Severity of nausea and vomiting in pregnancy and early childhood neurobehavioural outcomes: The Growing Up in Singapore Towards Healthy Outcomes study

Abstract Background Nausea and vomiting of pregnancy (NVP) affects 50 to 80 per cent of women. The existing literature has examined NVP from the perspective of the mother, and relatively less is known about offspring outcomes. Objectives To study the relationships of NVP with social‐emotional, behavioural, and cognitive outcomes of the offspring in a multi‐ethnic Asian cohort. Methods In the Growing Up in Singapore Towards Healthy Outcomes prospective mother‐offspring cohort study, mothers responded to a structured NVP questionnaire at 26‐28 weeks’ gestation (n = 1172) and participants with severe NVP were confirmed using medical records. Children underwent multiple neurodevelopmental assessments throughout childhood. We conducted multivariable regressions with post‐estimation predictive margins to understand the associations of NVP with offspring neurobehavioural outcomes, which included 1‐year Infant‐Toddler Social and Emotional Assessment, 1.5‐year Quantitative Checklist for Autism in Toddlers, 2‐year Bayley Scales of Infant and Toddler Development, 2‐ and 4‐year Child Behavior Checklist, and 4.5‐year Kaufman Brief Intelligence Test. Analyses were adjusted for household income, birth variables, maternal mental health, and other relevant medical variables. Cohen's d effect sizes were calculated using standardised mean differences (μd). Results Mothers were categorised into no (n = 296, 25.3%), mild‐moderate (n = 686, 58.5%), and severe NVP (n = 190, 16.2%), of whom 67 (5.7%) required admission. Compared to children of mothers who had no or mild‐moderate NVP, children with exposure to severe NVP exhibited more externalising behaviours (μd 2.0, 95% CI 0.3, 3.6; Cohen's d = 0.33) and social communication difficulties before 2 years (μd 4.1, 95% Cl 0.1, 8.0; Cohen's d = 0.38), both externalising (μd 1.5, 95% CI 0.4, 2.6; Cohen's d = 0.43) and internalising behaviours at 2 years (μd 1.2, 95% CI 0.1, 2.2; Cohen's d = 0.35), and only internalising behaviours after 2 years (μd 1.1, 95% CI 0.4, 2.0; Cohen's d = 0.37). Conclusions Severe NVP is highly prevalent in this Asian cohort and may be adversely associated with multiple offspring neurobehavioural outcomes.


| INTRODUC TI ON
Nausea and vomiting of pregnancy (NVP) is characterised by a spectrum of severity and affects 50 to 80 per cent of women during pregnancy. [1][2][3] Symptoms typically peak during the first trimester and remit by the 20th week of gestation. [4][5][6] The pathophysiology of NVP remains incompletely understood but is likely multifactorial with hormonal, environmental, psychological, evolutionary, and genetic etiologies. [4][5][6] Hyperemesis gravidarum represents the most morbid clinical presentation across the spectrum of NVP and has been variably defined as persistent nausea and vomiting leading to weight loss in excess of 5% of pre-pregnancy weight 2,7 and/ or warranting hospital admission based on the Fairweather criteria.
The Fairweather criteria define hyperemesis gravidarum as "vomiting occurring in pregnancy before the 20th week of gestation, not associated with coincidental conditions as appendicitis, pyelitis, etc, and of such severity as to require the patient's admission to hospital." 8 The reported incidence of hyperemesis gravidarum is 0.3 to 1.0 per cent in Caucasian populations 9,10 and up to 3.0 to 10.8 per cent in Asian populations, 4,6 hinting at potentially distinct biological mechanisms governing NVP in different races, ethnic aggregation, or genetic susceptibility. 11 Hyperemesis gravidarum and its impact on maternal physical and mental well-being have been extensively studied. Hitherto, much of the existing literature on the psychosocial burden of NVP has been examined from the perspective of the mother, [12][13][14][15][16] and relatively less is known of pregnancy and offspring outcomes. Whereas milder symptoms have been considered inconsequential or even protective of adverse outcomes such as miscarriage, [17][18][19][20] accumulating evidence has suggested a link between hyperemesis gravidarum and placental dysfunction disorders, including preeclampsia, abruption, stillbirth, small for gestational age, and preterm birth. 13,21 Considering that weight loss, poor diet quality, nutritional deficiencies, and ketonuria are complications after protracted vomiting, hyperemesis gravidarum imposes physiological and metabolic stresses which 0.1, 2.2; Cohen's d = 0. 35), and only internalising behaviours after 2 years (μ d 1.1, 95% CI 0.4, 2.0; Cohen's d = 0.37).
Conclusions: Severe NVP is highly prevalent in this Asian cohort and may be adversely associated with multiple offspring neurobehavioural outcomes.

K E Y W O R D S
behaviours, child development, hyperemesis gravidarum, Nausea in pregnancy, vomiting in pregnancy

Study question
What is the relationship of nausea and vomiting in pregnancy (NVP) with the social-emotional, behavioural, and cognitive development of the offspring?

What's already known
The existing literature demonstrates the psychosocial burden of NVP on pregnant mothers. A few studies have shown the contribution of NVP to emotional behavioural difficulties in school-age and adult offspring.
Neurobehavioural outcomes in early childhood have not been thoroughly explored.

What this study adds
Severe NVP in this multi-ethnic Asian cohort correlates with a variety of neurobehavioural impairments in the offspring. The behavioural phenotype shifts from predominantly externalising symptoms in the first 2 years to internalising symptoms after 2 years of life. No associations are found between NVP and cognitive intelligence at 4.5 years. recapitulate a state of starvation or malnutrition in utero, which in turn has been postulated to account for the developmental origins of many adult diseases. 22 Prenatal exposure to hyperemesis gravidarum has been reportedly associated with mental health disorders in adult offspring, as well as a variety of attention, learning, and language problems in children. 1,20,[23][24][25][26][27] Two studies have reported differences in developmental outcomes only for protracted NVP that continued into the mid-late second trimester. 1,24 Other studies have also emerged recently that described the relationship between NVP and autism spectrum disorders. [28][29][30][31] Accordingly, this study was conducted in a prospective mother-offspring cohort to examine a wider range of early childhood outcomes, consisting of social-emotional, behavioural, and cognitive outcomes, after exposure to different severity of NVP in utero.

| Study population
The Growing Up in Singapore Towards Healthy Outcomes (GUSTO) cohort is a population-based, prospective longitudinal study of pregnancies and children with the aim of studying how antenatal and early childhood conditions influence the long-term health of women and children. Briefly, women aged 18 years or older across all socioeconomic backgrounds were recruited from KK Women's and Children's Hospital (KKH) and National University Hospital (NUH) in Singapore between 2009 and 2010 during their first trimester of pregnancy, as described previously. 32 GUSTO inclusion criteria required both biological parents to be one of the three major ethnicities in Singapore (ie Chinese, Malay or Indian) and are Singaporeans (ie not recent immigrants). Participants were followed up through and beyond delivery, and their children tested regularly over childhood.

| Exposures
Severity of vomiting in the first and second trimester was recorded in a structured interview-administered questionnaire at 26-28 weeks' gestation. Electronic medical records maintained by the hospitals were utilised to confirm the severity of vomiting, including prescription of anti-emetic medications and hospital admissions. Seventy-four women were admitted due to non-infectious nausea and vomiting and all of them had "hyperemesis gravidarum" as the discharge diagnosis. Congruence between mothers' reports and medical records for severe NVP was 90.5%. Sixty-seven participants with congruent records were included. As weight loss was not systematically obtained, we did not categorise pregnant mothers into hyperemesis gravidarum. Instead, the severity of nausea and vomiting was classified into none, mild-moderate (defined as nausea and/or vomiting occasionally), and severe (defined as regular vomiting with inability to retain meals).

| Outcomes
The outcomes analysed in this study were scores from the 1-year Infant-Toddler Social and Emotional Assessment (ITSEA; n = 542) 33

| Statistical analyses
Multivariable linear regression models were used to delineate the relationships between NVP and child outcomes, and analyses were adjusted for the following covariates: gestational age, birthweight (World Health Organization (WHO) gestational age and sex-specific Z-scores), maternal age at delivery, monthly household income,  45 Briefly, this EFA was fitted with the individual items of the mental health scales as manifest variables, and factors were retained if the eigenvalue from observed data was larger than eigenvalues from parallel analysis (1000 randomly generated correlation matrices). 45 An oblique rotation (bi-geomin) was used to permit correlation between subfactors, and estimates were obtained under maximum likelihood with robust standard errors. 45 Unadjusted mean scores and their 95% confidence intervals for the three categories of NVP were obtained using the post-estimation predictive margins command immediately after fitting regression models. To further facilitate interpretation regarding the clinical significance, we reported Cohen's d for effect sizes. Statistical analyses were completed in Stata version 16.0 (StataCorp). 46

| Sensitivity analyses
Since women with severe NVP were grouped together in the multivariable regression regardless of admission into the hospital, we conducted sensitivity analyses to test the uncertainty around whether there were differences in child outcomes between mothers with severe NVP who sought care in the hospital versus those who did not. We found that the 123 mothers with severe NVP who were not hospitalised accounted for our results in Tables 2 and 3, and not the 67 women with severe NVP who were hospitalised (Table S1).

| Missing data
Our missing data tabulation showed that 37% of children born to mothers with NVP data did not have data related to neurodevelopment, while only 15% had dropped out entirely from the GUSTO cohort at 4.5 years. This was because approximately 22% of the children in the cohort were not asked to attend the neurodevelopment visits, as a way to reduce time burden on the families. Those who attended the neurodevelopmental visits were randomly selected from the 1172 mothers, except for the 4-year visit, when children with more consistent attendance in prior visits were prioritised to ensure a higher sample of complete data. To assess for potential bias due to missing data, we compared the demographic characteristics and the NVP distribution of mothers whose children had missing or non-missing outcomes. We did not find any differences between the two groups, except for a small difference in ethnicity at the 1-year follow-up visit (Table S2).
Missing data were handled by multiple imputations (m = 50) on the independent and dependent variables using chained equations.
Before re-calculating the multivariable regression estimates from the imputed data, we deleted imputed data of the dependent variables and left only the observed data to minimise bias introduced by a misspecified imputation model for the outcomes. 47

| Ethics approval
The study protocol was approved by the SingHealth Centralized Institutional Review Board and National Healthcare Group Domain Specific Review Board, and all participants gave written informed consent prior to study enrolment.  (Table 1).

| Description of cohort
However, there were differences in child's sex, ethnicity, gestational age, and breast feeding across the three categories of vomiting severity (Table 1). Mothers who conceived sons were, as a group, less likely to report severe NVP than mothers who conceived daughters. In our multi-ethnic cohort, Indian mothers were more likely to report severe vomiting as compared with Chinese mothers, while Malay ethnicity did not appear to be associated with severity of vomiting. No inter-group differences were found in maternal mood from antenatal to postnatal period based on ethnicity of the women. Finally, mothers with severe NVP were less likely to report TA B L E 1 Characteristics of participants in the Growing Up Towards Healthy Outcomes (GUSTO) study and relative risks between mothers with severe vomiting and no vomiting breast feeding for at least 3 months as compared with mothers with no, mild, and moderate NVP, which was likely related to unmeasured confounders.

| Child outcomes after exposure to nausea and vomiting in pregnancy
In our multivariable regression models, we observed more pronounced externalising behaviours at 1 year and more social communication difficulties at 1.5-year in children born to mothers who reported severe vomiting as compared to those who reported no or milder NVP (Table 2). Dysregulated behaviours were also elevated at 1-year, along with negative emotionality (ie tendency to react to stressful situations with unpleasant emotions) and sleep disturbances in infants born to mothers with severe NVP ( Table 2). The models for data deleted listwise and for imputed data produced the same results.
It is important to note that regression models were consecutively run for all domains and subscales (over 30 in total) in the ITSEA and CBCL without selective searching for notable associations. We elected to display only domains with meaningful effect sizes based on inspection of 95% confidence intervals. The reader may therefore interpret all ITSEA and CBCL outcomes not shown in the tables as having 95% confidence intervals that crossed zero.
To afford a more comprehensive and longitudinal assessment, we used the Childhood Behavior Checklist (CBCL) at 2-and 4-year time points (Table 3). At age 2, we found more externalising and internalising problems in children with exposure to severe NVP as compared to those whose mothers reported no or mild-moderate NVP during pregnancy. The most prominent areas of difficulty were related to regulation of emotions and attention, for example emotional reactivity and attention problems. In addition, consistent with age 1, higher scores in sleep problems were reported. In the DSM-oriented scales, children exposed to severe NVP had higher symptoms of attention-deficit hyperactivity disorder and affective disorder.
At 4 years, children exposed to severe NVP exhibited mainly internalising behaviours such as anxiety and depressive symptoms. It is notable that both mother and father corroborated this finding in separate reports. In addition, when compared to children born to mothers without or with milder NVP, those born to mothers who experienced severe vomiting during their pregnancy had more impairments based on several DSM-oriented scales, including in affective and anxiety conditions (Table 3). We compared the models obtained from our imputed data with the observed data and found minimal changes in the estimates.
Severe vomiting was not associated with language and motor scores in 2-year-old children based on the Bayley-III scale (Table 4).
However, reductions of 5.4-and 5.1-points in the non-missing and imputed Bayley cognitive scores, respectively, were found, which corresponded to 1/3 of a standard deviation in this instrument.

| Principal findings
In this large Asian prospective mother-offspring cohort, children born to mothers with severe NVP during pregnancy showed worse emotional and behavioural functioning, as compared to children without exposure to NVP.

| Strengths of the study
To date, there has been no other study using a non-clinical cohort with such a wide range of outcomes in early childhood and in an Asian population, which is important considering potential biological and aetiological differences between women from different ethnic backgrounds. Our present study has important policy, clinical, and research implications, as our findings suggest that the in utero environment of fetuses exposed to severe NVP may set a trajectory of increased risk for poorer developmental outcomes. Other strengths include the prospective longitudinal study design, rigorous phenotyping of mother-child dyads, adjustments for multiple covariates, and the use of validated tests for assessing early childhood neurocognitive and behavioural profiles.

| Limitations of the data
Limitations of the study include self-reporting of vomiting severity, which is subjective and may be associated with recall bias.
However, medical record details have been used to ensure accuracy of the women's self-report with a congruence of 90.5% in this study.

| Interpretation
A novel finding from this study is that children born to mothers with severe NVP initially demonstrate mainly externalising symptoms in infancy and toddlerhood, while children present with more internalising symptoms by preschool age. Although the association between NVP and internalising symptoms in adult offspring has been consistently reported, 2,9,25 little is known about internalising symptom profile from 1 to 4 years of age. In this study, the association of NVP and internalising symptoms can be observed as early as 2 to 4 years of age.
Consistent with recent studies linking NVP and autism spectrum disorder, we have replicated similar results at 1.5-years. Instead of using an ASD diagnosis as the outcome, this study uses social communication skills from the Q-CHAT, which allows for a more granular understanding of the relationship between NVP and social communication as a continuum.
Severe NVP is highly prevalent in our cohort (16.2%), which has been suggested to be epidemiologically associated with Asian ethnicity. 4,6 Nevertheless, it is notable that rates have varied widely across studies. Possible explanations for between-study heterogeneity in prevalence estimates include differences in the criteria used to define hyperemesis gravidarum, insufficient sample sizes, self-reporting and recall bias, differences in the specific ethnic composition, and country-specific sociocultural attitudes, guidelines and accessibility to health care institutions. It is intriguing to note that in our sensitivity analyses (Table S1), mothers with NVP admitted into the hospital and received treatment have children with similar outcomes as those born to mothers with no, mild, or moderate NVP.
Consistent with two prior studies, poorer outcomes are identified in children whose mothers have not sought aggressive treatment for NVP. 26,27 As further described below, lack of treatment for NVP likely entrenches the condition of persistent ketonuria and starvation during pregnancy, while early rehydration or use of antiemetics may help to restore a normal in utero environment.
In this study, mothers with severe NVP also have infants born at earlier gestational age. This corroborates previously established risk factors and reinforces the need for awareness regarding the psychosocial burden of NVP. 10,[12][13][14][15][16]49 The exact mechanisms which drive NVP remain an area of investigation. Emerging evidence suggests a multifactorial pathogenesis involving an interaction between genetic, hormonal, and gastrointestinal factors. 50 Previously, NVP was attributed to hCG production; however, a review of 31 papers showed conflicting evidence. 51 There has been more support for the role of genetic factors in the pathogenesis of NVP. A genome-wide association study (GWAS) of European mothers identifies GDF15 (encoding growth/differentiation factor 15), which delays gastric emptying and contributes to nausea, as a risk factor. GFRAL, the receptor of GDF15, in the vomiting centre of the brainstem (ie postrema), which signals loss of appetite and taste aversion, is another risk factor. 52 The GWAS also implicates IGFBP7 (encoding insulin-like growth factor-binding protein 7), a placental protein and biomarker of cachexia, and PGR, a hormone receptor involved TA B L E 2 Severe NVP and its relationships with 1-year ITSEA subscales and 1.5-y Q-CHAT total score  53 To paint a more conclusive picture regarding the epigenetic alterations induced by NVP in the offspring, epigenome-wide association studies should be undertaken to identify differences in genome methylation between children with or without exposure to NVP.

TA B L E 3
Severe NVP and its relationship with 2-y mother-rated behaviours, 4-y father-reported behaviours, and 4-year mother-reported behaviours from CBCL  The CBCL questionnaire comprises numerous subscales; only domains with meaningful effect sizes based on inspection of 95% confidence intervals are displayed in this table. b Covariates included gestational age, birthweight Z-scores, maternal age, household monthly income, gestational diabetes, hypertensive disorders of pregnancy, history of smoking exposure and/or positive plasma cotinine, sex of child, maternal general mood factor from pregnancy to 3-month post-partum, and parity.
In contrast to the observation that hyperemesis gravidarum is associated with worse placental dysfunction 13,21 and negatively influences offspring development, 20,[23][24][25] there is growing acceptance of the idea that milder symptoms may be protective of pregnancy outcomes such as miscarriage and congenital malformation. 17,19,54 These ostensibly opposing effects on perinatal outcomes allude to the complex biology of NVP. Further studies are warranted to understand the possibility of NVP providing feto-protection while in utero and simultaneously setting an unfavourable development trajectory in the offspring after birth.

| CON CLUS IONS
Severe NVP is highly prevalent in this Asian cohort and is associated with a multitude of unfavourable neurobehavioural outcomes in the offspring.

ACK N OWLED G EM ENTS
We would like to thank Dimple Rajgor for helping with formatting and submission of this manuscript. We also acknowledge the GUSTO study group, which includes Allan Sheppard, Amutha

CO N FLI C T O F I NTE R E S T
S-Y Chan, KM Godfrey, and Y-S Chong are part of the Epigen Academic Consortium that has received funding for research studies outside of this submitted work. All other authors declare no conflicts of interest.