Background: The impact of feeding a slightly sweet nutrient supplement early in life on later sweet taste preference is unknown. Objective: We tested the hypothesis that the level of sucrose most preferred by 4-6-y-old children exposed to a slightly sweet lipidbased nutrient supplement (LNS) early in life would not be higher than that of children never exposed to LNS. Design: We followed up children born to women (n = 1,320) who participated in a randomized trial in Ghana. In one group, LNS was provided to women on a daily basis during pregnancy and the first 6 mo postpartum and to their infants from age 6 to 18 mo (LNS group). The control groups received daily iron and folic acid or multiple micronutrients during pregnancy and the first 6 mo postpartum, with no infant supplementation (non-LNS group). At age 4-6 y, we randomly selected a subsample of children (n = 775) to assess the concentration of sucrose most preferred using the Monell 2-series, forced-choice, paired-comparison tracking procedure. We compared LNS with non-LNS group differences using a noninferiority margin of 5% weight/volume (wt/vol). Results: Of the 624 children tested, most (61%) provided reliable responses. Among all children, the mean ± SD sucrose solution most preferred (% wt/vol) was 14.6 ± 8.6 (LNS group 14.9 ± 8.7; non-LNS group 14.2 ± 8.4). However, among children with reliable responses, it was 17.0 ± 10.2 (LNS group 17.5 ± 10.4; non-LNS group 16.5 ± 10.0). The upper level of the 95% CI of the difference between groups did not exceed the noninferiority margin in either the full sample or those with reliable responses, indicating that the LNS group did not have a higher sweet preference than the non-LNS group. Conclusion: Exposure to a slightly sweet nutrient supplement early in life did not increase the level of sweet taste most preferred during childhood. This trial was registered at clinicaltrials.gov as NCT00970866.
The iLiNS DYAD-Ghana trial was a community-based, partially double-blind, individually randomized controlled trial conducted between 2009 and 2014 in 2 semi-urban districts (Yilo and Manya Krobo) in the Eastern region of Ghana, located about 70 km north of the capital, Accra. The trial was designed to examine the efficacy of a small quantity of LNS (20 g) for the prevention of malnutrition in pregnant and lactating women and their infants. Details of the study design, randomization, and recruitment have been reported elsewhere (12). Briefly, pregnant women at ≤20 weeks of gestation were randomly assigned to 1 of 3 groups. One group (LNS group) received 20 g of LNS daily during pregnancy and the first 6 mo postpartum; women typically ate the palatable, slightly sweet supplement (containing 22 micronutrients and 118 kcal) mixed with any food of their choice. From 6 to 18 mo of age, mothers were instructed to feed their infants 20 g LNS daily either mixed with other foods or alone (11). Each 20 g of LNS supplement contained 4 g of total sugars with 1.2 g and 1.6 g of added sugar in the maternal and child versions, respectively. The children in the other 2 groups are combined herein (non-LNS group) because neither they nor their mothers received any food supplements. Their mothers ingested nonflavored capsules containing either iron and folic acid during pregnancy and a low dose of calcium for 6 mo postpartum, or MMN during pregnancy and the first 6 mo postpartum. Although all mothers received basic nutritional advice, none of the mothers in either group received instructions to limit the added sugar intake of their infants. As reported previously, LNS was added to the food of the children in the LNS group an average of 73.5% of the days between 6 and 18 mo of age (11). Sociodemographic information at the time of enrollment into the parent trial was collected by means of a questionnaire and included details of: maternal age, education, marital status and nulliparity, household food insecurity, and household assets. We constructed a household assets score based on ownership of a set of assets (radio, television, refrigerator, cell phone, and stove), lighting source, drinking water supply, sanitation facilities, and flooring materials. The household asset score was created using principal components analysis and had a mean of 0 and SD of 1, with higher values representing a higher socioeconomic status. We assessed the feeding practices of infants and young children, including caregiver reports of the child’s consumption of food and beverage items, at multiple time points between birth and 18 mo. This was achieved through a guided free recall of liquids and foods consumed by the child on the day before the interview and a list‐based recall of the number of days that each food group was consumed in the 7 d preceding the interview (20). The questionnaire had a total of 32 food and beverage items; each item was part of a list of items belonging to a food group or food category (e.g., fruits, dark green vegetables, sugary foods). There were 6 sugary food and beverage items in total (i.e., fruit juice or any juice drinks; chocolate or cocoa drink without milk; chocolate or cocoa drink with milk; yogurt; soft drinks or any other sweet drink; sugary foods such as chocolates, sweets, candies, pastries, cakes, or sweet biscuits). In this article, we report the proportion of children who consumed a sugary food or a sugary beverage the day preceding the day of interview when the child was 9 and 18 mo of age. When children were 4-6 y of age, we conducted a follow-up (January to December 2016) of the participants of the iLiNS DYAD-Ghana trial to determine the long-term impact of early nutritional supplementation on health and nutrition outcomes, which included sweet taste preferences, food and beverage preferences, and consumption among the children. All children whose mothers participated in the parent trial were eligible for the follow-up study regardless of whether or not they or their mothers were lost to follow-up before the end of the parent trial. Excluding child deaths, miscarriages, and stillbirths, 1,222 children were eligible to participate in the follow-up study (Figure 1). The present study on sweet taste preference includes a randomly selected subsample of children (n = 775) selected from the 1,222 who were eligible. Study profile. LNS group, women received 20 g LNS daily during pregnancy and 6 mo lactation. Infants received 20 g LNS daily from 6–18 mo of age; non-LNS group, women received either IFA during pregnancy and placebo for 6 mo postpartum or multiple micronutrient (MMN) capsules during pregnancy and 6 mo lactation. Infants did not receive any supplement. LNS, lipid-based nutrient supplement; IFA, iron and folic acid. *Details reported in (12). We contacted mothers or caregivers (in cases where the child was in the care of someone other than the mother) mostly by phone to inform them of the follow-up study. If they were interested, study personnel went to the home to provide more details of the study procedures and to obtain informed consent. If consent was obtained, the test visit was scheduled. We reimbursed each mother (or caregiver) for transportation costs on the day of testing. Mothers and study personnel were not informed of the study hypothesis and study personnel were blind to group assignment of the children. All study protocols were approved by the Institutional Review Board of the University of California, Davis, the Ethics Committee for the College of Basic and Applied Sciences at the University of Ghana, and the Ghana Health Service Ethical Review Committee. The follow-up study was part of the iLiNS DYAD-Ghana trial, which was registered at clinicaltrials.gov as NTC00970866. We used the Monell 2-series, forced-choice, paired-comparison tracking procedure to determine the concentration of sucrose most preferred. This tool is ideal for testing in pediatric populations because it requires a short time to complete, does not require verbal communication of responses, and controls for position bias (19, 21, 22). Five concentrations of sucrose solution were prepared [3%, 6%, 12%, 24%, 36% weight/volume (wt/vol), which is equivalent to 0.09, 0.18, 0.35, 0.70, 1.05 M]. These sucrose solutions were prepared every 1–2 d and stored refrigerated; the refrigerated solutions were brought to room temperature prior to testing. During taste testing, the sucrose solutions were presented to the child in small disposable medicine cups; distilled water was used for rinsing the mouth between pairs and between series and a bucket was provided for spitting; a stopwatch was used to monitor inter-pair and inter-series intervals; and data were recorded on a tracking grid, as described in Mennella et al. (19, 22). The test instructions on the grid were translated into 3 common local languages (i.e., Krobo, Ewe, and Twi). Testing took place in a closed room at our testing center which was easily accessible by public transportation. The room was ventilated and eating was not allowed in the room to minimize food odors. The room was partitioned into sections, one of which was used to familiarize the children with the testing procedures in the presence of the mother/caregiver. The mother/caregiver remained silent in this room during the testing and was out of view of the child (to prevent any distraction). Testing of children occurred in the other sections of the room, each of which was staffed by a trained research assistant. The child sat on a chair behind a small table designed for children. One research assistant conducted the study and the other monitored the timing of inter-pair and inter-series intervals using a stopwatch and, based on the child’s response, selected the pairs of sucrose solutions for testing (see below). As a check for following correct procedures, each assistant monitored the child’s responses and both had to agree on the next pair of samples given to the child or when the child reached criterion. Details of the test have been described elsewhere (19, 21). In brief, following fasting for at least 1 h, participants were presented with pairs of solutions (5 mL each) in medicine cups that differed in sucrose concentration. The first pair presented was from the middle range of concentrations (6% and 24% wt/vol). The child tasted each solution within a pair for 5 s without swallowing and then pointed to the solution they preferred, without instruction on how the stimuli differed. They rinsed their mouths once between each sample and twice between each pair during an enforced 1-min interval. Each subsequent pair of solutions presented contained the concentration selected by the child in the preceding pair and an adjacent concentration stimulus. This pattern continued until the child chose the same concentration when paired with both a higher and lower concentration in two consecutive pairs or chose the highest or lowest concentration twice consecutively. After a 3-min break, we repeated the entire task but stimulus pairs were presented in the reverse order (in the protocol, for series 1 the lower concentration was presented first; for series 2 the higher concentration was presented first). This controls for position bias and enables researchers to determine objectively whether the child understands the task or is responding by pointing to whatever is presented to their right or left (19, 21). The geometric mean of the concentrations selected in series 1 and 2 provides the estimate of the most preferred concentration of sucrose. Height was measured in duplicate to the nearest 0.1 cm using a stadiometer (Seca 217; Seca) and weight was measured in duplicate to the nearest 50 g using a Seca scale (Seca 875; Seca). Using the WHO Anthro software (23), we calculated BMI-for-age z-scores of the children. Originally, our sample size was calculated based on detecting a small effect size of 0.2 (24) between groups in the concentration of sucrose most preferred and in other outcomes in the follow-up study (to be reported elsewhere). This yielded a minimum sample size of 775 (388 per group) assuming an α = 0.05 and 80% power, an SD of 8.2% (obtained from a pilot study to test the feasibility of the tool to examine sweet taste preference in 30 children who did not participate in the parent trial, conducted prior to the start of the follow-up study), and up to 25% attrition. We posted a statistical analysis plan on the project website (www.ilins.org) prior to data analysis. We used a noninferiority approach to compare the sweet taste preference between the intervention and control groups. The primary outcome was the concentration of sucrose most preferred. We chose 5% wt/vol as our noninferiority margin based on one study that reported a mean difference of 6% wt/vol in the concentration of sucrose most preferred at age 6–10 y between children who were routinely fed sugar water during infancy when compared with similarly aged children who were rarely or minimally exposed to sugar water during infancy (6). A 5% wt/vol difference in sucrose solution between groups translates to 5 g sugar in 100 mL water. To detect this mean difference of 5% between groups, we needed a minimum sample size of 336 (168 per group) (α = 0.05, power = 0.9), calculated based on published data (19) from which we determined the most preferred concentration of sucrose (17.1% ± 11.0% wt/vol; mean ± SD) among only those participants who were aged between 5 and 7 y (J Mennella, Monell Chemical Senses Center, Philadelphia. Personal communication, 2017). The actual target sample size of 775 allowed for the possibility that some children might not understand the sweet taste test instructions and their data might have to be excluded in sensitivity analyses. We first determined how many children understood the task by categorizing children based on the reliability of their responses between series 1 and 2. Responses were considered reliable if the choice made in series 2 was the same as or ≤2 steps away from the choice made in series 1, and unreliable if the choice made in series 2 was ≥3 steps away from the choice made in series 1. Second, we determined whether the geometric mean of the two series differed between the groups by conducting 2 separate analyses. The first analysis included all children and the second included only children who exhibited reliable responses. We examined differences between treatment groups using both negative binomial and linear regression modeling techniques. Since results from the analytical methods were similar, we present the results from the linear regression models for ease of interpretation. Noninferiority was deemed to be established if the 95% CI of the difference between the treatment groups fell below the noninferiority margin. ANOVA or chi-squared tests were used to determine whether the groups differed in maternal, child, or household characteristics. Potential prespecified covariates were considered for covariate adjustment if they were significantly associated with the outcome (P < 0.10). These included child sex as well as maternal characteristics assessed at baseline (prior to enrollment into the parent trial) (12): years of formal education, marital status, age, estimated prepregnancy BMI, and nulliparity; a household assets index derived from a principal components analysis (25); household food insecurity access scale (26); distance in meters to the nearest weekly market; and main language spoken at home. All models included child age at the time of testing. We examined the potential interaction between child age and intervention group with regard to level of sweetness most preferred. Data analysts were fully blinded to group assignments until analyses were completed. All analyses were conducted using SAS for Windows Version 9.4 (SAS Institute).