Participants and setting
Study participants were recruited from a prospective observational cohort of pregnant women conducted at a public health care center in the city of Rio de Janeiro, Brazil. Enrollment was open from November 2009 to October 2011 and the last follow-up visit occurred in July 2012. The cohort eligibility criteria were defined as follows: (i) being between 5 and 13 weeks of gestation at the time of enrolment; (ii) aged between 20 and 40 years; (iii) be free from any known chronic diseases (other than obesity) such as hypertension and diabetes; (iv) residing in the study catchment area; and (v) intending to continue prenatal care in the public health center.
At the first prenatal visit, a researcher invited women to participate in the cohort study. Ninety percent of eligible women agreed to participate and 299 women were enrolled in the cohort. The cohort study included four waves: at first (T0, baseline = 5–13 weeks), second (T1 = 22–24 weeks) and third trimesters (T2 = 30–32 weeks), and postpartum (T3 = 4–6 weeks’ postpartum). Baseline visit consisted of blood collection followed by a psychiatric interview using the Mini International Neuropsychiatric Interview (MINI; version 5.0.0) , general questionnaire (socioeconomic data, obstetric history, lifestyle), mental health scales (including the Edinburgh Postnatal Depression Scale - EPDS), dietary intake (food frequency questionnaire, FFQ), and anthropometric assessment (weight and height). All interviews were conducted on the same day of the routine prenatal appointment at the health care center. More details on data collection for the cohort are published elsewhere .
Trial inclusion and exclusion criteria
After cohort baseline data collection those women who reported a past history of depression or presented a depressive symptom score at baseline ≥9 based on the EPDS were considered as being at risk for PPD and potential candidates for the randomized clinical trial (RCT). Women that were depressed or presented psychotic symptoms, or had past history of mania or hipomania, were at suicidal risk, taking any psychiatric medication (as anti-depressive and anxiolytic) or being seen by a psychologist or psychiatrist were excluded. Furthermore, we also considered the exclusion of women taking any oil supplementation (as fish oil, flaxseed oil or cod liver oil) and those with fish or seafood allergy. However, no such cases were identified.
Trial design and randomization
The design comprised a RCT nested to the observational cohort. Eligible women identified after the cohort baseline were randomly assigned to receive either fish oil or placebo capsules. The supplementation started after the second follow-up visit (T1) and lasted for 16 weeks. The randomization was performed by a researcher not involved in the data collection using the participant identification (subject ID) after stratification for EPDS score and previous history of depression. Participants and all research assistants and technicians responsible for running both the cohort study and the RCT were blinded to the study allocation.
Supplementation and compliance
Participants in the RCT received 6 gelatin capsules (1 g each) per day containing either fish oil (intervention) or soybean oil (control) for 16 consecutive weeks. The capsules were identical regarding color, shape, size and packing. The fish oil capsules contained a total dose of 1.8 g of n-3 per day (1.08 g of EPA and 0.72 g of DHA). Both capsules were deodorized, and contained 0.2 mg/g of vitamin E as antioxidant. The fish oil supplement and its placebo were manufactured by Galena Nutrition® Química Industrial, São Paulo, Brazil.
Women received individual packages containing the daily dose of supplementation and were advised to take three capsules at lunch and three capsules at dinner. Within the first week, a research assistant contacted participants from both groups by phone to verify compliance and possible side effects using a standardized questionnaire. Supply of capsules was provided at week-2, week-8 (3rd trimester of follow-up visit), and week-12. Week-2 and week-12 appointments were schedule solely to provide batch of supplements and assess compliance. All participants consumed 400 μg/d of folic acid from the beginning of pregnancy, and 60 mg/d ferrous sulphate from the 2nd trimester until delivery, as provided during standard prenatal care in Brazil. Participants were asked to not alter their usual dietary habits and not consume any supplements other then the ones provided by the research team and prenatal care service.
To assess compliance, women were advised to return all empty supplement packages every visit, and a research assistant estimated the number of capsules women had taken during the interval. Additionally, serum fatty acid composition analysis was assayed as a biological marker of compliance.
Depressive symptom measurements
The first cohort follow-up (T1) was considered baseline to the RCT. At this time point, the EPDS scale was administered and repeated after 8 weeks (at third trimester visit – T2) and 4–6 weeks’ postpartum (T3 - end-point). This scale has been previously validated for use in pregnancy . Despite EPDS being considered as a screening tool rather than a diagnostic instrument for depression, this scale is widely used and it enables comparison with previous trials [23,24,25, 35]. The Portuguese version of the EPDS was validated in a sample of mothers from Pelotas, southern Brazil . According to the validation study, the cutoff of ≥9 presents 91.3% of sensitivity and 54.7% of specificity for detecting depression compared to a semi-structured interview based on the International Statistical Classification of Disease and Related Health Problems (10th Revision) administered by mental health care professionals (gold standard) . At the study end-point, for screening of moderate and severe cases of PPD, the cutoff point of ≥11 (83.8% of sensitivity and 74.7% of specificity)  was used.
Psychiatric interviews were conducted at T0 using the MINI . This instrument consist of a standard model of a brief (15–30 min) structured interview for the evaluation of the existence of Axis I psychiatric disorders according to the Diagnostic and Statistical Manual of Mental Disorders . The interview is divided into 16 modules (A-P), each one containing questions that represent different psychiatric disorders. The patients must answer ‘yes’ or ‘no’ to each of the questions; at the end of each module, and the diagnostic (yes/no) is provided. The interviews were performed by physicians and postgraduate students trained for this purpose.
Fasting blood samples (5 mL) were collected by a trained technician throughout gestation (T0, T1 and T2). Serum concentrations of EPA and DHA were used to assess compliance to the fish oil supplementation, as both fatty acids measurements are used as biological markers for dietary intake . The serum fatty acids composition was determined using a high-throughput robotic direct methylation method coupled with fast gas-liquid chromatography. The method was developed and validated by the Section of Nutritional Neurosciences, Laboratory of Membrane Biochemistry and Biophysics of the National Institute of Health (NIH, Bethesda, MD, USA), where the samples were analyzed [40, 41].
A standardized and pre-piloted questionnaire was applied at the cohort baseline to collect the following information: age (years), schooling (years), marital status (married or stable partnership/single), smoking habit (no/yes), history of depression (yes/no), alcohol consumption (no/yes), monthly per-capita family income, and parity (number of parturition).
The early pre-pregnancy body mass index (BMI, kg/m2) was determined from self-reported pre-pregnancy weight (kg), and height (m) measured at baseline (Seca Ltd., Hamburg, Germany). Anthropometric measurements were collected by trained interviewers according to standardized procedures .
A validated semi-quantitative FFQ consisting of 81 food items was administered at baseline (at 1st trimester - referring to prior 6 months of habitual diet) and at the third trimester of pregnancy (referring to dietary habits from second to third trimester) [43, 44]. At baseline, predominant dietary sources of long-chain n-3 PUFAs identified were fresh fish, canned tuna/sardines, eggs and red meat .
Gestational age was initially determined based on the reported date of the last menstrual period, and lately confirmed with the first ultrasound performed before 24 weeks of gestation. Obstetric outcomes such as length of gestation and birth weight were extracted from obstetric records.
The original sample size calculation was based on a prevalence of postpartum depression ranging from 26.9 to 39.9% according the previous studies conducted in similar population [17, 46]. We designed this study to detect a 5% reduction in the prevalence of EPDS ≥11 from baseline of 27% to 22% after the study intervention, compared with controls. The required sample size ranged from 22 to 43 for each treatment group (fish oil and control). Assuming an attrition rate of 15% due to dropouts and gestational complications, 25 to 51 participants were needed in each treatment group.
The prevalence of depressive symptoms (EPDS score ≥ 11) was the primary outcome, while the mean and absolute changes in EPDS score and length of gestation, and birth weight were the secondary outcomes.
The data distribution was analyzed according to the Shapiro-Wilk test. Socio-economic, demographic, nutritional and gestational outcome characteristics of women in both fish oil and control groups were described using means and standard deviations or medians and interquartile ranges for symmetric and asymmetric variables, respectively. Normal distributed variables were analyzed using Student t-test or qui-square test and asymmetric variables were analyzed using Mann-Whitney U test.
We compared the serum fatty acid composition between fish oil and control groups at T0 and T1, and tested the compliance with fish oil supplementation at T2. Comparisons were performed using Mann Whitney U test.
The prevalence of depressive symptoms and mean score of EPDS were compared at T0, T1, T2 and T3 between fish oil and control groups using Mann-Whitney U test, Fisher exact test or chi-square. The absolute change (Δ) in EPDS score between end-point (T3) and T0 (cohort baseline) and between end-point (T3) and T1 (trial baseline) were tested using Mann-Whitney U test.
We performed two different longitudinal regression procedures. First, a random-intercept and slope logistic regression model having longitudinal changes in depressive symptoms (EPDS score ≥ 11/<11) as the outcome and the intervention groups as the main exposure was performed. Secondly, linear mixed-effect (LME) regression models were used to evaluate the effect of fish oil supplementation on EPDS continuous score change during pregnancy and postpartum. In the LME models, the time (weeks elapsed after conception) was included in all the models as random effects to adjust for the overall and individual variations in the EPDS score over time.
The longitudinal regression model 1 included interactions terms between groups of intervention (fish oil vs. placebo) and time [random-intercept logistic regression = 2nd trimester (reference category) vs. 3rd trimester or postpartum; LME = continuous weeks elapsed after conception] in order to evaluate if the EPDS scores (continuous or categorical) change over time were different depending on the supplement group. Results from model 2 were further adjusted for the following confounders: parity, education, skin color and early pre-pregnancy BMI.
All analyses were performed using both the pre-protocol and intention-to-treat (ITT) procedures. The ITT population included all women randomized, regardless of whether or not they dropped out or fully adhered to the intervention, while the per-protocol population included all women who completed the trial at each time point.
The analyses described above were repeated to compare results considering a subgroup composed of women with previous history of depression.
Statistical analyses were performed using STATA version 12.0 (Stata Corp., College Station, Texas, USA). Values were considered statistically significant when the p-value was lower than 0.05.