We conducted this analysis using maternal plasma samples collected from a prior investigation, “The Mothers, Omega-3, and Mental Health Study,” a 3-armed prospective, double-blinded, randomized controlled trial designed to test whether EPA- or DHA-fish oil supplementation would prevent perinatal depressive symptoms among women at risk. The full details of the trial have been previously described.15,16 In brief, participants with singleton pregnancies between 12- and 20-weeks gestation were recruited. Participants were selected based on an elevated risk for depression, defined as a past history of major depressive disorder, a history of postpartum depression, or an Edinburgh Postnatal Depression Scale (EPDS) score between 9 and 19. Exclusion criteria included age <18, current major depressive disorder diagnosis, bipolar disorder or schizophrenia, substance abuse disorder, history of bleeding disorder, or clotting disorder requiring anticoagulation.15,16 Potential subjects taking antidepressant medications or omega-3 fatty acid supplements at baseline were excluded from participation in the study. However enrolled participants were allowed to initiate antidepressant medications during the course of the trial, if indicated. 16Women who met eligibility criteria and who consented to participate were randomized to one of three arms: 1) an EPA-rich fish oil supplement (1060 mg EPA plus 274 mg DHA), 2) a DHA-rich fish oil supplement (900 mg DHA plus 180 mg EPA), or 3) a soy oil placebo. Because the DHA- and EPA-rich fish oil capsules were not identical in appearance, all participants took some placebo capsules (double-dummy design). Full details of the randomization procedure and the supplements are described elsewhere. 15,16
Maternal venous blood samples were collected after a three hour fast; baseline samples were collected at the time of study enrollment at 12-20 weeks gestation (visit 1), and again between 34- and 36-weeks’ gestation (visit 3). After delivery, a sample of fetal blood was collected from the umbilical cord (visit 4). Samples were centrifuged within 12 hours of collection and plasma aliquots were stored at -70 degrees Celsius.
Stored maternal and umbilical cord plasma samples from participants of the study were analyzed using commercially-available sandwich enzyme-linked immunosorbent assay kits (EMD Millipore, St. Charles, MO) to quantify leptin and adiponectin levels according to the manufacturer’s protocols.17,18 The manufacturer’s reported within- and between-assay coefficients of variation were 3.3% and 5.7% respectively for adiponectin and 2.6% and 3.75% respectively for leptin.17,18 The only variation from the manufacturer’s recommended protocol was a dilution of fetal adiponectin samples to 1:1000 as opposed to 1:500 for maternal samples, based on higher fetal levels of adiponectin. Absorbance of both adiponectin and leptin were measured at 450 nm and 690 nm and the results compared. 25-hydroxyvitamin D levels, as well as DHA and EPA fractions in maternal and cord blood serum had been previously assayed.16,19
ETHICS:
The procedures followed were approved and conducted in accordance with the institutional review boards of the University of Michigan Medical Center, Ann Arbor, MI, St. Joseph Mercy Hospital in Ypsilanti, MI, and The University of New Mexico Health Sciences Center Human Research Protections Office. All participants gave written informed consent to participate in the study. The trial was registered on 7/7/2008 at clinicaltrials.gov: NCT00711971 under the title: “Does Fish Oil Prevent Depression in Pregnancy and Postpartum”. The secondary blood sample analyses described in this manuscript were judged exempt by the University of New Mexico Health Sciences Center Human Research Protections Office.
STATISTICS:
Demographic variables were compared among three randomized groups (EPA-rich fish oil, DHA-rich fish oil, and placebo) as means and standard deviations if continuous or ordinal scale using 1-way ANOVA and as frequencies using Fisher’s exact test. We computed medians, and interquartile ranges for each outcome parameter, adiponectin, leptin and the adiponectin:leptin ratio (ALR). Using raw (unadjusted) data, in order to test whether these parameters changed over time or were different between groups, we performed repeated measures ANOVA with group as grouping factor and visit as repeated factor. The analysis of the ALR was also performed using repeated measures ANOVA with time (enrollment before supplementation, and after supplementation) as a repeated factor and 3 groups as a grouping factor. This was an intent-to-treat analysis in that all available data were used. Outcomes adiponectin, leptin, and constituents of ALR, were analyzed similarly. Because box plot analysis of the data revealed non-normality, we also analyzed these data after square root transformation. In addition, each outcome after supplementation waswere analyzed among the 3 groups using linear regression methods (multivariable regression) with predictors being baseline measures of the given outcome and covariates (BMI at enrollment, maternal weight gain, age, and ethnicity).
As an alternative analysis we also computed the change in the maternal adiponectin, leptin, and the adiponectin:leptin ratio, and compared the change among the three treatment groups. One- way ANOVA were followed by Fisher’s post hoc t tests. We used ANCOVA to evaluate initiation of antidepressant medications during the trial as a predictor of the change in adiponectin, leptin, and the ALR.
We performed post-Posthoc analyses of the relationship of adiponectin, leptin, and the ALR with the following predictive variables: serum DHA fraction, serum and EPA fraction, plasma fractions and 25-OH vitamin D concentrations, maternal BMI at study entry, maternal weight gain, and age at study enrollment were performed using multivariable linear regression. The analysis of the effect of maternal weight gain on the outcome measures was adjusted for maternal weight at study entry.
For these exploratory analyses, the three treatment groups and two maternal time points were pooled unless otherwise specified. We calculated the effect of a one unit increase in each of the predictive variables on the ALR using the two-tailed student's t distribution.
We calculated the medians and interquartile ranges for adiponectin, leptin and the ALR in umbilical cord blood. We used raw unadjusted data, as well as square root transformation to evaluate difference according to group assignment using ANOVA. Covariate adjustments were done for birth weight and length of gestation. The association of DHA and EPA fractions, as well as 25-OH vitamin D concentrations in cord blood with measures of insulin sensitivity were analyzed by multivariable linear regression. P-values ≤ 0.05 were considered statistically significant.
SAMPLE SIZE CONSIDERATIONS:
The current study is a secondary analysis of data from a randomized controlled trial for which the primary outcome was depressive symptoms as measured by the Beck Depression Inventory.15,16 The sample size was chosen based on a hypothesized difference in the BDI score . For this secondary analysis, we performed a post- hoc power analysis to determine the power of the current study to detect a difference in maternal and cord blood adiponectin concentrations based on adiponectin values available in the literature. From Luo et al.20 we assumed that baseline adiponectin level will have a baseline standard deviation of approximately 3 micrograms/ml. Assuming a modest within subject correlation of 0.7, change scores from the enrolment (12-20 weeks) to post supplementation (34-36 weeks) should have a standard deviation of approximately 2.3 microg/ml. For this standard deviation the minimum detectable mean difference between 2 groups from a two-sample t-test with significance of 0.05 and power of 0.8 is 1.5 micrograms/ml where the sample sizes for the two groups both are 36.
For the comparisons of adiponectin, leptin, and the ALR in cord blood, we assumed an adiponectin standard deviation of approximately 8.0 micrograms/ml, based on Luo, et al.20 For a two sample t test with this standard deviation a sample size of 32 per group gives significance level of 5% and power of 80%;% the minimum detectable difference is 5.7 micrograms/ml. The proposed analysis should have superior power and a smaller detectable difference.