Study design and recruitment. We conducted a longitudinal birth-cohort study from mid-to-late pregnancy through the children’s first two years of life. Mother-child pairs were recruited from antenatal clinics of 2 hospitals in Shanghai from February to September 2010. Study data were collected via in-person interviews with the pregnant women during mid-to-late gestation (at 28-36 weeks of gestation) and with the mothers at 24-36 months postpartum, and via medical history records at the 12th week of gestation (pre-pregnancy height and weight information) and at delivery (birth information). Singleton pregnant women who had no mental disorders and who were without severe complications of pregnancy were considered eligible. The mother-infant pairs were also excluded if the infant had a clinical diagnosis of perinatal asphyxia (Apgar scores at 5 minutes were 7 or lower). A total of 398 pregnant women who visited the antenatal clinics were invited to participate, and 173 mother-child pairs were out of contact or declined to participate in the follow-up interview. Therefore, the study sample included 225 mother-child pairs. There were no significant differences in maternal age, ethnicity, education or economic status between the mother-child pairs who were followed up and those who were not(28).
The study was approved by the Medical Research Ethics Committee of Shanghai Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, and written informed consent was obtained from all study participants.
Exposures. The pre-pregnancy BMI and GWG were the main exposure variables. Pre-pregnancy weight was self-reported, and maternal height and weight were also measured at the first prenatal visit at the 12th week of gestation. The height was measured twice by the same investigator to the nearest 0.1 cm, using medical height meter (SECA 799, Hangzhou, China), with the subjects standing without shoes. The weight was also measured twice by the same investigator to the nearest 0.1 kg, using a medical calibrated weighing scale (SECA 799, Hangzhou, China), with subjects only wear underclothing. If readings differed by more than 1cm and 0.2 kg for the height and weight, respectively, a third measurement was taken and the average of all three readings were used. If the difference between self-reported and measured weight was over 2kg, the woman would be inquired again to assure that her self-reported pre-pregnancy weight was correct. Pre-pregnancy BMI was calculated based on pre-pregnancy weight and maternal height. Pre-delivery weight was obtained from medical record at delivery. GWG was calculated as pre-delivery weight minus pre-pregnancy weight.
Outcome. The primary outcome was the duration of breastfeeding (the number of months breastfeeding lasted). An in-person interview was performed by trained researchers at 24-36 months postpartum, and women were asked questions about their infants’ feeding type (formula-feeding, breast-milk-feeding or mixed feeding) and the duration of breastfeeding ("did you ever breastfeed your child?", "at what age of the infant did you stop breastfeeding?", "did you ever introduce another kind of milk than breast-milk to your child?", "at what age of the infant did you introduce another kind of milk than breast-milk to your child?"). In this study, breastfeeding duration referred to the duration of any breastfeeding, that is to say, the number of months the infant received any breast-milk, irrespective of the concomitant introduction of other milk and foods. The breastfeeding duration among mothers who never breastfed their children was defined as "zero" month.
Covariates. Information on covariates was collected from the interviews during mid-to-late pregnancy and at 24-36 months postpartum, and from medical record at delivery. Because of possible changes between the pregnancy and postpartum periods, information on marital status, family monthly income, housing area per capita, maternal occupation, employment, workloads (daily working hours), smoking and alcohol consumption, and stress was repeatedly collected during pregnancy and at 24-36 months postpartum. Other information collected prenatally included maternal age at enrollment, ethnicity and maternal education. Information collected at birth included children’s gender, birth weight, length, head circumference, parity, gestational weeks, and delivery mode. Whether or not the women had caesarean-section (C-section) delivery without medical indications was based on the medical records at delivery, and would be confirmed by the mother herself at 24-36 months postpartum. Information collected at 24-36 months postpartum also included days the mother living with her child per week, family interpersonal relationship, and family atmosphere.
Maternal emotional stress levels were assessed using Symptom Checklist-90-Revised (SCL-90-R) during mid-to-late pregnancy and toddlerhood. The SCL-90-R is widely used, valid and reliable in assessing maternal psychological symptoms (29). The global severity index (GSI) is calculated by dividing the total score of SCL-90-R by 90. Higher GSI scores indicated higher emotional distress levels (30).
Statistical analyses. We first analyzed maternal and infant demographic characteristics and their unadjusted association with breastfeeding duration. Chi-square test was used to compare the difference in feeding types among mothers with different delivery modes.
Because the applicability of GWG grouping based on the 2009 IOM recommendation may be not appropriate for Chinese pregnant women (30), we did not define excessive or inadequate GWG according to current grouping criteria recommended by the IOM, but used GWG and pre-pregnancy BMI as continuous variables. We first applied smoothing splines to visually inspect the shape of the associations of pre-pregnancy BMI and GWG with breastfeeding duration, then according to the shapes of the associations (non-linear associations were observed), two-piece-wise linear regression models were fitted to describe the effects of maternal pre-pregnancy BMI and GWG on the breastfeeding duration with the package ‘segmented’ in R software (http://www.R-project.org). The turning point of BMI and GWG where the relationship between BMI, GWG and breastfeeding duration started to change and became eminent was determined using trial method, which was to move the trial turning point along the pre-defined interval and picked up the one which gave maximum model likelihood. For all models using breastfeeding duration as the outcome, confounders were selected based on the observed bivariate associations with the outcome (breastfeeding duration) or the exposures (pre-pregnancy BMI or GWG), or from the literature. Because only five mothers were from ethnic minorities, and all mothers were married and didn’t smoke, therefore, ethnicity, marital status and smoking were not controlled in the models.