The sample was taken from the Norwegian Mother and Child Cohort Study (MoBa), a prospective population-based pregnancy cohort study performed by the Norwegian Institute of Public Health . Except two hospitals, all hospitals and maternity units in Norway with more than 100 births annually, altogether 50 units, were included . There were no exclusion criteria for the women to be included into the study; all pregnant women were eligible. Expecting mothers were invited to join the study through postal invitation in connection with a routine ultrasound examination offered to all pregnant women in Norway at gestation weeks 17–18, and 42.7% agreed to participate. The assessment points were at 17 weeks gestation, 30 weeks gestation, and six months postpartum. At these time points the mothers were sent questionnaires containing questions on their physical health, mental health, nutritional status, and demographic status. For those invited to the study, the response rate during pregnancy ranged from 92% to 95%, and the response rate at six months postpartum was 87%. The MoBa is ongoing, and the MoBa-study group annually releases quality-assured data files. The current study was based on version three of the quality-assured data files released for research in 2007. Informed consent was obtained from each participant. The MoBa has been granted by the Norwegian Parliament  has a license from the Norwegian National Data Inspectorate (license 01/4325). Furthermore, the study was considered by the regional committee for ethics in medical research for South-eastern Norway, and received a positive ethical consideration (S-97045; S-95113). In addition to questionnaire data, we retrieved medical information on parturition for the present study from the Medical Birth Registry of Norway (MBRN). This registry contains information about all births in Norway .
At the time of the present study, 47 659 mothers had returned the questionnaires at both 30 weeks gestation and six months postpartum. This number included only the last enrollment of a mother in the study and excluded previous participation with earlier pregnancies. Moreover, in the case of plural births, we included only one of the twins to avoid dependence between observations. Among the 47 659 mothers 5434 had missing data on one or more variables. Missing data among these mothers was distributed as follows: maternal age, 3; breast milk and solids, 3508; plural birth, 240; preterm birth, 33; daily smoking, 2045. The 5434 mothers with missing data were excluded from the analysis.
The variables included in the data analysis were primiparity, plural births, cesarean sections, and gestational age (based on ultrasound examination) from the MBRN. When information was missing in the MBRN, we used self reported data from the questionnaire at six months postpartum. We collected information on symptoms of anxiety and depression, breastfeeding, introduction of solids, and daily smoking from the questionnaires. We defined daily smoking as smoking one or more cigarettes per day during the first six months postpartum. Preterm birth was defined as parturition before 37 weeks gestation.
We assessed maternal anxiety and depression symptoms at 30 weeks gestation and six months postpartum using a short version of the Hopkins Symptom Checklist (SCL-8) . The SCL-8 is an 8-item self report instrument designed to assess psychological distress, in particular anxiety and depression. The response categories range from one to four (not bothered to very much bothered). The SCL-8 had internal consistencies of α = 0.84 at 30 weeks gestation and α = 0.86 at six months postpartum. We computed the average score across the eight items. For the multivariate analyses, we standardized the scores. Each interval then represents a shift of one standard deviation.
The introduction of and sustainment of breastfeeding, bottle feeding, and solids was reported by the mothers month by month at six months postpartum. Breastfeeding was categorized into three groups: predominant breastfeeding, mixed breastfeeding, and bottle-feeding. This is largely in accordance with the classification system of the World Health Organization. This categorization is described in detail in an earlier study on breastfeeding derived from this cohort . Predominant breastfeeding is when the infant’s predominant source of nutrition is breast milk. Mixed breastfeeding is continued breastfeeding up to six months postpartum, supplemented by formula or solids. Bottle-feeding referred to those mothers who stopped breastfeeding completely and used only milk supplementation and solids.
We imputed missing data on the SCL-8 by applying the estimation-maximization algorithm . The imputed data was used across all analyses.
To investigate main effects and interaction effects, we used linear regression with several blocks, always keeping variables from previous blocks. Mixed breastfeeding and bottle-feeding comprise together the alternatives to full breastfeeding. Therefore they were added together in each block. The first block was the main effects of mixed breastfeeding and bottle-feeding on symptoms of anxiety and depression at six months postpartum. In the second block we investigated the effect of mixed breastfeeding and bottle-feeding on the change in symptoms of anxiety and depression by introducing the anxiety and depression symptom score assessed at 30 weeks gestation. In the third block, we investigated any additional interactive effects over and beyond the additive effects, and introduced the interaction term between mixed breastfeeding and symptoms of anxiety and depression at 30 weeks gestation and the interaction term between bottle-feeding and symptoms of anxiety and depression at 30 weeks gestation. In the forth block, we introduced adjusting variables related to events happening from parturition to six months postpartum (i.e. cesarean sections, primiparity, plural births, preterm births, and daily smoking). All adjusting variables were entered in a single block to investigate if they as a whole could account for the associations in the previous blocks. The blocks were expanded by the following theoretical rationale: First block, to establish that breastfeeding cessation is associated with post partum symptoms of anxiety and depression; second block, to be the cause of increase in symptoms of anxiety and depression, breastfeeding cessation must be associated with post partum symptoms of anxiety and depression after adjusting for prepartum symptom level; third block, if women with high prepartum levels of anxiety and depression are more vulnerable to the detrimental effects of breastfeeding cessation, there must be an interaction effect between prepartum symptom level and breastfeeding cessation; fourth block, if the interaction effect could merely be a consequence of other contextual factors known to influence breastfeeding, therefore a handful of these factors were adjusted for.
To limit multicollinearity, increase normality, and enhance interpretability, the SCL-8 scores were centered, log-transformed, and standardized. We used an alpha of 0.05 across the analyses.