Escherichia coli Ingested via Food May Overshadow the Positive Effects of Clean Drinking Water: An Example from Dhaka

ABSTRACT.

The minimal health impact observed in large-scale water sanitation and hygiene (WASH) intervention studies motivated us to investigate the contribution of contaminated food and drinking water to the total daily Escherichia coli load ingested by the average adult in a low-income, urban area. Leftover food (food left at room temperature for more than 6 hours) from 32 households was collected eight times at 6-week intervals in 2014–2015 in the low-income area of Arichpur, Dhaka, Bangladesh. In total, 117 samples were obtained from four food types: fish, lentils, rice, and vegetables, which comprise approximately 85.2% of the average adult’s personal daily food consumption. Samples were analyzed for E. coli using selective chromogenic media. For an average adult, the daily consumption of the four food types at mean contamination levels of E. coli can contribute 4.45 log colony-forming units (cfu)/day (95% confidence interval 4.06–4.84). Drinking water quality was measured 211 times at the point of drinking, with a mean, median, and maximum contamination of 1.9, 1.2, and 2.82 log E. coli cfu/100 mL, respectively. If the typical adult in Arichpur was able to drink water with 0 E. coli cfu/100 mL, it would only remove < 5.2% of the total E. coli ingested per day with a mean-contaminated diet. These approximations may suggest why insignificant effects have been observed for water quality interventions in similar, low-hygiene settings. In Arichpur, the E. coli contribution from drinking water to the total E. coli load was insufficient to exert a substantial effect.

INTRODUCTION

The global death toll from diarrheal diseases has been declining over the last decades. However, mortality remains high, with an estimated 1.57 million deaths annually, of which one-third occur in children under 5 years of age. ¹ Diarrheal diseases are primarily linked to the ingestion of fecal pathogens transmitted by routes illustrated in the F-diagram proposed in 1958 by Wagner and Lanoix. ² This simple diagram has served as a road map for reducing the burden of diarrheal diseases by breaking different fecal–oral transmission routes. Until recently, it has been assumed that there was a good understanding of the transmission routes of enteric pathogens and how relatively simple interventions, such as providing clean drinking water, hygiene education, and toilets can break these transmission routes and prevent the spread of diarrheal diseases. ^{3 – 5} However, three highly rigorous, randomized controlled trials of water quality (drinking water chlorination), sanitation (upgraded sanitation), and hygiene (promotion of hand washing with soap) in low-income, rural households (WASH-Benefits parallel studies in Bangladesh and Kenya and the similar SHINE study in Zimbabwe) ^{6 – 10} showed surprisingly no effects on child linear growth; moreover, only the Bangladesh study showed a modest effect on pediatric diarrhea. The complexity of transmission routes was modeled in the Bangladesh WASH-Benefits study, where Kwong et al. ¹¹ concluded that the daily ingestion of Escherichia coli in children aged between 6 and 35 months primarily originated from mouthing their hands, which led to direct soil ingestion and the ingestion of contaminated food. Additionally, Mattioli et al. ¹² modeled hand–mouth contact as more important than drinking water for microbial transfer in children in Tanzania. Wang et al. ¹³ modeled similar results for Ghanaian children, which indicated that hands played a pivotal role in fecal microbe transfers with food as a primary contributor (99.9%).

As children suffer the most significant burden of diarrheal disease, they have been the primary focus of transmission studies. However, if food is a dominant source of E. coli ingestion in adults, where we assume hand–mouth transmission plays a lesser role, the ingestion of microbes via food and water would be dominant.

Traditionally, E. coli transmission from food and water has been based on samples of raw and cooked food, and the direct monitoring of tap water or inside drinking water storage containers. However, few studies have quantified the E. coli load from food and water at the exact time and place of consumption—that is, quantifying the amount of E. coli in food while it is being eaten and in water while it is being drunk from a glass. This point-of-consumption food sampling strategy is essential in an area such as Arichpur in Bangladesh, a low-income community with limited access to safe food storage. This is an area where food is typically cooked once a day, and people rely on eating stored food for their remaining meals. Previous studies have shown that in the low-income community of Arichpur, in-house stored drinking water is further contaminated in drinking glasses at the point of consumption ¹⁴ and that stored food can be heavily contaminated by flies. ¹⁵

To investigate the transmission routes that are dominant in low-income urban households in Arichpur, Bangladesh, this prospective study aimed to estimate the average daily consumption of E. coli via food and water by adults living in the community. A comparative understanding of the relative E. coli load distribution between food and water experienced by adults living in low-income households in Dhaka may be useful for redefining priority areas for WASH interventions.

MATERIALS AND METHODS

This work represents a smaller study conducted under the more extensive longitudinal study Combating Cholera Caused by Climate Change (C5). ¹⁶

Study area.

The study area, Arichpur, Tongi, is located in the outskirts of greater Dhaka in Bangladesh (GPS coordinates 23°53'03.9″N 90°24'31.5″E). Arichpur is a low-income, urban community as defined by the Center for Urban Studies. ¹⁷ Most of the population in this area live in semi-pacca houses (constructed of concrete walls and a roof made of tin or wood), share water sources, and kitchen and toilets facilities are located in a compound (a cluster of households that share the same yard and other facilities), which are all considered characteristics of a low-income setting. ¹⁶ The area is surrounded by several industrial establishments and garment factories; consequently, this 1.2-km² area is densely populated with 100,000 people per km². We used the formal definition of household from the Bangladesh Bureau of Statistics (BBS), ¹⁸ which defines it as “persons, either related or unrelated, living together and taking food from the same kitchen constitute a household.” Typically, an average of 10 to 15 nuclear families share common cooking areas, latrines, and water supplies. ¹⁸ In this community, people typically cook food once a day and keep the food at room temperature to eat later (unpublished observations). Kitchen tables are not used in this community, and vegetables, fish, and meat are prepared on the kitchen floor (concrete or mud) by a large wooden or metal-mounted, sickle-shaped metal knife called a boti (Figure 1). After preparation, the food is kept on a plate, and the waste is left on the floor. Pots, plates, glasses, and utensils are typically washed under a tap after use (although sometimes these are washed in a bucket depending on the household’s water supply) and left to dry. Most people use their hands to eat, and the food is served from a shared food plate.

Figure 1.

Traditional knife used for food preparation. This figure appears in color at www.ajtmh.org.

Data collection.

This study consisted of 32 purposely selected households that were identified from a group of 477 randomly selected households participating in the C5 study. A detailed description of the baseline data collection from these 477 households is described in Sultana et al. ¹⁶ The selection of the 32 households was based on the availability and willingness of the participants to be involved in the study. Trained research team members from the International Center for Diarrheal Disease Research, Bangladesh (icddr,b) conducted baseline data collection from the households. After gathering this data, routine visits to each household were conducted once every 6 weeks between November 2014 and December 2015. As part of these visits, the type of food consumed by household members and their water-use information were collected. Information concerning the quantity of water used by each household member for hygiene purposes was collected, and samples of food and water were also collected. Details of the methods used to collect this data have been described elsewhere. ¹⁶

Food and water sample collection.

During each visit, leftover food (food that had been resting at room temperature for more than 6 hours) was collected, if present. Six hours was chosen because this would be the duration of time food would be left out in households that cook only once a day. A total of 137 leftover food samples were collected from the 32 households during the study period, with the total number of samples per household ranging from 1 to 13 due to the inconsistency in the presence of leftover food. Solid food samples of approximately 25 g were collected in sterile plastic bags using sterile spoons during each visit. For liquid foods, 25 mL samples were collected in sterile centrifuge tubes.

Two hundred and eleven water samples were collected during the visits from drinking vessels (glasses, mugs, bottles, or jugs) used by household members to drink water; method and collection are described elsewhere. ¹⁴ During the study, volumes of 150 to 200 mL of water were sampled five to six times from each household. Water samples were collected using presterilized, wide-mouth sampling bottles (SPL Life Sciences, Gyeonggi-do, Korea).

All samples were maintained at a low temperature in a sample box containing gel ice packs and transported to the Environmental Microbiology Laboratory, University of Dhaka, within 2 to 4 hours of collection where sample processing and microbiological analyses were performed.

Processing of food samples.

Solid food samples (approximately 25 g from each household) were weighed aseptically, and 100 mL of phosphate buffer saline (PBS) were added and blended in a Stomacher Laboratory Blender (Stomacher 80; Steward Limited, West Sussex, UK). Additional PBS was added for a 1:10 final dilution. Liquid food samples were first homogenized and then 10 mL were aliquoted and added to 90 mL of PBS. All samples were examined for thermotolerant E. coli contamination by quantifying colony-forming units (cfu) in chromogenic selective media (mTEC ChromoSelect Agar; Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany) that were incubated at 44.5 ± 0.5°C for 18 to 24 hours. Typical reddish-purple or magenta colonies of E. coli were enumerated and recorded as cfu per gram. ¹⁶

Processing of water samples.

Basic water quality was analyzed via the membrane filtration culture method for detecting E. coli in all collected water samples. Aliquots of 100 mL of water were filtered through 0.45-µm-pore, 47-mm-diameter white gridded membrane filters (S-Pak, Merck Millipore, Darmstadt, Germany). The membrane filters were placed on mTEC agar plates (Thermo Fischer Scientific Inc., Oxoid Limited, Hampshire, United Kingdom) and incubated at 44.5 ± 0.5°C for 18 to 24 hours. Typical reddish-purple or magenta colonies of E. coli were enumerated and recorded as cfu per 100 mL of water. ¹⁶

E. coli concentrations on both solid and liquid samples were log₁₀ transformed. The lower limit of detection (LOD) was considered equal to the concentration of the sample with the lowest concentration detectable in all assayed samples. For food samples, the LOD was equal to 10 cfu/g and for water samples the LOD was equal to 1 cfu/100 mL. Solid and liquid samples with observed zero counts were substituted with half of the LOD, that is, 5 cfu/g and 0.5 cfu/100 mL respectively, before log₁₀ transformation. It was assumed that this censoring approach would have a minimal effect on the results given the relatively low LOD.

Data analysis.

Estimating the average daily E. coli consumption.

Average food consumption per adult person per day was based on national data concerning food consumption among Bangladeshi residents in low-income households taken from the Household Income and Expenditure Survey 2016 conducted by the BBS. ¹⁹ Based on this survey, it was estimated that 85.2% of the average 802.34 g of food eaten daily by an adult in Arichpur comes from only four main food types: rice (369.9 g), pulse/lentils (10.7 g), vegetables/potatoes and onions (263.1 g), and fish (40.2 g). Other types of food were excluded in the estimate calculations because they were either rarely eaten or there were a limited number of samples available in the households. The average quantity of drinking water consumed per person per day was estimated to be 1.9 L per person (range = 1.1–2.7 L) based on 71 days of observations of 59 individuals from a subset of the households. ¹⁶

An estimate of an individual’s daily E. coli intake from food was calculated by multiplying the average food consumed from each food type (as reported by the BBS) with the mean E. coli concentration measured in the leftover food found in the households. Monthly averages of each food type were calculated to identify possible monthly variation throughout the year.

To estimate the daily E. coli ingestion from drinking water, 211 water samples were collected at an individual household’s point of drinking (cup, mug, etc.), and these results were multiplied by the average daily consumption of 1.9 L of drinking water.

An alternative estimate of the daily E. coli consumption via water was calculated by substituting the drinking water contamination we measured with three theoretical contamination levels of 999, 99, and 9 E. coli cfu/100 mL. This was done to compare our results to similar scenarios with different drinking water qualities.

Ethical clearance.

Informed written consent was obtained from all study households. This study was approved by the Ethical Review Committee of icddr,b.

RESULTS

E. coli contamination in food.

The mean E. coli contamination of the 137 leftover food samples was 1.51 log cfu/g, with a range of 0.7 to 3.63 log cfu/g. Eight types of cooked food were collected. Rice, vegetables/potatoes and onions, fish, and lentils were the most frequent, and therefore the only included in Table 1. Meat dishes were the rarest (i.e., three beef, two chicken, and two mutton) and thus not included in the estimates. The same is true for cooked rice with water, known as Panta vat (leftover rice soaked in water overnight and eaten on the first day in the Bengali year in April), which showed the highest contamination with a mean of 3.51 log cfu/g (95% confidence interval [CI]: 3.42–3.61) in the four samples analyzed. Plain rice had an average contamination of 1.79 log cfu/g (95% CI: 1.39–2.20). Other types of food not included in the estimates due to a limited number of samples included honey, milk, bread, and various traditional dishes (N = 6, 2.85 log cfu/g).

Table 1.

Concentrations of Escherichia coli in collected food samples

Food type	Mean (log E. coli cfu/g)	95% Confidence interval	N
Fish	1.72	1.37–2.07	40
Rice	1.79	1.39–2.20	29
Vegetables/potatoes and onions	0.99	0.74–1.24	34
Pulse/lentils	1.56	0.97–2.15	14

E. coli contamination of individual samples for the most frequent types of food are shown in Table 1 and Figure 2, with the respective means and 95% CIs.

Figure 2.

Individual Escherichia coli concentrations (log₁₀ cfu) on the samples collected from the four most frequent types of food: fish, rice, vegetables/potatoes and onions, and pulse/lentils.

The seasonal distribution of E. coli contamination revealed a maximum contamination in May (mean count: 3.09 log cfu/g) and a minimum in November (mean count: 0.6 log cfu/g).

Daily E. coli ingestion via food.

The estimated daily consumption of E. coli with mean food contamination levels and the respective lower and upper 95% CIs are shown in Table 2.

Table 2.

Daily Escherichia coli contribution from rice, lentils, vegetables, and fish at different contamination levels in an average, low-income household diet

Food type	Daily consumption of food fractions (g)*	E. coli consumption at mean food contamination (log cfu/day)	E. coli consumption at lower CI (log cfu/day)	E. coli consumption at upper CI (log cfu/day)
Fish	40.2	3.32	2.97	3.67
Rice	369.9	4.36	3.96	4.77
Vegetables, potatoes, onions	263.1	3.41	3.16	3.66
Pulse/lentils	10.7	2.59	2.00	3.18
Total	683.72	4.45	4.06	4.84

CI = confidence interval.

^*Information taken from the Bangladesh Bureau of Statistics based on a low-income household. ¹⁹

Effect of drinking water on total daily E. coli consumption.

Water samples from the 211 individual household’s point of drinking (cup, mug, etc.) revealed mean drinking water contamination levels of 1.90 E. coli log cfu/100 mL (95% CI: 1.80–1.99). The total daily E. coli ingested via food versus drinking water is shown in Figure 3.

Figure 3.

The estimated personal, daily Escherichia coli (log₁₀ cfu) ingested from food (only 85.2% of the daily food consumption included) and 1.9 L of drinking water in Arichpur, Dhaka.

These results indicate that the drinking water fraction constituted < 5.2% of an individual’s daily E. coli consumption given that an adult drinks 1.9 L of highly contaminated water with an E. coli load of 1.9 log cfu/100 mL (the mean value found in Arichpur) and a daily diet of rice, vegetables, and fish at the mean E. coli contamination level. At the lower and upper 95% CI of the mean this fraction would be < 11.6% and 2.1%, respectively.

To place these results into perspective and allow comparisons with similar settings that have different drinking water qualities, the charts in Figure 4 have been divided into three theoretical drinking water contamination scenarios of 999, 99, and 9 E. coli cfu per 100 mL, based on 1.9 L of daily drinking water consumption. The figure illustrates daily E. coli consumption via drinking water presented as a percentage in relation to the total daily E. coli consumption (water and food) and calculated at the mean (a) food contamination, and respective lower (b) and upper (c) 95% CI levels from Table 2.

Figure 4.

The relative contribution of drinking water to the total Escherichia coli intake. Consumption of 1.9 L of drinking water at three theoretical drinking water contamination levels (999, 99, and 9 E. coli cfu/100 mL) compared with the total daily E. coli intake (only 85.2% of the daily food consumption is included: fish, rice, lentils and vegetables). Results are presented at the mean (A) food contamination, lower 95% confidence interval [CI] (B) and upper 95% CI (C).

For an adult drinking highly contaminated water (999 E. coli cfu/100 mL) and consuming a daily diet of rice, vegetables, and fish at a mean E. coli contamination level, the drinking water fraction would constitute < 40.3% of that person’s daily E. coli intake. At the lower 95% CI, it would constitute 62.1% and at the upper 95% CI 21.4%.

DISCUSSION

These estimates show that < 5.2% of the daily total E. coli an average adult in Arichpur was exposed to was ingested via drinking water. To further illustrate the effect of food contamination on overall E. coli consumption, the weight of different water contamination levels against the level of food contamination found in Arichpur was evaluated. These numbers show that even if a person only drank highly contaminated drinking water (999 E. coli cfu/100 mL), this would, at best, constitute only 40.3% of their total daily E. coli intake.

This study included only four major food categories comprising 85.2% (683.9 g) of an individual’s daily food consumption. The remaining 14.8% (i.e., chicken and beef) was not found as leftovers at a sufficient quantity during household visits to justify its inclusion in our estimates. However, the inclusion of the remaining 14.8%, would only increase the food contribution to the overall amount of E. coli ingestion. At times, leftover food was reheated before consumption, which would cause bacterial cell death to some extent. In a worst-case scenario, with artificial water contamination levels at 999 E. coli cfu/100 mL and maximum food contamination (upper 95% CI), a total load of 4.28 log E. coli per person per day ingested from drinking water would constitute < 21.4% of the almost 89 thousand or 4.95 log cfu E. coli ingested per day from water, fish, lentils, rice, and vegetables. Unfortunately, the additional contribution of E. coli ingested directly from fingers to mouth (eating with the fingers) could not be measured in this study. However, the process of eating with one’s fingers may add further pathogens to the food and deposit pathogens directly from the fingers into the mouth.

It is assumed that newly cooked rice, or any other food in the Bengali kitchen that is normally cooked, baked, or fried is free of contamination; however, when this food is stored for hours it will become contaminated. The E. coli present in stored food can originate from various sources, such as contaminated kitchen utensils ²⁰ or the dirty hands of food handlers or children, ²¹ but it may also come from the public domain, which may influence contamination levels. Hossain et al. ²² described how the gutting and cleaning of Hilsa fish can pollute a kitchen environment (including cleaned utensils) with E. coli contained in the fish slime, scales, and intestine. For a kitchen in Arichpur, with low hygiene and water availability, and where all cutting processes involve the same uncleaned Boti and the food is prepared on the kitchen floor, cross-contamination can have a significant effect. In Bangladesh, Hilsa fish is traditionally eaten with Panta vat for the Bengali New Year.

Flies are an overlooked but potentially important contributor to domestic contamination. A separate study conducted in Arichpur demonstrated that half of the flies landing on a bowl of rice deposited more than 2.78 log cfu E. coli per landing, with an average frequency of 1.13 fly landings per minute on exposed rice (based on 4290 observational minutes). ¹⁵ Therefore, a 20 g sample from the top of a rice bowl exposed to fly landings for 71.5 minutes would have an average concentration of 3.49 log E. coli/g (95% CI: 3.34–3.60). ¹⁵ Cooked rice preserved at warm temperatures may further provide an ideal growth environment for E. coli and Vibrio cholerae. ^{23, 24}

Kitchen environments and utensils may also be contaminated by polluted washing water. However, for Arichpur, with more than 50% of water samples containing less than 1.20 log cfu E. coli per 100 mL, it is unlikely that domestic water sources were solely responsible for the contamination levels we observed.

A previous study conducted in the households of rural Bangladesh found a mean of 1.1 log most probable number (MPN)/dry g E. coli contamination in food samples stored at room temperature for ≥ 3 hours. ²⁵ Another study conducted in rural Bangladesh showed that 12% of the stored, complementary food for children was highly contaminated with E. coli (≥ 100 MPN/dry g). ²⁶ The food samples from the Arichpur area showed a higher mean E. coli contamination compared with the aforementioned studies. This could be explained by the fact that study results from rural communities may not be generalizable to urban settings. However, in the low-income, urban environments of Accra, Ghana, food substantially contributed to exposure to fecal contamination according to a quantitative assessment by Wang et al. ¹³

If the drinking water in Arichpur, with an average E. coli contamination contribution < 5.2% of an adult’s total E. coli intake, the effect of water without any E. coli would probably be slight. This may help to explain why the effects of drinking-water quality interventions are difficult to detect in low-income settings; they are overwhelmed by the total E. coli contamination inside the household.

Limitations.

This study was conducted to investigate the potential ingestion of E. coli via food and water by an average, adult inhabitant of Arichpur. Therefore, these results do not represent exact exposures. Rather, they should be seen as potential exposures that require further, in-depth investigations. Further, the participating households volunteered to be in the study, which may have introduced selection bias. Our results are not based on individuals’ reported daily consumption of food but, rather, on an average diet as calculated by the BBS and based on a low-income household. Likewise, the measurements of E. coli do not represent the average daily E. coli contamination of the entire diet because samples were only taken when food had been left out at room temperature for more than 6 hours. Because meals are usually cooked only once a day, one can argue that a freshly prepared meal (time = 0) would contain a significantly smaller E. coli contamination than subsequent meals. However, it has been previously shown that several factors can influence the contamination of a freshly prepared meal. Fomites, such as food plates, can serve as a reservoir of bacteria. We have seen extensive contamination of washed food plates within the 32 households (results to be published elsewhere). Freshly prepared meals can therefore be contaminated with E. coli present in the plates. In addition, as discussed earlier, flies can heavily contribute to the contamination of meals. Lindeberg et al. observed an average of 1.1 fly landings per minute on a single bowl of rice in the same study area. ¹⁵ Lastly, it has also been shown that insufficient hand hygiene can also lead to meal contamination. ²¹ The authors assume that in a kitchen with good hygiene (no fomite or hand contamination and no contamination via flies), the daily estimates presented could be overestimated by approximately 33% because the first meal eaten at Time 0 would be without contamination. However, further studies are needed to estimate the actual contribution of the fomites, hands, and contamination via flies from Time 0, as well as regrowth as a function of time.

Because this study was based on the average food consumption of adults and not children, our results are also difficult to compare with previous studies. However, we believe that the E. coli contribution from food and water would not be substantially different in children eating the same diet as adults.

Previous studies looking into diarrheal disease transmission have in bigger part focused on water contamination. Nevertheless, our study clearly illustrates the role that food can have as a pathogen reservoir and emphasizes the need for future studies and interventions to look closer at the contamination of food, as well as kitchen hygiene within low-income households.