Epidemiology and Impact of Campylobacter Infection in Children in 8 Low-Resource Settings: Results from the MAL-ED Study

Background. Enteropathogen infections have been associated with enteric dysfunction and impaired growth in children in low-resource settings. In a multisite birth cohort study (MAL-ED), we describe the epidemiology and impact of Campylobacter infection in the first 2 years of life. Methods. Children were actively followed up until 24 months of age. Diarrheal and nondiarrheal stool samples were collected and tested by enzyme immunoassay for Campylobacter. Stool and blood samples were assayed for markers of intestinal permeability and inflammation. Results. A total of 1892 children had 7601 diarrheal and 26 267 nondiarrheal stool samples tested for Campylobacter. We describe a high prevalence of infection, with most children (n = 1606; 84.9%) having a Campylobacter-positive stool sample by 1 year of age. Factors associated with a reduced risk of Campylobacter detection included exclusive breastfeeding (risk ratio, 0.57; 95% confidence interval,. 47-.67), treatment of drinking water (0.76; 0.70-0.83), access to an improved latrine (0.89; 0.82-0.97), and recent macrolide antibiotic use (0.68; 0.63-0.74). A high Campylobacter burden was associated with a lower length-for-age Z score at 24 months (-1.82; 95% confidence interval, -1.94 to -1.70) compared with a low burden (-1.49; -1.60 to -1.38). This association was robust to confounders and consistent across sites. Campylobacter infection was also associated with increased intestinal permeability and intestinal and systemic inflammation. Conclusions. Campylobacter was prevalent across diverse settings and associated with growth shortfalls. Promotion of exclusive breastfeeding, drinking water treatment, improved latrines, and targeted antibiotic treatment may reduce the burden of Campylobacter infection and improve growth in children in these settings. The MAL-ED study design and methodology have been described elsewhere [18]. The study was conducted at 8 sites: Dhaka, Bangladesh; Vellore, India; Bhaktapur, Nepal; Naushero Feroze, Pakistan; Venda, South Africa; Haydom, Tanzania; Fortaleza, Brazil; and Loreto, Peru. Children were enrolled from November, 2009 to February, 2012 and followed up through 24 months of age. Monthly anthropometry was performed [21]. Stool samples were collected in the absence of diarrhea at 1–12, 15, 18, 21, and 24 months of age as well as from each diarrhea episode, defined as maternal report of ≥3 loose stools in 24 hours or visible blood in stool and identified through twice-weekly home visits. Caregivers were surveyed biannually from 6 months of age, including questions about maternal income and education, the home environment, drinking water source and treatment, and the presence of animals. Crowding was defined as >2 persons per room living in the home. An improved latrine and water source were defined following World Health Organization guidelines [22]. Treatment of drinking water was defined as boiling, filtering, or adding bleach. Poor access to water was defined as having a primary drinking water source more than a 10-minute walk from the home. Twice-weekly home surveillance assessed breastfeeding in the prior day as exclusive (no consumption of other food or liquid), partial, or none, identified the introduction of specific foods, and recorded antibiotic use. All sites received ethical approval from their respective governmental, local institutional, and collaborating institutional review boards. Written informed consent was obtained from the parent or guardian of each child. The laboratory methods used in the MAL-ED study have been described elsewhere [10, 23]. Pertinently, EIA was performed for Campylobacter (ProSpecT) as well as Giardia and Cryptosporidium (TechLab). Monthly surveillance stool samples were also tested for myeloperoxidase (MPO; measured in nanograms per milliliter), a marker of neutrophil activity in the intestinal mucosa (Alpco); neopterin (NEO; measured in nanomoles per liter), a marker of T-helper cell 1 activity (GenWay Biotech); and α-1-antitrypsin (AAT; measured in milligrams per gram), a marker of intestinal permeability (Biovendor). Blood samples collected at 7, 15 and 24 months were tested for α-1-acid glycoprotein (AGP; measured in milligrams per deciliter), a marker of systemic inflammation. To identify factors associated with Campylobacter detection in surveillance stool samples, we used generalized estimating equations to fit a generalized linear model with a first-order autoregressive working correlation matrix and robust variance to account for nonindependence of stool testing within each child. To estimate risk ratios for Campylobacter detection, we used Poisson regression as an approximation of log-binomial regression since the log-binomial models did not converge [24]. First, we estimated the association for each factor of interest with Campylobacter detection, adjusting for age (using a natural spline with knots at 6, 12, and 18 months), sex, site, and season via the terms sin(2mπ/12) + cos(2mπ/12), where m is the month of the year as well as—given possible variation in seasonality between sites—an interaction between these terms and site [25]. Then, based on statistical significance, model fit based on the quasi-likelihood information criterion and an assessment of covariance between individual factors, we fit a multivariable model. We also fit site-specific models, excluding those variables from the multivariable model that did not vary within specific sites. Finally, to further describe any association with recent antibiotic use, we fit 2 multivariable models, adjusted as described above, which included (1) class-specific antibiotic use in the prior month and (2) class-specific use in 15-day windows over the prior 60 days. For the analysis of linear growth, included individuals were required to have a length-for-age Z (LAZ) score at 24 months of age. We excluded children from the Pakistan site, owing to bias noted in a subset of length measurements at this site. To estimate the association between the burden of detection of an individual pathogen and 24-month LAZ score, we calculated the enteropathogen burden for each subject using the surveillance stool samples (namely, samples positive/samples tested). Similar burden indices were calculated for the 0–6, 7–12, and 13–24 month intervals. Persistent infection was defined as detection of Campylobacter from all surveillance stool samples tested during a 3-month period. We then fit a multiple linear regression, including enrollment LAZ score, sex, site, and Campylobacter burden and further adjusted for possible confounders, including factors associated with Campylobacter infection in the multivariable model as well as highly-correlated pathogens. To calculate model-predicted 24-month LAZ scores, we calculated predicted population marginal effects using least-squares means [26]. Overall and site-specific high and low burdens of Campylobacter were defined as the 90th and 10th percentile of Campylobacter burden, respectively. To estimate the association between Campylobacter detection and fecal markers of intestinal permeability and inflammation collected in surveillance stool samples (MPO, NEO, and AAT), we used generalized estimating equations to fit a generalized linear model, as described previously but using a gaussian distribution. These models adjusted for age, sex and site, as previously described, and we also fit site-specific models. Finally, to describe the association between Campylobacter burden and systemic inflammation, we fit multiple linear regression models and calculated predicted population marginal effects using least-squares means, both for the entire cohort and for each site, as described for the analysis of linear growth, but with the mean AGP value for each individual as the response variable instead of 24-month LAZ score. All statistical analysis was performed using R software, version 3.2.2 (Foundation for Statistical Computing).