Effects of a lipid-based nutrient supplement during pregnancy and lactation on maternal plasma fatty acid status and lipid profile: Results of two randomized controlled trials

It is unknown whether a novel small-quantity lipid-based nutrient supplement (SQ-LNS) containing alpha-linolenic (ALA) and linoleic acids impacts maternal plasma lipids and fatty acid status. We measured plasma fatty acids (wt%) and lipid concentrations at 36 wk gestation and breast milk fatty acids (wt%) at 6 months postpartum in a subsample of women enrolled in a randomized controlled trial studying the effects of SQ-LNS on birth outcomes and child growth. Women≤20 wk gestation in Ghana (n=1,320) and Malawi (n=1,391) were assigned to receive daily either: 1) iron-folic acid (pregnancy); 2) multiple micronutrients (pregnancy and lactation); or 3) SQ-LNS (pregnancy and lactation). At 36 wk, plasma ALA levels were higher in those receiving SQ-LNS. SQ-LNS increased breast milk ALA in Ghana but not Malawi. There was no effect on plasma lipids or other selected fatty acids. SQ-LNS may impact plasma and breast milk ALA levels depending on the population. This was a sub-study of participants from two randomized controlled trials conducted in Malawi and Ghana as part of the International Lipid-Based Nutrient Supplements (iLiNS) Project (www.ilins.org). The primary objective of these trials was to determine the effect of SQ-LNS, provided during pregnancy, lactation, and early childhood, on child growth at 18 months of age, as compared with IFA provided during pregnancy or MMN provided to the mother during pregnancy and the first six months postpartum. Details of the study methods have been reported elsewhere [18], [20], [21]. Briefly, the study teams in Ghana and Malawi recruited women attending prenatal care visits at four health facilities in semi-urban areas of the Yilo Krobo and Lower Manya Krobo districts about 70 km north of Accra, Ghana between December 2009 and December 2011, and four health facilities in the rural Mangochi district in southern Malawi between February 2011 and August 2012. While the trials were similar in design, each trial operated independently and there were some differences in inclusion/exclusion criteria. In Ghana, women were eligible if they were: ≤20 wk gestation (confirmed by ultrasound), ≥18 y of age, had a completed antenatal health card, and signed or thumb-printed informed consent. We excluded women if they were: HIV positive, had asthma, epilepsy, tuberculosis, a chronic disease that required medical attention, did not reside in the defined catchment area, had a milk or peanut allergy, or were participating in another clinical trial. The Malawi trial had similar exclusion and inclusion criteria, however women in Malawi were eligible if they were ≥15 y of age and were not excluded if they were HIV positive. A daily iron and folic acid capsule is the standard of care during pregnancy in both countries. In Malawi, nutritional supplementation before or after pregnancy is not common. In Ghana, women continue to receive the iron and folic acid capsule 6 wk after delivery and receive high dose vitamin A within 8 wk after delivery, and the use of dietary and herbal supplements in the study setting is common. Women were enrolled in the trials at a mean of 16–17 wk gestation and received supplementation through pregnancy and until 6 months postpartum, which represents an average of 56 wk of supplementation. In both trials, pregnant women were randomized to receive one of the following three daily treatments: 1) iron and folic acid (IFA, a capsule consisting of 60 mg iron and 400 µg folic acid, received during pregnancy only); 2) multiple micronutrients (MMN, a capsule consisting of 18 vitamins and minerals [including 20 mg iron], received during pregnancy and the first 6 months postpartum); or 3) SQ-LNS (a 20 g sachet which included the same 18 micronutrients as the MMN capsule plus four additional minerals: calcium, phosphorus, potassium, magnesium) (see Table 1 for nutrient content of each supplement). Both the IFA and MMN groups were considered control groups. Group allocations were determined by a statistician who used a computer-generated randomization scheme in blocks of 9 (3 codes for each of the 3 interventions) and codes were placed in sealed opaque envelopes. A woman chose an envelope from a stack of envelopes (6 and 9 envelopes per stack in Malawi and Ghana, respectively) to determine her group allocation and received her first 2-wk supplement ration at this time. Field workers made home visits biweekly, during which they delivered the supplements and collected information on the participant’s adherence to the study intervention. Adherence was assessed by maternal report as well as by counting the numbers of unconsumed capsules or sachets. This was a partially-blinded trial, as it was not possible to blind the fieldworkers and study participants to those consuming capsules vs. SQ-LNS (because of the starkly different characteristics). Nutrient composition of supplements used in the study: iron and folic acid (IFA) capsule, multiple micronutrient (MMN) capsule, and small-quantity lipid-based nutrient supplement (SQ-LNS). In Ghana, from the 1,320 women enrolled in the trial, 510 were excluded from the present analysis due to an error in the labeling of the IFA and MMN supplements, resulting in mixed exposure [18]. From the remaining 810 women, enrolled from October 2010 to December 2011, 369 women were randomly selected for analysis of blood lipids and fatty acids (Supplementary Fig. 1). In Malawi, from the 1,391 women enrolled in the trial, 315 women were randomly selected for analysis of fatty acids and all 1,391 for blood lipids (Supplementary Fig. 2). The institutional review boards at the College of Medicine Research, University of Malawi and the Ethics Committee of Pirkanmaa Hospital District, Finland approved the study protocol for the trial in Malawi. The institutional review boards at the University of California, Davis; the Noguchi Memorial Institute for Medical Research, University of Ghana; and the Ghana Health Service approved the study protocol for the trial in Ghana. Both trials were registered at www.clinicaltrials.gov (IDs: {“type”:”clinical-trial”,”attrs”:{“text”:”NCT01239693″,”term_id”:”NCT01239693″}}NCT01239693, {“type”:”clinical-trial”,”attrs”:{“text”:”NCT00970866″,”term_id”:”NCT00970866″}}NCT00970866). At both sites, study nurses collected venous blood samples from women at enrollment and 36 wk gestation into a heparin-treated, trace element-free, Sarstedt Monovette tube. Because of the difficulty of obtaining fasting blood samples from pregnant women, non-fasting blood samples were collected. However, we collected information regarding time between last meal and sample collection. Lab technicians centrifuged the blood samples at 4000 rpm for 15 min to obtain plasma. Breast milk was collected at 6 months postpartum. In Ghana, study nurses assisted women in collecting a 10–20 mL mid-stream breast milk sample during a follow-up clinic visit. In Malawi, women expressed a full milk sample from a single breast during a home visit. A trained field worker then mixed the breast milk and collected a 10 mL sample, with the remaining milk provided to the infant by spoon. All samples were stored at −20 °C within 24 h of collection and moved to −80 °C for longer term storage. Plasma samples from the Ghana trial were analyzed in Accra, Ghana for total cholesterol, HDL-C, and triglyceride concentrations using a Flexor Junior Chemistry Analyzer (Vital Scientific, Dieren, Netherlands). LDL-C was calculated using the Friedewald equation: LDL-C=total cholesterol–(HDL-C)–(triglycerides/5), mg/dL [25]. This is accepted as an accurate method of determining LDL-C concentration, as long as triglyceride concentration is not ≥400 mg/dL [25], [26]; all of our samples were below this level. Plasma samples from the Malawi trial were shipped on dry ice to Davis, CA, where lab technicians quantified total cholesterol and triglyceride concentrations by enzymatic colorimetric assay using a Cobas Integra 400 plus automatic analyzer (Roche Diagnostic Corp., Indianapolis, IN). All lab technicians were blinded to the intervention groups. Plasma and breast milk samples from both trials were shipped to OmegaQuant Analytics (Sioux Falls, SD) for analysis of fatty acids. Fatty acid composition was analyzed by gas chromatography with flame ionization detection. Plasma or breast milk was added to a mixture of solvents (methanol containing 14% boron trifluoride: toluene: methanol; 35:30:35 v/v/v, all from Sigma-Aldrich, St. Louis, MO). The tube was vortexed and heated in a hot bath at 100 °C for 45 min. After cooling, hexane (EMD Chemicals, USA) and distilled water were added. The sample was vortexed and centrifuged, and then an aliquot of the hexane phase was analyzed by gas chromatography using a GC-2010 (Shimadzu Corporation, Columbia, MD) equipped with a SP-2560, 100-m fused silica capillary column (0.25 mm internal diameter, 0.2 µm film thickness; Supelco, Bellefonte, PA). Fatty acid composition was expressed as a percent by weight (wt%) of total identified fatty acids. All lab technicians were blinded to the intervention groups. At enrollment, anthropometrists measured weight and height using high-quality scales and stadiometers. We calculated body mass index (BMI = kg/m2). Lab technicians assessed malaria and HIV infection status with rapid tests. Concentrations of inflammatory biomarkers C-reactive protein (CRP, mg/L) and alpha-1 glycoprotein (AGP, g/L) were measured in plasma by immunoturbidimetric assay using the autoanalyzers specified earlier. Trained interviewers collected socioeconomic and demographic information at a follow-up home visit. For analysis of fatty acids and lipid concentrations, we used a minimum effect size (Cohen’s d) of 0.5 to calculate sample size, assuming a two-sided α=0.05% and 80% power, requiring a subsample of 79 per group (total n=237). Allowing for attrition and women missing a sample at either baseline or 36 wk gestation, 369 women were randomly selected from the 810 women in Ghana enrolled after October 1, 2010. In Malawi, 315 women were randomly selected from women with blood samples at both baseline and 36 wk gestation and a breast milk sample at 6 months postpartum. The effect of the intervention on cholesterol and triglyceride concentrations was examined in the full sample in Malawi and in the subsample in Ghana. Statistical analysis was performed according to intention-to-treat and focused on examining differences between the three intervention groups in plasma fatty acid levels and lipid concentrations at 36 wk gestation and breast milk fatty acid levels at 6 months postpartum. The primary fatty acids of interest were ALA and LA. Secondarily, to determine if there was any impact on LCPUFAs, we also examined the effect of supplementation on DHA, EPA, AA, the sum of DHA and EPA, the sum of all long chain omega-3 fatty acids (DHA, EPA, and DPA), and the following ratios: LA:AA, ALA:DHA, AA:EPA, and omega-6 fatty acids:omega-3 fatty acids. We limited our analyses to these fatty acids and ratios as indicators of LCPUFA metabolism and due to their potential effects on infant growth and neurodevelopment [1], [3]. Cholesterol and triglycerides were analyzed as concentrations, and fatty acids were analyzed as percentage of total fatty acids (by weight). Logarithmic transformation of all fatty acid variables and triglyceride concentration was applied to approximate a normal distribution of the data which was evaluated using the Shapiro-Wilk test. Fatty acids were dichotomized into high or low values using a median cut-point and low cholesterol was defined as <10th percentile of the IFA group at 36 wk gestation. We also used the following clinical definitions: high total cholesterol (≥240 mg/dL), high LDL-C (≥160 mg/dL), and low HDL-C (<50 mg/dL) [27], [28]. We used the Household Food Insecurity Access Scale [29] to estimate food insecurity and created scores using standard criteria adjusted for the month of collection. An asset index was created using principal components analysis [30] based on household ownership of a set of assets (radio, television, cell phone, bed, mattress, bed net, and bicycle), lighting source, drinking water supply, sanitation facilities, and flooring materials. We evaluated the effect of the nutritional intervention on lipids and fatty acids in plasma using ANCOVA (continuous outcomes) and logistic regression (binary outcomes) models, using the Tukey-Kramer adjustment for multiple comparisons and p<0.05 indicating statistical significance. We performed analyses both with and without covariates, as guidelines for best statistical practices support the use of covariates in analyses of randomized controlled trials [31]. All models included the baseline value for the outcome variable. This is mathematically the same as testing for a difference in the change in the outcome (between baseline and 36 wk) between the three groups. We used similar statistical analyses for fatty acids in breast milk, although those models could not include a baseline value, and we only had one time point for breast milk sample collection. Pooled analyses included trial site in all models. We considered covariates for inclusion in the model if they were associated with the outcome variable at p<0.1. Potential covariates were selected from previous literature and stated in a predefined analysis plan; they included baseline measurement of the outcome variable, gestational age, maternal age, education level, BMI and height, season of enrollment, malaria infection, HIV status (Malawi only), inflammatory markers, household food insecurity, asset index, parity, site of enrollment, and time since last meal. To determine whether a pooled data analysis could be conducted, we tested for interaction between study site and intervention group. We also tested for interaction between the intervention group and maternal age, parity, and baseline BMI. All interactions were evaluated in linear regression models and interaction term p-values<0.1 were considered to be statistically significant. We compared baseline characteristics between women in each of the three intervention groups, as well as between those with complete data and those with missing data at 36 wk gestation, as this time point has the largest proportion of missing data. In sensitivity analyses, we performed unadjusted analyses using the same sample that was used in adjusted analyses, and considered a >10% change in results to be indicative of bias. We also performed another sensitivity analysis to compare those receiving LNS with those not receiving LNS (IFA group and MMN group combined) for all outcomes. Model assumptions were checked using standard regression diagnostics for linearity, normality, leverage, and influence. All analyses were performed using SAS 9.4 (SAS Institute, Cary, NC).