Associations between Neonatal Serum Bilirubin and Childhood Obesity in Term Infants

Inverse correlations between serum bilirubin level and obesity had been reported in adults. We aimed to investigate the associations between neonatal hyperbilirubinemia and childhood obesity. Data was obtained from the U.S. Collaborative Perinatal Project (CPP), a multicenter study from 1959 to 1976. Data of serum bilirubin in term newborns were used to observe the association with obesity at age of 7 years. Logistic regression models were performed to calculate adjusted odds ratios (aORs) for obesity. For children from the same mother sharing similar factors, Generalized Estimating Equation (GEE) model was used to correct for intracluster correlation. Relative to newborns with total serum bilirubin (TSB) < 3 mg/dl, there are lower risks for obesity in those with 3 mg/dl ≤ TSB < 6 mg/dl (aOR 0.91; 95%CI 0.81, 1.02), 6 mg/dl ≤ TSB < 9 mg/dl (aOR 0.88; 95%CI 0.78, 0.99), 9 mg/dl ≤ TSB<13 mg/dl (aOR 0.83; 95%CI 0.71, 0.98). By stratifying for subtypes of bilirubin, the inverse correlations only existed in exposure to unconjugated bilirubin. By using the GEE model correcting for intracluster correlations, the results are consistent. In summary, exposure to bilirubin up to 13 mg/dl is inversely associated with obesity at the age of 7 years in term infants.

previous studies on this topic [13][14][15][16][17][18][19] , our previous study demonstrated that neonatal bilirubin concentrations positively associated with childhood obesity among preterm infants 21 . So the question rise whether term-newborns with hyperbilirubinemia has the same associations with obesity in later life? Basing the U.S. Collaborative Perinatal Project (CPP) data, the purpose of this study is to exposure the potential associations between serum bilirubin in neonatal period and childhood obesity among term-newborns.

Methods
Study population. The data using in the current study obtained from the US. CPP, a multicenter birth cohort study from 1959 to 1966 contained 56 990 pregnancies within 46 021 women at 12 centers. By using standardized procedures, the trained observer measured and recorded anthropometric parameters of offspring (including weight, height, and etc) at each follow-up visit during 7 years follow-up time. At age of 7 years, the fullscale score (IQ) of Wechsler Intelligence Scale for Children (WISC) was administered 22 . The standard questionnaire in the cohort study recorded general characteristics of mothers and infants, including socioeconomic status, infant's races, gestational ages, delivery methods and maternal ages, smoking, educational levels, prepregnancy body mass index (BMI), marital status, etc. The data have been publicly available at the U.S. National Archives (www.archives.gov/), with detailed description provided in other study 23 . The study protocol was approval by the Institutional Review Board of the Affiliated Hospital, Southwest Medical University. Due to the publicly available de-identified data, this study without informed consent.
In this analysis, after stillbirths or terminations (n = 2195) and multiple births (n = 1148) were excluded, a total of 53 647 single births were enrolled. Next, we excluded newborns with gestational age at birth <37 weeks (n = 8809), ≥43 weeks (n = 2610) or unknown (n = 41). Infants with unknown serum bilirubin levels during neonatal period (n = 2719) or BMI at age 7 (n = 8535) have been also excluded. After excluding the infants with chromosomal congenital and severe structural anomalies, the final study subjects were left to be 28 489 term infants (Fig. 1).

Bilirubin measurements.
Except for direct spectrophotometry used in one center, the diazo method were used to measure bilirubin in the other 11 centers. Among the CPP laboratories, the coefficient of variation of standard specimens was about 10%; the intralaboratory coefficient of variation was about 2% 24 . This level of reliability is the same as the results in two more recent surveys of bilirubin measurements.  In the study protocol of the CPP, total serum bilirubin (TSB) levels were measured between 36 and 60 hours for each infant, as close to 48 hours as possible. If the value of TSB was exceeded 10 mg/dl (171μmol/L), the next test was performed 24 hours later. If the second TSB also higher than 10 mg/dl, a third measurement was obtained at 4 to 5 days of age. Additional bilirubin measurements were obtained at the decisions of physicians at each study site. We primarily used the maximum TSB as exposure factor and divided the babies into five groups based on the bilirubin concentrations: TSB < 3 mg/dl, 3 mg/dl ≤ TSB < 6 mg/dl, 6 mg/dl ≤ TSB < 9 mg/dl, 9 mg/dl ≤ TSB < 13 mg/dl, and TSB ≥ 13 mg/dl (as pathological jaundice 25 ). In addition, the values of conjugated bilirubin (CB) could be obtained from 17150 infants in the CPP. Thus, we also used CB and unconjugated bilirubin (UCB) as independent factors to explore the correlations between the different type of bilirubin and pediatric obesity. outcomes and confounders. Based on the childhood growth standards of WHO, obesity was referred as BMI for age and sex greater than 2 standard deviations above the WHO Growth ref. 26 . The validity and the reliability of the Wechsler Intelligence Scales for children have been confirmed elsewhere 27 . In this study, low IQ was referred to IQ < 70 at 7 years, and was considered as impaired neurodevelopment.

Statistics analysis.
The differences of the maternal and infantile demographic characteristics among different groups were compared by Cochran-Mantel-Haenszel Chi-square. Relative to the babies with neonatal TSB < 3 mg/dl, the crude odds ratios (ORs) for obesity and low IQ at 7 years old in each other group were calculated by the univariate logistic regression model (model 1). The adjusted ORs were calculated by a multivariable logistic regression model (model 2) to adjust the potential confounders. Because 3515 women, with 7619 pregnancies, contributed more than one birth, and the offspring of the same mother sharing the similar genetic and household factors, these intracluster correlations was corrected by using the Generalized Estimating Equation (GEE) model (model 3). All the statistical analyses were performed in SAS version 9.2 (SAS Institute, Cary, North Carolina).

ethical approval and informed consent. The study protocol was approval by the Institutional Review
Board of the Affiliated Hospital, Southwest Medical University. Use of publicly available de-identified data is exempt from informed consent. Table 1 showed the maternal and infantile baseline characteristics in different groups. Significant differences were observed in birthweight, gestational age, feeding methods, maternal smoking among term babies with different levels of TSB concentrations.
In order to consider the influence of severe cholestatic and hemolytic pathology, in second step, we excluded infants with positive direct Coombs' test, need exchange transfusion, CB ≥ 2 mg/dl, and definite hepatic disease at birth. Hence, the level serum TSB and UCB negatively correlated with body mass index at age 7 years (β = −0.18 and −0.24, respectively, P all <0.0001). To define possible negative influences of TSB concentrations on cognitive development we correlated TSB to IQ at age 7 years not finding any association (Table 4).

comments
To our best knowledge, this is the first study reporting that, in term newborn babies, exposure to high concentrations of bilirubin is inversely associated with obesity in children at age 7 years, without increasing the risk of low IQ.
It was well-known that bilirubin has toxic effects on developing neuronal tissues, and high serum bilirubin concentrations are associated with neurological dysfunction in newborn babies 6,7 . However, recent investigations reported on mild hyperbilirubinaemia having positive health effects by exerting anti-oxidant, anti-inflammatory activities 8-10 , increasing insulin sensitivity 10,18 and regulating lipid synthesis 18,28 . Obesity, increasing all over the (2019) 9:14575 | https://doi.org/10.1038/s41598-019-51043-w www.nature.com/scientificreports www.nature.com/scientificreports/ world, is a low-grade inflammatory disease 11,12 . Population based studies have documented that serum bilirubin levels are inversely associated with abdominal obesity in adults and adolescent 22,27,29,30 . However, these cross-sectional studies regarding obesity did not find associations to neonatal bilirubin levels. By using the prospective data from the U.S. CPP, we could document that exposure to bilirubin in the neonatal period was inversely associated with obesity in later life. The previous studies reported that the elevated bilirubin could   www.nature.com/scientificreports www.nature.com/scientificreports/ increase insulin sensitivity 10,18 , reduce lipid synthesis 18,28 , and regulate the serum level of leptin, PPAR α and adiponectin 18 , then reduce the body weight. Thus, we refer that these factors may involve in the potential mechanism of the inverse correlation between bilirubin concentrations and childhood obesity.
Our results showed that the two subtypes of bilirubin have opposite associations to childhood obesity in term babies. These findings are consistent with one previous study reporting on the inverse relation of direct bilirubin/ CB, rather than indirect bilirubin/UCB, to metabolic syndrome 15 . Furthermore, regarding our results, it seems that exposure to UCB, even on physiological levels (<13 mg/dl), is inversely associated with childhood obesity in our study subjects, while serum CB seems to have a positive dose-response relationships with childhood obesity. As the underlying causes of jaundice in neonates may also be non-physiological, we excluded infants presenting pathological factors, still confirming the negative correlations between serum bilirubin levels in newborns and body mass index at age 7 years.
For the anti-oxidative and anti-inflammatory properties, bilirubin has been suggested as possible target to intervene cardio-metabolic disorders in future 31,32 . Regarding the concentrations of bilirubin presented in previous population based studies they refer to the normal ranges. In an experimental study infusing bilirubin directly to mice, Vera et al. found that moderate hyperbilirubinemia (1 mg/dl) resulted in a decrease in vascular oxidative stress 33 . Another experimental study using a Gunn rat model, inducing severe hyperbilirubinemia, showed that bilirubin was protecting for cardiac reperfusion injury. The bilirubin concentrations in this model were much higher than those observed in population based studies 34 . However, there was no evidence on this topic from natural exposure to high bilirubin level. The physiological exposure to bilirubin in the neonate is much higher than in any other period of life in humans. Using the data set of the CPP study period (1959)(1960)(1961)(1962)(1963)(1964)(1965)(1966), when phototherapy was not used routinely for the treatment of neonatal jaundice, made it particularly suitable to study the exposure-effect relationships of untreated neonatal serum bilirubin and childhood obesity. Thus, our findings emphasizes the existence of a negative dose-response correlation between physiological jaundice in term babies and later childhood obesity, what may generate new hypothesis for interventional studies in the future to intervene obesity or obesity related diseases.
Regarding the role of lifelong exposure to increased serum bilirubin in Gilbert syndrome patients 31,32,35 , one may speculate on protective effects of hyperbilirubinemia. However, in an experimental study in mice beneficial health effects of high bilirubin levels could not be determined 33 . In contrast our results, indicating the beneficial effect of bilirubin on childhood obesity in the neonatal period could be interpreted as a potentially epigenetic effect determined only in early life.
Our study has several limitations as it is a historic study performed 40 years ago. For the more active intervention to neonatal jaundice in current guideline, the relevance of our finding to the contemporary populations may be questioned. However, for the same reasons, the same study is unlikely to be duplicated nowadays. Thus, we believe our findings have useful implications in the future to intervene obesity.
In conclusion, in term newborn babies exposure to physiological level of bilirubin is inversely associated with childhood obesity in later life without increasing the risk of low IQ. Our findings may shed some light on investigations and interventions regarding the role of bilirubin in the prevention and treatment of obesity or obesity related diseases.

Data Availability
The CPP data are publicly available through the U.S. National Archives (www.archives.gov/).