Breastfeeding and Once-Daily Small-Volume Formula Supplementation to Prevent Infant Growth Impairment

PRIMES eligibility criteria and trial methods have been published.³⁹  In Bissau, Guinea-Bissau, and Kampala, Uganda, infants eligible for PRIMES were healthy singleton breastfeeding newborns. Key eligibility criteria included a weight of 2000 g to 2885 g at <6 hours after birth and negative maternal HIV testing; eligible infants were randomly assigned if their weight was <2500 g at <6 hours after birth (Day 0) or <2600 on Day 4. This study was approved by the Guinea-Bissau Comite Nacional de Etica na Saude (Guinea-Bissau National Committee on Ethics and Health, 075/CNES/INASA/2020), the Higher Degrees, Research and Ethics Committee of Makerere University (871), the Uganda National Council of Science and Technology (HS1226ES), and the University of California San Francisco Institutional Review Board (19-29405).

Recruitment, enrollment, and randomization occurred from February to December 2021. Local clinicians informed mothers about the study, asked them about potential interest in study participation, and referred those who expressed interest to study staff, who then described the study in detail, including all study procedures, the randomization process, and the voluntary nature of participation, and encouraged mothers to ask questions. Structured items were then used to ascertain maternal comprehension of the nature of randomization and the ongoing voluntary nature of participation, and written informed consent was then obtained by signature or thumbprint. After obtaining informed consent, trained study staff obtained treatment assignment generated by the online software application, randomize.net, with randomly permuted blocks of 4 and 6 infants stratified by site (Guinea-Bissau or Uganda) and weight (2000–2249 g on Day 0, 2250–2499 g on Day 0, 2250–2424 g on Day 4, and 2425–2599 g on Day 4).

For enrolled infants randomly assigned to the ESVFS (intervention) group, trained study staff provided breastfeeding support and taught the mothers to breastfeed the infant frequently on demand and at least 8 times per day, and to follow 1 completed breastfeeding per day with the offer of a premixed, commercially sterile, single-use 59 mL bottle of Similac Pro Advance (Abbott Nutrition, Columbus, OH), allowing the infant to first consume breast milk until the infant discontinued breastfeeding and then consume the supplement until the infant discontinued feeding. Powdered formula was not used, and no formula was mixed by the study staff or participants. Trained study staff used a structured tool to teach mothers to assemble each ESVFS feeding in a hygienic manner from a commercially sterile, single-use container and a single-use sterile nipple and then observed 4 daily feedings of breastfeeding followed by ESVFS feeding, providing feedback on breastfeeding, ESVFS assembly, and ESVFS feeding technique when needed. After these directly observed feedings, trained study staff delivered two 59 mL formula bottles and 2 single-use sterile nipples to infants in the ESVFS group every other day, with instructions to offer only 1 bottle each day, only once per day and to save any formula remaining after the daily ESVFS feeding to return to the study team for measurement. Breastfeeding support was provided on an ongoing basis, and mothers were also taught that they should provide no other food or fluid to enrolled infants other than vitamins, minerals, and medication. For infants randomly assigned to the EBF (control) group, trained study staff provided ongoing breastfeeding support and taught mothers to breastfeed the infant frequently on demand, at least 8 times per day, and to provide no other food or fluid other than vitamins, minerals, and medications.

Infants and their mothers were enrolled on the day of birth and assessed for baseline characteristics. Data on gestational age were not available because first-trimester ultrasound was rare and pandemic restrictions precluded accurate gestational age assessment in the absence of first-trimester ultrasound.^40,41  Infants were assessed for outcomes at 4, 14, 30, 60, and 180 days of age.

Infant weight and length were obtained by blinded staff at least twice at each study visit with a Seca 334 scale (Seca Inc, Wandsbek, Germany) and a Seca 416 infant stadiometer, respectively. If the initial 2 weights or 2 lengths differed by >10 g or >0.5 cm, respectively, a third and fourth weight or length were obtained; this occurred for weight and length at 8 and 28 visits, respectively. The interclass correlation coefficient was 1.00 for both weight and length from intra-class correlation coefficient (ICC) (3,k) for test-retest reliability.⁴²  Measurements were averaged to obtain final weights and lengths. Infant mid-upper arm circumference (MUAC) and head circumference, maternal height, weight, and MUAC, household dietary intake, infant dietary intake, any infant WHO danger signs or AEs, and hemoglobin were measured by using standard instruments and point-of-care Hemocue (Hemocue, Danaher Inc, Washington, DC).⁴³  For infants receiving ESVFS, formula intake volume was assessed by using prespecified markings (<7.5 mL, 7.5–15 mL, 15–22.5 mL, 22.5–30 mL, 30–45 mL, and 45–59 mL) on the single-serving formula container. Infant deaths were assessed by using the WHO verbal autopsy instrument.

Weight-for-age z score (WAZ), weight-for-length z score (WLZ), and length-for-age z score (LAZ) were calculated by using the WHO Anthro Survey Analyzer⁴⁴ ; z scores <−2 were defined as underweight, wasting, and stunting for WAZ, WLZ, and LAZ, respectively. Conservatively estimating 20% loss to follow-up by 180 days of age, the predetermined sample size of 324 randomized infants was selected to obtain outcomes for 258 infants. This number of outcomes was needed to attain at least 80% power to detect as statistically significant a difference between the treatment groups in WAZ at 30 days of age of 0.20 or larger (SD = 0.6, [Cohen’s d] d= 0.33), a magnitude that would have important public health relevance.

A detailed statistical analysis plan has been published at https://osf.io/px7ct/. Linear mixed models were performed by using repeated measures of data collected at all study visits, with an unstructured covariance matrix across time points to test the effect of ESVFS on WAZ, WLZ, and LAZ at 30 days, as well as at other time points, with contrasts between treatment groups at each time point. A permutation test comparing groups on day 30 was also performed. When the null hypothesis was not rejected (P < .05), Bayes Factors⁴⁵  were calculated by comparing models with and without a factor for group on outcomes at day 30.

Primary analyses were performed by using an intention-to-treat approach including all randomized infants. In sensitivity analyses, the primary analyses were repeated by using a per protocol approach comparing infants randomly assigned to ESVFS who consumed more than half the formula provided to infants randomly assigned to EBF who did not receive any formula. Missing data due to dropout was handled by using linear mixed models.^46,47  Model assumptions were considered satisfied as residuals were approximately symmetric (skewness <2) and with similar variance between groups (SD² ₁/SD² ₂ = 1.20).^48,49  We used linear regression to evaluate the association between the average volume of the supplement consumed by infants randomly assigned to ESVFS and their WAZ change through 30 days of age. χ² testing and Fisher’s exact tests were used to compare the intervention and control groups with respect to AEs, serious AEs (SAEs), and breastfeeding cessation. All analyses were conducted by using SAS version 9.4 (SAS Institute Inc, Cary, NC).

Of 904 infants initially identified as potentially eligible, PRIMES enrolled 618, of whom 324 (52.1%) met the criteria for randomization and were randomly assigned (Fig 1). Each site randomly assigned 162 infants, of whom 77 (48%) and 72 (44%) were randomly assigned on the day of birth and 85 (52%) and 90 (56%) were randomly assigned at 4 days of age in Guinea-Bissau and Uganda, respectively. The mean (± SD) birth weight for those randomly assigned on the day of birth was 2338 ± 135 g, and the mean birth weight for those randomly assigned at 4 days of age was 2620 ± 82 g (Table 1). Among randomly assigned infants, 144 (44%) were male (P < .05); of these 144 males, 64 (44%) were randomly assigned to ESVFS (P = .091). No other baseline characteristics differed by treatment group.

Among infants in the intervention group, the median (interquartile range [IQR]) daily volume of ESVFS feeding ranged from 14.79 (7.39–22.18) mL at 4 days of age to 36.97 (14.79–51.75) mL at 30 days of age and averaged 893 ± 331 mL in total over all feedings (Fig 2). The average volume of the supplement consumed was not associated with WAZ change in the intervention group (P = .710). Only 4 (2%) mothers in the control group reported formula use in the first 30 days (Table 2). Nonformula supplementation, including milk, water, sweetened water, juice, tea, and other liquids, occurred in the first 30 days for 19 infants (12%) in the intervention group and 35 (21%) controls (P = .02). Of 304 (94%) infants with follow-up at 180 days of age, 4 infants in the ESVFS group and 3 infants in the control group had stopped breastfeeding (P = .697).

Intention-to-treat and per protocol analyses both revealed that treatment assignment did not significantly affect WAZ, WLZ, or LAZ or the proportion with growth impairment at day 30 based on linear mixed models or permutation tests (Table 3). Bayes Factors of 7.2 to 8.4 revealed “moderate evidence” of the null hypothesis of no group differences at day 30. Tests of effect modification did not identify significant interactions for treatment by site, baseline weight, or sex. The most common AEs in the first 30 days were fever, jaundice, and difficulty breathing, which occurred for 33 (10%), 27 (8%), and 27 (8%) infants, respectively, and did not differ by treatment assignment. The most common AEs in infants >30 days of age were fever, diarrhea, and vomiting, which occurred for 113 (36%), 57 (18%), and 27 (9%) infants, respectively, and did not differ by treatment assignment. In the ESVFS group, 1 infant experienced sudden unexpected infant death at 1 month of age, and another infant died of congenital heart disease diagnosed at 2 months of age, whereas in the EBF group, 2 infants died of fever (one at 7 days of age and one at 4 months of age), and 1 additional infant experienced sudden unexpected infant death at 2 months of age. Overall, in this cohort, the prevalence of growth impairment remained high at 180 days, including stunting (27%), wasting (4%), and being underweight (14%).

As shown in Fig 1, ESVFS intake increased over the first 30 days but remained highly variable between individual infants. Growth outcomes did not differ between infants who consumed >30 mL formula daily and those who consumed <30 mL. Female sex was associated with relative risks of 0.66 (0.42–1.04) and 0.50 (0.34–0.73) for being underweight and stunting, respectively, compared with male infants.

In this cohort of otherwise healthy, breastfeeding newborns at risk for growth impairment in Guinea-Bissau and Uganda, the use of ESVFS together with breastfeeding support in the first 30 days did not impact WAZ, WLZ, LAZ, breastfeeding cessation, AEs, or SAEs through 180 days of age. Provided as once-daily feeding immediately after a completed breastfeeding, ESVFS intake was modest, with the average newborn consuming ∼30 mL/day in addition to breastfeeding. Infants randomly assigned to ESVFS were less likely to receive sweetened water and other liquids during the first 30 days, suggesting that the provision of ESVFS may reduce the use of nonformula, non-breast milk feedings. Taken together, these results suggest that the scientifically quantifiable risks of ESVFS are relatively small and have the potential to be balanced by benefits for some populations.

Designed as a proof-of-concept pilot study with a relatively small sample size of 324 randomly assigned infants, our study reveals 1 method by which early supplemental nutrition can be fed in low-income countries without interfering with ongoing breastfeeding. A technique such as ESVFS could potentially be adapted to allow infants in LMIC access to nutritional support when initial growth during exclusive maternal breastfeeding is inadequate. Unique elements of the ESVFS method used in this trial were the small volume of formula supplement provided, the instruction to supplement formula immediately after completing a breastfeeding, the use of a single-serving, commercially sterile bottle and single-use nipple along with recommendations for hygienic assembly, and the limited time course for ESVFS supplementation. Using this unique intervention, neither efficacy nor harm was demonstrated in this group of at-risk newborns. It is unknown whether more frequent use or higher volumes of ESVFS, in addition to breastfeeding, might demonstrate efficacy for this population, or whether larger sample sizes or the use of reconstituted powdered infant formula might demonstrate harm. Importantly, it is also unknown whether ESVFS impacts growth for infants whose characteristics differ from PRIMES eligibility criteria. Effect size estimates from this study can be used to inform further research to explore these questions, as well as to determine the effect on the growth of other approaches, including specialized breastfeeding support, donor milk, and maternal milk fortification. In this way, it may be possible to learn how infants in LMIC can attain the same healthy growth as infants in HIC.

This study has important limitations. First, the daily formula intake of infants randomly assigned to ESVFS was ∼30 mL/day. This small volume may account for the lack of demonstrated efficacy, the lack of detrimental impact, or both. To improve scientific understanding of the relationship between dietary intake and growth and AEs, the authors of future studies in this area should continue to carefully and frequently assess both growth and adverse outcomes. Second, given that actual intake of ESVFS was directly observed by study staff only for the first 4 ESVFS feedings and that subsequent intake was measured from amounts remaining in bottles when returned by mothers, some recorded formula intake might represent the diversion of formula to non-study uses. However, few bottles were fully consumed, suggesting that the extent of diversion was not large. Third, because reliable data on gestational age was not available, we were not able to assess the impact of prematurity on study outcomes. Although this limited our ability to assess differences between infants born premature and infants born term, it may improve the generalizability of our results to clinical settings without reliable gestational age estimation. Fourth, we did not study breast milk intake volumes during the period of formula supplementation. Thus, although mothers were taught to breastfeed immediately before providing formula and were encouraged to continue each breastfeeding until the infant discontinued the feeding, we are unable to report whether ESVFS displaced the intake of maternal breast milk in this trial. Emerging technologies for measuring milk intake might facilitate such measurement for future projects.⁵⁰  Fifth, the carefully managed, structured formula supplementation provided in this trial might be difficult to duplicate outside of study conditions. This limitation reduces the generalizability of our study results. Sixth, although the study findings did not reveal an effect of treatment assignment on AEs, it is important to consider that, because of a limited sample size, the available power was not sufficient to ensure that there were no clinically meaningful risk differences between the groups. Although risk can be calculated scientifically, no intervention is ever entirely free of risk, so the safety of any intervention is a social or policy judgment made in the context of current knowledge regarding risks and benefits.⁵¹  Risk estimates from this study may inform such decisions in the future.

In this pilot study of breastfeeding newborns weighing 2000–2499 g at birth or <2600 g at 4 days of age, once-daily ESVFS for 30 days did not alter growth or cause breastfeeding cessation or AEs. Whether different eligibility criteria or more frequent supplement administration might increase growth or affect AEs is unknown. When testing any future interventions, outcomes, including AEs, SAEs, and breastfeeding cessation, should continue to be carefully assessed. In this way, it may be possible to identify a safe, feasible intervention effective at reducing the risk of newborn growth impairment and improving health and development for this important and vulnerable population.

We are grateful for the participation of the infants and their families, the clinical staff at the birthing facilities and clinics, and the team of research nurses and enumerators in Guinea-Bissau and Uganda.

AE

adverse event

EBF

exclusive breastfeeding

ESVFS

early small-volume formula supplementation

HIC

high-income countries

IQR

interquartile range

LAZ

length-for-age z score

LMIC

low- and middle-income countries

MUAC

mid-upper arm circumference

PRIMES

Preventing Infant Malnutrition with Early Supplementation

PROBIT

Promotion of Breastfeeding Intervention

SAE

serious adverse event

WAZ

weight-for-age z score

WHO

World Health Organization

WLZ

weight-for-length z score