (P = 0.0930).

Discussion
To our knowledge, this is a large study to research association between GDM and LGA and preschool-aged children overweight after adjusting for covariates, and the first to associate BMI z-score with children overweight in preschool-aged children in China. We observed that offspring born to m-GDM-LGA pregnancy showed BMI z-score acceleration from 5 to 6 years of age compared with offspring born to m-GDM-AGA pregnancy. Meanwhile, annual BMI z-score had accelerated from 3 to 6 years of age. These results are consistent with a previous study that found BMI standard-deviation score increased with age, with particularly high acceleration during preschool years 13 . Several studies have also reported that a relationship exists between adolescent and adult overweight or obesity and high BMI during childhood [14][15][16] . In addition, only 20 percent of young children who are overweight or obese will become a normal weight during adolescence. Previous studies have tracked BMI trajectories, found that high-growth BMI trajectory groups diverge from stable BMI trajectory groups starting at around 3 years of age 17,18 . These results are consistent with our analysis.
Offspring born to m-GDM-LGA pregnancy retained higher BMI z-score during preschool years and had a higher risk of overweight compared with offspring born to m-GDM-AGA pregnancy after adjusting for covariates. However, the association between m-GDM-LGA pregnancy and offspring overweight was attenuated after adjusting for pre-pregnancy BMI. Several studies on offspring born to a GDM pregnancy have also observed similar results on overweight 19,20 . In contrast, considering that pre-pregnancy BMI may significantly affect GDM 21 , offspring who were born to a GDM pregnancy cannot be directly compared to mother who have a higher pre-pregnancy BMI, on account of different methodology. Unexpectedly, there was no difference in BMI between LGA and non-LGA offspring born to GDM pregnancy in a study of 6 to 7-year-old infants 11 .
With regard to childhood overweight or obesity, it is becoming increasingly clear that GDM can affect foetal metabolism and growth, which results in higher adiposity in the offspring 1 . The 'fetal programming theory'    www.nature.com/scientificreports www.nature.com/scientificreports/ emphasizes the significance of long term effect of suboptimal intrauterine exposure on offspring 22 . Hyperglycaemia in uterus can result in fetal overnutrition and elevated oxidative stress that induces pro-inflammatory response 23 , methylation modifications 24 , insulin secretion increasing 25 , which can lead to hypothalamus epigenetic and neurohormonal changes 26 , and high adiposity at birth 3 . Ultimately, these may lead to offspring overweight or obesity later in life 1 . A number of studies have investigated the association between GDM and offspring overweight or obesity [27][28][29] , but have inconsistent results. A Swedish study suggested birth weight only played a minimal role in the relationship between GDM and offspring overweight or obesity. Overall, maternal GDM may predispose offspring to overweight or obesity later in life via mechanisms beyond excessive fetal growth, as illustrated by birth weight.
The major strength of this population-based study is a large sample size. Moreover, detailed information was obtained on pre-pregnancy BMI, which was no adequately adjusted for in previous studies. However, this study also has several limitations. Firstly, we did not have BMI z-score data for every offspring from 1 to 6 years of age. Secondly, study participants were all from the Chinese population, and therefore future studies among other ethnicities are needed. Thirdly, data on risk factors for offspring overweight or obesity such as maternal smoking or alcohol consumption would have improved the study.
In conclusion, this is a large population-based retrospective study on Chinese mother-child pairs. Offspring of mothers with GDM and LGA had a higher BMI z-score and increased risk of overweight from 1 to 6 years of age. However, the later association was attenuated after adjusting for maternal pre-pregnancy BMI, suggesting maternal pre-pregnancy BMI may play a key role in offspring overweight for children born to mothers with GDM and LGA. Therefore, we should focus on maternal pre-pregnancy weight to reduce the prevalence of childhood overweight or obesity. Further researches needed to expand these results to other populations, and to identify the underlying biological mechanisms.

Study design.
We conducted a population-based retrospective study using the healthcare records data from Data sources of MBRX. All women in Xiamen are registered at their community health centres when they get pregnant, and were then referred to a secondary hospital or a tertiary hospital for healthcare from the 32nd gestational week till delivery. All children were given health examinations every year from birth (<3 days after birth) until the age of 6. Women and children were linked by individual record linkages to the Xiamen citizen health information system using the person-unique identification number assigned to each Xiamen citizen. Every child was also linked to his/her biological mother's maternal identification number.

Study population.
A total of 33,157 mother-child pair healthcare records were available. All women over the age of 18 performed a 75-g OGTT between 24 and 28 weeks of gestational in this study. Eligibility criteria were as follows: (1) an OGTT was conducted between 24 and 28 weeks of gestation; (2) gestational age at delivery ≥37 weeks, with no major neonatal malformations or fetal/neonatal death; and (3) the offspring was followed-up through 6 years of age. Exclusion criteria included: (1) missing a mother's weight or height information at pre-pregnancy; (2) medical history of diabetes (diagnosed before the index pregnancy); and (3) fasting glucose level ≥7.0 mmol/L before 12 gestational weeks, as this could indicate an under-diagnosed diabetes cases prior to pregnancy.
Maternal and offspring characteristics. Information from MBRX on maternal factors included age, education, weight in 12 weeks before of pregnancy, occupation, first visit date, numbers of pregnancy/infants, last menstrual period, expected delivery date, smoking habits, drunk status, medical history, family history of disease, hypertension history, pregnancy reactions, as well as labour status. Furthermore, GDM, gestational weight gain, gestational age at delivery, height, weight, blood pressure, fasting glucose, gynaecological examinations, ultrasonography, gestational diabetes screening results, other lab tests results, complications during pregnancy, and pregnancy outcomes were also included in the MBRX system.
MBRX also included information from newborns to preschool-aged children on date of birth, sex, gestational week of birth, weight, Apgar score, names of the child and his/her parents, family history of diseases, feeding modalities (exclusive breast feeding, mixed breast and formula feeding, and exclusive formula feeding) during the first 6 months, date of examination, weight, height, number of teeth, and blood pressure.
Offspring measurements and data transformations. During each health examination, each child's height and weight was measured by a trained clinician. Body weight was measured in kilograms using regularly calibrated electronic scales.
Children were classified into three groups according to GDM and gestational age status during pregnancy: 1) children born to m-nonGDM-AGA; 2) children born to m-GDM-AGA; and 3) children born to m-GDM-LGA. Body mass index (BMI) was calculated as weight (kg) / height 2 (m 2 ). BMI z-score for age was used to present the trajectory tracking of offspring BMI. We calculated sex-adjusted and age-adjusted z-score of childhood BMI referred to Chinese reference growth charts 30 . Childhood overweight or obesity were also defined by age-specific and sex-specific Chinese criteria 30 .

Large for gestational age.
LGA was defined as birth weight was above 90 percentile for gestational age, according to gestational age and gender-specific intergrowth-21 st curves 31 .

Statistical analysis.
Mean±SD was showed for continuous variables, and discontinuous variables were presented as n (%). To evaluate our hypothesis that GDM and LGA were associated with offspring BMI growth trajectories, we performed several analysis. Firstly, mean BMI z-score were visually compared in yearly time intervals between m-nonGDM-AGA, m-GDM-AGA, and m-GDM-LGA. Secondly, the growth trajectories of offspring BMI were established by the longitudinal analyses that fitted flexible and smooth curves with a random-effects model 32 . The square model was used in BMI z-score points. The values of children from the GDM and LGA group and from children in the non-GDM and LGA group were modelled as shown below: Y = Intercept + β 0ij + β 1ij (age) + β 2ij (age) 2 . In addition, considering the previous studies and clinical relevance 33 , we assessed the joint effect of GDM and pre-pregnancy BMI (a major risk factor for offspring overweight or obesity), using a model with a combination of GDM (yes, no) and BMI categories 34 (BMI < 23.9 kg/m 2 , 24-27.9 kg/m 2 , or ≥28 kg/m 2 ) at 6 years of age. Thirdly, a logistic regression model was used to access the significance of offspring overweight between m-GDM-AGA and m-GDM-LGA groups. Three multivariable-adjusted models were included in this study. Model 1 adjusted for offspring sex, maternal age, education, and infant feeding; model 2 adjusted for the variables in model 1 plus maternal gestational weight gain; model 3 adjusted for the variables in model 2 plus maternal pre-pregnancy BMI. Statistical significance was two-tailed with a P-value < 0.05. All calculations were carried out using SAS 9.4 (SAS Institute Inc, Cary, North Carolina, USA).

informed consent from participants
Informed consent of this retrospective study was waived by the ethics committee of the First Affiliated Hospital of Xiamen University (KYH2018-007).

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.