Estimation of human feces-associated extended spectrum beta-lactamase-producing E. coli discharge into sanitation systems: a first global- and regional-scale calculation

Abstract

Background:

Improving management of and treatment within sanitation waste streams may slow the development and transmission of antimicrobial resistant (AMR) organisms, but the magnitude of impact has not been quantified. Extended-spectrum beta-lactamase-producing (ESBL) E. coli are an important cause of AMR infections and are frequently detected in fecal waste streams, making them model indicators of the distribution of fecal-oral AMR organisms.

Methods:

We used publicly-available sanitation coverage data and literature on human fecal production and ESBL E. coli carriage by healthy individuals to estimate the feces discharged containing ESBL E. coli by region through levels of the Sustainable Development Goal sanitation ladder.

Findings:

About 66% of the prospective global fecal biomass containing ESBL E. coli was managed in at least basic sanitation systems, 12% in limited sanitation, and 25% in unimproved sanitation (14%) or openly defecated (11%). Open defecation of ESBL E. coli was common in Southeast Asia (29%) and Africa (22%). Fecal biomass containing ESBL E. coli totaled 1.9–190 billion kg by carriage scenario, with 1.2–120 billion kg managed by at least basic sanitation and 220 million-22 billion kg openly defecated, and the remainder in limited or unimproved sanitation systems. Southeast Asia, Western Pacific, and Africa produced 63% of the total global fecal biomass, but 90% of feces containing ESBL E. coli.

Interpretation:

Although safely-managed and basic sanitation do not guarantee effective removal or inactivation of AMR organisms; our results indicate the need for mitigating AMR transport via unsafely-managed sanitation, including open defecation, which may result in direct environmental discharge and subsequent risk of transmission back to humans.

Funding:

No funding associated with this study.

Background

Antimicrobial resistance (AMR) is one of the most pressing global public health threats. Antimicrobial use increased by 65% worldwide, predominantly in low- and middle-income countries, from 2000–2010.¹ AMR is conservatively estimated to contribute to 700,000 deaths per year currently, a figure estimated to grow to 10 million annually and cost 10,000 US dollars per person by 2050 as currently-effective therapies wane and AMR infections spread.^2,3

Extended-spectrum β-lactamase (ESBL) mediated resistance is a growing AMR concern. Identification of ESBL resistance mechanisms have increased rapidly, both inside and outside of healthcare settings, as shown by its prevalence in the human gut, wastewater, and fecal sludge in high and low-income contexts.^4–7 ESBL-producing enteric organisms, such as ESBL E. coli, have also demonstrated the potential for rapid spread and colonization via global travel,⁸ with prevalent reservoirs in water, sewage, and soil.^5,9,10 Additionally, though a cause of significant clinical symptoms in patients, ESBL E. coli is also associated with a relatively high prevalence of asymptomatic carriage—and estimated transmission—among otherwise healthy individuals with community-based estimates available.^11,12 All of these characteristics make ESBL E. coli a model indicator for AMR enteric bacteria that are transmitted through the fecal-oral route.

Efforts to combat AMR have mainly focused on reducing and optimizing antibiotic use and on advocating for new antibiotics.² Beyond these two approaches, and in line with strategic objective three of the AMR Global Action Plan,¹³ it is understood that improving water, sanitation, and hygiene (WASH) infrastructure may reduce both rates of infection (and therefore the need for antibiotics) and, in areas where water used for drinking or bathing may be contaminated with feces, overall exposure to AMR organisms.^2,14,15 Still, existing wastewater and fecal sludge treatment technologies in high and low-income settings may not eliminate AMR organisms and their genes.^16–18 Although recent reports indicate access to WASH may be key in the fight against AMR,^19,20 the potential impact of WASH interventions on the frequency of infections caused by AMR organisms, or even the magnitude of AMR organisms transmitted through the environment via fecal waste streams, has not been quantified.

In order to estimate the global scale at which sanitation infrastructure currently serves as a vehicle for dissemination of AMR organisms (and therefore could reduce dissemination of AMR organisms if safely managed) and understand how these estimates may change with growing population and carriage rates, we calculated the total global human fecal discharge of ESBL E. coli (defined as PCR detection of a CTX-M, SHV- or TEM-type ESBL gene) into sanitation systems both currently (2015) and as predicted in 2030. We estimated discharge by World Health Organization (WHO) region and at different levels of the Sustainable Development Goals (SDG) sanitation ladder²¹ to calculate region-specific and sanitation technology-specific loads, given most recent (2015) coverage levels. These estimates provide a first accounting of the dissemination of AMR organisms through sanitation systems globally, and allow us to better understand the impact of increasing coverage with safely-managed sanitation systems that would more effectively limit dissemination of these organisms.

Methods

We applied approaches to estimating global and region-specific production of human feces²² to coverage and population estimates at each level of the SDG sanitation ladder produced by the Joint Monitoring Program (JMP) of the WHO and United Nations Children’s Fund (UNICEF) in 2015.²³ Briefly, as described in Berendes et al.,²² we estimated current (2015) annual human fecal production by combining the current population estimates from the World Bank,²⁴ region-specific estimates of average human body mass,²⁵ and a daily body mass-fecal production equation for mammals, including humans²⁶ that we extended to annual (yearly) production. Estimates of the population using safely-managed sanitation, basic sanitation, limited sanitation, unimproved sanitation, and open defecation were obtained from recent JMP data.²³ We estimated the proportion of humans in a given region discharging ESBL E. coli using previous region-specific ESBL carriage estimates from healthy individuals estimated by the Karanika et al. (2016) meta-analysis of 66 studies selected after screening 17,479 studies for eligibility.¹² ESBL E. coli is limited to isolates with PCR detection of a CTX-M, SHV-, or TEM-type ESBL gene.¹² Table 1 provides greater detail of the specific values (listed as [A], [B], [C], and [D]) used from each data source.

Table 1:

Values and sources for calculations

Value	Source	Notes
[A] Percent of population (by region or globally) carrying ESBL E. coli	Karanika et al. 201612 using clinical and molecular studies of healthy individuals	Used region-specific mean estimates for calculation and region-specific lower and upper bounds of 95% confidence interval for ‘low’ and ‘high’ estimates
[B] Population (by region or globally) using a particular sanitation system	2015 JMP data23	Categories23: Safely-managed sanitation (grouped with ‘Basic sanitation’ to form ‘At least basic sanitation’ category where safely-managed estimates were unavailable) Basic sanitation Limited sanitation Unimproved sanitation Open defecation (no sanitation facility)
[C] Per capita kg human feces produced annually (by region or globally)	Berendes et al. 201822 (using mammalian mass-based calculations26 and region-specific estimates of human biomass25)	Used region-specific human fecal production estimates for calculation, including region-specific lower and upper estimates (based variability in regional human biomass25) for ‘low’ and ‘high’ estimates
[D] Percent of E. coli shed in feces of a carrier that is ESBL (or ESBL with CTX-M)	Johnson et al. 200828	Because shedding of E. coli clones in health individuals may vary from 1–11 (median: 2), we presented 3 scenarios: Scenario A-assumes 100%, i.e. only ESBL E. coli clones, shed Scenario B-assumes 10%, i.e. ESBL E. coli is one of 10 clones shed Scenario C-assumes 1%, i.e. ESBL E. coli is one of 100 clones shed

We calculated fecal biomass per person based on adults, given limited data on human body biomass for children, changing age distributions,²⁵ and pediatric carriage of ESBL E. coli. We also assumed region-specific prevalence of ESBL E. coli from previous literature¹² applied uniformly across users of different sanitation systems in that region, though sub-regional and sub-national variation may occur but are, as yet, unquantified. We also did not extend management of feces to include estimates of the effectiveness of treatment, given significant uncertainty in estimates of safe management and final treatment^21,27 as well as unresolved research into the efficacy of treatment systems at removing or mitigating antimicrobial resistance mechanisms in excreta.¹⁸ Healthy adults shed up to 6 unique clones of E. coli in their stool, with the majority (77%) shedding a single clone.²⁸ Given limited data on the extent to which carriers of ESBL E. coli (or other types of E. coli) excrete clonal vs. multiple types of E. coli,²⁸ we modeled log₁₀-ordered scenarios where ESBL E. coli excreted by carriers represented 100% (Carriage Factor A), 10% (Carriage Factor B), and 1% (Carriage Factor C) of all fecally excreted E. coli. Using these data sources, we calculated—by region, and summed globally—the following estimates (letters indicate values from Table 1, which are described briefly here and in greater depth in the table):

1. Percentage of feces containing ESBL E. coli entering different levels of the SDG sanitation ladder in a given region or globally

   • Main estimate: ([A_region] × [B_region] × [C_region] × [D_{carriage factors}])/([A] × [B] × [C] × [D_scenarios])

     • Where A is the percent of population (regionally or globally) carrying ESBL E. coli, B is the total population using a particular level of the sanitation ladder (for a given region), C is the mass of human feces produced annually in a region, and D is the different Carriage Factors (A-C) describe above
   • Low estimate: [A_{lower 95% bounds for region}] × [B_region] × [C_{low estimate for region}] × [D_{carriage factors}]/[A_{lower 95% bounds}] × [B] × [C_{low estimate}] × [D_scenarios]
   • High estimate: [A_{higher 95% bounds for region}] × [B_region] × [C_{high estimate for region}] × [D_{carriage factors}]/[A_{higher 95% bounds}] × [B] × [C_{high estimate}] × [D_scenarios]
2. Kg/year of feces entering sanitation systems (by type) that contained ESBL E. coli (“kg of feces with ESBL E. coli”)

   • Main estimate: [A_region] × [B_region] × [C_region] × [D_{carriage factors}]
   • Low estimate: [A_{lower 95% bounds for region}] × [B_region] × [C_{low estimate}] × [D_{carriage factors}]
   • High estimate: [A_{higher 95% bounds for region}] × [B_region] × [C_{high estimate}] × [D_{carriage factors}]

As described above, high and low estimates are not 95% confidence intervals for the mean value, but rather more conservative upper and lower bounds of the range, given current areas of uncertainty.

For predictions of global fecal biomass containing ESBL E. coli in 2030, we used World Bank projected population figures for 2030 at regional levels as population estimates²⁹ and estimated growth in concomitant fecal biomass in 2030,²² accounting for growth in carriage rates of ESBL E. coli by projecting estimates of annual growth from systematic reviews of carriage of ESBL E. coli globally¹² (5.38% increased carriage/year) or ESBL Enterobacteriaceae regionally³⁰ (ranging from 0.5% - 7.7%/year by region) forward to 2030. Current and projected carriage rates estimates from literature are summarized in Table S1. Using the most conservative global carriage trends, we also estimated the following differing interventions:

1. current estimated growth in population with 50% reductions in the estimated annual rate of change of ESBL E. coli carriage globally;

2. current estimated growth in population only (i.e. current ESBL E. coli carriage rates held constant at present day rates);

3. current estimated growth in population with 50% reductions in the estimated annual rate of increase of ESBL E. coli carriage in Africa, Southeast Asia, and the Western Pacific (where WASH efforts are current most focused), but rates of increase as per #1 in other regions;

4. current estimated growth in population only in Africa, Southeast Asia, and the Western Pacific (current ESBL E. coli carriage rates held constant at present day rates in those regions), but rates of increase as per #1 in other regions.

All analyses were performed in Microsoft Excel 2016 (Microsoft, Redmond, WA, USA) and R Version 3.4.3 (R Foundation for Statistical Computing, Vienna, Austria).³¹

Role of the funding source:

There was no specific funding for this project. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication

Results

Globally, we estimate that 190 billion kg of feces (scenario A), 19 billion kg (scenario B), and 1.9 billion kg (scenario C) carrying ESBL E. coli were excreted in 2015 (Figure 1, representing scenario B in biomass estimates). The Western Pacific (53%) and Southeast Asia (24%) accounted for the majority of all annual ESBL E. coli-carrying fecal biomass globally, followed by Africa (13%), the Eastern Mediterranean (6%), Europe (2%), and the Americas (1%, Figure 2A). Comparatively, the Western Pacific accounted for 26% of all annual fecal biomass, followed by Southeast Asia (24%), the Americas (15%), Europe and Africa (13% each), and Eastern Mediterranean (9%).

Figure 1:

Fecal biomass containing ESBL E. coli entering levels of the JMP sanitation ladder under Scenario B (10% ESBL E. coli discharge by carriers), globally and by region. Under Scenarios A and C (100% and 1% ESBL E. coli discharge), absolute estimates in Figure 1 would increase or decrease (respectively) evenly by 10-fold, respectively, but relative ratios of production would remain constant. AFRO: African region; AMRO: Americas region; EMRO: Eastern Mediterranean region; EURO: European region; SEARO: Southeast Asian region; WPRO: Western Pacific region.

Figure 2:

Regional global fecal biomass and fecal biomass containing ESBL E. coli based on Karanika et al. (2016) global estimates of ESBL E. coli carriage trends¹² (A) and Woerther et al. (2013) regional estimates of ESBL Enterobacteriaceae carriage trends³⁰ (B). Estimates of regional composition of the global fecal biomass for both (A) and (B) are adapted from Berendes et al.²² Total estimates of feces containing ESBL E. coli are calculated under scenario B, 10% clonal ESBL E. coli (see Table 3).

Percentages and total biomass estimates of feces containing ESBL E. coli discharged into sanitation systems are shown in Tables 2 and 3, respectively. Almost two-thirds (66%, Table 2) of the total fecal biomass containing ESBL E. coli (1.2–120 billion kg by scenario, Table 3) were discharged into at least basic sanitation (Table 2). About 8% (160 million-16 billion kg), 14% (270 million-27 billion kg), and 11% (220 million-22 billion kg) of feces carrying ESBL E. coli were discharged into limited sanitation systems, unimproved systems, or openly defecated, respectively. The regions with the highest proportion of openly defecated feces containing ESBL E. coli were Southeast Asia (29%), Africa (22%), and the Eastern Mediterranean (8%) (Figure 1, Table 2). Africa had the most even distribution of feces containing ESBL E. coli across different levels of the sanitation ladder—31% to at least basic sanitation, 18% to limited sanitation, 30% to unimproved sanitation, and 22% to open defecation—while in other regions the majority (>50%) was discharged into at least basic sanitation.

Table 2:

Percentage of feces containing ESBL E. coli discharged into sanitation systems for each region/globally

Region	Safely-managed1	Basic	Limited	Unimproved	Open defecation
Africa	30.5%		17.6%	30.1%	21.8%
America	42.7%	48.1%	3.23%	4.04%	1.96%
Eastern Mediterranean	72.9%		6.87%	12.5%	7.78%
Europe	67.5%	28.5%	1.11%	2.81%	0.04%
Southeast Asia	50.9%		13.0%	6.72%	29.4%
Western Pacific	57.4%	20.6%	4.8%	15.1%	2.15%
World	65.8%		8.38%	14.4%	11.4%

¹Safely-managed sanitation estimates for Africa, Eastern Mediterranean, and Southeast Asia regions could not be quantified in 2015, thus estimates represent ‘at least basic’ sanitation (combined basic and safely-managed categories)

Table 3:

Kg of feces/year containing ESBL E. coli (all) in sanitation systems by ESBL E. coli discharge scenario^a

	Kg of feces/year carrying ESBL E. coli (low estimate, high estimate)
Region	Safely-managed	Basic	Limited	Unimproved	Open defecation
Scenario A (100%)
Africa	7.3 × 109 (1.5 × 109, 1.7 × 1010)		4.2 × 109 (8.7 × 108, 9.9 × 109)	7.2 × 109 (1.5 × 109, 1.7 × 1010)	5.2 × 109 (1.1 × 109, 1.2 × 1010)
Americasb	1.1 × 109 (0, 3.0 × 109)	1.2 × 109 (0, 3.3 × 109)	8.3 × 107 (0, 2.2 × 108)	1.0 × 108 (0, 2.8 × 108)	5.0 × 107 (0, 1.4 × 108)
Eastern Mediterranean	8.8 × 109 (2.2 × 109, 2.0 × 1010)		8.3 × 108 (2.0 × 108, 1.9 × 109)	1.5 × 109 (3.7 × 108, 3.4 × 109)	9.4 × 108 (2.3 × 108, 2.1 × 109)
Europe	3.1 × 109 (1.4 × 109, 4.2 × 109)	1.3 × 109 (6.0 × 108, 1.8 × 109)	5.1 × 107 (2.3 × 107, 6.8 × 107)	1.3 × 108 (5.9 × 107, 1.7 × 108)	1.6 × 106 (7.5 × 105, 2.2 × 106)
Southeast Asia	2.3 × 1010 (6.6 × 109, 5.0 × 1010)		5.8 × 109 (1.7 × 109, 1.3 × 1010)	3.0 × 109 (8.6 × 108, 6.6 × 109)	1.3 × 1010 (3.8 × 109, 2.9 × 1010)
Western Pacific	5.8 × 1010 (3.3 × 1010, 8.6 × 1010)	2.1 × 1010 (1.2 × 1010, 3.1 × 1010)	4.8 × 109 (2.8 × 109, 7.1 × 109)	1.5 × 1010 (8.7 × 109, 2.3 × 1010)	2.2 × 109 (1.2 × 109, 3.2 × 109)
World	1.2 × 1011 (5.7 × 1010, 2.2 × 1011)		1.6 × 1010 (5.5 × 109, 3.2 × 1010)	2.7 × 1010 (1.2 × 1010, 5.0 × 1010)	2.2 × 1010 (6.3 × 109, 4.7 × 1010)
Scenario B (10%)
Africa	7.3 × 108 (1.5 × 108, 1.7 × 109)		4.2 × 108 (8.7 × 107, 9.9 × 108)	7.2 × 108 (1.5 × 108, 1.7 × 109)	5.2 × 108 (1.1 × 108, 1.2 × 109)
Americasb	1.1 × 108 (0, 3.0 × 108)	1.2 × 108 (0, 3.3 × 108)	8.3 × 106 (0, 2.2 × 107)	1.0 × 107 (0, 2.8 × 107)	5.0 × 106 (0, 1.4 × 107)
Eastern Mediterranean	8.8 × 108 (2.2 × 108, 2.0 × 109)		8.3 × 107 (2.0 × 107, 1.9 × 108)	1.5 × 108 (3.7 × 107, 3.4 × 108)	9.4 × 107 (2.3 × 107, 2.1 × 108)
Europe	3.1 × 108 (1.4 × 108, 4.2 × 108)	1.3 × 108 (6.0 × 107, 1.8 × 108)	5.1 × 106 (2.3 × 106, 6.8 × 106)	1.3 × 107 (5.9 × 106, 1.7 × 107)	1.6 × 105 (7.5 × 104, 2.2 × 105)
Southeast Asia	2.3 × 109 (6.6 × 108, 5.0 × 109)		5.8 × 108 (1.7 × 108, 1.3 × 109)	3.0 × 108 (8.6 × 107, 6.6 × 108)	1.3 × 109 (3.8 × 108, 2.9 × 109)
Western Pacific	5.8 × 109 (3.3 × 109, 8.6 × 109)	2.1 × 109 (1.2 × 109, 3.1 × 109)	4.8 × 108 (2.8 × 108, 7.1 × 108)	1.5 × 109 (8.7 × 108, 2.3 × 109)	2.2 × 108 (1.2 × 108, 3.2 × 108)
World	1.2 × 1010 (5.7 × 109, 2.2 × 1010)		1.6 × 109 (5.5 × 108, 3.2 × 109)	2.7 × 109 (1.2 × 109, 5.0 × 109)	2.2 × 109 (6.3 × 108, 4.7 × 109)
Scenario C (1%)
Africa	7.3 × 107 (1.5 × 107, 1.7 × 108)		4.2 × 107 (8.7 × 106, 9.9 × 107)	7.2 × 107 (1.5 × 107, 1.7 × 108)	5.2 × 107 (1.1 × 107, 1.2 × 108)
Americasb	1.1 × 107 (0, 3.0 × 107)	1.2 × 107 (0, 3.3 × 107)	8.3 × 105 (0, 2.2 × 106)	1.0 × 106 (0, 2.8 × 106)	5.0 × 105 (0, 1.4 × 106)
Eastern Mediterranean	8.8 × 107 (2.2 × 107, 2.0 × 108)		8.3 × 106 (2.0 × 106, 1.9 × 107)	1.5 × 107 (3.7 × 106, 3.4 × 107)	9.4 × 106 (2.3 × 106, 2.1 × 107)
Europe	3.1 × 107 (1.4 × 107, 4.2 × 107)	1.3 × 107 (6.0 × 106, 1.8 × 107)	5.1 × 105 (2.3 × 105, 6.8 × 105)	1.3 × 106 (5.9 × 105, 1.7 × 106)	1.6 × 104 (7.5 × 103, 2.2 × 104)
Southeast Asia	2.3 × 108 (6.6 × 107, 5.0 × 108)		5.8 × 107 (1.7 × 107, 1.3 × 108)	3.0 × 107 (8.6 × 106, 6.6 × 107)	1.3 × 108 (3.8 × 107, 2.9 × 108)
Western Pacific	5.8 × 108 (3.3 × 108, 8.6 × 108)	2.1 × 108 (1.2 × 108, 3.1 × 108)	4.8 × 107 (2.8 × 107, 7.1 × 107)	1.5 × 108 (8.7 × 107, 2.3 × 108)	2.2 × 107 (1.2 × 107, 3.2 × 107)
World	1.2 × 109 (5.7 × 108, 2.2 × 109)		1.6 × 108 (5.5 × 107, 3.2 × 108)	2.7 × 108 (1.2 × 108, 5.0 × 108)	2.2 × 108 (6.3 × 107, 4.7 × 108)

Colors are coded by order of magnitude (darkest gray = 10¹¹, lightest gray = 10⁷, white = ≤10⁶). Numbers in parentheses represent conservative higher and lower bounds of the range, given current areas of uncertainty, as described in the methods.

^aScenario A assumes that 100% of E. coli discharged by an individual carrying ESBL E. coli are ESBL, while Scenario B assumes 10% (minimum value from Johnson et al. 2008 analysis of E. coli clonal carriage), and Scenario C assumes 1% as a conservative minimum value.

^bLower bounds of 95% confidence interval for estimate of prevalence of ESBL E. coli carriage in Karanika et al. (2016)¹² was 0, therefore lower bounds are 0).

Southeast Asia discharged 130 million-13 billion kg via open defecation alone, followed by Africa (50 million-5 billion kg), and the Western Pacific (20 million-2 billion kg). The Western Pacific had the largest discharge of ESBL E. coli-containing feces into at least basic systems (790 million-79 billion kg), followed by Southeast Asia (230 million-23 billion kg).

When projected to 2030 assuming 10% of E. coli discharged are ESBL (see scenario B in Table 3), global feces discharged carrying ESBL E. coli would almost double from 2015 estimates to 37.6 billion kg when using global trends in ESBL E. coli from Karanika et al.¹² (Figure 2A). Under less conservative region-specific trends for ESBL Enterobacteriaceae from Woerther et al.³⁰, estimates increase about 2.5-fold to 49.6 billion kg (Figure 2B). The Western Pacific (Figure 2A) or Southeast Asia (Figure 2B) contribute a higher proportion of the global fecal biomass containing ESBL E. coli than to the global fecal biomass in general, but contributions from other regions, such as Africa, increase due to both larger population and more ESBL E. coli carriage.

When projected to 2030 using the most conservative trend estimates (Karanika et al.)¹² under interventions 1–4 and assuming 10% of E. coli discharged are ESBL (Figure 3), global feces discharged carrying ESBL E. coli is estimated to be 37.2 billion kg under Intervention 1 (50% reduction in global projected rates of increases in ESBL E. coli carriage), representing a 1.1% decrease from our projection in Figure 2. Under Intervention 2 (100% reduction in global projected rates of increases in ESBL E. coli carriage, i.e. carriage rates remain constant at present day levels), we estimate 21.8 billion kg (32% decrease). Under Intervention 3 (Intervention 1 conditions focused in Africa, Southeast Asia, and the Western Pacific only), we estimate 37.3 billion kg (0.8% decrease). Under Intervention 4 (Intervention 2 conditions focused in Africa, Southeast Asia, and the Western Pacific only), we estimate 27.6 billion kg (27% decrease). Interventions 1–4 estimates under the less conservative trend estimate (Woerther et al.),³⁰ which ranged from a 0.1% decrease (under Intervention 3) to a 24% decrease (under Intervention 2), are presented in Table S2.

Figure 3:

Estimates of kg of feces containing ESBL E. coli in 2030 by region and intervention, assuming 10% of E. coli discharged by an individual carrying ESBL E. coli are ESBL.

Discussion

We combined estimates of fecal biomass production with estimates of sanitation coverage and ESBL E. coli carriage to present calculations used to derive the first estimations of the distribution of human-associated feces containing ESBL E. coli that needs to be treated by sanitation systems currently and projected to 2030. These methods and results represent an initial estimation of the global distribution of fecally-excreted AMR organisms and their associated load in human waste treatment systems. Out of the total current fecal biomass containing ESBL E. coli estimated to be discharged, two-thirds was discharged into at least basic sanitation systems but one quarter was discharged into unimproved systems or via open defecation, mostly in low- and middle-income countries (LMICs). It is important to note that even technologies on the highest SDG sanitation ladder level—’safely-managed’—do not guarantee effective treatment and removal of AMR organisms, such as ESBL E. coli, and their AMR genes.^32,33 Projections to 2030 reveal that major, comprehensive changes are needed to combat the combination of high, increasing carriage rates and poor WASH in critical, high population and low-income areas of the globe.

Both prevalent antibiotic use in young children—98% of children are thought to be exposed to antibiotics by 6 months of age in LMICs³⁴—and poor WASH conditions leading to frequent bacterial infections³⁵ will likely contribute to continued increases in carriage rates in these settings by 2030. Our projected results and intervention scenarios suggest that even relatively large (50%) reductions in the rate of change of ESBL E. coli carriage through environmental (WASH) preventive measures and antibiotic stewardship or other interventions would only have minimal (<2%) impact on total feces containing ESBL E. coli in 2030. Rather, there is a need for large-scale efforts to prevent new ESBL E. coli colonization, especially in low- and middle-income countries, to avert up to one-third of the fecal-derived ESBL E. coli hazards.

Though our data suggest ESBL E. coli are shed in significant quantities of feces in most of the world, this analysis also suggests low-income settings—particularly Southeast Asia, the Western Pacific, and Africa—as regions of focus for environmental efforts to curb its spread via fecal discharge. These regions comprised about two-thirds of the world’s population (65%) and fecal discharge (62%) in 2015,^22–24 yet produced 90% of the global feces containing ESBL E. coli. Southeast Asia and Africa had the largest proportions of feces containing ESBL E. coli that were openly defecated, and most sanitation facilities in these regions are onsite, decentralized systems (e.g. pit latrines with emptying).²² Correspondingly, these regions continue to experience among the highest global burdens of diarrhea, enteric infections, and other sanitation-related morbidities,³⁶ such that the addition of AMR in fecal exposures may spread very quickly and be particularly devastating. These populations, and their associated human and animal fecal biomass—with and without ESBL E. coli—will also continue to grow, especially in Africa, which is projected to have the largest human population increases by 2030.^22,29 Notably, our data were limited to ‘healthy’, non-care-seeking individuals as per previously-defined criteria,¹² which may be an underestimate of true discharge as it omits feces from hospitals and other institutions caring for non-healthy individuals.³⁷ Our estimates may also vary due to variation in human fecal production (e.g. regionally by biomass of the individual, diet, and other factors^25,38) and increased carriage rates beyond those estimated in 2015.^37,39 Of note, modeled estimates of human feces production²⁶ we used were slightly larger than those estimated empirically from sporadic studies.³⁸

While we cannot directly estimate the treatment efficacy associated with onsite, decentralized sanitation facilities in many of these countries,²¹ safe emptying, management, and treatment of pathogens remains a significant challenge in low-income settings.⁴⁰ Onsite systems such as latrines without slabs and proper lining that are classified as ‘unimproved’ are not considered to safely separate users from excreta and associated risks.²¹ Among those ‘improved’ systems that are shared (‘limited’ systems), concerns of elevated transmission risk for diarrhea exist from the multiple, daily users,⁴¹ which could be extended to transmission of ESBL E. coli. However, mechanistic evidence around exact transmission pathways for fecal contamination (including AMR fecal contamination) in shared sanitation settings is needed.⁴² Beyond onsite systems, a significant proportion of fecal waste from sanitation facilities connected to drains or sewers in these settings may still go untreated,²⁷ therefore unsafe management of feces with ESBL E. coli is likely underestimated in our calculations.

We note, however, that use of safely-managed sanitation, such as sewered sanitation connected to a functional wastewater treatment plant, does not—by itself—imply total removal of ESBL E. coli or other AMR organisms, or their resistance genes, from feces, fecal sludge, and wastewater before discharge into the environment. A recent analysis indicated that moving from open defecation and unimproved sanitation to improved sanitation can significantly reduce the environmental load of AMR genes, with additional reductions when moving from secondary to tertiary treatment.⁴³ There is growing evidence that viable AMR bacteria, including ESBL E. coli, is detected in effluent from wastewater treatment plants,^32,33,44,45 which may represent a risk to public health if these waters are used for drinking, personal hygiene, irrigation, or recreation. Therefore, further research into effective technologies for removing these pathogens in both wastewater treatment plants and decentralized sanitation systems (fecal sludge treatment plants or onsite treatment) is urgently needed. While studies have hypothesized selection pressure by residual antibiotics in the wastewater treatment stream to be the major contributor to AMR organism discharge, recent evidence suggests that risk of detecting AMR organisms in the environment is more highly correlated with amount of fecal discharge in an area,⁴⁶ with the exception of waters downstream of antimicrobial manufacturing sites. These uncertainties suggest a need for further monitoring and evaluation of discharge limits either based on risk evaluation or the ‘best available techniques’ approach. Modifications to onsite treatment systems may also be warranted in high-AMR environments as understanding of pathogen die-off in latrines under different conditions increases.⁴⁷

Insufficient treatment of AMR organisms in safely-managed systems combined with the potential discharge of untreated feces with ESBL E. coli into the environment (including via open defecation) highlights a policy gap in linking WASH infrastructure to reductions in exposures to ESBL E. coli and similar AMR organisms. These findings represent a dual research and policy need. Efforts to end open defecation and improve use of safe sanitation facilities are only beginning to be recognized in the global effort to combat AMR, which have broadly focused on clinical interventions to-date.^15,48 Environmental contributions to clinical AMR, including risk factor analyses examining the role WASH may play in increasing or decreasing transmission risks, are needed. Fate and transport of AMR organisms, and their genes, in the environment (and associated environmental exposure pathways) and in sanitation systems of all types are beginning to be better-described.¹⁵

Notably, we have not accounted for animals, which comprise a major source of fecal biomass and potentially ESBL E. coli,⁵ because we could not systematically quantify their contribution. Animal feces is often not safely managed in the onsite (household) and offsite environments and may be directly applied to agricultural land (without treatment).²² While increasing antibiotic use in animals makes contact with them, including preparing and consuming their meat, an important exposure risk,⁴⁹ animal feces may be an equally important, underappreciated, environmental source of ESBL E. coli.^50,51 Among food animals globally, β-lactam drugs are given both prophylactically and therapeutically and the prevalence of ESBLs among commensal gut bacteria is estimated to range from <1–41%.⁵⁰ Animal feces makes up two-thirds of onsite feces worldwide, a figure that is expected to increase in coming decades.²² The global absence of safe management of animal feces is beginning to be highlighted for its role in enteric pathogen transmission, especially in low-income settings.⁵² Thus, the global prevalence of ESBL E. coli in animal feces in these settings should be a focus of future study and included in future estimates.

There are additional limitations to these calculations, which represent research gaps that could improve estimates of the scale, load, and environmental discharge of ESBL E. coli. Critically, the percentage of discharge that ESBL E. coli represent in feces of ‘carriers’ or ‘infected’ individuals was a key point of uncertainty that we addressed in our model through scenarios informed by a smaller carriage study.²⁸ Additionally, although Karanika et al. 2016 did not observe significant differences in carriage between studies focusing on adults compared with those focusing on children,¹² they represent a single meta-analysis of existing ESBL E. coli carriage data and other evidence suggests that children’s carriage of AMR organisms differs from carriage in adults.^53,54 Improved understanding of carriage differences within individuals and across age groups can inform assumptions about applying generalized carriage estimates to children. Urban/rural differences in sanitation are well-documented,²¹ but additional investigation into differences in ESBL E. coli carriage between those populations would allow for more precise modeling of disparities in carriage and treatment. Additionally, we did not account for short-term temporal variation in carriage of ESBL E. coli (e.g. acquisition or loss of organisms/genes over time among those exposed), and instead assumed a static prevalence. For example, during the 2011 outbreak of ESBL-carrying Shiga toxin-producing E. coli (STEC), serotype 104:H4, the median duration of shedding after hospital discharge was less than 20 days,⁵⁵ whereas household transmission was rather rare.⁵⁶ A recent modeling study suggests an individual’s carriage status is acquired in 3 years (95% confidence interval: 1.6–6.3 years) and lost in 1.1 years (95% CI: 0.8–1.6),⁵⁷ thus future models at household- or individual-scales should account for individual changes in carriage due to prevention of exposure, or other measures, when conducting longitudinal assessments.¹² Beyond the recognized need to determine the effectiveness of current wastewater treatment plants in removing ESBL E. coli and similar organisms and genes,^{15,32,33,44,45} there is also a need to understand die-off of AMR pathogens in onsite, decentralized systems (e.g. pit latrines), which collect the majority of human feces (both with and without ESBL E. coli).²² For example, specific gaps may include whether the AMR status of an enteric pathogen alters within-latrine bacterial communities and time scales from those known for susceptible pathogens and how onsite latrines may contribute to transfer of AMR genes and emergence of new AMR pathogens.^58,59

In summary, we present a first accounting of ESBL E. coli discharge through varying sanitation systems globally and by region. At least basic sanitation systems receive about two-thirds of feces containing ESBL E. coli globally (about 1.2–120 billion kg/year); however, these sanitations systems do not guarantee total removal of AMR organisms. Importantly, more than 10% of feces containing ESBL E. coli—about 220 million-22 billion kg/year—were openly defecated, and another 14% (270 million-27 billion kg/year) were discharged into unimproved systems. WASH, and specifically sanitation, has an under-acknowledged role to play in mitigating the fate and transport of AMR organisms such as ESBL E. coli, and efforts to improve WASH and reduce environmental loads and exposure to feces should be further integrated with those to combat AMR.

Research in context.

Evidence before this study:

Human carriage of extended-spectrum beta-lactamase resistance, including extended-spectrum beta-lactamase-producing E. coli, has been on the rise over the past several decades. As confirmed by exploratory review of the literature, this rise is indicative of but one of many antimicrobial resistant organisms of clinical importance with environmental transmission pathways. To-date, improvements in water, sanitation, and hygiene systems globally are commonly implicated as important preventative measures for slowing antimicrobial resistance development (e.g. proper treatment of drinking water to interrupt environmental transmission of many of these organisms, handwashing at key times to prevent infections). Sanitation systems in high-income settings have been found to be breeding grounds for antimicrobial resistant organisms and their genes, and current scientific efforts aim to improve the understanding of how to treat these genes and organisms in high-income settings, but also understand the efficacy of onsite treatment technologies common in lower-income settings.

Added value of this study:

This study provides, to our knowledge, the first global estimates of feces containing ESBL E. coli passing through sanitation systems by region and quality of sanitation according to Sustainable Development Goal 6 indicator criteria. It also calculates future projections of feces containing ESBL E. coli as populations continue to grow.

Implications of all of the evidence available:

Sanitation systems in low-income settings, which are already disproportionately poorer in treatment of susceptible organisms, receive the largest fecal biomass containing ESBL E. coli. When combined with available evidence of increasing human carriage of ESBL E. coli, the importance of sanitation systems—especially in low-income settings—grows exponentially. Improving sanitation systems in these settings should be included within efforts to prevention and combat antimicrobial resistance, as reducing infections and reducing antimicrobial resistant organisms discharged into the environment can offset and reduce already strained clinical preventive and therapeutic measures.