Impact of Early Life Factors on Metabolic Phenotypes of Obesity in Preschool Children

The prevalence of obesity in children has been markedly growing worldwide and reaching an epidemic level, with an estimated 5.7% or 38.9 million children aged <5 years around the world affected by overweight in 2020, which was a rapid increase of nearly 6 million since 2000 (1). It is a major clinical and public health concern. Obesity alters the state of metabolism and physiology, leading to dyslipidemia, insulin resistance, and inflammation, and exacerbates metabolic and cardiovascular comorbidities, such as type 2 diabetes, fatty liver disease, and cardiovascular diseases (2).

Although compared with healthy-weight individuals, obese children are more likely to develop metabolic disturbances and cardiovascular abnormalities, but not all children with obesity possess equivalent health risks (3). Individuals in the same BMI category have been reported to have substantial heterogeneity of metabolic prognoses, such as insulin sensitivity, blood pressure, glucose tolerance, and lipid and hormonal profiles (4). A subset of obese children seems to be protected against obesity-related metabolic disorders and is considered as the metabolically healthy obese (MHO) group (5), contrary to the group referred to as having metabolically unhealthy (one or more metabolic risk factors) obesity (MUO) condition (6). On the other hand, besides the individuals who are metabolically healthy with normal weight (MHNW), those who are not obese shoulder increased risks of metabolic abnormalities, such as higher susceptibilities to insulin resistance or hypertension, and they are reported as the metabolically unhealthy with normal weight (MUNW) (7).

As the Developmental Origins of Health and Disease (DOHaD) hypothesis suggests, it is well recognized that events occurring in the earliest stages of life, even before birth, may influence the occurrence of diabetes, cardiovascular disease, and other diseases (8). The first 1,000 days of life—the period from conception to the child’s second birthday—are the most crucial, considering the irreversible or only partly reversible impact on development and function in this period (9,10). Albeit a number of factors and conditions in children have been suggested to impact their obesity and metabolic disorders, respectively, whether the early life factors are linked to the dissociation of obesity into metabolic phenotypes needs to be elucidated. Additionally, we found no studies that estimated the relationship between BMI growth trajectories in the specific time frame of early childhood and metabolic phenotypes of obesity in preschool children. We hypothesized that early life exposures are associated with childhood metabolic phenotypes of obesity. Better understanding of which factors contribute to the particular metabolic phenotypes of obesity may provide clues for the early prevention of obesity and metabolic disorders in children.

Therefore, leveraging the ambidirectional cohort study, we aimed to investigate the impact of factors during the first 1,000 days, including the trajectories of BMI growth, on metabolic phenotypes of obesity (i.e., MHNW, MUNW, MHO, and MUO) in preschool children.

This report followed the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) reporting guidelines for cohort studies (11).

The ambidirectional cohort study was undertaken in Tianjin municipality, China, from 2017 to 2020. The study recruited 3,822 children with measurements of body mass and composition in their 1st year of kindergarten (age 3 years) during 2017–2018 from 42 kindergartens, selected by the stratified cluster sampling method, and followed up at their 2nd and 3rd years. Data from before the children were recruited in the cohort were extracted from maternal and child health (MCH) records, which covered information on routine health care visits from mothers’ pregnancy until children’s 6th year, and were also collected by questionnaires (12–14).

In this study, children’s exclusion criteria include 1) inability to give informed consent; 2) any conditions or chronic diseases or use of any drug known to affect growth and development; 3) acute illnesses that prohibit children from participating in the physical examination; 4) multiple births; or 5) unobtainable or no collected blood samples.

The study was conducted in accordance with the Declaration of Helsinki and approved by the Tianjin Women’s and Children’s Health Center Institutional Review Board (BGI-IRB 17116-201711). Written informed consent to participate in this study was provided by the participants’ parents.

We extracted exposures in early life from MCH records, including maternal age, maternal education, maternal height, prepregnancy weight, weight before delivery, gravidity, parity, delivery mode, gestational age, child sex, birth weight, and breastfeeding duration. Prepregnancy weight was self-reported at the mother’s first health care visit during the 1st trimester of pregnancy. Prepregnancy BMI was calculated as prepregnancy weight in kilograms divided by square of height in meters. Gestational weight gain was calculated as weight before delivery minus prepregnancy weight. The data of weight and length or height of children measured using standardized procedures at their routine health care visits from age 1 month to 2 years were also extracted from MCH records. Exposures of maternal history of gestational diabetes mellitus, paternal education, annual family income, and family history of metabolic disorders were collected through an interviewer-administered questionnaire, which was reported by parents when children were in their first year of kindergarten.

Potential covariates were also assessed through the parent-reported questionnaire when children were in their first year of kindergarten. Children’s patterns of junk food consumption and physical activity, sleep duration, and sleep quality over the past 3 months, as well as the paternal height and current weight, were collected. Junk food consumption pattern was about the frequency of the consumption of fast food/snacks/carbonated beverages and divided into five categories: less than once a month, one to three times a month, one to three times a week, four to six times a week, and every day. Physical activity pattern in terms of duration of regular physical exercise per weekday was categorized as <1 h, 1–2 h, 2–3 h, 3–4 h, and >4 h of physical exercise per day. Sleep duration per weekday was grouped into <10 h and ≥10 h.

When children were in their first year of kindergarten (age 3 years), their height was measured without shoes to the nearest tenth of a centimeter by trained staff with wall-mounted stadiometers according to the standardized protocol (14,15). BMI was calculated as weight in kilograms divided by the square of height/length in meters. BMI z score (zBMI) was calculated from age and sex standardized to World Health Organization (WHO) measures using the igrowup package for SAS based on the WHO growth standards (16–18). Children’s weight status categories (normal weight, overweight, and obesity) were defined according to the WHO reference, which classified children at 0–5 years above +2 SD as overweight and above +3 as obese (17–19).

Children were required to be fasting, and venous blood samples were collected into evacuated tubes, with and without anticoagulant agents. The samples were centrifuged, aliquoted, and stored at −80°C until use. Serum triglycerides (TG), total cholesterol, LDL cholesterol (LDL-C), and HDL cholesterol (HDL-C), uric acid (UA), and fasting plasma glucose (FPG) were analyzed using the Hitachi 7060C Automatic Biochemistry Analysis System (Hitachi, Tokyo, Japan) according to the protocol (15). Metabolic risk factors were defined as high TG levels: TG concentration of ≥1.24 mmol/L, low HDL-C levels: HDL-C concentration of ≤1.03 mmol/L, impaired fasting glucose: FPG level of ≥5.6 mmol/L; and high UA levels: UA concentration of >420 μmol/L in boys and >360 μmol/L in girls (15,20). According to children’s BMI (normal weight, overweight/obesity) and metabolic status (healthy without metabolic risk factors, unhealthy with one or more metabolic risk factors), they were classified into four metabolic phenotypes of obesity: MHNW (as the reference), MHO, MUNW and MUO (21).

In sensitivity analyses, we chose UA ≥5.5 mg/dL (327 μmol/L) to define the high UA levels, from which the probability of primary hypertension increases significantly (22); moreover, we used the cutoff points established by the World Obesity Federation to categorize childhood BMI status (23,24).

Mean (SD) or median (interquartile range) are used to present continuous variables, and frequencies (%) show categorical variables. Overall differences among groups of metabolic phenotypes of obesity were compared by ANOVA for means, the Kruskal-Wallis test for medians, and the χ² test for frequencies. When significant overall differences among the four groups were demonstrated, pairwise comparisons of groups were conducted, with Bonferroni corrections for multiple comparisons. The coefficients of variation of biochemical measurements were also calculated.

zBMI growth trajectories in children from 0 to 2 years of age were estimated by piecewise linear mixed models, modeling the trajectory with several linear splines and assessing the linear associations by different periods (25,26). In this study, the knot point was fit at 6 months, considering the approximate age in months when the slope of growth changed and the coincidence with the regular child health care visit (12,26). The growth rates for each growth period were estimated using the cumulative sum of estimated regression coefficients from all previous growth periods (26). Metabolic phenotypes of obesity were analyzed as the principal fixed effects to explore the associations of zBMI growth trajectories with metabolic phenotypes of obesity (MHNW as the reference group). Multivariate models were further conducted by adjusting covariates, including maternal age at delivery (<35, ≥35 years), maternal education (high school or less, college, above college), annual family income (<10,000, 10,000–20,000, ≥20,000 RMB), prepregnancy BMI (<18.5, 18.5–23, 23–27.5, ≥27.5 kg/m²) (27), gestational weight gain (inadequate, appropriate, excessive) (28), gravidity (one, two or more), maternal history of gestational diabetes mellitus (without, with), gestational age (preterm, full-term), delivery mode (vaginal delivery, cesarean delivery), paternal education (high school or less, college, above college), paternal current BMI (<23, 23–27.5, ≥27.5 kg/m²) (27), family history of metabolic disorders (without, with), child sex (male, female), birthweight (<2,500, 2,500–3,999, ≥4,000 g), breastfeeding duration (<6, ≥6 months), junk food consumption frequency (less than once a month, one to three times a month, one to three times a week, four to six times a week, every day), physical activity duration (<1 h a day, 1–2 h a day, 2–3 h a day, 3–4 h a day, >4 h a day), sleep duration (<10 h, ≥10 h), and sleep quality (good, medium, poor). Unadjusted and adjusted associations between early life risk factors, other than zBMI growth rate, and metabolic phenotypes of obesity were assessed by logistic regression models, and the same covariates were in adjustments as in the piecewise linear mixed models. Missing values of a covariate were coded as a category in multivariate models.

A two-tailed P value was considered significant at <0.05. Statistical analyses were performed in Stata 15.0, SAS 9.4 (SAS Institute, Cary, NC), and R 4.0 (R Foundation for Statistical Computing, Vienna, Austria) software programs.

Among the 3,822 children recruited, 252 (6.6%) were excluded due to being part of multiple births (n = 74) or missing data regarding the numbers of fetuses (n = 178), and 555 (14.5%) were further excluded because blood samples were unobtainable. Finally, 3,015 children remained in the analyses. The coefficients of variation for TG, HDL-C, FPG, and UA concentrations were 40.74%, 19.03%, 8.67%, and 22.92%, respectively. In the primary analyses, 2,211 (73.3%), 688 (22.8%), 75 (2.5%), and 41 (1.4%) children were classified as MHNW, MUNW, MHO, and MUO, respectively (Supplementary Fig. 1). In sensitivity analyses when UA ≥5.5 mg/dL (327 μmol/L) was used to define the high UA levels, 2,118 (70.2%), 781 (25.9%), 69 (2.3%), and 47 (1.6%) children were classified as MHNW, MUNW, MHO, and MUO, respectively; when the cutoff points established by World Obesity Federation were used to categorize childhood BMI status, 2,145 (71.1%), 659 (21.9%), 141 (4.7%), and 70 (2.3%) children were classified as MHNW, MUNW, MHO, and MUO, respectively. Maternal and child characteristics in early life according to metabolic phenotypes of obesity are shown in Table 1. The overall tests show significant differences among groups of metabolic phenotypes of obesity in maternal prepregnancy BMI, gestational weight gain, gravidity, and children’s birth weight (P < 0.05); in contrast, the other characteristics were comparable for the four groups. According to pairwise comparisons, mothers in the MHO and MUO groups had higher prepregnancy BMI and higher proportions with excessive gestational weight gain compared with those in the MHNW and MUNW groups, the proportion of mothers with gravidity of two or more in the MHO group was higher than in the MHNW group, and the proportions of children with birth weight ≥4,000 g were higher in the MHO and MUO groups than in the MUNW group (P < 0.05 after Bonferroni corrections).

Children’s growth rates of zBMI were diverse in groups of metabolic phenotypes of obesity (Fig. 1). Table 2 shows the differences in growth rates of zBMI by metabolic phenotypes of obesity in each growth period (MHNW group as the reference). After covariates adjustments, the results remained similar as in the crude analyses, with the accelerated zBMI growth rate between 6 and 24 months positively associated with MHO (β = 0.022; 95% CI 0.009, 0.036 SD units/month) and MUO (β = 0.037; 95% CI 0.018, 0.056 SD units/month).

Unadjusted associations between other early life risk factors and metabolic phenotypes of obesity (MHNW as the reference group) are presented in Supplementary Table 1. After adjustments of covariates, maternal overweight (odds ratio [OR] 3.16; 95% CI 1.55, 6.42) and obesity (OR 8.14; 95% CI 3.73, 17.76) before pregnancy and macrosomia (OR 2.47; 95% CI 1.32, 4.59) significantly increased the odds of being MHO, and maternal obesity before pregnancy (OR 6.35; 95% CI 2.17, 18.52) increased the risk of MUO (Table 3).

In sensitivity analyses, although different classification criteria influenced the effect sizes, the associations of interest remained similar. Maternal overweight and obesity, macrosomia, and accelerated zBMI growth rate between 6 and 24 months were still linked to MHO, and maternal obesity and accelerated zBMI growth rate between 6 and 24 months were associated with MUO (P < 0.05, Supplementary Tables 2–5).

This study comprehensively examines the impacts of multiple factors during the first 1,000 days on different metabolic phenotypes of obesity in preschool children. Using different classification criteria, we found that the early life risk factors for MHO included maternal overweight and obesity before pregnancy and macrosomia, and for MUO included maternal obesity before pregnancy; additionally, accelerated zBMI growth rate between 6 and 24 months was significantly correlated not only to MHO but also to MUO.

The increasing rate of pediatric obesity highlights the importance of distinguishing phenotypes and considering different metabolic conditions to better inform intervention options (29). Moreover, life course perspective and prenatal and childhood growth data are paramount (30). Few researchers have studied the impacts of BMI trajectory in specific windows during early childhood. In our study, we found accelerated zBMI growth rate between 6 and 24 months increased the risks of MHO and MUO, but not MUNW; the adjusted mean difference of growth rates was higher between the groups of MUO and MHNW compared with the difference between MHO and MHNW. We also found that accelerated zBMI growth rate between 0 and 6 months was connected with MHO in sensitivity analyses. Previously, MHO children were reported to have less risk of adverse metabolic consequences than those with the MUO phenotype; additionally, MHO has been considered a transient state, and the individuals will convert to MUO in all probability (31,32). We speculate that accelerated BMI growth of children in the earliest stage leads to overweight and obesity rather than metabolic morbidities, considering the capacity to store excess fat mass, so the risk of MHO rather than MUNW might increase. Continually accelerated BMI growth in the later stage over time would increase the risk of negative metabolic effects and contribute not only to MHO but also to MUO, and the greater extent of accelerated BMI growth was more likely to result in MUO. Our results were in accordance with prior studies that reported a longer duration of obesity in childhood was associated with metabolic risk later in life (33). Our results also manifested that even a relatively slight accelerated zBMI growth rate would increase the risks of MHO and MUO and the effects should be tracked in later childhood.

Our study supports that maternal overweight and obesity before pregnancy increased the risk of MHO and that maternal obesity is a risk factor for MUO. It was reported that maternal overweight and obesity before pregnancy impacted the offspring’s BMI growth pattern and increased the risk of childhood obesity (34). Our results also indicate the effect of maternal overweight and obesity on child obesity, and that maternal obesity rather than overweight increases children’s risk of metabolic disorders. The association of birth weight with metabolic and cardiovascular comorbidities was thought to reflect fetal nutrition’s effect on prenatal programming, which affects metabolism in the long-term (35). Our study demonstrates that macrosomia increased the risk of MHO, following a study that reported that high birth weight possibly programmed insulin sensitivity and adipose tissue metabolism and contributed to MHO (36). In our research, there were also some factors with significant effect sizes under a specific classification standard. For example, breastfeeding duration ≥6 months had a protective effect on MUO phenotype in the primary analyses; in the sensitivity analyses, gravidity of two or more increased the risk of MHO, and inadequate gestational weight gain was a protective factor for MHO. Nevertheless, most of these factors were borderline significant, which needs to be explored in future research.

Our study has several strengths. First, we comprehensively investigated a wide variety of factors during the first 1,000 days for metabolic phenotypes of obesity, which provided additional impetus for early interventions for children’s obesity and metabolic morbidities. Second, prior studies for metabolic phenotypes of obesity have mostly been limited to adolescents and adults. Given the dramatically increased prevalence of overweight and obesity among preschool children (1), it is essential to focus on them. Third, the nature of the cohort study and the exposures extracted from MCH records allowed our study to elucidate causal relationship and insusceptibility to reporting bias, and we used different classification criteria and adjusted lifestyle factors to reduce residual bias and transient results.

Some limitations need to be acknowledged. First, to date, there is no consensus on the definition of metabolic health (29,37), and the study did not measure children’s blood pressure. Therefore, we used UA, an independent risk factor for cardiovascular diseases and strongly associated with blood pressure (6,38), as one of the markers for metabolic health classification; when defining the high UA levels, in addition to using the criterion for hyperuricemia definition (20), we also used UA ≥5.5 mg/dL (327 μmol/L), a cutoff point from which the probability of primary hypertension significantly increases (22), and we found the associations of interest remained similar.

Second, when collecting data on breastfeeding duration, we did not distinguish between partial and exclusive breastfeeding, two ways with different protein intake (12,39), and we did not collect information on the history of maternal smoking, formula feeding, and the time of introducing complementary foods, which might lead to a residual bias.

The validity and reliability of the questionnaire in this study needs to be further checked. We adjusted the duration of physical activity, and it would be better to use the PREFIT (Assessing levels FITness in PREschoolers) fitness test battery to access physical fitness in future follow-ups, given that fitness, especially cardiorespiratory fitness, is a good marker of metabolic phenotypes (21,40). The impact of physical activity during the first 1,000 days of life on metabolic phenotypes of obesity warrants exploration in future research. Although groups of MHO and MUO, in which children were overweight and obese, had small sample sizes, some early life factors showed significant associations with the two groups, meaning these factors significantly influenced the children’s overall health. Still, our results had wide 95% CIs and limited power. We did not estimate children’s metabolic health in the follow-ups, during which the MHO phenotype might translate into MUO (31,32). Thus, future studies with larger sample sizes and multiple follow-up visits are warranted to examine the effect of early life factors on long-term metabolic health. Our cohort recruited children from kindergarten, who might not represent all preschool children, likely resulting in a selection bias. Whether significantly accelerated growth rates of zBMI, of which 1 SD unit is equivalent to 1 major percentile band (25 percentiles) on the WHO growth curves (17,26), appeared clinically meaningful needs to be researched, and calculating the specific value of corresponding BMI change according to sex and age for clinical practice in the future would be beneficial.

The increasing prevalence of pediatric obesity and possible metabolic changes highlight the need to study whether factors during the first 1,000 days impact metabolic phenotypes of obesity in order to intervene in childhood obesity and metabolic abnormalities in advance. We found that the early life risk factors for MHO included maternal overweight and obesity before pregnancy, macrosomia, and accelerated zBMI growth rate between 6 and 24 months, and for MUO, included maternal obesity before pregnancy and accelerated zBMI growth rate between 6 and 24 months. These results underscore the importance of early screening of metabolic phenotypes of obesity among children with the related risk factors and preventing the onset of obesity and metabolic morbidities in children.

Acknowledgments. The authors would like to give special thanks to all the staff and participants of the study.

Funding. This research was supported by The National Key Research and Development Program of China (2016YFC1300100, 2016YFC0900600), the National Natural Science Foundation of China (72104148), the Chinese Academy of Medical Sciences (CAMS) Initiative for Innovative Medicine (2016-I2M-1-008), the Public Service Development and Reform Pilot Project of Beijing Medical Research Institute (BMR2019-11), the Special Fund of the Pediatric Medical Coordinated Development Center of Beijing Hospitals Authority (XTZD20180402), Beijing Hospitals Authority Youth Programme (QML20211301), and Beijing Municipal Administration of Hospitals Incubating Program (Px2022052).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Z.L. conducted the analysis and drafted the manuscript. Z.L. and F.C. designed the current study. F.C., J.W., Y.C., T.Z., G.L., and X.X. critically revised and reviewed the manuscript. J.W and G.L. designed the larger study. J.W. and G.L. conducted the investigation. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work. F.C. and G.L. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.