Investigating One Health risks for human colonisation with extended spectrum β-lactamase-producing Escherichia coli and Klebsiella pneumoniae in Malawian households: a longitudinal cohort study

Abstract

Background: Low-income countries have high morbidity and mortality from drug-resistant infections, especially from enteric bacteria such as Escherichia coli. In these settings, sanitation infrastructure is of variable and often inadequate quality, creating risks of extended-spectrum β-lactamase (ESBL)-producing Enterobacterales transmission. We aimed to describe the prevalence, distribution, and risks of ESBL-producing Enterobacterales colonisation in sub-Saharan Africa using a One Health approach.

Methods: Between April 29, 2019, and Dec 3, 2020, we recruited 300 households in Malawi for this longitudinal cohort study: 100 each in urban, peri-urban, and rural settings. All households underwent a baseline visit and 195 were selected for longitudinal follow-up, comprising up to three additional visits over a 6 month period. Data on human health, antibiotic usage, health-seeking behaviours, structural and behavioural environmental health practices, and animal husbandry were captured alongside human, animal, and environmental samples. Microbiological processing determined the presence of ESBL-producing E coli and Klebsiella pneumoniae, and hierarchical logistic regression was performed to evaluate the risks of human ESBL-producing Enterobacterales colonisation.

Findings: A paucity of environmental health infrastructure and materials for safe sanitation was identified across all sites. A total of 11 975 samples were cultured, and ESBL-producing Enterobacterales were isolated from 1190 (41·8%) of 2845 samples of human stool, 290 (29·8%) of 973 samples of animal stool, 339 (66·2%) of 512 samples of river water, and 138 (46·0%) of 300 samples of drain water. Multivariable models illustrated that human ESBL-producing E coli colonisation was associated with the wet season (adjusted odds ratio 1·66, 95% credible interval 1·38-2·00), living in urban areas (2·01, 1·26-3·24), advanced age (1·14, 1·05-1·25), and living in households where animals were observed interacting with food (1·62, 1·17-2·28) or kept inside (1·58, 1·00-2·43). Human ESBL-producing K pneumoniae colonisation was associated with the wet season (2·12, 1·63-2·76).

Interpretation: There are extremely high levels of ESBL-producing Enterobacterales colonisation in humans and animals and extensive contamination of the wider environment in southern Malawi. Urbanisation and seasonality are key risks for ESBL-producing Enterobacterales colonisation, probably reflecting environmental drivers. Without adequate efforts to improve environmental health, ESBL-producing Enterobacterales transmission is likely to persist in this setting.

Funding: Medical Research Council, National Institute for Health and Care Research, and Wellcome Trust.

Translation: For the Chichewa translation of the abstract see Supplementary Materials section.

Figure 1

Households recruitment, visits, and loss to follow-up

Figure 2

Regional differences in the prevalence of ESBL-producing Enterobacterales colonisation and environmental contamination (A) Proportion of samples positive for ESBL-producing Enterobacterales at urban, peri-urban, and rural households. p values indicate significant regional differences in the prevalence of ESBL-producing Enterobacterales for each sample type. (B) Breakdown of urban, peri-urban, and rural proportions of ESBL-producing Enterobacterales. (C) Proportion of the household human stool, animal stool, and environmental samples positive for ESBL-producing Escherichia coli or ESBL-producing Klebsiella pneumoniae, stratified by sample type, bacterial species, and region. Significant variations in the proportion of ESBL-producing E coli vs ESBL-producing K pneumoniae by sample type, assessed by χ² test, are shown. ESBL=extended spectrum β-lactamase.

Figure 3

Confidence ellipses of regional effects for the (A) individual-level, (B) household-level, and (C) environmental contamination datasets, from the first two principal components Points in (A) represent individuals and points in (B) and (C) represent households. PCA=principal component analysis.

Figure 4

Parameter estimates for the fixed effects used in a multivariable logistic regression model of (A) ESBL-producing Escherichia coli and (B) ESBL-producing Klebsiella pneumoniae colonisation Data are expressed as odds ratios (ORs) with 95% credible intervals (CrIs). The distribution of random effects is visualised in appendix 2 (p 33). Separate variable selection and model fit procedures were carried out for ESBL E coli and ESBL K pneumoniae. Individual and household variables identified through pre-screening were included alongside region and a random intercept per individual, nested within a random intercept per household. The wet season is defined as Nov–April and dry as May–Oct. ESBL=extended spectrum β-lactamase.