Maternal cortisol and stress are associated with birth outcomes, but are not affected by lipid-based nutrient supplements during pregnancy: An analysis of data from a randomized controlled trial in rural Malawi

Background: Prenatal micronutrient supplements have been found to increase birth weight, but mechanisms for increased growth are poorly understood. Our hypotheses were that 1) women who receive lipid-based nutrient supplements (LNS) during pregnancy would have lower mean salivary cortisol concentration at 28 wk and 36 wk gestation compared to the multiple micronutrient (MMN) and iron-folic acid (IFA) supplement groups and 2) both salivary cortisol and perceived stress during pregnancy would be associated with shorter duration of gestation and smaller size at birth. Methods: Women were enrolled in the trial in early pregnancy and randomized to receive LNS, MMN, or iron-folic acid (IFA) supplements daily throughout pregnancy. At enrollment, 28 wk and 36 wk gestation, saliva samples were collected and their cortisol concentration was measured. Self-report of perceived stress was measured using questionnaires. Gestation duration was indicated by ultrasound dating and newborn anthropometric measurements (weight, length, head circumference) provided indicators of intrauterine growth. Results: Of the 1391 women enrolled in the trial, 1372, 906 and 1049 saliva samples were collected from women at baseline, 28 wk and 36 wk, respectively. There were no significant differences in mean cortisol concentrations by intervention group at 28 wk or 36 wk gestation. Cortisol concentrations were negatively associated with duration of gestation (Baseline: β = -0.05, p = 0.039; 36 wk: β = -0.04, p = 0.037) and birth weight (28 wk: β = -0.08, p = 0.035; 36 wk: β = -0.11, p = 0.003) but not associated with length-for-age or head circumference-for-age z-scores. Perceived stress at 36 wk was significantly associated with shorter newborn LAZ (p = 0.001). There were no significant associations with the risk of small for gestational age, preterm birth, or low birth weight. Conclusions: Maternal salivary cortisol concentration was strongly associated with birth weight and duration of gestation in rural Malawi, but these data do not support the hypothesis that LNS provision to pregnant women would influence their salivary cortisol concentrations. Trial registration: Clinicaltrials.gov identifier NCT01239693. To evaluate these questions, we nested a substudy within the iLiNS trial in Malawi. Details of the overall study design methods and primary outcome results have been published previously [15]. In brief, the target population included pregnant women attending antenatal care through one of four hospitals or health facilities in Mangochi District in southern Malawi. Inclusion criteria were as follows: ≤20 wk gestation confirmed by ultrasound, residence in the defined catchment area, availability during the period of the study, and signed or thumb-printed informed consent. Exclusion criteria were: age <15 years old, need for medical attention due to a chronic or severe illness, diagnosed and medically treated asthma, history of peanut allergy, history of anaphylaxis or serious allergic reaction to any substance, pregnancy complications evident at enrolment visit (moderate to severe edema, blood Hb concentration < 50 g/l, systolic blood pressure (BP) > 160 mmHg or diastolic BP > 100 mmHg), earlier participation in the trial during to a previous pregnancy or concurrent participation in any other clinical trial. A statistician independent of the research group generated randomization codes by creating four unique lists (one for each enrollment site) in blocks of nine (3 codes for each of the 3 interventions). The codes were inserted into individual opaque envelopes and eligible participants selected one from a shuffled stack of 6 envelopes. This code determined both the participant’s group allocation as well as her identification number. Based on their randomization code, women received one of three supplements to be consumed daily throughout pregnancy: 1) small quantity LNS; 2) MMN; or 3) IFA. The nutrient content for each of the supplements can be found in Table 1. Data collectors delivered supplements every two weeks to participants and they advised women to consume the supplements daily either after a meal (IFA or MMN groups) or mixed with a meal (LNS). Data collectors monitored adherence at each distribution visit by counting any unused supplements from participants. Nutrient content of the supplements At enrollment, trained nurses confirmed pregnancies and gestational age estimates using ultrasound imagers (EDAN DUS 3 Digital Ultrasonic Diagnostic Imaging System, EDAN Instruments, Inc., Shekou, Nanshan Shenzhen, China). Study nurses were trained in ultrasound assessment by two study physicians and they conducted all measurements in duplicate. Anthropometrists measured in triplicate maternal height using a stadiometer (Harpenden stadiometer, Holtain Limited, Crosswell, Crymych, UK), weight using a flat scale (SECA 874 flat scale, Seca GmbH & Co., Hamburg, Germany), and mid-upper arm circumference (MUAC) using a non-stretchable tape (Weigh and Measure, LLC, Maryland, USA). Research nurses tested for malaria using rapid diagnostic tests (Clearview Malaria Combo, British Biocell International Ltd., Dundee, UK), haemoglobin concentration (HemoCue AB, Angelholm, Sweden), and HIV (Alere Determine HIV-1/2, Alere Medical Co., Ltd., Chiba, Japan). Positive HIV tests were repeated using another whole blood antibody test (Uni-Gold HIV, Trinity Biotech plc, Bray, Ireland). During a follow-up home visit, trained interviewers asked mothers about demographic and socioeconomic characteristics, including questions on household food insecurity. Mothers were asked to return to the clinic for repeat visits at 32 and 36 wk gestation. A follow-up home visit was also conducted at 28 wk gestation. At enrolment, 28 wk, and 36 wk gestation, interviewers asked women about stress during the previous month using the 10-item Perceived Stress Scale [40], a tool that has been used in other low-income settings [41, 42]. The research nurses collected saliva samples during the clinic visits at baseline and 36 wk gestation and during the 28 wk home visit between 8 am and 4 pm after a 30 minute fast. Time of collection, time of waking, and time of last food or drink were recorded by the nurse. Saliva collection occurred before any other measurements or sample collection. Nurses asked each woman to place an inert polymer cylindrical swab (Salimetrics Oral Swab) under her tongue for approximately two minutes, while moving her tongue and jaw as if she were chewing to stimulate saliva. The woman removed the swab and placed it in a capped tube and then it was refrigerated or placed on ice packs. Swabs were brought to room temperature, then centrifuged for 15 min at 3,000 RPM (1500 x g) to extract saliva, which then was frozen at −20 °C. After a maximum of 2 days, samples were transferred to a −80 °C freezer for longer term storage. Samples were shipped to Davis, CA for analysis. Lab technicians measured cortisol concentrations in duplicate using an ELISA method (expanded range high sensitivity salivary cortisol kit, Salimetrics, State College, PA), which can detect cortisol concentrations ranging from 0.193 to 82.77 nmol/L (0.007-3.0 μg/dL). The intra- and inter-assay coefficient of variability is 3.5 % and 5.1 %. The mean of each duplicate measure was used for analysis. Research nurses collected venous blood samples at baseline and 36 wk gestation using 7.5 mL trace mineral-free syringe (Sarstedt Monovette, Nh4-heparin, Sarstedt Inc., Newton, NC). Lab technicians measured zinc protoporphyrin in washed red blood cells within 30 hr of collection using a hematofluorometer (Aviv Biomedical, Lakewood, NJ). They also measured soluble transferrin receptor, c-reactive protein (CRP), and alpha-1-acid glycoprotein (AGP) by immunoturbidimetry on the Cobas Integra 400 system (F. Hoffmann-La Roche Ltd, Basel, Switzerland). Research assistants measured infants’ weight as soon as possible after birth either at home or in the health center. Of the recorded birth weights, 89 % were measured within 48 h of delivery while the remainder were back-translated from a measurement within 14 days. They also collected early neonatal measurements, including length to the nearest 1 mm using an infantometer (Harpenden Infantometer, Holtain Limited, Crosswell, Crymych, UK), weight to the nearest 10 g using an infant scale (SECA 381 baby scale, Seca GmbH & Co., Hamburg, Germany), head circumference and arm circumference using non-stretchable tape. Women provided written informed consent or indicated their consent to participate in the study with a thumbprint. In Malawi, individuals ≥15 y are able to provide consent themselves and so parental consent was not obtained. The institutional review boards at the College of Medicine Research and Ethics Committee (COMREC), University of Malawi and the Ethics Committee of Pirkanmaa Hospital District, Finland reviewed and approved the trial protocols at all of the hospitals and health facilities. Sample size estimates for the main trial were calculated to be 370 per group, based on an effect size (difference between groups divided by the pooled SD) of 0.23, assuming a two-sided α = 0.05 and β = 0.2. That would correspond to a detectable difference of 0.83 nmol/L in cortisol and a 1.2 point difference in the perceived stress score (PSS). We used standard scoring methods to calculate the PSS [40]. We checked the salivary cortisol and PSS for normality using the Shapiro-Wilk test and cortisol was log transformed. We calculated Pearson’s correlation coefficients to compare log cortisol and the PSS at each time point. We also analyzed PSS and cortisol categorically. PSS was dichotomized into high or low values using a median cut-point and cortisol concentrations were grouped into quartiles based on the distributions at each measurement point (enrolment, 28 wk, 36 wk). We used the Household Food Insecurity Access Scale [43] to estimate food insecurity and created the scores using standard criteria. An asset index was created using principal components analysis [44] based on household ownership of a set of assets (radio, television, cell phone, bed, mattress, bednet, and bicycle), lighting source, drinking water supply, sanitation facilities, and flooring materials. For all analyses, participants were included if they had non-missing data on either cortisol or the perceived stress score at any time-point. We compared characteristics for those with complete data vs. those who were missing data on cortisol at 28 wk gestation. We also compared baseline characteristics between women in each of the three intervention groups. To evaluate the effect of the nutritional interventions on cortisol and PSS, we tested group-wise differences using ANOVA and ANCOVA models, using the Tukey-Kramer adjustment for multiple comparisons, and p-values <0.05 were considered statistically significant. We considered covariates for inclusion in the model if they were significantly (p < 0.1) associated with salivary cortisol. These included baseline cortisol, age, gestational age, maternal BMI and height, season, malaria infection, HIV status, hemoglobin, iron status, inflammatory markers, household food insecurity, asset index, parity (primiparous or multiparous), infant sex, site of enrollment, and maternal PSS. We included time since waking and time since last meal in all models, regardless of their association with the outcome variables. Interaction terms were created by the cross product of the intervention group and maternal age, parity, baseline BMI, and infant gender and these were evaluated in linear regression models. Interaction term p-values <0.1 were considered to be statistically significant. To examine the associations between cortisol or PSS and birth outcomes (duration of gestation, weight-for-age z-score [WAZ], length-for-age z-score [LAZ], head circumference z-score [HCZ]), we used linear regression models and present standardized regression coefficients. We used Poisson regression models with robust estimation of the standard errors to estimate relative risk for dichotomous birth outcomes, including preterm birth (<37 wk gestation), low birth weight (<2.5 kg, LBW), stunting (LAZ < −2), small head circumference (HCZ < −2), and small for gestational age [45]. We considered covariates for inclusion into the models based on previous literature and tested as described above. Because cortisol and the inflammatory markers are likely related to each other, but the causal pathways are unclear, we have analyzed models both with and without adjustment for the two inflammation variables. Missing data were considered in two ways. We first compared baseline characteristics between those with complete data and those with missing data. Secondly, we imputed missing values [46] for 28 wk and 36 wk cortisol concentrations and re-analyzed the data as a sensitivity test on the primary models. Model assumptions were also checked using standard regression diagnostics for linearity, normality, leverage, and influence. All analyses were performed using SAS 9.3 (SAS Institute, Cary, NC).