Maternal plasma lipid levels across pregnancy and the risks of small-for-gestational age and low birth weight: a cohort study from rural Gambia

This study is a secondary analysis of the Early Nutrition and Immune Development (ENID) trial; ISRCTN49285450]), a randomized, partially blinded trial investigating the impact of prenatal and infancy nutritional supplementation on infant development in the West Kiang region of The Gambia. The published ENID trial protocol provides complete details of the trial [21] while an overview of relevant information to the current secondary analysis is included here (participant selection outlined in Fig. 1). All non-pregnant women aged 18 to 45 years registered in the West Kiang Demographic Surveillance System (DSS) were invited to participate in the trial [22]. After written consent, all women were visited monthly. Between January 2010 and June 2013, 2798 women were recruited for monthly surveillance of pregnancy. Women who missed their last menses and had a positive urine pregnancy test were invited to the Medical Research Council (MRC) Keneba clinic for an ultrasound examination of pregnancy status and stage. Of the 1195 women with a positive pregnancy test, those confirmed as pregnant but with a gestational age ≥ 20 weeks or a multiple pregnancy or those confirmed as HIV positive or with severe anemia (hemoglobin (Hb) < 7 g/dL) were excluded from the study. A total of 875 pregnant women who met the inclusion criteria were randomized into the antenatal supplementation phase of the trial, yielding 800 live births. For the current analysis, 227 infants were excluded due to missing birth visit (n = 105) or birth weight (n = 41) or not born at term with gestational age at delivery below 37 weeks or over 42 weeks (n = 81), resulting in 573 mother-newborn pairs (Fig. 1). Of note, preterm infants were not included in this analysis due the low numbers of preterm births within the ENID cohort (n = 14).

The randomization procedure allocated eligible pregnant women to one of four prenatal nutritional supplements; (i) iron folic-acid (FeFol) tablets, the standard of care as per Gambian Government guidelines; (ii) multiple micronutrients (MMN) tablets, a combination of 22 micronutrients designed for use during pregnancy by UNICEF/WHO/UNU [23]; (iii) protein-energy and iron-folate (PE + FeFol), a lipid-based nutritional supplement (LNS); and (iv) protein-energy and multiple micronutrients (PE + MMN). Supplement composition is described in Additional Table A1. Field staff visited enrolled women weekly for supplement provision. During weekly visits, compliance to supplement was evaluated by an assessment of the quantity of supplements not used, and a record of maternal morbidity was also collected.

At enrolment (mean gestational (SD) = 13.9 (3.3) weeks gestation), gestational age was assessed using a Siemens ACUSON Antares Ultrasound Imaging System (Siemens Medical Solutions USA, Inc., California, USA) with a CH6–2 (5.71 MHz) transducer). Further clinic visits were conducted at 20 and 30 weeks gestation. At all visits, data were collected on maternal anthropometry, Hb, blood pressure, and a urinary analysis was performed. A sample of venous blood (10 mL) was also collected following an overnight fast and plasma samples were stored on ice until processing within 1 h of collection. Standardized and validated equipment and standard operating protocols were used and applied for all measurements. A study midwife visited women and their newborns within 72 h of delivery for a standard maternal and newborn health check. Infant weight and length were measured using digital infant scales (Seca mobile digital baby-scale 334; UK) with 10 g precision and a portable infant roll meter (Rollameter 100; Harlow Healthcare, UK) to the nearest 0.1 cm, respectively.

All maternal plasma samples were collected after an overnight fast and analyzed in the MRC Keneba laboratory for TC, HDL-c, low-density lipoprotein cholesterol (LDL-c), and TG levels using enzymatic colorimetric assays with Roche/Hitachi reagents and COBAS INTEGRA® 400 plus analyzer (Roche Diagnostics, Indianapolis, IN). The few samples that were physiologically aberrant, such as samples with detectable HDL-c and LDL-c levels but with a TC level near zero, were excluded from the analysis.

Low birth weight (LBW) was characterized as a birth weight below 2500 g and compared to normal birth weight (NBW) (≥2500 g) [9]. Small-for-gestational-age (SGA) was defined as a birth weight-for-gestational-age below the 10th percentile of the INTERGROWTH-21ST standards for birth weight in comparison to adequate-for-gestational (AGA) (≥10th percentile of the INTERGROWTH-21ST standards for birth weight) [24]. The West Kiang DSS was used to verify mother's birth date and age at enrolment [22]. A questionnaire collected at enrolment was used to determine maternal parity (i.e. numbers of stillbirths and live births). School attendance was defined as a binary variable (yes/no) based on whether enrolled women reported 1 year or more of English or Arabic school and not as the number of school years attended due to the low attendance mean (SD) (0.31 (0.62) years). Maternal BMI was calculated as weight (kg)/height (m)² at each measurement time point in gestation. Maternal gestational weight gain was calculated by subtracting maternal weight between; enrolment and 20 weeks gestation, 20 and 30 weeks’ gestation, and enrolment and 30 weeks gestation, and by dividing each subtraction by the number of weeks between weight measurements and expressed in kg/week. Maternal morbidity was determined based on the number of self-reports of morbidity episodes (e.g. fever, nausea/vomiting, dysuria, bleeding and abdominal pain) assessed by questionnaire and divided by the number of weeks enrolled in the trial (n/week). Compliance to supplementation from enrolment to delivery (%) was computed by dividing the number of jars of LNS products (PE and PE + MMN) (empty, half-empty, and full) or count of tablets (MMN and FeFol) the women consumed by the number received and multiplying by 100. Birth season of the infant was defined as dry (November to May) or rainy (June to October).

Maternal and infant variables were compared between LBW and NBW infants, and between SGA and AGA infants by Student’s t-test with Welch correction for unequal sample sizes. Unadjusted mean TC, HDL-c, LDL-c, and TG levels at enrolment and 20 and 30 weeks’ gestation were calculated with 95% confidence intervals (95% CI) and compared by Student’s t-test. Changes in maternal lipid levels across gestation were examined using a Wilcoxon rank-sum test for ordered groups and a paired sample t-test. Maternal lipid levels were analyzed as continuous variables and grouped into low (<10th), referent (10th–90th), and high (>90th) percentiles. Binary regression models were used to measure the relative risks (RR) of LBW and SGA associated with maternal lipid levels included as continuous or categorical variables (by percentile groups). Reduction in relative risk (RRR) expressed as a percentage was calculated as RRR% = (1-RR) × 100. The RR values for the analyses of the risks of LBW and SGA are presented in Additional Tables A2 and A3, respectively. Linear regression models were used to investigate associations between maternal lipids and birth weight (g) as an outcome. Models were adjusted for confounding factors selected based on previous research and on whether they significantly impacted on the models. For LBW and birth weight outcomes these included enrolment maternal age, parity, supplement groups (FeFol, MMN, PE or PE + MMN), gestational age, BMI, Hb level at lipid measurements, and compliance to supplement during pregnancy and infant birth sex and birth season. For SGA, the same confounding factors were used, but excluding gestational age. A previous analysis of the ENID dataset has shown a complex relationship between season and gestational weight gain on birth outcomes [25]. Given the potential relevance of changes in maternal BMI over pregnancy on gestational lipid levels, we also fitted BMI and changes in BMI within the models presented here. Statistical analyses were conducted with Stata version 15 (StataCorp LP Texas, USA).

A total of 573 pregnant women with singleton term infants were included in this study. There were 45 (7.9%) cases of LBW and 186 (32.5%) cases of SGA. At enrolment, 20.6% of women were underweight (BMI < 18.5 kg/m²), 68.2% were normal weight (BMI 18.5─24.9), 9.1% were overweight (BMI 25─29.9) and 2.1% were obese (BMI ≥ 30 kg/m²). Underweight women were significantly more likely to have a LBW (12.7% vs 6.4%, P = 0.023) or SGA (41.5% vs 29.7%, P = 0.014) infant compared to women with a BMI > 18.5 kg/m². Table 1 compares the descriptive characteristics of the participants, split according to LBW versus (vs) NBW and SGA vs AGA. Women with LBW infants were enrolled later in the trial, were more likely to be nulliparous, had lower gestational age at delivery, a lower BMI at all time points of gestation and a lower gestational weight gain between enrolment (mean gestational (SD) = 13.9 (3.3) weeks gestation) and 20 weeks gestation compared with women with NBW infants (all, P < 0.05). LBW infants were all born SGA and were more likely to be female (P = 0.012). Women with SGA infants also had a lower BMI at all gestational time points (all, P < 0.01), a lower gestational weight gain between enrolment and 30 weeks gestation (P = 0.028) and higher gestational age at delivery (P = 0.012) compared with women with AGA infants. SGA infants were more likely to be born with a normal birth weight than with a LBW (75.8% vs 24.2%, P < 0.001). There were no significant differences in risk of LBW or SGA by maternal nutritional supplement group.

Mean levels of TC, LDL-c and TG increased from enrolment to 30 weeks gestation (all, P < 0.001) while mean HDL-c levels increased slightly from enrolment to 20 weeks gestation (51.4 vs 53.1 mg/dL, P = 0.003) before decreasing back, by 30 weeks gestation, to similar levels observed at enrolment (53.1 vs 51.2 mg/dL, P < 0.001). Longitudinal changes in maternal lipid levels were compared by LBW vs NBW and SGA vs AGA (Fig. 2). Women with LBW infants had lower mean TC levels at 20 weeks gestation compared to those with NBW infants (133.6 vs 149.6 mg/dL, P = 0.048) (Fig. 2a). Women with SGA infants had lower mean LDL-c (99.5 vs 105.9 mg/dL, P = 0.033), and TG (94.2 vs 99.4 mg/dL, P = 0.048) levels at 30 weeks gestation compared to women with AGA infants (Fig. 2b).

Links between maternal lipid levels during pregnancy and the risk of delivering a LBW infant were investigated using regression models (Table 2 and Table 3). In the adjusted analyses, every unit increase in HDL-c at enrolment was associated with a 2.7% (P = 0.011) reduction in the relative risk of LBW and every unit increase in TC levels at 20 weeks gestation were associated with a 1.3% reduction in the relative risk of LBW (P = 0.002) (Table 2). Women with low (<10th percentile) HDL-c at enrolment or at 20 weeks gestation had a 2.6 (P = 0.007) and 3.0 (P = 0.003) higher risk of delivering a LBW infant compared to women with referent (10th─90th percentile) HDL-c level, respectively (Table 3).

Tables 4 and 5 present associations between maternal lipid levels and risk of SGA. In the unadjusted analyses, an association between increased LDL-c levels at 30 weeks gestation and a reduced risk of SGA was detected (RRR = 0.38%, P = 0.048) (Table 4) but no other associations were observed. In both the unadjusted and adjusted analyses, women with high (>90th percentile) LDL-c at 30 weeks gestation had a 55% (P = 0.017) lower risk of delivering an SGA infant compared to women with referent (10th─90th percentile) LDL-c levels (Table 5). In the unadjusted analyses, women with high (>90th percentile) TG at 30 weeks gestation had a 48% (P = 0.035) lower risk of delivering an SGA infant compared to women with referent (10th─90th percentile) TG, however, this association was lost following adjustment for confounding factors (Table 5).

Linear regression models were used to examine associations between maternal lipid levels during pregnancy and birth weight (Table 6 and Table 7). In the unadjusted analyses, increased levels of TC and LDL-c at 20 and 30 weeks gestation and of TG at 30 weeks gestation were all associated with a higher infant birth weight (all, P < 0.05) (Table 6). After adjustment with confounding factors, only increased levels of TC at 20 weeks gestation (β = 1.3, P = 0.027) and of TC (β = 1.2, P = 0.006) and LDLc (β = 1.5, P = 0.002) at 30 weeks gestation remained associated with a higher birth weight (Table 6). In both the unadjusted and adjusted analyses, women with high (>90th percentile) levels of TC, LDL-c or TG at 30 weeks gestation had infants with higher birth weights compared with women with referent (10th─90th percentile) levels (P < 0.05) (Table 7).

In a sub-analysis, we explored potential interactions between maternal nutritional supplement groups, BMI, lipid levels and infant birth weight outcomes (Additional Tables A4 to A8). Associations were observed between maternal nutritional supplement groups and maternal lipid levels during pregnancy (Table A4). Compared to FeFol (referent/control group), supplementation with PE + MMN was associated with lower HDLc levels at 20 (β = − 3.7, P = 0.028) and 30 (β = − 4.3, P = 0.013) weeks gestation, and supplementation with PE was associated with higher LDL-c levels (β = 9.3, P = 0.034) at 30 weeks gestation (Table A4). We also detected associations between maternal lipid levels and BMI during pregnancy (Table A5). At enrolment, a lower BMI was associated with increased HDL-c levels (β = − 0.025, P = 0.021) whereas a higher BMI was associated with increased LDL-c levels (β = 0.012, P = 0.034). At 20 and 30 weeks gestation, higher BMIs were linked to increased TG levels (β = 0.011, P = 0.018 and β = 0.009, P = 0.048, respectively) (Table A5). However, there were no associations between maternal nutritional supplement groups and maternal BMI at any timepoint across pregnancy (Table A6). Additionally, maternal supplement groups were not associated with the risks of LBW or SGA (Table A7) or birth weight (Table A8).