Supplementary feeding and infection control in pregnant adolescents—A secondary analysis of a randomized trial among malnourished women in Sierra Leone

Undernutrition during pregnancy in adolescence confers a high risk of maternal morbidity and adverse birth outcomes, particularly in low-resource settings. In a secondary analysis, we hypothesized that younger undernourished pregnant adolescents (<18 years) would benefit more than undernourished pregnant adults (>20 years) from the intervention of supplementary food and anti-infective treatments. The original trial in Sierra Leone enrolled 236 younger adolescents (<18 years), 454 older adolescents (aged 18–19 years), and 741 adults (≥20 years), all with a mid-upper arm circumference ≤23 cm. Younger adolescents had lower final fundal height as well as smaller newborns (−0.3 kg; 95% confidence interval [CI], −0.3, −0.2; p < 0.001) and shorter newborns (−1.1 cm; 95% CI, −1.5, −0.7; p < 0.001) than adults. The intervention's effect varied significantly between maternal age groups: adults benefited more than younger adolescents with respect to newborn birth weight (difference in difference, 166 g; 95% CI, 26, 306; interaction p = 0.02), birth length (difference in difference, 7.4 mm; 95% CI, 0.1, 14.8; interaction p = 0.047), and risk for low birth weight (<2.5 kg) (interaction p = 0.019). The differences in response persisted despite adjustments for maternal anthropometry, the number of prior pregnancies, and human immunodeficiency virus status. Older adolescents similarly benefited more than younger adolescents, though differences did not reach statistical significance. In conclusion, newborns born to younger adolescent mothers had worse outcomes than those born to adult mothers, and adults and their newborns benefited more from the intervention than younger adolescents. This was a secondary analysis of clinical outcomes from a trial of malnourished pregnant girls and women in Sierra Leone (Hendrixson et al., 2021). Participants were categorized into three age groups: younger adolescents (<18 years), older adolescents (18–19 years), and adults (≥20 years). This categorization was chosen based on the WHO definition of an adolescent as <20 years old. The distinction between younger and older adolescents was made because marriage to a child <18 years old is illegal in Sierra Leone and adolescents <18 years old are more likely to have ongoing physical maturation than older adolescents. Maternal age was also modeled as a continuous variable to visualize its association with key maternal and infant outcomes. Undernutrition was defined by a mid‐upper arm circumference (MUAC) ≤ 23 cm. Eligible participants had a fundal height <35 cm and attended one of 43 health centers in Pujehun and Western Area Rural Districts. Exclusion criteria were known gestational diabetes, hypertension, or severe anemia. Informed consent was obtained and documented by a signature or thumbprint. Women older than 16 years were eligible to consent for themselves, and girls younger than 16 years desiring to participate required consent from a parent or guardian. Ethical approvals were obtained from the appropriate local and international review committees. The original trial was registered at ClinicalTrials.gov. The intervention group received a package of care including two doses of 1 g azithromycin by mouth and monthly sulfadoxine‐pyrimethamine (SP: 1500 mg/75 mg) given during the second and third trimesters. Participants were tested for vaginal dysbiosis, and if positive, they were treated with 500 mg metronidazole bid by mouth for 7 days. They also received a daily ration of ready‐to‐use supplemental food (RUSF). The RUSF was composed of skimmed milk powder, whey protein isolate, vegetable oil, sugar, peanut paste, and pearl millet, providing 2180 kJ (520 kcal). The RUSF provided 18 g of protein and over 100% recommended daily allowance for most micronutrients during pregnancy (Hendrixson et al., 2018). The control group received standard care, 250 g/d of corn‐soy blended flour (SuperCereal), and 25 g of palmolein oil daily, providing 2474 kJ (589 kcal) and 17.5 g of protein, as well as a ration for sharing per World Food Programme standards. In addition, the control group received three doses of malarial chemoprophylaxis, 60 mg/d iron, and 400 µg/d folic acids during the second and third trimesters. Nutritional supplementation was initiated at the time of enrolment and continued until delivery. A standardized questionnaire assessed adherence to the nutritional intervention at each visit. Azithromycin and SP were given under direct observation. The study was conducted in conjunction with government‐provided antenatal clinics. Upon enrolment, demographic information, time of last menses, an estimated delivery date, and clinical symptoms were recorded. Weight, height, MUAC, blood pressure and fundal height were measured. Fundal height was measured in the supine position with a nonelastic tape to the nearest 0.5 cm and used as a proxy for gestational length (WHO, 2007). Participants returned for follow‐up every 2 weeks for anthropometric assessment and provision of the study foods and medications until delivery. Participants were considered lost‐to‐follow‐up after missing three consecutive visits. Home visits were attempted for any patient lost to follow‐up. Clinic staff and participants were provided a telephone number and credit card to call the study coordinator at delivery. A birth measurement team was dispatched to conduct measurements of the infants within 48 h of delivery. Infant survival, weight, length, head circumference, MUAC, morbidity and feeding practices were assessed at each child visit. Maternal weight, MUAC and morbidity were also assessed at these visits. Full details of the intervention, food supplementation, and study procedures have been previously published (Hendrixson et al., 2018, 2021). The primary outcomes of this secondary analysis were maternal rate of weight gain and birth weight. Key secondary outcomes were birth length, rates of low birth weight (<2.5 kg), and neonatal mortality. Data were first recorded on clinic management cards. Data from these cards were double‐entered into a database (Microsoft Access) and cross‐checked for discrepancies. All discrepancies were resolved by examination of the original data card. Once the content of the database was determined, it was locked for analysis. Anthropometric indices were calculated using WHO 2006 growth standards (Anthro version 3.2.2). Baseline characteristics were summarised as n (%) if categorical. The distributions of continuous variables were visualized with histograms and Q–Q plots and summarised as mean (SD) unless they were skewed, in which case they were summarised as median (interquartile range). The maternal rate of weight gain was calculated by subtracting the baseline weight from the final visit weight and dividing it by the time elapsed. No missing data were imputed. Continuous outcomes, irrespective of the intervention group, were compared across the three maternal age cohorts using analysis of variance or Kruskal–Wallis test, depending on their distributions, while binary outcomes were compared using χ 2, or Fisher's exact test when any subgroup n < 5. For unadjusted comparisons between intervention groups, Student's t‐test or the Mann–Whitney U test was used for continuous variables, while the test for equality of proportions was used for binary variables. To test for effect modification of the intervention by maternal age group, an interaction term between the intervention group and maternal age group was created within regressions. After assessing model assumptions and deeming them not violated, linear regression was used for continuous outcomes, while modified Poisson regression with robust variance estimates was used for binary outcomes. In both cases, a difference‐in‐difference analysis comparing the effect of intervention between different age groups was done, with an estimation of 95% confidence intervals (CIs) and p values for the interaction terms. Difference‐in‐difference results were reported as positive if the intervention was more beneficial in the adult or older adolescent age groups than the intervention's effect in the younger adolescent age group. These analyses were done both without adjustment and with adjustment for (i) baseline maternal anthropometry (body mass index, height and MUAC), (ii) the number of prior pregnancies, and (iii) human immunodeficiency virus (HIV) status. These adjustments were chosen based on previously identified associations with maternal and infant outcomes as well as differences at baseline between maternal age groups in the parent study, which, therefore, might have explained differential intervention effects. Associations between maternal age and singleton infant birth weight and length and maternal rate of weight gain during pregnancy were also modeled with baseline maternal age as a continuous variable using linear regression. The associations between maternal age and these three outcome variables were visualized using loess curves; in all cases, the associations were nonlinear. As a result, restricted cubic splines were used to model age nonlinearly and to estimate the associations between age, intervention group, and outcomes (Harrell, 2016). Within these regressions, p values were computed using partial F‐statistics (Harrell, 2022). All analyses were conducted using R version 4.1.2 (R Foundation for Statistical Computing).