Full breastfeeding protection against common enteric bacteria and viruses: Results from the MAL-ED cohort study

Background: Breastfeeding is known to reduce the risk of enteropathogen infections, but protection from specific enteropathogens is not well characterized. Objective: The aim was to estimate the association between full breastfeeding (days fed breast milk exclusively or with nonnutritive liquids) and enteropathogen detection. Methods: A total of 2145 newborns were enrolled at 8 sites, of whom 1712 had breastfeeding and key enteropathogen data through 6 mo. We focused on 11 enteropathogens: adenovirus 40/41, norovirus, sapovirus, astrovirus, and rotavirus, enterotoxigenic Escherichia coli (ETEC), Campylobacter spp., and typical enteropathogenic E. coli as well as entero-aggregative E. coli, Shigella and Cryptosporidium. Logistic regression was used to estimate the risk of enteropathogen detection in stools and survival analysis was used to estimate the timing of first detection of an enteropathogen. Results: Infants with 10% more days of full breastfeeding within the preceding 30 d of a stool sample were less likely to have the 3 E. coli and Campylobacter spp. detected in their stool (mean odds: 0.92-0.99) but equally likely (0.99-1.02) to have the viral pathogens detected in their stool. A 10% longer period of full breastfeeding from birth was associated with later first detection of the 3 E. coli, Campylobacter, adenovirus, astrovirus, and rotavirus (mean HRs of 0.52-0.75). The hazards declined and point estimates were not statistically significant at 3 mo. Conclusions: In this large multicenter cohort study, full breastfeeding was associated with lower likelihood of detecting 4 important enteric pathogens in the first 6 mo of life. These results also show that full breastfeeding is related to delays in the first detection of some bacterial and viral pathogens in the stool. As several of these pathogens are risk factors for poor growth during childhood, this work underscores the importance of exclusive or full breastfeeding during the first 6 mo of life to optimize early health. The primary goal of the MAL-ED study was to describe relations among enteric infections, diet, gut function, and the growth and development of infants, and a detailed description (24) and primary results are described elsewhere (25–27). Briefly, infants were enrolled at 8 sites in low- and middle-income settings and followed to 24 mo of age. Inclusion criteria were enrollment within 17 d of birth (median: 7 d; IQR: 4–12 d), born from a singleton pregnancy to a mother at least 16 y of age, birth weight or enrollment weight >1500 g and no major morbidities, and with a family planning to stay in the community for at least 6 mo. Each site chose a target for enrollment with the aim to have data on approximately 200 children per site at 24 mo. Enrollment was staggered over 2 y. The analyses presented here were restricted to the period from birth to 6 mo, covering the recommended period of exclusive breastfeeding. Households were visited twice weekly to inquire about illness symptoms since the prior visit (28). Stools were collected when mothers reported diarrhea and also collected monthly when children were considered free of diarrhea (separated from symptoms by at least 2 d) (28). The original study protocol utilized standard techniques to identify enteropathogens in stools, but then quantitative PCR using custom-designed TaqMan Array Cards (ThermoFisher) was used to re-analyze the stool samples for the presence of 29 enteropathogens (29–31). Here, we focus on the pathogens that accounted for the majority of attributable diarrhea in the first year of life (29): adenovirus 40/41, norovirus, sapovirus, astrovirus, rotavirus, enterotoxigenic Escherichia coli (ETEC), Campylobacter spp. (pan-genus), typical enteropathogenic E. coli (tEPEC), Shigella, and Cryptosporidium. We additionally considered entero-aggregative E. coli (EAEC) as it was both frequently detected and associated with growth deficits (26). Three countries (Brazil, Peru, and South Africa) had national rotavirus vaccinations at the time of data collection and were excluded from models of rotavirus because vaccination (at ∼2 and 4 mo of age) alters the likelihood of infection and/or detection and thereby any association with breastfeeding. Following Rogawski McQuade et al. (32), pathogen presence was defined as a qPCR cycle threshold of <35. Coinfections were also identified when more than 1 pathogen was detected in a stool sample. An interview at enrollment asked for specific details about the timing of breastfeeding initiation, whether or not colostrum was given, and prelacteal feeding (33). During the twice-weekly surveillance visits, mothers were asked if they had breastfed the child on the previous day and whether or not other liquids or foods had been given and what foods or liquids they were. Infants who were fully breastfed were identified based on these reports. For analysis, we considered the proportion of visits that a child was fully breastfed in 2 ways. First, to determine whether full breastfeeding was associated with a lower likelihood of pathogen detection in stool, we focused on the 30-d period prior to each stool sample collection. We also continued to disaggregate time from the stool collection back to the child's enrollment in 30-d periods to evaluate period-specific associations with full breastfeeding. Second, to determine whether full breastfeeding was associated with delays in the detection of pathogens, we considered time since birth that a child was fully breastfed (exclusive of prelacteal feeding). Full breastfeeding as an exposure variable was described either as the proportion of visits between birth and when a given stool was collected or the proportion of time from birth to the age when a pathogen was first detected. In both cases, the proportion of time was multiplied by 10 to give a per 10%-time interpretation to coefficients. At enrollment, and then monthly, anthropometric assessments (weight, length) were performed by trained workers using standardized protocols (34). Building on risk factors associated with specific pathogens (35–39), we controlled for child sex and weight-for-age z score (WAZ) at enrollment, the latter evaluated here as a continuous z score following the WHO growth standards (40). Some pathogen detections were also associated with aspects of lower household socioeconomic status (SES) (35, 37, 41), which was evaluated by questionnaire twice yearly. The SES metric is described in detail elsewhere (42), but briefly was defined using an index (with a range of 0, low SES, to 1, high SES) that included access to improved water and sanitation, maternal education, average monthly household income, and a range of assets or household attributes (e.g., household crowding). For the purposes of these analyses, the mean SES index across all sampling points was multiplied by 10 to examine a per 10% change in SES. In sensitivity analyses, the raw components of the metric were also examined to determine whether they had greater explanatory power than the combined construct. The study was conducted in accordance with the Declaration of Helsinki. Field workers explained the study protocol and obtained written informed consent from a parent or guardian for the children enrolled in the original study. The study was approved by the following institutional review boards that correspond to each site and to collaborating institutions: Institutional Review Board for Health Sciences Research, University of Virginia, Charlottesville, VA, USA; the Committee for Ethics in Research, Universidade Federal do Ceara; National Ethical Research Committee, Health Ministry, Council of National Health in Brasília and Fortaleza—Brazil (Brazil site; BRF); Institutional Review Board, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA; PRISMA Ethics Committee; Health Ministry, in Loreto, Peru (Peru site; PEL); Health, Safety, and Research Ethics Committee, University of Venda; Department of Health and Social Development, Limpopo Provincial Government, in Venda, South Africa (South Africa site; SAV); Medical Research Coordinating Committee, National Institute for Medical Research; Chief Medical Officer, Ministry of Health and Social Welfare in Haydom, Tanzania (Tanzania site; TZH); Ethical Review Committee, Aga Khan University (Pakistan site in Naushahro Feroze; PKN); Ethical Review Committee, icddr,b in Dhaka—Bangladesh (Bangladesh site; BGD); Institutional Review Board, Christian Medical College, in Vellore, India, and the Health Ministry Screening Committee, Indian Council of Medical Research (India site; INV); Institutional Review Board, Institute of Medicine, Tribhuvan University; Ethical Review Board, Nepal Health Research Council; and Institutional Review Board, Walter Reed Army Institute of Research in Bhaktapur, Nepal (Nepal site; NEB). Site-specific descriptive characteristics of the study sample at enrollment were calculated, as well as site- and pathogen-specific distributions of child age at first detection. The proportion of visits at which full breastfeeding was reported was also calculated by site. Separate modeling approaches were used to address the following 2 questions. First, we hypothesized that full breastfeeding would be associated with a lower likelihood of pathogen detection in stool samples. To test this, a Bayesian multivariable logistic regression model was constructed for each pathogen, with the presence or absence of the pathogen in stool as the outcome. The proportion of visits with a report of full breastfeeding during the current month was the primary exposure variable. Also included were variables for the proportion of full breastfeeding during multiple prior 30-d periods (up to 120 d). Characterized in this way, we capture influences of current full breastfeeding and a history of earlier full breastfeeding. Covariates included infant sex, WAZ at enrollment, household SES, and the count of other pathogens detected in the same stool. Infant age (months) at stool collection was included in the model. Models were further adjusted for site and individual using random effects to account for repeated measurements (an error term to account for the correlation between measurements from the same individual). Second, we conducted a survival analysis to test the hypothesis that full breastfeeding would be associated with a delay in the time to first detection of a given pathogen. In this model, the proportion of time fully breastfed (proportion of visits from enrollment to any given stool sample) was treated as a time-varying variable (assuming a log-transformation of age) to account for potential changes with infant age in the association of full breastfeeding with first detection. Site was included as a frailty term, equivalent to a random effect in a linear regression to account for clustering in the repeated observations of each site, and covariates included sex, enrollment WAZ, SES, and the number of coincident enteropathogens. All analyses were conducted in R 4.1.0 (R Foundation for Statistical Computing).