Study design and participants
The data used in this study were derived from a large cross-sectional survey of a population in Shaanxi Province, Northwest China, between August and November 2013. The survey was designed to investigate the risk factors affecting adverse birth outcomes. Considering the different proportions and fertility rates in the urban and rural residents, this study adopted a stratified multistage random sampling design to select women of childbearing age (15 ~ 49) who gave birth, lived in the research area and had definitive pregnancy outcomes in 2010 ~ 2013. Women who experienced serious illnesses, such as cardiovascular disease or cancer, during the investigation, were excluded. The sampling method was comprehensively presented in the previous literature [27]. Briefly, 10 districts and 20 counties were randomly selected from the urban and rural strata, respectively. Then, we randomly sampled three streets from each sampled district and six communities from each street in the urban area, and six townships from each sampled county, and six villages from each township in the rural area. Finally, we randomly selected 60 eligible newborns and their mothers in each sampled community and 30 in each sampled village.
Data collection
After acquiring written informed consent, well-trained investigators conducted in-person interviews to collect the information about women using a structured questionnaire, including their place of residence, date of birth, educational level, family economic status, pregnancy history, prenatal care, lifestyle, and health status, and information was also collected about children, including gender and fetal number. BWs were measured to the nearest 10 g, and birth dates were obtained by reviewing the birth certificates. Before the investigation, the investigators received unified training, passed the examination, and mastered the inquiry skills and questionnaire filling abilities. To guarantee the quality of the survey, we reviewed the questionnaires at three levels: on-site review was performed by investigators; scrutiny after the day of investigation was performed by investigators who exchanged and checked the questionnaires; and the supervisors checked each questionnaire to see if there were any missing values or logical errors and returned to reinvestigation if any errors were found. At the end of each district (county) survey, 5% of the respondents were randomly selected for repeated surveys to ensure the authenticity and credibility of the data.
There were 30,027 women enrolled in the survey. After excluding non-live births and multiple gestations (n = 1116), and unknown BW and iron supplementation status (n = 702), 28,209 single live newborns and their mothers were finally included as the sample population.
Study variables
Newborn BW was used as the outcome variable in this analysis. The BW, which was measured within 1 h after birth, was collected through browsing the birth certificates and operationalized as a continuous variable. For the comparison between groups, the BW was separated into three categories: normal BW (2500 g ~ 4000 g), low birth weight (< 2500 g), and macrosomia (> 4000 g). Iron supplementation was considered as the exposure variable. Women were divided into two groups: the iron supplementation group (iron supplementation, such as ferrous sulfate tablets, iron dextran tablets, from 3 months before pregnancy to the end of pregnancy) and the non-users (no iron supplementation). One to three months before pregnancy was defined as before pregnancy, and 1 ~ 3, 4 ~ 6, and 7 ~ 10 months of pregnancy were defined as the first trimester, the second trimester, and the third trimester, respectively.
Covariates
Several factors were defined as covariates in this study. The covariates were classified as follows: maternal age (< 25, 25 ~ 29 or > 30 years); residence (urban area or rural area); family economic status (low, middle or high); mother’s education (primary school or less, junior high school or senior high school or above); passive smoking (yes or no); folate supplementation (yes or no); pregnancy-induced hypertension (PIH; yes or no); anemia (yes or no); parity (1 or ≥ 2); medication use (yes or no); number of antenatal care (ANC) visits (< 5 or ≥ 5); preterm status (yes or no); and neonatal sex (boy or girl). Maternal age was calculated from the birth date of the woman to the birth date of the child. Family economic status was calculated as follows: family adjusted adult number = family adults number + 0.5 × family children number, per capita annual household income = annual household income/family adjusted adult number. Family economic status was classified as low, middle, and high according to the upper quartile and lower quartile of per capita annual household income. Passive smoking was referred to as exposure to tobacco smoke from others for at least 15 min/d [28]. Anemia and PIH were self-reported by participants, and evaluation of anemia and PIH was performed according to whether women were diagnosed by their physicians during pregnancy. Preterm birth was defined as birth at < 37 weeks of gestation. Medication mainly included cold medicine, antibiotics, antidepressants, salicylic acid drugs, and antitussives.
Statistical analysis
All data were double-entered into a database established using EpiData3.1 (EpiData Association, Denmark). The characteristics of the subjects were expressed as the mean ± standard deviation for the continuous variables and counts (proportions) for the categorical variables. In the univariate analysis, a t-test or χ2 test was employed for comparisons between groups.
Ordinary least squares (OLS) and the QR were employed to assess the association of iron supplementation with the mean and the different quantiles (0.05, 0.10, 0.15, 0.20, 0.25, 0.30, 0.35, 0.40, 0.45, 0.50, 0.55, 0.60, 0.65, 0.70, 0.75, 0.80, 0.85, 0.90, and 0.95) of the BW, adjusted for the covariates. The graphs were plotted to visualize the changes in the effect of iron supplementation on the BW at different percentiles. The 95% CI of the regression coefficient was expressed by the shading around the curve. Considering the low intake rate of iron before pregnancy, we combined the group of women supplemented with iron before pregnancy and in the first trimester into a single group when analyzing the association of iron supplementation with the BW at different periods. Moreover, as a sensitivity analysis, the stratified analysis was conducted based on the anemia status. Stata 15.1 software (StataCorp) was used for all analyses.