1. Introduction
The importance of maternal nutrition during pregnancy for fetal brain development has been well documented. Beginning from around 18 days post-fertilization, the embryo undergoes a coordinated process of nerve proliferation and migration, synaptogenesis, myelination, and apoptosis to develop and form the fetal brain [
1], but the brain is more vulnerable to nutritional deficiencies at this time. Additionally, the hypothesis of the Developmental Origins of Health and Disease (DOHaD) suggests that during this period of rapid development, the brain becomes more sensitive to the environment, and this is a vulnerable and critical period of perturbation that may predispose the fetus to postnatal neuropsychological disorders [
2,
3,
4].
Human epidemiological evidence has identified an association between maternal nutrient deficiencies during pregnancy and cognitive development of their offspring. Prenatal vitamin A, folic acid, and vitamin D deficiencies are associated with subsequent suboptimal neuropsychological development [
4,
5,
6], such as susceptibility to autism and delayed language development, but several nutrients are not sufficient to assess the nutritional status of pregnant women, and fortunately, birth cohorts on the association of maternal dietary patterns during pregnancy with neuropsychological development of offspring have bridged this gap. Cohort findings have implied that unhealthy dietary patterns during pregnancy are associated with reduced executive function, delayed language development, and lower IQ scores in offspring, with the unhealthy diet including lower Mediterranean diet scores [
7,
8,
9,
10,
11]. However, the western Mediterranean diet (higher intake of fruits, vegetables, fish, pasta, and rice, and lower intake of meat, sugar, and fat) differs from the eastern dietary pattern (predominantly carbohydrates, vegetables, fruits, pork, etc.). Furthermore, another difference is inland and coastal diets in China because of the higher intake of aquatic products in coastal diets. Therefore, it is necessary to investigate the association between dietary patterns of pregnant women and behavioral problems of offspring in China’s inland. In addition, most studies use principal component analysis (PCA) [
9,
11,
12,
13] or cluster analysis [
14,
15] to classify food groups intakes, but some studies prefer to use latent class analysis (LCA), which is recommended for food intakes to study the effects of mutually exclusive categories [
16,
17].
Accordingly, we hypothesize that children of mothers with healthy dietary patterns during pregnancy have fewer behavioral problems. This paper aims to classify maternal food group intake into appropriate categories during pregnancy in inner-city China by the LCA method and then analyze the association of dietary patterns with behavioral problems of offspring at the preschool age.
2. Methods and Materials
2.1. Study Population
This study is based on the China-Anhui Birth Cohort Study (C-ABCS), which has been established in six municipal health institutions between November 2008 and October 2010 with 5084 pregnant women and their offspring recruited. Specific inclusion and exclusion criteria are described in the team’s previously published literature [
18]. After excluding maternal loss (202), spontaneous abortions (92), stillbirths, fetal death, induced labor (55), and twin pregnancy (66), 4669 pairs of mothers and singleton live births have been included in the child follow-up cohort. Between April 2014 and April 2015, we have accessed cognitive and behavioral development at early childhood (4.25 ± 0.41 years) using assessment tools that include Strengths and Difficulties Questionnaire (Edition for parents, SDQ), Clancy Autism Behavior Scale (CABS), and Conner’s Abbreviated Symptom Questionnaire (C-ASQ). However, the team consists of several groups, and our group has participated in a survey of the former 1783 mothers, thus obtaining survey data from 1783 mother-child pairs. Among them, 171 mothers have been excluded for no food intakes data, and the data of 1612 mother-child pairs is finally included in the analysis.
Figure 1 provides a more visual description.
2.2. Measurements
2.2.1. Food Groups Intakes Assessment during Pregnancy
Based on collected literatures and consultation with experts, a semi-quantitative food frequency questionnaire (FFQ) has been composed by selecting food items that represent the dietary intakes of pregnant women in Anhui province, China. The questionnaire is administered at 12.13 ± 3.82 and 30 ± 2.11 gestational weeks, asking about dietary intakes during the first and second trimester. A total of 19 food items have been included, which are rice, wheaten food, vegetables, fruits, beef and mutton, poultry, pork, animal fishery products, eggs, dairy products, beans, nuts, fried foods, pickles, animal innards, and garlic. For each food entry, pregnant women are asked about the frequency of intake in a week, and the options are divided into 5 levels: 1 = no intake, 2 = 1 to 3 times per week, 3 = 4 to 5 times per week, 4 = 6 to 8 times per week, and 5 = more than 9 times per week. The data of food intakes are a skewed distribution and we would regroup it. Referring to the relevant literature [
16,
17] and the actual distribution of the intake frequency, the criteria for regrouping are as follows. The percentage of non-consumers (option was no intake) is less than 7.5%, and variables are transformed into binary variables: above median and below median. The percentage of non-consumers is higher than 7.5% and lower than 45%, and variables are transformed into triple variables: non-consumed, below median, and above median. The percentage of non-consumers is higher than 45%, and variables are transformed into binary variables: non-consumed and consumed.
2.2.2. Outcomes
The behavioral problems in early childhood are assessed by the SDQ, C-ASQ, and CABS, which is fulfilled by a familiar caregiver and then reviewed by trained investigators.
SDQ refers to children’s behavior and emotions over the previous six months. The scale provides balanced coverage of emotional symptoms, conduct problems, hyperactivity/inattention, peer relationship problems, and prosocial behavior. The former four scales are added together to generate a total difficulties score. The higher the score of total difficulties score, the more serious the objective difficulty is, and the delineation criteria of boundary values refer to the scoring rules [
19].
The 10-item C-ASQ is derived from the revised Conners Parent Rating Scale. The widely used scale is used to assess attention-deficit hyperactivity disorder (ADHD) symptoms in children. The options in this scale range from 0 (never) to 3 (frequently) according to the frequency of symptoms. ADHD symptoms have been defined as a total score of ≥15 [
20].
CABS is used as a screening tool to identify children with autism. The scale consists of 14 items, with scores of 0, 1, and 2 assigned to “never”, “occasionally”, and “often”, respectively, and a total score of ≥14 is considered positive for potential autism [
21].
2.2.3. Covariates
Socio-demographic variables have been investigated in a self-administered maternal and child health record form, including mainly maternal age, education level, place of residence, monthly income, type of work, secondhand smoke exposure, and home renovation at the time of inclusion. We have also extracted children’s birth date, sex of the child, birth weight, and gestational weeks of delivery from hospital birth records at the time of delivery. It should be noted that child age is calculated as the date of examination minus the date of birth, and for preterm infants, age is calculated as the date of testing minus the expected date of delivery. As pregestational body mass index (BMI) = pregestational weight (kg)/maternal height
2 (m
2), BMI categories specific for adult Chinese female are assigned as follows: BMI < 18.5 (underweight), BMI = 18.5–24 (normal), BMI ≥ 24 (obesity or overweight). Maternal depression is assessed by the center for epidemiological survey depression scale (CES-D) [
22], with a score above 16 indicating possible depression.
2.3. Statistical Analysis
To identify mutually exclusive groupings, we have used LCA to derive dietary patterns. A trivial 1-class model is first fitted in which all individuals belong to the same category, and then 2 to 5-category models are fitted. The optimal model is selected based on BIC and AIC values while considering the same number of categories and reasonable category probabilities in different trimesters to ensure substantial dimensionality reduction in food intakes, ease of model understanding, and further analysis. The names of the clusters are chosen based on the conditional distribution of food intakes. We have four clusters of food intakes, called “High-consumed pattern (HCP)”, “Southern dietary pattern (SDP)”, “Northern dietary pattern (NDP)”, and “Low-consumed pattern (LCP)”.
Afterward, the response probabilities of the potential classes are described. The covariates in the models include mainly maternal age at inclusion, pregestational BMI, maternal education, residential region, monthly income, maternal depression, child gender, and child age at the visit. We also explore differences in the distribution of dietary patterns across covariates using χ2 test. Finally, we used a logistic regression model to access the association of dietary patterns with behavioral problems in early childhood. To verify the stability of the regression analysis results, three analytical models with different covariates have been constructed. Model 1 only includes the first trimester and second trimester dietary patterns. Secondarily, model 2 includes covariates such as child gender and age, maternal education, residence, maternal age, pregestational BMI, and monthly income, and model 3 takes into account maternal depression during pregnancy on the basis of model 2. All above analyses have been performed in Mplus 7.4 and SPSS 23.0.
4. Discussion
In this birth cohort study in inland China, we have observed that maternal SDP (characterized by higher vegetable and fruit intakes), NDP (characterized by higher meat intakes), and HCP (characterized by high food groups intakes) during early pregnancy are associated with lower incidence of emotional symptoms in preschool-age children compared with LCP (characterized by lower food groups intakes), and maternal SDP at mid-pregnancy is associated with reduced conduct problems in children. In addition, we have detected associations between maternal SDP in early pregnancy and lower emotional symptoms, and between maternal SDP in mid-pregnancy with decreased peer relationship problems in boys. In girls, total difficulty scores are lower with maternal SDP in mid-pregnancy. Overall, these findings supported our hypothesis, and maternal SDP in both early and mid-pregnancy may predict fewer childhood behavioral problems, but maternal HCP and NDP during pregnancy unlikely result in reduced behavioral problems.
Several birth cohort studies have reported an association between unhealthy maternal dietary patterns during pregnancy and decreased behavioral problems in the offspring. A few of these studies have had several dietary patterns that are slightly similar to the dietary patterns in this paper. The Avon Longitudinal Study of Parents and Children in the United Kingdom shows that 8-year-old children of mothers in the “vegetables and fruits” dietary cluster have higher IQs, while an unhealthy maternal diet during pregnancy (processed and junk foods) is associated with lower cognitive function in 7- and 8-year-old children [
23]. The EDEN mother-child cohort in France reports a positive association between maternal “low health diet (characterized by low intake of fruits, vegetables, fish and whole grains)” and “high Western diet (processed foods and snacks)” during pregnancy and a high symptomatic ADHD-attention trajectory in children aged 3 to 8 years [
7]. The Generation R Study and The Norwegian Mother and Child Cohort Study (MoBa) find similar conclusions: unhealthy dietary patterns during pregnancy are positively associated with externalizing behavior (inattention, aggression) in the offspring [
8,
11]. A US cohort has reported that maternal intake of a higher quality diet during pregnancy (higher Mediterranean diet score or Alternative Health Diet Index) is associated with better visuospatial skills in offspring at 3.2 years of age and better intellectual and executive functioning in offspring at 7.7 years of age [
10]. Similarly, a birth cohort study in the coastal city of Jiangsu, China, identifies a high intake of dietary fiber and high-quality protein (aquatic products, fresh vegetables) during mid and late pregnancy as predictive of better gross motor and receptive communication development in 1-year-old children [
9]. However, not all studies have found an association between diet during pregnancy and behavior problems in offspring, and data from The Southampton Women’s Survey does not reveal an association between vegetarian consumption during pregnancy and poorer cognitive development in children aged 6–7 years [
24]. In addition, two meta-analyses have implied that higher maternal diet quality is associated with a lower risk of poorer cognitive development in offspring [
5,
25]. The above studies confirm the plausibility of the findings of this study.
This paper represents the children of mothers with a southern dietary pattern of higher vegetable and fruit intake that have reduced behavioral problems compared to mothers with a low intake diet. Vegetables and fruits provide the macronutrients (vitamin A, vitamin C, carotenoids, and small amounts of B vitamins) and key minerals (calcium, magnesium, potassium, and iron) that the fetal brain needs to develop in utero. Moreover, low levels of food intake imply inadequate nutrient intakes. Sub-optimal macronutrient balance and micronutrient deficiencies can lead to poor maternal body composition and metabolism, which in turn can affect maternal health and lead to intrauterine programming of the fetus, altering fetal brain morphogenesis, brain neurochemistry, and neurophysiology long-term metabolic and cognitive health consequences [
2,
26,
27].
This study has several strengths. To our knowledge, this study is one of the few studies to use the LCA method to categorize food groups intakes during pregnancy and then explore the association with reduced behavioral problems of offspring in early childhood. LCA is applicable to a wide range of variable types (categorical and continuous variables) and provides higher classification accuracy, for it is based on probabilistic mixture modeling. It is also suitable for missing data [
28,
29]. This study is an inland Chinese prospective birth cohort study that investigates maternal diets at both early and mid-pregnancy visits. However, this study also has several limitations. Firstly, the main drawback of this study is that dietary patterns are assessed based on frequency of intake rather than the actual amount consumed. Secondly, the data is part of the cohort study data and not all data are available. Thirdly, it is unable to investigate the food intakes in late pregnancy and could not assess dietary information throughout pregnancy. Fourthly, the evaluation results of children’s behavioral problems, especially ASD and ADHD, are obtained through questionnaires and have not been diagnosed by special clinicians, which makes the evaluation results reliable. In addition, C-ABCS is not a national cohort and the dietary patterns in the paper are only representative of the diet of people living in central China, which makes extrapolation of the results of this study limited. Finally, although the design of the food frequency questionnaire has been discussed and modified several times, it could not cover all types of foods consumed by pregnant women, such as the lack of root and tuber crops. The deficiencies of the study do not deny its value.