Data resources
Research data were extracted from the birth defect monitoring sub-module in the Liuzhou Maternal and Child Health Information Management System between January 2019 and December 2019. This includes records from perinatal babies, including live birth, stillbirth and infant death within 7 days, reported by 114 midwifery agencies in Liuzhou for a total of 32,549 births. The data excluded twin and multiple births. Among this cohort, there were 635 cases of BDs. All data was derived under the supervision of the health administration, and this study was approved by the Institutional Review Board of Liuzhou Maternal and Child Health Hospital.
Additional perinatal and maternal data were derived from the China Maternal and Child Health Monitoring Data Direct Reporting System (https://zhibao3.mchscn.org/) and the Maternal and Child Health System of Liuzhou City. Using the “Birth Defects Registration Card” and “Quarterly Report on Number of Perinatal Births” from the Maternal and Child Health System of Liuzhou City, relevant data were collected, including maternal status, birth status, birth defect diagnosis and family history. Maternal data included ethnicity, age, education, family income, date of last menstrual period, residential address, registration address and parity. Infant data included date of birth, sex, gestational age, fetal number, weight, and outcome (including stillbirth, fetal death or live birth between 20 weeks of gestation through 7 days after birth).
Family history included abnormal fertility history and family genetic history. The diagnosis of BDs was based on the “International Statistical Classification of Diseases and Related Health Problems, Tenth Edition” (ICD-10) and Chinese National Criteria of BDs [32]. We investigated a wide range of birth defects including 11 major types of BDs: hydrocephaly, congenital heart disease, cleft lip with or without cleft palate, urinary system abnormalities, ear anomalies, congenital clubfoot, polydactyly and congenital syndactyly, severe thalassemia, Cystic hygroma, and Bart’s Syndrome. Any birth defects not included in these 11 diagnoses were categorized as “other”.
Quality controls
In order to ensure the accuracy of the report, a physician at each registered hospital was required to complete a quarterly form, in addition to the Birth Defects Registration Card. Each quarterly table contained 3 months of data, including ethnicity, date of last menstrual period, parity, education, family income, date of birth, gestational age, weight, number of births, whether labor was inducted after diagnosis of a birth defect, diagnostic basis, diagnosis of malformation, and birth defect diagnosis for each birth occurring in the hospital. Birth defect registration cards and quarterly tables were reviewed and audited by maternal and child health hospitals and health administrative departments. Quality control measures were monitored regularly in respective hospitals, quarterly at the county level and every two years at the municipal or provincial level. The quality requirements for BDs monitoring data included: 100% completion rate of form, form items error rate less than 1%, input error rate less than 1%, and a rate of missed birth defects less than 1%.
Exposure assessment
The ambient air pollution data used in this study came from the weather information data collected by the Liuzhou Environmental Protection Bureau between January 2018 and December 2019, which includes six state-controlled air automatic monitoring points in Liuzhou (HX Waterworks, Liuzhou Fourth Middle School, GTS, Environmental Monitoring Station, Liudong Primary School, Liuzhou Ninth Middle School), two district control stations (Liuzhou Second middle school, LW), and six city and county control stations (LJ District Experimental High School, LC County Middle School, LZ County Youth Activity Center, RA County Quality Supervision Bureau, RS County Health School, the SJ County Guyi Town Center) (Fig. 1). Monitored pollutants included particulate matter less than 2.5 μm in aerodynamic diameter (PM2.5), particulate matter less than 10 μm in aerodynamic diameter (PM10), sulfur dioxide (SO2), Carbon monoxide (CO), Nitrogen dioxide (NO2), and ozone (O3). O3 was calculated as a maximum daily average over 8 h. The daily concentration values of other pollutants were calculated based on the average of 24 h measured at 14 monitoring points.
The assessment was performed according to the national standard “Ambient Air Quality Standard” (GB 3095–2012). The average annual secondary concentration limits of PM2.5, PM10, SO2 and NO2 are 35 μg/m3, 70 μg/m3, 60 μg/m3 and 40 μg/m3, respectively. The average 24-h secondary concentration limits of PM2.5, PM10, SO2, CO, NO2 and O3 are 75 μg/m3, 150 μg/m3, 75 μg/m3, 4 mg/m3, 80 μg/m3, and 160 μg/m3, respectively.
Maternal ambient air exposure was determined through block kriging, a statistical technique which predicts average exposure concentrations based on spatial variation. The daily concentrations of PM2.5, PM10, SO2, CO, NO2 and O3 on the dates between last menstrual period and date of delivery were determined based on maternal residential address. Monthly concentrations for the 3 months prior to pregnancy and the first trimester of pregnancy were estimated for each participant [33, 34].
Covariates
Potential covariates considered from the birth defects registration cards included maternal age (< 20 years, 20–24 years, 25–29 years, 30–34 years, ≥35 years), family income (<2000RMB, 2000–3999 RMB, 4000–7999 RMB, ≥8000 RMB), highest education levels (classified as primary school or below, middle school, high school/technical school, college or above), birth weights (< 1500 g, 1500 g–2499 g, 2500 g–3499 g, ≥3500 g), number of pregnancies (1, 2, and ≥ 3), total previous live births (0, 1, and ≥ 2), and gender of infant (male, female). We also adjusted for additional maternal covariates according to the existing literature and the study population characteristics. We collected variables suspected as potential confounders from the China Maternal and Child Health Monitoring Data Direct Reporting System. These potential confounders included premature rupture of membrane, syphilis, ethnicity, hyperthyroidism, gestational diabetes, preeclampsia, gestational hypertension, HIV, hypothyroidism, infection, medication and in vitro fertilization-embryo transfer. Maternal smoking and alcohol use during pregnancy were not controlled for because < 0.3% of the mothers reported smoking or drinking alcohol.
Statistical methods
Chi-square test was performed to examine the differences in social demographics between infants with BDs and infants without BDs. Prevalence rates of overall BDs and each specific BD were also examined. Distributions of air pollutant concentrations were presented by quartile and interquartile range (IQR) averaged over three months preconception and during the first trimester of pregnancy. Logistic regression was used to assess the association of air pollution exposure on BDs. BDs were the dependent variable, and the individual exposure concentration of air pollutants during three months prior to pregnancy and the first trimester of pregnancy were the independent variables. Important covariates were controlled, in order to explore the influence of exposure concentration and exposure time on BDs. To assess the role of ambient air pollutant exposure at different stages of pregnancy, we constructed exposure variables for the 3 months before pregnancy and the first trimester of pregnancy. We studied both the impact of a single pollutant on BDs and the combined effects of multiple air pollutants. Corresponding odds ratio (OR) and 95% confidence interval (95% CI) were calculated for birth defects and air pollutant exposure at different stages of pregnancy. In order to evaluate the robustness of the estimated effects, we performed sensitivity analyses measuring the associations between ambient air pollution and birth defects in Han and Zhuang ethnic groups (sample sizes for other ethnic groups are relatively small). Statistical test significance level is 0.05 (two-tailed). Data processing and statistical analysis were performed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA) statistical software package.