The Association Between Prenatal Infection and Adolescent Behavior: Investigating Multiple Prenatal, Perinatal, and Childhood Second Hits

Objective: Exposure to infections during pregnancy may be a potential risk factor for later psychopathology, but large-scale epidemiological studies investigating associations between prenatal infection and long-term offspring behavioral problems in the general population are scarce. In our study, we aimed to investigate the following: (1) the association between prenatal infection and adolescent behavior, (2) putative underlying pathways (mediation), and (3) “second hits” interacting with prenatal infection to increase the risk of adolescent behavior problems (moderation). Method: Our study was embedded in a prospective Dutch pregnancy cohort (Generation R; n = 2,213 mother–child dyads). We constructed a comprehensive prenatal infection score comprising common infections for each trimester of pregnancy. At age 13 to 16 years, we assessed total, internalizing, and externalizing problems, and autistic traits using the Child Behavioral Checklist and the Social Responsiveness Scale, respectively. We investigated maternal lifestyle and nutrition, perinatal factors (placental health and delivery outcomes), and child health (lifestyle, traumatic events, infections) as mediators and moderators. Results: We observed associations of prenatal infection with adolescent total behavioral, internalizing, and externalizing problems. The association between prenatal infection and internalizing problems was moderated by higher levels of maternal psychopathology, alcohol and tobacco use, and a higher number of traumatic childhood events. We found no association between prenatal infection and autistic traits. Yet, children exposed to prenatal infections and maternal substance use, and/or traumatic childhood events, had a higher risk of autistic traits in adolescence. Conclusion: Prenatal infection may be a risk factor for later psychiatric problems as well as a disease primer making individuals susceptible to other hits later in life. Study preregistration information: Prenatal maternal infection and adverse neurodevelopment: a structural equation modelling approach to downstream environmental hits; https://osf.io/cp85a; cp85a.


Preregistration deviations
As per our preregistration, we attempted to fit structural equation models (SEM).However, the latent variables reported bad fit measures for all modification indices.In line with the alternative strategies outlined in our preregistration, we conducted linear regression analyses to study the effects of prenatal infection on adolescent behavior and performed individual moderation and multiple mediation analyses to investigate the role of various cofactors.There were six deviations.First, the variable maternal age, a factor in one of our preregistered SEM latent variables, was included as an observed covariate in all analyses.Second, due to the large number of individual tests carried out to replace the originally planned SEM, we investigated the effects of infection timing and sex in the linear regression analyses only.Third, we clustered self-reported prenatal infection symptoms into the clinically meaningful categories 'upper respiratory tract infection' (rhinitis, pharyngitis, ear infection and sinusitis), 'lower respiratory tract infection' (pneumonia and bronchitis), and 'gastrointestinal tract infection' (enteritis or diarrhea).Fever as well as other infections, such as cystitis/pyelitis, dermatitis, eye infections, herpes zoster, flu, and sexually transmitted diseases (STD), were treated individually as described in the preregistration.We also investigated 'fever' as a separate severity marker.The results with the preregistered sum score can be found directly below in Tables S1-3.Fourth, we also investigated infection-specific effects on child behavioral problems as a post-hoc analysis.Fifth, we excluded lowfrequency mediators (<5%) from the final analyses due to power issues.The following mediators were omitted: gestational diabetes, preeclampsia, and pregnancy-induced hypertension.Sixth, we added the following mediators and moderators based on biological plausibility.Moderators added were 'postnatal life events' and 'postnatal direct victimization'.Mediators added were '5-minute Apgar score' and 'umbilical cord blood pH'.

Preregistration results
Supplementary Tables 1-4 show the results of the preregistered infection sum score (https://osf.io/cp85a)using a False Discovery Rate -Benjamini Hochberg (FDR-BH) correction applied for 104 tests.Similar to the main manuscript, we investigated the direct associations between prenatal infection and adolescent behavior, the moderating effect of perinatal factors and the mediating effects of various maternal and child factors.The results of the preregistered sum score are similar to the results of the main analysis presented in the manuscript.analyses are socioeconomic and non-modifiable variables potentially confounding the association for which we wanted to adjust.Postnatal direct victimization 1 A mediator is a variable that explains the relationship between two other variables.It helps to clarify the mechanism through which the independent variable affects the dependent variable. 2A moderator is a variable that affects the strength and direction of the relationship between two other variables.It helps to identify under what conditions or for whom the relationship between the independent variable and dependent variable may be stronger or weaker. 3A covariate is a variable that is controlled or adjusted for, to assess the relationship between two other variables.It helps to account for potential confounding variables that could affect the relationship between the independent variable and dependent variable.

Supplement 5: Additional information: mediating variables
Standardized fetal ultrasound examinations were performed by trained staff in mid pregnancy (18-25 weeks) 17,18 with high intra-and inter-observer reproducibility. 19More information on pregnancy dating can be found elsewhere. 18,20To evaluate the uteroplacental and fetoplacental circulation, flow velocity wave forms from the arteria umbilicalis and arteria uterine were recorded in mid pregnancy. 21,22We used two indirect measures of placental vascular resistance, namely the umbilical artery pulsatility index and the uterine artery mean resistance index.The umbilical artery pulsatility index was defined as (peak systolic velocity -end diastolic velocity) / divided by mean peak systolic velocity. 22The uterine artery mean resistance index was defined as: (peak systolic velocity -end diastolic velocity) / by systolic peak velocity. 22An increase in the uterine artery resistance index and the umbilical artery pulsatility index can be an indicator of increased placental resistance.More information on this measurement can be found elsewhere. 22formation on the placental weight (in grams) was collected at birth through medical records.
Placental growth factor (PlGF) was measured in early pregnancy.Details of processing procedures can be found elsewhere. 10,23In brief, analyses were performed in non-fasting venous blood samples.
The Department of Clinical Chemistry of the Erasmus Medical Center analyzed PlGF concentrations using an immunoelectrochemoluminence assay on the Architect System (Abbott Diagnostics B.V., Hoofddorp, the Netherlands). 23stational age at birth was established via ultrasound examinations during prenatal visits at the research center. 3Details of the standardized method for fetal ultrasonographic measurements can be found elsewhere. 18Information on the child's birthweight was obtained from midwives and hospital registries.
The 5-minute Apgar score was obtained from medical records.It is a quick assessment tool to evaluate the physical condition of the newborn baby based on the following five components (each of which is scored on a scale of 0-2): heart rate, respiratory effort, muscle tone, reflex irritability, and skin color.
The maximum score is 10.A score of 7 or above is generally considered normal, while a score of 4-6 is considered moderately abnormal, and a score of 0-3 is considered severely abnormal.
Samples of cord blood treated with heparin were collected from an umbilical cord segment that was immediately clamped and isolated.The cord blood gases were assessed using automated gas check machines, which analyzed pH, pO2, and pCO2, and calculated base excess.The interval between the collection of cord blood samples and their analysis was under 60 minutes.The blood gas analyzers were calibrated and standardized utilizing pH solutions and reference gases.

Supplement 6: Additional information: moderating variables
The level of vitamin D was assessed by measuring 25-hydroxyvitamin D in the unit nmol/L in the mother's serum during the first eighteen weeks of gestation. 725-hydroxyvitamin D was defined as the sum of 25-hydroxyvitamin D2 (25OHD2) and 25-hydroxyvitamin D3 (25OHD3). 8We will use this variable as a continuous variable.To quantify the samples, isotope dilution liquid chromatographytandem mass spectrometry was used.The analytical system consisted of a Shimadzu Nexera UPLC coupled to an AbSciex 5500 QTRAP equipped with an atmospheric-pressure chemical ionisation source.To evaluate the accuracy of the assay, we used certified reference materials, which were purchased from the National Institute of Standards and Technology (NIST SRM 972a Levels 1-4).From July 2013 to August 2014, samples were analyzed at the Queensland Brain Institute in Brisbane, Australia.Further information of the assay methodology have been described elsewhere. 9Maternal blood serum samples were collected during early pregnancy.More information about the transportation and storage method can be found elsewhere. 10 used iron biomarker ferritin as a measure of iron store. 11Ferritin was determined by electrochemiluminescence immunoassay on the Cobas e411 analyzer (Roche) in the unit micrograms/L.
To evaluate the mother's diet in early pregnancy, a semi-quantitative 293-item food frequency questionnaire (FFQ) was used at enrollment.The FFQ comprises frequently consumed Dutch food and was modified for use during pregnancy. 2Calculation of the energy and nutrient intakes was based on the 2006 Dutch food composition table. 12Three 24-hour recalls of 71 pregnant women, who lived in Rotterdam, were used to validate the FFQ. 13 With the national dietary guidelines 14 , a predefined diet quality score was constructed for pregnant women.The next 15 components and corresponding cutoffs were included for the diet quality score: vegetables (≥200 grams/day), fruit (≥200 grams/ day), whole grains (≥90 grams/ day), legumes (≥135 grams/week), nuts (≥15 grams/day), dairy (≥300 grams/ day), fish (≥100 grams/week), tea (≥450 grams/ day), grain quality (ratio whole grains of total grains), soft fats and oils (ratio of total fat), red meat (≤375 grams/week), sugar-containing beverages (≤150 grams/ day), alcohol (yes/no), salt (≤6 grams/ day), and folic acid supplements in early pregnancy (periconceptional/first ten weeks/not).The scores for all the individual components were summed up, which consequently led to an overall continuous score with a range from 0 to 15.A higher score indicates a healthier diet.More information on the construction of the score can be found elsewhere. 15e pre-pregnancy body mass index (BMI) was assessed by self-report using a questionnaire at enrollment.Mothers were asked to report their pre-pregnancy weight in kilograms and their height in centimeters, which were used to calculated BMI in kg/m 2 .Previous studies in Generation R have shown the correlation between self-reported pre-pregnancy weight and weight measured at enrollment to be 0.97 (P < 0.01). 16formation on maternal psychoactive substance use during pregnancy was assessed with a questionnaire at enrollment.Maternal psychoactive substance usage (i.e., marihuana, hashish, cocaine, heroin, or ecstasy) was categorized into: 'no', 'yes, until pregnancy was known,' and 'yes, continued during pregnancy.'Information on maternal tobacco use during pregnancy was assessed with a questionnaire at enrollment.Maternal tobacco use was categorized into: 'no', 'yes, until pregnancy was known,' and 'yes, continued during pregnancy.'Information on maternal alcohol consumption during pregnancy was assessed using a questionnaire at enrollment.Maternal alcohol consumption encompassed the categories: 'none during pregnancy,' 'drank until pregnancy was known,' 'continued drinking occasionally,' and 'continued drinking frequently' (one or more glass/week for at least two trimesters).
Prenatal maternal psychopathology was measured with a validated self-reported questionnaire (Brief Symptom Inventory) at enrollment.From this 53-item questionnaire 1 , a Global Severity Index was calculated that served as a continuous score of prenatal maternal psychopathology with higher scores indicating more problems.
Information about breastfeeding was obtained from delivery reports and postnatal questionnaires at the ages of 2, 6 and 12 months after birth. 4Mothers were asked whether they ever breastfed their child, with the answer possibilities were: 'yes' or 'no'.In this study, breastfeeding was included as a binary variable (ever/never).
Information on childhood body mass index (BMI) was collected when the children visited the research center at 9-12 years of age.The height and weight of the child were measured by trained staff.Height was measured with a stadiometer (Holtain Limited) and weight was measured with an electronical scale (SECA). 24When measuring weight, any heavy clothing or shoes were taken off.BMI (kg/m 2 ) was calculated using height and weight.We used a measure of BMI which is adjusted for age and sex based on Dutch growth curves 25 , also called BMI-SDS.BMI-SDS is an index of the nutritional status of the child. 26More information can be found elsewhere. 24formation about the child's substance use (alcohol and tobacco) was obtained using a self-report questionnaire at age 13-16 years.The children were asked if they ever drank alcohol and if they had ever smoked tobacco.Both variables were coded as 'yes', 'no', and 'don't know'.
A sum score for childhood infection was created based on information at seven different time points.
In Table S5, the time point, the question, the answer possibilities, and the scoring can be found.A total of 15 points could be scored, with a higher score representing a higher number of infections.

Figure S1 .
Figure S1.Distribution different infection types.Figures S1A-C show the distribution of the infection for trimesters 1, 2 and 3, respectively.

Figure S2 .
Figure S2.Frequency plot for exposure and outcome variables.

Supplement 3 :Figure S3 .Supplement 4 :Figure S4 .
Figure S3.Correlation between C-reactive protein levels <18 weeks of gestation (1 fixed time point measurement available in Generation R) and total and trimester-based prenatal infection scores.

Yes, 3 times or more 3 Supplement 7 :
Frequency mediators and moderators

Table S3 . Mediating effects of prenatal maternal infection on adolescent behavior Outcome
Supplement 2: Distribution infection types and behavioral problems Individually tested moderators: maternal psychopathology, maternal diet food score, maternal iron levels, maternal vitamin D levels, pre-pregnancy BMI, maternal substance use, maternal alcohol use, maternal tobacco use, child BMI, child tobacco use, child alcohol use, breast feeding, childhood infections, postnatal life events, postnatal direct victimization *

Table S6 .
Standardized coefficients for indirect effects of mediators.

Table S7 .
Standardized coefficients for direct associations of fever on adolescent behavior

Table S8 .
Testing interaction term for sex (exposure: prenatal infection)

Table S9 .
Testing different infection types and adolescent behavior income, maternal national background, maternal IQ, child sex, child IQ, and child age at measurement.