Adapting Enzyme-Based Microbial Water Quality Analysis to RemoteAreas in Low-Income CountriesAdam Abramson,* Maya Benami, and Noam Weisbrod

Department of Environmental Hydrology & Microbiology, Zuckerberg Institute for Water Research, The Jacob Blaustein Institutesfor Desert Research, Ben-Gurion University of the Negev, Sede Boqer Campus, Midreshet Ben-Gurion, 84990 Israel

*S Supporting Information

ABSTRACT: Enzyme−substrate microbial water tests, origi-nally developed for efficiency gains in laboratory settings, arepotentially useful for on-site analysis in remote settings. This isespecially relevant in developing countries where water qualityis a pressing concern and qualified laboratories are rare. Weinvestigated one such method, Colisure, first for sensitivity toincubation temperatures in order to explore alternativeincubation techniques appropriate for remote areas, and thenin a remote community of Zambia for detection of totalcoliforms and Escherichia coli in drinking-water samples. Wesampled and analyzed 352 water samples from source,transport containers and point-of-use from 164 random households. Both internal validity (96−100%) and laboratory trials(zero false negatives or positives at incubation between 30 and 40 °C) established reliability under field conditions. We thereforerecommend the use of this and other enzyme-based methods for remote applications. We also found that most water samplesfrom wells accessing groundwater were free of E. coli whereas most samples from surface sources were fecally contaminated. Wefurther found very low awareness among the population of the high levels of recontamination in household storage containers,suggesting the need for monitoring and treatment beyond the water source itself.

■ INTRODUCTION

Routine monitoring of microbial water quality is one of themost effective means of quantifying, and thereby reducing, risksto public health in remote areas of low-income countries. Thedisease burden, measured in disability-adjusted life years,associated with diarrhea from consuming contaminateddrinking water is over 30 times greater in low- vs high-incomeregions.1 These areas are also the most difficult and costly tomonitor. Standard laboratory-based microbial water tests,including multiple-tube fermentation, membrane-filter techni-que,2 and polymerase chain reaction (PCR),3 have certaintechnical requirements that make them particularly unsuitablefor analyzing water samples from rural areas in developingcountries. Each of these techniques requires (1) expensiveequipment; (2) careful use of aseptic technique duringcollection, analysis and storage; (3) highly trained personnel;(4) a temperature-controlled incubator; and (5) a 24 h timeconstraint for analysis initiation.4

Distance and low levels of infrastructure greatly inhibit thefeasibility of conducting these standard procedures in off-sitelaboratories, both technically and due to increased costs. As aresult, a trade-off has traditionally existed between theremoteness of a monitoring effort and cost-effectiveness.5−7

As costs of off-site analytical techniques increase withremoteness, technical feasibility may reach a threshold. At acertain point, conventional laboratories become useless asremoteness pushes water sampling and transport time beyond

the 24 h limit. Coincidentally, remote water-testing efforts arelimited in scope and depth. Comprehensive routine monitoringis, in practice, practically nonexistent in low-income countries.Along with the approaching culmination of the MillenniumDevelopment Goals is the emerging acknowledgment thatefforts must be expanded beyond the development of improvedsources to include monitoring of drinking water quality both atthe source and point-of-use (POU).8,9

As Bain et al.10 demonstrate, there are 44 microbial watertests currently available for meeting compliance, surveillance,operational, or other monitoring objectives in low- or medium-resource contexts. In this study, we focus on the need toconduct compliance monitoring for ensuring that samples meetregulatory standards. In this case, it is important for a test to beable to detect contamination corresponding to the WHO’sthreshold of 1 CFU E. coli per 100 mL. In addition, regulatoryapproval of the methodology is necessary, such as by the U.S.EPA or through inclusion in the Standard Methods for theExamination of Water and Wastewater.4 When these require-ments are considered along with the technical challengesassociated with remote areas, only 6 of the 44 methodologiesare found feasible for conducting compliance testing in remote

Received: May 15, 2013Revised: August 11, 2013Accepted: August 16, 2013

Article

pubs.acs.org/est

© XXXX American Chemical Society A dx.doi.org/10.1021/es402175n | Environ. Sci. Technol. XXXX, XXX, XXX−XXX

areasall of which are enzyme-based: Colisure, Colilert,Colilert-18, Colitag, E*Colite, and Readycult.Enzyme-based methodologies, accepted as the industry

standard for microbiological water testing in laboratories,possess untapped potential for use as microbial field tests forlow-cost, rapid assessments of water quality in remote areas.11

Unlike other field tests, they are inexpensive and simple toimplement. Through simultaneous chemical targeting ofcoliform-specific enzymatic processes with inhibition of non-coliform species, the entire process can be conducted in a singlesterile container, internalizing the aseptic environment whilesimultaneously allowing for the simple analysis of an entire 100mL sample. Neither dilution nor filtration, and only simpleaseptic procedureswearing latex gloves and ensuring propersampling technique, are required.Nevertheless, the recent emergence of such methods,

primarily for the benefit of in-lab convenience and gains intime efficiency, has largely overlooked the rural water sector,where public health gains from reliable water-quality data arehigh. Most studies investigating enzyme-based methods havefocused on the advantages of Colisure or its sister technique,Colilert, in detecting or enumerating E. coli in microbiallaboratories.12−18 However, their application for remote on-sitemonitoring in developing countries has been overlooked: toour knowledge, only three studies have applied theseadvantages to water sampling in developing countries,19,20

and only one of those conducted a remote on-site analysis.21

No studies to date have investigated enzyme-based method-ologies for on-site compliance testing relative to the WorldHealth Organization (WHO) guideline for drinking water.1

While enzyme−substrate tests remove most obstacles, thechallenge of achieving and maintaining standardized incubationregimes remains a central obstacle to expanding water testinginto low resource settings. Bain et al., for example, deem thesesix approaches ‘not ideal’ for analysis without a basic laboratoryor clean space with electricity, presumably due to this need forlab-grade incubation procedures.10 Attempts have been made toovercome this by alternative incubation mechanisms. TheAquatest, a recently developed Most Probable Number test,uses boiled water to maintain incubation temperatures.10

Chuang et al. investigated the EC-Kita combination ofColilert and Petrifilmusing a waistbelt incubator operating onbody heat alone.21,22 Brown et al. explored the feasibility ofambient temperature incubation in tropical areas with threemicrobial tests, including Colilert, and found that test accuracywas only marginally compromised.23 These studies provideevidence that the incubation temperature ranges may beexpanded without sacrificing the accuracy of enzyme-basedmethods.The purpose of this study was to (1) test the accuracy of one

enzyme-based method under a wider range of incubationtemperatures than currently approved; (2) develop a simple,low-cost incubator for field use; (3) assess the technicalfeasibility of this adapted approach in a remote community forcompliance testing relative to the WHO guideline for drinkingwater; (4) use this method to trace the fecal-contaminationpathway and its relationship to user perceptions of waterquality; and (5) compare the cost-effectiveness of this methodto other feasible options for remote analysis.Of the six enzyme-based methods mentioned above, two

(Readycult and E*Colite) were demonstrated to performinsufficiently in laboratory tests conducted by Olstadt et al.24

We present further details of a comparative analysis of enzyme-

based methods in Supporting Information S1. Thus, four testsmay be considered potentially appropriate: Colisure, Colilert,Colilert-18, and Colitag. These tests all share the followingcharacteristics: (1) they are chromogenic and fluorogenicpresence−absence (P-A) enzyme−substrate tests for fecal andtotal coliforms; (2) they test standard, 100 mL samples; (3)they consist of only a sampling container and a powdered agar,and require incubation and a hand-held UV light; (4) the costof consumables is between $4.50 and $5.00 per test; and (5)they are approved for incubation at 35 ± 0.5 °C.10,25−27 Thesetests differ only in the specific enzymes and substrates used (tometabolize coliforms, Colilert uses ONPG and MUG, Colisureuses CPRG and MUG, and Colitag uses ONPG and MUGalong with a buffer medium, TMAO), storage temperature andshelf life, and the duration of the test.Of these techniques, Colisure (IDEXX Technologies,

Westbrook, ME) was used in this study because it is highlycited in the literature. Most of these citations are for testingdrinking water samples with low levels of contamination. Inthese cases, its accuracy has been previously documented,28,29

and it has been shown to be superior to certain acceptedreference methods in the detection of chlorine-injuredcoliforms and E. coli under conditions that resemble those ofcontaminated drinking water.30,31 Colisure (and other enzyme-based approaches) is also emerging as an important tool fortesting environmental water samples.32−35 For example, it is themost common approach for testing recreational waters in theGreat Lakes region of the US.32 The test has been shown topossess acceptable ability to suppress non-coliform bacteria,24

although false readings are more common in environmentalsamples. This is true for both enzyme-based as well as somestandard methodologies; Wohlsen, et al. reported a coefficientof variation of 38% for Colisure, and 44.8% with membranefiltration when testing environmental samples.36 This suggeststhat its accuracy when testing environmental samples may becompromised, but no more than that of conventional methods.

■ EXPERIMENTAL SECTIONTemperature and Incubator Testing. The sampling and

analytical procedures for Colisure are simple (see SupportingInformation S2). To find an appropriate solution for theincubation challenges mentioned above, we conducted apreliminary investigation of the method’s sensitivity totemperature. This was conducted in a microbial laboratory bytesting triplicate samples of sterile bottles with 100 mL sterilewater inoculated with a 1 μL loop from frozen cultures ofStaphylococcus aureus (ATCC 49834), Enterococcus faecalis(ATCC 10100), Klebsiella pneumonia (ATCC 10031), orEscherichia coli (MG 1655). A fourth triplicate of 100 mL of rawtoilet water was tested in order to investigate the methodologywith water resembling samples encountered in the field.Although exact contamination levels were not determined,the inoculation ensured that at least one viable cell was presentin each sample while allowing for a degree of variation commonin environmental water samples. These were incubated in ascientific-grade incubator in the laboratory at 25, 27, 30, 35, 37,40, and 42 °C.In order to demonstrate the feasibility of other incubation

techniques, a low-cost alternative was constructed from a small,temperature-sensitive space heater and a large cardboard boxwith insulating blankets. At the study site, this incubator wascalibrated with a CR10X data logger (Campbell Scientific,Logan, UT) recording in situ water-sample temperatures by

Environmental Science & Technology Article

dx.doi.org/10.1021/es402175n | Environ. Sci. Technol. XXXX, XXX, XXX−XXXB

placing three thermocouples in different Colisure water-sampling bottles located at different distances from the heatsource to determine the temperature range during incubation.The incubator was also used to incubate the samples. The mainadvantage of this device is that it is lightweight (consisting onlyof a blanket, cardboard box, and small space heater) andinexpensive.Study Site. The field study was performed in Simango

Ward, a rural municipal subdistrict consisting of 15 villages inthe Kazungula District of Southern Province, Zambia. It islocated 57 km north of Livingstone municipal center. Meanvillage size is 51.7 households, and for an estimated ward size of300 km2, population density is roughly two households or 13people per km2. Recent surveys indicate that 77.2% of thehouseholds in this catchment use improved water services (inthis case, boreholes with hand pumps), mean water-fetchingtime is 30.4 ± 1.29 min, and per capita water use averages 57.8± 5.3 L per day.37 Average yearly precipitation in the region is650 mm, 90% of it falling between December and February.38

Most of the residents conduct rain-fed farming of maize, whichis the staple food, and supplement their subsistence with lowlevels of cash income: average income levels of $0.22 ± 0.078per capita per day are well below the global poverty line of$1.25 per capita per day.37

Survey Instrument. Two sampling efforts were conducted:one during the dry season (July 2011) in which 171 watersamples38 sources and 133 containerswere taken from101 households, and one during the rainy season (February2012) in which 181 water samples37 sources, 80 containers,and 64 drinking cupswere taken from 63 households. These164 households were randomly visited by a paired team ofvisiting and local Tonga-speaking surveyors who administeredtwo separate sections of the survey instrument. The first sectionincluded household questions related to water use, sanitation,and water-quality perceptions. User perceptions of waterquality for source, transport, and storage points were ratedaccording to four statements ranging from “I fear very much thatsomeone may get sick f rom it” to “I am sure that no one will getsick f rom it.”The second section was a mixed-methods observation

activity consisting of photographs of storage and transport (ifdifferent from storage) containers, written notes of respondentbehavior, and testing household water samples using theColisure kit. Water samples were taken by visiting studentsusing latex gloves and, if it was not possible to pour the watersample directly into the Colisure container, a sterile 50 mLcontainer was used to draw water from the source. For point-of-use (POU) drinking samples, heads of household were asked topour water into the sample cup for testing in the same way theywould pour water into their own cup, to ensure fidelity toregular water-drawing practices. Similarly, hand-washingpractices were observed when the respondent complied.Samples were stored in a cooler for not more than 5 h beforethe sampling teams returned to the village center, whereColisure reagent agar was added and samples were placed in thefield incubator.At the end of the visit, surveyors asked to see the household

water source, which was sampled and photographed. If thehousehold used different containers to transport and store thewater, a sample was taken from the transport container aftercarrying water in it for several minutes. Thus, a total of two orthree water samples were associated with each householdinterview. After review of the data from the first sampling event,

an additional sampling point was added to the second (rainyseason) survey: the household’s drinking cup.

■ RESULTS AND DISCUSSION

Laboratory Testing of Colisure. As expected, samplesspiked with E. faecalis and S. aureus did not change color orfluoresce under any of the incubation-temperature conditions.K. pneumonia, representing total coliform bacteria, wasdetected, as expected, at incubation temperatures between 25and 37 °C, but not at 40 °C. E. coli was detected at alltemperatures between 27 and 40 °C. Fecal contamination intoilet water was detected between 30 and 40 °C. Thus, recoveryof E. coli was possible at incubation temperatures ranging from30 to 40 °C with the Colisure method. The results arecompiled in Table S3. These results suggest that the accuracy ofthis method may not be as dependent on precise incubationtemperature regimes as standard protocol requires. If this is thecase, alternative incubation techniques appropriate for low-resource areas may be used. Our findings concur with those ofBrown et al., who found that analysis of samples using Colilertmay be achieved with low levels of false negatives (4%) andfalse positives (3%) when incubated at ambient temperatures(i.e., without an incubator).23 It is especially remarkable that theambient temperatures in their study were all below 30 °C.These findings demonstrate the need for a more robust study ofthe impact of incubation temperature on test accuracy todefinitively determine acceptable temperature ranges ofappropriate methods.

Incubator Performance. Results of the makeshift fieldincubator demonstrated that its operating temperature range iswithin the range of reliability for incubation of total coliformsand E. coli. The in situ temperature range of means from 10 minintervals was between 28.85 and 42.4 °C, with a 7-day averagetemperature of 35.4 ± 1.3 °C. A detailed schematic is presentedin Figure S1 and discussed in SI section S3.This suggested that a simple, low-cost, and readily available

incubation device can be used in a makeshift manner for ruralapplications of Colisure microbial analysis. This particulardevice produced an incubation regime within acceptedtemperature specifications as the manufacturer recommendsfor some countries of 36 ± 2 °C.25 The only requirement for itsimplementation is a reliable source of electricity. There arecertainly other possible techniques that would provide atemperature regime within the 30−40 °C range. Fire-heatedrocks placed in an insulated box, for example, would removethe requirement for electricity. Preliminary trials using thismethod produced the same results as samples under standardincubation, although a more appropriately designed experimentwould need to be performed for confirmation. The possibilityof incubation at ambient temperature in tropical climatesshould also be considered.

Internal Validity of Water Samples. A total of 352 watersamples were taken from 164 households and analyzed withColisure. Figure 1 presents the consistency of results fromsources downstream of positive results.The small discrepancy in the E. coli results, mostly located

among transport containers, may be partly due to respondenterror, in which transport and storage water samples were statedto have been drawn from the same source, when in fact theywere not. This is especially plausible for the 34% of householdsin the area regularly using multiple water sources.38 Thissuggests that a small fraction (<5%) of unreliable results may be

Environmental Science & Technology Article

dx.doi.org/10.1021/es402175n | Environ. Sci. Technol. XXXX, XXX, XXX−XXXC

present for E. coli analyses, with total coliform analyses beingcompletely consistent.Water Samples. Of the 352 samples, 75 were source

samples, most either boreholes or riverbed sources. Theremaining 277 samples were taken from household storage(170), transport containers (43), or drinking cups (64). Thetypes of sources investigated, with contamination results, aresummarized in Figure 2. Results sorted by type of container arepresented in SI section S4 and Figure S2.

As expected, both E. coli and total coliform contaminationwere highly source-specific, with almost all of the improvedsources (boreholes) being free of E. coli, and almost the exactopposite for unimproved sources. Of the 164 householdsvisited, 96 used boreholes, 50 used a riverbed well, 8 used theriver directly, 5 used a dam or pond, 4 used an unlined hand-dug well, and 1 used a lined hand-dug well.A description of the first point of fecal contamination is

summarized in Figure 3. Source-based policy is clearlyinadequate for ensuring consumption of clean water amongthe study site population. Of those obtaining clean water at the

source, only 22% kept it clean up until the time it was sampled,suggesting that meeting the Millennium Development Goals(MDGs) can be alarmingly irrelevant to whether or not thepopulation consumes water deemed clean by the WHO.While 58% of the households obtained contaminant-free

water at the source, that figure dropped to 14% by the timetheir water was consumed. Transport, storage, and drinkingcups contributed varying proportions of this after-sourcecontamination. About one-third of the clean water samplestransported from the source to the household becamecontaminated, possibly due to contact with unwashed handsor other objects, or due to the container itself. About two-thirdsof the remaining clean samples were contaminated while instorage, and all remaining clean samples that were tested werecontaminated by the transfer from storage to the drinking cup.

Household Water and Sanitation Practices. Of the99.4% of households who cleaned their storage and transportcontainers, most used sand (48%) and/or soap (68.9%), with aminority using ash from a firepit (6.1%). The mean number ofdays between cleaning was 3.6 (SEM 0.377), i.e., about twice aweek. Drinking cups, however, were cleaned at least once a dayby 95% of the respondents, using soap (87.8%), ash (15.2%), orsand (3.0%).Of the 25.8% of respondents who treated their storage water,

the only method used was addition of chlorine (which issporadically available at no cost at the health clinic). Our resultsfailed to establish a significant difference between treated anduntreated water samples with respect to risk of contamination.This may be at least partially due to procedural discrepancies inthe administration of chlorine, which were not investigated.Some 22% of respondents reported experiencing sickness fromtheir water, with diarrhea as the main symptom reported (54%of cases). Almost half of the respondents (46%) used a latrine.Most (80%) reported always washing their hands after usingthe toilet, while some (16%) washed sometimes and a minority(4%) never washed their hands. The use of soap for hand-washing was more evenly distributed: 37% always, 39%sometimes, and 24% never used soap.

Regression Analysis of Contamination Factors. Otherstudies have shown that risks of recontamination can bereduced by various behaviors: pouring from instead of dippinginto vessels,39 securing tight lids on storage containers,40,41

using latrines, practicing hand-washing, and treating water with

Figure 1. Consistency of Colisure results between water samplesyielding positive results and results of subsequent points along thewater procurement pathway. The assumption is that positive resultsshould be positive at all subsequent points since contamination istransferred, providing a measure of the test’s internal consistency. Intotal, 96.5% and 100% of 172 and 130 test results were consistent forE. coli and total coliforms, respectively.

Figure 2. Primary water sources used by sampled households withcorresponding sample contamination prevalence. Households (N =164) are represented by bars and water samples (n = 75) by pie charts.The majority of households (57.8%) used boreholes and thereforeobtained fecal-free water at the source. All other sources were foundcontaminated with E. coli.

Figure 3. Tracing household water contamination along the waterprocurement pathway. The proportion of households (n = 164)experiencing the first introduction of E. coli are indicated in pie chartssized relative to the number of households in each category. After thesource water, only households with negative results are considered(indicated by arrows), in order to outline the first point ofcontamination, if any. The total number of such households at eachprocurement point is indicated, as well as the resulting fecalcontamination prevalence at the point of consumption.

Environmental Science & Technology Article

dx.doi.org/10.1021/es402175n | Environ. Sci. Technol. XXXX, XXX, XXX−XXXD

boiling, solar heating, chlorine, or other techniques.42−45

However, when we performed regression tests with the abovefactors against the presence of E. coli, we did not find any suchrelationship: the only factor significantly related to the presenceof E. coli in storage or transport containers was whether it wasfound in the source water. The investigated factors appeared tohold little importance in predicting whether fecal contami-nation would be present in a household storage or transportcontainer. Results of this nominal logistics model are presentedin Table S4.These results suggest that there may be factors other than

those investigated here which more strongly determine whetheror not a clean water source will be recontaminated by thetransport or storage process. One possible factor is that thepouring and handling techniques associated with the watersampling of storage and transport containers were notseparated from the data; training participating households inaseptic techniques was intentionally omitted from this study sothat contamination would be measured in the context of anactual glass of water to be consumed. As such, it is impossibleto separate the observed contamination into what wasintroduced into the sampling process by the survey participantand what was already present in the water containers. A morerigorous examination of the hand-to-water fecal-contaminationpathway, as implemented by Pickering et al.,46 or introducing arandomized aseptic technique element into the surveyinstrument, might provide a more precise line of causationthan that appearing here. If the sampling technique wereprimarily responsible, we would expect hand-washing with soapto be a significant determinant, which it is not.Another possible factor is that these stated behavioral results

may not necessarily reflect actual behaviors. As such, it is notpossible to determine the effectiveness of the current water-treatment practice of adding chlorine at preventing deterio-ration in quality, or of soap applied in washing containers or tohands after defecation, due to the uncertain extent of thisinterview bias.Other potentially important contamination factors that were

present but not measured included (1) the presence or absenceof livestock manure in the proximity of the household, watersource, or storage container; (2) household use of manure forheating or cooking; (3) the extent of seasonal factors such asflooding near water sources or the birthing of farm animals.Animal manure is prevalent in these communities as 94% ofhouseholds own livestock. Charcoal is the primary cooking fuel,although some households may use manure. Despite the highlyseasonal rainfall regime and the fact that households were 12times more likely to experience high incidents of sickness in thewet season, the season does not have a significant effect on thelikelihood of fecal contamination (p = 0.637).Regression Analysis of User Perceptions. Least squares

regression of user perception of water quality with severalfactors is presented in Table S5. The presence of E. coli wasfound to relate significantly (p < 0.1), and in the expecteddirection, to user perceptions of source water. However, thissignificant relationship breaks down for transported waterquality. Furthermore, user perception responded inversely toactual presence of E. coli in storage containers at a significant (p< 0.01) level. These relationships are shown by the estimates ofthe interaction term Waterpathpoint*E. coli. Whether or nothouseholds got sick from their water did significantly deflatetheir overall perception of quality, but this relationship was notsensitive to water point, suggesting that users experiencing

water-related sickness do not hold a certain point of the waterpathway responsible for causing the sickness. In short, users failto discern the health risk associated with recontaminatedstorage containers.

Fecal Contamination and Reported Diarrhea. Therelationship between fecal contamination and health is morestrongly supported in studies investigating water-source qualitythan POU quality.47,48 Some argue that POU contamination isless of a risk due to acquired immunity.39 Results of a chi-square test of independence between reported diarrhea andfecal contamination in source and POU water samples arepresented in Table S6. As indicated by Pearson’s chi-square forboth source (p = 0.0823) and POU (p = 0.128) water samples,although the evidence raises doubt as to the independence ofexperienced diarrhea and presence of fecal contamination, it isinsufficient to reject the null hypothesis. The relationship isstronger at the source than at the POU. There are certainlystatistical limitations to this presence/absence test that may beremoved by enumerating the bacterial contamination.

Comparative Cost-Effectiveness of Water-MonitoringApproaches. We compared the technical feasibility ofColisure as well as of several technologies currently availablefor rural water applications: two leading portable field kits(PFKs), Potatest and Potaflex,49 and a recently developedcombination of Petrifilm and Colilert-10 mL known as the EC-Kit.21,50 This comparison is presented in SI section S6. Whilethe EC-Kit possesses many features that enhance its feasibilityin remote areas, it lacks the sensitivity necessary for compliancetesting. Colisure possesses similar advantages, except for itsrequirement of an electrical incubator. As discussed above, thisrequirement may be bypassed by simple means. Both Colisureand the EC-Kit are simple enough to be operated by untrained(i.e., local) personnel, which eases economic and technicalconstraints. Figure 4 demonstrates the cost-effectiveness ofthese alternatives.For compliance testing, Colisure was found to be the most

cost-effective approach with moderate sample sizes in remoteareas, largely because engaging professional staff for operating

Figure 4. Per-sample cost ranges by the number of rural microbialwater tests conducted for detection of E. coli using various appropriateon-site methodologies: PFKs using membrane filtration, the EC-Kit(not suitable for compliance testing), and Colisure. Cost ranges aredetermined by the cost of consumables as well as the operational costs,which are investigated for engaging local rural health practitioners orexternal (professional) staff, according to the requirements of eachmethod. They are determined by the number of staff workers required,their corresponding wage rates and transport costs, and may beexpected to vary by remoteness and local economic trends. Details ofthis comparison are given in SI section S7.

Environmental Science & Technology Article

dx.doi.org/10.1021/es402175n | Environ. Sci. Technol. XXXX, XXX, XXX−XXXE

PFKs entails considerable operational costs. With larger samplesizes (n > 1000), the method is not cost-effective compared toPFKs. This comparison assumes that there is no limit to theavailability of trained external technicians or to the logisticalresources necessary for implementing such approaches. Asdiscussed above, logistical difficulties may reduce the feasibilityof routine testing by outside experts in some remote areas,further enhancing the advantage of locally implementedapproaches. The EC-Kit is comparable to Colisure at allsample sizes, and may be the most suitable approach in areaswhere surveillance, rather than compliance, testing is desired.Application of Findings. The tremendous logistical and

economic challenges associated with monitoring water qualityin remote populations of low-income countries require carefuldesign of monitoring efforts. These results support previousfindings that a significant disparity exists between theconformity of unimproved and improved water sources tomicrobial quality guidelinesmost improved sources areuncontaminated, and vice versa. With global water-develop-ment efforts recently reaching the MDGs for drinking water,most (86%) remote populations in low-income countriesalready have access to improved water sources.51 As such,monitoring efforts in most remote areas need not focus onranking health risks associated with unimproved sourcesaspromoted by the WHO’s ‘risk-level analysis’.1 Improved sourcesexist, and households should be using them. Instead, thehighest priority for most remote areas lacking water-quality datais to monitor the long-term performance of improved watersources vis-a-vis their compliance to drinking water guidelines.For these sources, compliance testing could be implementedthrough either P-A or enumeration formats, as long as theypossess the adequate sensitivity to detect 1 CFU E. coli per 100mL.The second priority is to monitor the level of recontamina-

tion occurring from improved sources to the POU. Despitesignificant progress, water networks remain out of reach for asignificant proportion of remote populations; only 46% in low-income countries have access to household taps, while the restcarry water from either improved (40%) or unimproved (14%)water sources to their households.51 Tracing household-levelrecontamination is important for identifying countermeasures,such as POU treatment,42−45 which could be more effectivelytargeted where appropriate data are available. This studyoutlines the need for such tests to boost low communityawareness of this recontamination. Again, compliance testing ofPOU water samples could be implemented through either P-Aor enumeration formats. An important advantage of promotingthese two applications is that the inaccuracy associated withtesting heavily contaminated or environmental water sampleswould be minimized.As this study demonstrates, enzyme-based methods offer a

promising, cost-effective alternative for achieving these twopriority goals, especially where small sample sizes are needed.They are capable of obtaining microbial quality data forcompliance testing using a simple, low-cost format that can belocally implemented. This may apply, for example, to localwater service programs needing reliable measures of drinkingwater compliance. Likewise, nongovernmental organizations(NGOs), which are responsible for a high proportion of remotewater improvements in low-income countries, often have smallbudgets, if any, for routine monitoring. As policy shifts continuetoward community ownership of water services, local village

water user committees may also find enzyme-based methodsattractive tools for monitoring their drinking water.These results provide evidence that one such method is

capable of accurately detecting E. coli in a wider temperaturerange than required by a laboratory incubator, which, alongwith demonstrated internal consistency, implies that accuracy isnot sacrificed under an alternative field setup. This field studydemonstrates the potential for further laboratory trials toestablish an approved incubation regime that extends beyondthe temperature range of standard practices.Since Colisure belongs to a larger family of methods, our

findings are expected to apply to other similar EPA-approvedmethodologiesnamely, Colilert, Colilert-18, and Colitag.However, further studies are needed to determine the feasibilityof applying these other approaches in remote areas. Theenumeration vs detection of indicator bacteria is also animportant issue for future work: it is possible to determine amost probable number (MPN) through only a marginalincrease in methodological complexity for Colisure. This mayexpand applicability to the 14% of the population withoutaccess to an improved source to achieve risk-level analyses asrecommended by the WHO.1

Regardless, these findings show that in the ongoing searchfor microbial water tests appropriate for remote areas,adaptation of existing, standard approaches should beinvestigated along with the development of new methodologies.

■ ASSOCIATED CONTENT*S Supporting InformationEnzyme-Based Approaches, Colisure Methodology, FieldIncubator Performance, Water Sample Results by Containers,Regression Analysis, Comparative Analysis of Microbial WaterTest Feasibility in Remote Areas, Cost Analysis of MicrobialTests. This material is available free of charge via the Internet athttp://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Author*Tel.: +972 8 6563433; fax: +972 8 6596909. E-mail address:adamab@bgu.ac.il.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThe authors wish to thank the graduate students of the 2011and 2012 Rural Water Development course as well as theZuckerberg Institute for Water Research for making this studypossible. We thank the Future Hope Church of Simango,Zambia, for the local help, and the Grace & Hope CharitableTrust, USA, for financial support.

■ REFERENCES(1) WHO (World Health Organization)Guidelines for Drinking-waterQuality, Fourth Edition, 2011; pp 1−564(2) Rompre, A.; Servais, P.; Baudart, J.; de-Roubin, M.-R.; Laurent, P.Detection and enumeration of coliforms in drinking water: currentmethods and emerging approaches. J. Microbiol. Methods 2002, 49,31−54.(3) Bej, A.; McCarty, S. Detection of coliform bacteria andEscherichia coli by multiplex polymerase chain reaction: comparisonwith defined substrate and plating methods for water qualitymonitoring. Appl. Environ. Microbiol. 1991, 57 (8), 2429−2432.

Environmental Science & Technology Article

dx.doi.org/10.1021/es402175n | Environ. Sci. Technol. XXXX, XXX, XXX−XXXF

(4) Hunt, M. E.; Rice, E. W. Microbiological examination. InStandard Methods for the Examination of Water and Wastewater, 21sted.; Eaton, A. D., Clesceri, L. S., Rice, W. R., Greenberg, A. E.,Franson, M. A. H., Eds.; American Public Health Association:Washington, DC, 2005.(5) Trevett, A. F.; Carter, R. C.; Tyrrel, S. F. Water qualitydeterioration: A study of household drinking water quality in ruralHonduras. Int. J. Environ. Health Res. 2004, 14, 273−283.(6) Hoque, B. A.; Hallman, K.; Levy, J.; Bouis, H.; Ali, N.; Khan, F.;Khanam, S.; Kabir, M.; Hossain, S.; Shah Alam, M. Rural drinkingwater at supply and household levels: Quality and management. Int. J.Hyg. Environ. Health 2006, 209, 451−460.(7) Islam, M. S.; Siddika, A.; Khan, M. N. H.; Goldar, M. M.;Sadique, M. A.; Kabir, A. N. M. H.; Huq, A.; Colwell, R. R.Microbiological analysis of tube-well water in a rural area ofBangladesh. Appl. Environ. Microbiol. 2001, 67, 3328−3330.(8) Bain, R.; Gundry, S. W.; Wright, J. A.; Yang, H. Accounting forwater quality in monitoring access to safe drinking-water as part of theMillennium Development Goals: lessons from five countries. Bull.World Health Org. 2012, 90, 228−235.(9) Onda, K.; LoBuglio, J.; Bartram, J. Global access to safe water:Accounting for water quality and the resulting impact on MDGprogress. Int. J. Environ. Res. Public Health 2012, 9, 880−894.(10) Bain, R.; Bartram, J.; Elliott, M.; Matthews, R.; McMahan, L.;Tung, R.; Chuang, P.; Gundry, S. A summary catalogue of microbialdrinking water tests for low and medium resource settings. Int. J.Environ. Res. Public Health 2012, 9, 1609−1625.(11) Hamilton, W. P.; Kim, M.; Thackston, E. L. Comparison ofcommercially available Escherichia coli enumeration tests: Implicationsfor attaining water quality standards. Water Res. 2005, 39, 4869−4878.(12) Merle, R.; Bleul, I.; Schulenburg, J.; Kreienbrock, L.; Klein, G.Drinking Water Test Methods in crisis-afflicted areas: Comparison ofmethods under field conditions. Foodborne Pathog. Dis. 2011, 8, 1185−1189.(13) Pitkanen, T.; Karinen, P.; Miettinen, I. T.; Lettojarvi, H.;Heikkila, A.; Maunula, R.; Aula, V.; Kuronen, H.; Vepsalainen, A.;Nousiainen, L.-L.; Pelkonen, S.; Heinonen-Tanski, H. Microbialcontamination of groundwater at small community water supplies inFinland. AMBIO 2010, 40, 377−390.(14) Mavridou, A.; Smeti, E.; Mandilara, G. Equivalency testing ofTTC Tergitol 7 agar (ISO 9308−1: 2000) with five culture media forthe detection of E. coli in water samples in Greece. Water Sci. Technol.2010, 61 (1), 67−75.(15) Pitkanen, T.; Paakkari, P.; Miettinen, I. Comparison of mediafor enumeration of coliform bacteria and Escherichia coli in non-disinfected water. J. Microbiol. Methods 2007, 68, 522−529.(16) Kloot, R. W.; Radakovich, B.; Huang, X.; Brantley, D. D. AComparison of bacterial indicators and methods in rural surfacewaters. Environ. Monit. Assess. 2006, 121, 273−285.(17) Fricker, E.; Illingworth, K. S.; Fricker, C. R. Use of twoformulations of Colilert and QuantiTray for assessment of thebacteriological quality of water. Water Res. 1997, 31 (10), 2495−2499.(18) Fricker, E. J.; Fricker, C. R. Use of two presence/absencesystems for the detection of E. coli and coliforms from water. WaterRes. 1996, 30 (9), 2226−2228.(19) Macy, J.; Dunne, E.; Angoran-Benie, Y. Comparison of twomethods for evaluating the quality of stored drinking water in Abidjan,Cote d’Ivoire, and review of other comparisons in the literature. J.Water Health 2005, 3 (3), 221−228.(20) Gundry, S.; Wright, J.; Conroy, R.; Preez, Du, M.; Genthe, B.;Moyo, S.; Mutisi, C.; Ndamba, J.; Potgieter, N. Contamination ofdrinking water between source and point-of-use in rural households ofSouth Africa and Zimbabwe: implications for monitoring theMillennium Development Goal for water. Water Practice andTechnology 2006, 1, 1−9.(21) Chuang, P.; Trottier, S.; Murcott, S. Comparison andverification of four field-based microbiological tests: H 2S test, Easygel®, Colilert ®, Petrifilm. Journal of Water, Sanitation and Hygiene forDevelopment 2011, 1, 68−85.

(22) Horman, A.; Hanninen, M. Evaluation of the lactose Tergitol-7,m-Endo LES, Colilert 18, Readycult Coliforms 100, Water-Check-100,3M Petrifilm EC and DryCult Coliform test methods for detection oftotal coliforms and Escherichia coli in water samples. Water Res. 2006,40, 3249−3256.(23) Brown, J.; Stauber, C.; Murphy, J. L.; Khan, A.; Mu, T.; Elliott,M.; Sobsey, M. D. Ambient-temperature incubation for the fielddetection of Escherichia coli in drinking water. J. Appl. Microbiol. 2011,110, 915−923.(24) Olstadt, J.; Schauer, J.; Standridge, J. A comparison of ten USEPA approved total coliform/E. coli tests. Journal of Water and Health2007, 2 (5), 267−282.(25) IDEXX Laboratories, Inc. Colilert, Colilert-18, Colisure; OneIDEXX Dr, Westbrook, ME 04092; 2004.(26) Colitag manufacturer literature, CPI International, 5580 SkylaneBlvd, Santa Rosa, CA 95403.(27) US EPA; Federal Register 40 CFR Part 141, Dec. 5, 1994;Colisure.(28) Edberg, S.; Allen, M.; Smith, D. Enumeration of total coliformsand Escherichia coli from source water by the defined substratetechnology. Appl. Environ. Microbiol. 1990, 56 (2), 366−369.(29) Edberg, S. Defined substrate technology test for E. coli. Appl.Environ. Microbiol. 1991, 57 (11), 3403−3404.(30) McFeters, G.; Pyle, B.; Gillis, S. Chlorine injury and thecomparative performance of Colisure (TM), ColiLert (TM) andColiQuik (TM) for the enumeration of coliform bacteria and E. coli indrinking water. Water Sci. Technol. 1993, 27 (3−4), 261−265.(31) McFeters, G.; Broadaway, S.; Pyle, B. Comparative performanceof Colisure. J. Am. Water Works Assoc. 1997, 89 (9), 112−120.(32) Sampson, R. W.; Swiatnicki, S. A.; Osinga, V. L.; Supita, J.;McDermott, C. M.; Kleinheinz, G. T. Effects of temperature and sandon E. coli survival in a northern lake water microcosm. J. Water Health2006, 4 (3), 389−393.(33) Gentry, R. W.; McCarthy, J.; Layton, A.; McKay, L. D.;Williams, D.; Koirala, S. R.; Sayler, G. S. Escherichia coli loading at ornear base flow in a mixed-use watershed. J. Environ Qual. 2006, 35 (6),2244−2249.(34) Kleinheinz, G. T.; McDermott, C. M.; Leewis, M.-C.; Englebert,E. Influence of sampling depth on Escherichia coli concentrations inbeach monitoring. Water Res. 2006, 40, 3831−3837.(35) Janezic, K. J.; Ferry, B.; Hendricks, E. W.; Janiga, B. A.; Johnson,T.; Murphy, S.; Roberts, M. E.; Scott, S. M.; Theisen, A. N.; Hung, K.F.; Daniel, S. L. Phenotypic and genotypic characterization ofEscherichia coli isolated from untreated surface waters. The OpenMicrobiology Journal 2013, 7, 9−19.(36) Wohlsen, T.; Bates, J.; Vesey, G.; Robinson, W. A.; Katouli, M.Evaluation of the methods for enumerating coliform bacteria fromwater samples using precise reference standards. Lett. Appl. Microbiol.2006, 42, 350−356.(37) Abramson, A.; Becker, N.; Garb, Y.; Lazarovitch, N. Willingnessto pay, borrow, and work for rural water service improvements indeveloping countries. Water Resour. Res. 2011, 47, 1−12.(38) The study on the National Water Resources Master Plan in theRepublic of Zambia: Final Report, Japan International CooperationAgency (JICA): Tokyo, 1995.(39) VanDerslice, J.; Briscoe, J. All coliforms are not created equal: Acomparison of the effects of water source and in-house watercontamination on infantile diarrheal disease. Water Resour. Res. 1993,29, 1983−1995.(40) Chidavaenzi, M. T., Jere, M.; Nhandara, C.; Chigunduru D.;Bradley, M. An evaluation of water urns to maintain domestic waterquality; 24th WEDC Conference; Islamabad, Pakistan, 1998.(41) Mazengia, E.; Chidavaenzi, M. T.; Bradley, M.; Jere, M.;Nhandara, C.; Chigunduru, D.; Murahwa, E. C. Effective and culturallyacceptable water storage in Zimbabwe: maintaining the quality ofwater abstracted from upgraded family wells. J. Environ. Health 2002,64 (15−8), 25.(42) Psutka, R.; Peletz, R.; Sandford, M.; Kelly, P.; Clasen, T.Assessing the microbiological performance and potential cost of

Environmental Science & Technology Article

dx.doi.org/10.1021/es402175n | Environ. Sci. Technol. XXXX, XXX, XXX−XXXG

boiling drinking water in urban Zambia. Environ. Sci. Technol. 2011, 45,6095−6101.(43) Du Preez, M.; Mcguigan, K.; Conroy, R. Solar disinfection ofdrinking water in the prevention of dysentery in South Africanchildren aged under 5 years: The role of participant motivation.Environ. Sci. Technol. 2010, 44, 8744−8749.(44) Schmidt, W.-P.; Cairncross, S. Household water treatment inpoor populations: Is there enough evidence for scaling up now?Environ. Sci. Technol. 2009, 43, 986−992.(45) Sobsey, M. D.; Stauber, C. E.; Casanova, L. M.; Brown, J. M.;Elliott, M. A. Point of Use household drinking water filtration: Apractical, effective solution for providing sustained access to safedrinking water in the developing world. Environ. Sci. Technol. 2008, 42,4261−4267.(46) Pickering, A.; Davis, J.; Walters, S.; Horak, H.; Keymer, D.;Mushi, D.; Strickfaden, R.; Chynoweth, J.; Liu, J.; Blum, A.; Rogers, K.;Boehm, A. Hands, Water, and Health: Fecal Contamination inTanzanian Communities with Improved, Non-Networked WaterSupplies. Environ. Sci. Technol. 2010, 44, 3267−3272.(47) Gleeson, C.; Gray, N. The coliform index and waterborndisease: Problems of microbial drinking water assessment; E & FNSpon: London, England, 1997.(48) Briscoe, J. The role of water supply in improving health in poorcountries (with special reference to Bangla Desh). Am. J. Clin. Nutr.1978, 31, 2100−2113.(49) Wagtech, Inc. Potatest WE-10005 Instructions; www.wagtech.co.uk(50) WHO (World Health Organization)Compendium of new andemerging health technologies, 2011; pp 1−54; www.who.int/medical_devices/en/index.html.(51) WHO (World Health Organization) and UNICEF (UnitedNations’ International Emergency Children’s Fund) Progress onDrinking Water and Sanitation, 2012; pp 1−66; www.who.int/water_sanitation_health/publications/2012/jmp_report/en/.

Environmental Science & Technology Article

dx.doi.org/10.1021/es402175n | Environ. Sci. Technol. XXXX, XXX, XXX−XXXH