Maternal exposure to life events during pregnancy and congenital heart disease in offspring: a case-control study in a Chinese population

The data in this study was based on an unmatched case-control study about CHD and its risk factors, which was conducted in Shaanxi province of China from January 2014 to December 2016. It was carried out in the six large hospitals responsible for monitoring birth defects in Xi’an, including First Affiliated Hospital of Xi’an Jiaotong University, Second Affiliated Hospital of Xi’an Jiaotong University, Xijing Hospital of the Fourth Military Medical University, Tangdu Hospital of the Fourth Military Medical University, People’s Hospital of Shaanxi Province and Northwest Women & Children’s Hospital. The cases were included according to the following inclusion criteria: the perinatal infants who were diagnosed with CHD according to the International Classification of Diseases-10 (ICD-10) from 28 weeks of gestation to 7 days after birth (including single live births and stillbirths), as well as the fetuses who were diagnosed with CHD by ultrasound examination in hospital less than 28 weeks of gestation. The controls were the newborn infants without any birth defects in corresponding hospitals. Briefly, a total of 2996 participants were recruited in the study. In this analysis, we excluded twins or multiple births, those diagnosed as other types of birth defects or CHD combined with other types of birth defects, those without a definite diagnosis, without a clear date of terminal pregnancy, unable to fill in a questionnaire and those lacked relevant variables data. Finally, 2280 participants were used as sample size for analysis after excluding those missing main variables or covariates. We conducted a post hoc analysis (G*Power program, version 3.1.9.4) [22] to assess the achieved power of our study. Based on our sample size and certain parameters of logistic regression in our analysis (OR = 1.64, α = 0.05, R²_{for covariates} = 0.21), the power about negative events was estimated at 99.7%. There was a similar power estimation conducted for positive events. It meant that the sample size included in our study could provide sufficient power to explore the association between exposure to life events (positive events and negative events) and CHD in offspring.

A self-designed structured questionnaire was used in the study to collect information, which has previously been published elsewhere [23]. The content included social demographic characteristics, maternal life behavior and environment during periconception, history of pregnancy and parturition, history of family disease, antenatal examination, disease and medication, nutrient supplementation, life events scale during pregnancy, neonatal information, etc. Maternal information was gathered by trained staff through face-to-face interviews with mothers, and newborn information was collected by the medical record system of hospital. The study was approved by the Human Research Ethics Committee of Xi’an Jiaotong University Health Science Center (No.2012008). All participants were aware of the content of study and signed the informed written consent before investigation.

The main variable life events were assessed by Life Events Scale for Pregnant Women (LESPW), a self-assessment scale compiled by Gao Yan et al. that has been published and widely used in China [21]. It included 53 representative life events, covering various changes that may occur in pregnant women’s daily life. A relevant study has suggested that the reliability and validity of LESPW could meet the requirement of statistical measurement (Cronbach’s alpha = 0.750) [24]. Unfortunately, these life events were not classified by their nature. Therefore, according to the potential influence of life events on pregnant women, we divided them into positive and negative events. Positive events (e.g., more care from family members) could have positive effect, and negative events could result in negative impact. Among all events, the description of 3 events ─ “house-moving”, “changing job-content or adjusting working hour and place” and “changing living habits such as sleep, diet and clothing” was equivocal (whether these changes were good or bad were unclear), so that it was difficult to determine the nature. Thus, the 3 events were excluded in our main analysis. We selected the remaining 50 life events, which included 9 positive events and 41 negative events (detail in Appendix 1). In addition, assuming that any changes may bring about stress on pregnant women, we regarded the 3 excluded events as negative events, i.e., 9 positive events and 44 negative events were included, for additional analysis to compare with the main analysis (Supplemental Table S1).

To clarify the association between life events and CHD more accurately, the relevant covariates should be considered for adjustment in analysis because they could confound the association. Reference to previous literature [6‐10] as well as the characteristics of participants, there were 16 possible covariates concerned in total. They covered socio-demographic characteristics (maternal age, residence, and paternal education), maternal exposures (maternal drinking, exposure to tobacco smoke, exposure to harmful substance, exposure to adverse environment, folic acid supplementation, infection, depression, medicine-taking and prenatal examination) and history of childbearing (history of parturition, history of abortion, family history of CHD). In order to reduce unnecessary covariates and excessive adjustment, the directed acyclic graph (DAG) [25] (DAGitty program, version 2.3) [26] was adopted to screen the minimal sufficient set of covariates. DAG was introduced by Greenland et al. for causal analysis, and widely used in the field of epidemiology. It could visualize the causal relationship between variables, identifying the minimal sufficient adjustment sets to minimize confounding bias in epidemiological studies [25]. After analysis of DAG, 8 covariates were selected as adjusted variables, including maternal age (defined as < 25, 25 ~ 30 or > 30 y), residence (defined as urban or rural), maternal and paternal education (defined as ≥college or < college), history of parturition, history of abortion, infection during periconception, and abnormal prenatal examination (Figure S1). Considering that the genetic factor also played an important role in the occurrence of CHD, we additionally included the family history of CHD as potential covariate. Finally, a total of 9 covariates were included as adjusted variables in our analysis. Among them, history of abortion contained spontaneous abortion, induced abortion and medical abortion; infection during periconception period meant that pregnant women suffered from colds, fever, urogenital system infection and some viral infection from 3 months before pregnancy to the whole period of pregnancy; abnormal prenatal examination referred to abnormalities detected by B-mode ultrasound or fetal cardiac ultrasound examination before delivery.

The categorical data was summarized by frequency and percentage, while the continuous data was summarized by mean ± standard deviation. The Chi-square (χ²) test and t-test were used to compare the differences between the case and control groups. Logistic regression models were constructed to estimate the odds ratio (OR) and 95% confidence interval (CI) for the effect of life events during pregnancy on CHD in offspring by not adjusting or adjusting for covariates. The adjusted covariates were 9 confounding factors as described above. Since both positive and negative life events occurred at the same time, we put them together in models for synchronous analysis. They were analyzed as both continuous variables and categorical variables classified by occurrence. For the latter, we also categorized them according to the frequency to explore the dose-response association. Moreover, based on the synthetical occurrence of positive and negative events, we created a new integrated variable, including four groups ─ “neither positive nor negative events group (P₀N₀)”, “only negative events group (P₀N₁)”, “only positive events group (P₁N₀)”, and “both of positive and negative events group (P₁N₁)”. Taking the P₀N₀ group as a reference, we calculated OR values of the other three groups. By comparing the OR values of the P₀N₁ and P₁N₁ groups, we aimed to explore the modifying effect of positive events under the circumstances that negative events occurred. In addition, subgroup analysis was employed to assess the consistency of the association across different categories of 8 selected covariates by DAG, and the remaining covariates were adjusted by multivariate logistic regression to estimate OR values of association between occurrence of life events (reference to non-occurrence) and CHD among each subgroup. Considering the smaller sample size for the family history of CHD group, we did not conduct subgroup analysis for this covariate. P < 0.05 was regarded as a statistical significance. SPSS version 22.0 (IBM SPSS Statistics, Chicago, IL, USA) was used for data analysis and R version 3.5.3 (R Development Core Team, Vienna, Austria) software was used to make plots.

A total of 2280 subjects were included in this study, 699 in the case group and 1581 in the control group (Fig. 1). Table 1 showed the characteristics of pregnant women between the case and control groups. The differences between the two groups were statistically significant in maternal age, education, residence, history of parturition, infection during periconception and abnormal prenatal examination, while there was no significant difference in history of abortion.

As described in Table 2, the average number of positive life events in the case group was 0.30 ± 0.63, while 0.58 ± 0.79 in the control group; the average number of negative life events in the case group was 1.12 ± 2.06, while 0.72 ± 1.46 in the control group. After adjusting for the confounders, the risk of CHD in the offspring could be reduced by 48% (OR _adj = 0.52, 95%CI: 0.44 ~ 0.62) for each additional positive life event during pregnancy, while each additional negative life event occurring in pregnant women could make the risk of their offspring increase by 20% (OR _adj = 1.20, 95%CI: 1.13 ~ 1.28).

According to the occurrence of life events, we analyzed them as binary variables. Among all participants, 852 pregnant women had ever experienced positive life events during pregnancy, 160 in the case group (22.89%) and 692 in the control group (43.77%); 849 pregnant women had experienced negative life events, 297 in the case group (42.49%) and 552 in the control group (34.91%). Both differences between the two groups had statistical significance. With confounders adjusted, the pregnant women with positive events experienced had a 62% lower risk of CHD in their offspring than those without (OR _adj = 0.38, 95%CI: 0.30 ~ 0.48). The pregnant women with negative events exposed were 1.62 times more likely to have CHD in their offspring than those without (OR _adj = 1.62, 95%CI:1.29 ~ 2.03).

According to the frequency of life events, we investigated the dose-response association between them and CHD in offspring. As shown in Table 3, a trend of dose-response association was observed (p _trend < 0.001). For positive events, the odds of CHD in offspring decreased with the increase of event frequency. In contrast with those who had not experienced positive events, those experiencing one event had 63% lower odds, and those experiencing two or more events had 66% lower odds (OR_1adj = 0.37, 95%CI: 0.28 ~ 0.48; OR_2adj = 0.34, 95%CI: 0.21 ~ 0.53). For negative life events, the risk increased along with the accumulation of frequency. After adjusting for the confounders, the odds of CHD in offspring increased 1.40 times for the pregnant women experiencing 1 ~ 2 negative events (OR_1adj = 1.40, 95%CI: 1.09 ~ 1.79), 2.08 times for those with 3 ~ 4 events (OR_2adj = 2.08, 95%CI: 1.34 ~ 3.25) and 3.15 times for those with 5 or more events (OR_3adj = 3.15, 95%CI: 1.82 ~ 5.43) respectively, in comparison with those without negative events.

Further, we divided the variable into four integrated groups to explore modification of positive events: P₀N₀, P₀N₁, P₁N₁ and P₁N₀. Compared with the P₀N₀ group, the odds of CHD in the P₀N₁ group increased by 38% (OR _adj = 1.38, 95%CI: 1.05 ~ 1.82), while that in the P₁N₁ group decreased by 33% (OR _adj = 0.67, 95%CI: 0.49 ~ 0.91) (Table 4). It indicated that positive events could weaken or even modify the risk impact of negative events on CHD in offspring.

The adjusted ORs of association between occurrence of positive or negative life events and CHD in offspring by selected covariates were shown in Fig. 2. The protective effects of positive events among pregnant women in each subgroup were observed significantly, demonstrating well stability and consistency of the association between life events and CHD. For negative events, except for two subgroups of those older than 30 years and those with an abnormal prenatal examination, the significant risk of negative events for CHD was found in the rest of subgroups, implying an overall robust association.

In addition, as mentioned above, 3 excluded events about life or work changes were supposed to be regarded as negative events for additional analysis. As shown in Supplemental Table S1, pregnant women with positive events experienced had 60% lower odds of CHD in offspring than those without (OR _adj = 0.40, 95%CI: 0.31 ~ 0.51), while pregnant women with negative events experienced had 21% higher odds than those without (OR _adj = 1.21, 95%CI: 0.97 ~ 1.51). Compared with the results of the primary analysis, the adjusted risk effect of negative events was reduced and not statistically significant, but the direction remained the same as before.