Are Presence/Absence Microbial Tests Appropriate for Monitoring Large Urban Water Supplies in Sub-Saharan Africa?

Screening for fecal contamination via microbial water quality monitoring is a critical component of safe drinking water provision and public health protection. Achieving adequate levels of microbial water quality testing, however, is a challenge in resource-limited settings. One strategy for addressing this challenge is to improve the efficiency of monitoring programs. In African countries, quantitative microbial testing methods are commonly used to monitor chlorinated piped water systems. However, presence/absence (P/A) tests may provide an appropriate alternative for water supplies that generally show negative fecal contamination results. This study compares 1048 water quality test results for samples collected from five African urban water systems. The operators of the systems conducted parallel tests on the 1048 samples using their standard quantitative methods (e.g., most probable number or membrane filtration) and the Colitag™ method in P/A format. Combined data demonstrates agreement rates of 97.9% (1024/1046) for detecting total coliforms and 97.8% (1025/1048) for detecting E. coli. We conclude that the P/A test offers advantages as a simpler and similarly sensitive measure of potential fecal contamination for large, urban chlorinated water systems. P/A tests may also offer a cost-effective alternative to quantitative methods, as they are quicker to perform and require less laboratory equipment.


Introduction
The ingestion of fecal pathogens transmitted by contaminated water, food, and hands are the primary bases of diarrhea, a leading cause of sickness and death among children under the age of five [1][2][3][4]. Increasing evidence suggests that exposure to fecal pathogens also contributes to malnutrition and childhood stunting, even in the absence of diarrhea [5][6][7]. Water quality improvements can reduce diarrheal disease levels [3,4,8], and microbial water quality testing is a critical component of water safety management [9]. 27,000 test results from piped water systems across Africa showed that fecal indicator bacteria were only detected in 4% of samples from water piped to plots and 2% of samples from water piped to public taps and standpipes [25].
Based on these findings, we hypothesized that P/A methods are appropriate alternatives to quantitative microbial water quality assays for chlorinated piped water supplies in Africa. To test this hypothesis, we compared P/A and quantitative test results for water samples collected in five African cities: Ouagadougou, Burkina Faso; Abidjan, Côte d'Ivoire; Nairobi, Kenya; Bamako, Mali; and Dakar, Senegal.

Materials and Methods
This study was implemented through a collaboration between The Aquaya Institute and urban water suppliers in five African countries: L'Office National de l'Eau et de l'Assainissement (ONEA) in Burkina Faso; Société de Distribution d'Eau de le Côte d'Ivoire (SODECI) in Côte d'Ivoire; Nairobi City Water and Sewerage Company (NCWSC) in Kenya; Société Malienne de Gestion de l'Eau Potable (SOMAGEP) in Mali; and Sénégalaise des Eaux (SDE) in Senegal. These water suppliers operate and monitor large, chlorinated piped distribution networks in varied geographic settings and were interested in evaluating P/A microbial testing for their operational monitoring needs. The water suppliers conducted parallel tests for total coliforms and E. coli on a total of 1048 water samples using their established quantitative methods and a P/A method. We have listed these suppliers as sites 1-5 in the results in order to ensure their confidentiality.

Sample Collection
Between June 2015 and August 2016, the five water suppliers participating in this study collected approximately 200 water samples from their distribution networks according to the locations, frequencies, and quality control procedures outlined in their routine water sampling plans. The samples were collected as volumes of at least 200 mL in sterile vessels to provide sufficient quantities for two parallel tests (quantitative and P/A) on 100 mL from the following sources: piped networks (700 samples), treatment plants (100 samples), treated water reservoirs (79 samples), untreated groundwater (27 samples), and untreated surface water (82 samples). The water suppliers also assayed 33 negative control and 28 positive (spiked) control samples, which are further described in Section 2.3. Along with the water sample source, sampling staff recorded information into a standardized data collection sheet that included characteristics on water conditions at the time of sample collection (i.e., turbidity and free chlorine residual), the sampling location, and the date of sample collection (see Supplementary Material).

Microbial Testing Methods
The Membrane Filtration (MF) technique is widely used for the enumeration of coliforms and fecal coliforms in drinking water. The method consists of filtering a measured water sample through a cellulose acetate membrane with 0.45-µm diameter pores. Bacteria are retained on the surface of the membrane, which is incubated on a growth medium that selectively promotes multiplication and colony formation by coliform species. Incubation at 37 • C allows for the growth of all coliforms and incubation at 44.5 • C selects for thermotolerant species. Some media also differentiate between total coliforms and E. coli colonies according to the colors that each group produces after metabolizing specific compounds in the growth medium. Four of the five water suppliers that collaborated in this study used MF as their quantitative testing method ( Table 1).
The multiple test tube method is a commonly employed Most Probable Number (MPN) assay for coliform indicator species in drinking water [26]. The method consists of inoculating a series of tubes containing selective growth medium with different dilutions of the water sample, screening for gas production resulting from lactose fermentation (a characteristic of coliform species) and conducting confirmatory tests that also check for E. coli in the samples. Results of the multiple test tube method are reported as a MPN: the statistical estimate of the mean number of coliforms present in the sample, which is based on the number of tubes (and corresponding dilution levels) that were confirmed for coliform growth [26]. In this study, only Site 1 used the MPN method (Table 1).
Colitag™ served as the P/A method and is approved by the U.S. EPA for the simultaneous detection of total coliforms and E. coli in drinking water for both P/A and MPN formats [27]. If coliforms are present in the sample, the enzyme β-galactosidase will hydrolyze the chromogenic indicator ortho-nitrophenyl-β-D-galactopyranoside (ONPG) to release a yellow-colored compound. If E. coli are present in the sample, the enzyme β-glucuronidase will hydrolyze the fluorogenic indicator 4-methylumbelliferyl-β-D-glucuronide (MUG) to release a compound that fluoresces when exposed to longwave ultraviolet light. Fluorescence differentiates E. coli from other coliforms.

Quality Control
Sterile deionized or distilled water samples were used as negative controls (blanks) in each round of water quality testing. Positive control samples were generated by spiking sterile water samples with verified cultures of E. coli. In some cases, study partners did not have access to E. coli control strains and utilized likely contaminated sources to confirm media reactions (e.g., untreated groundwater or untreated raw water).

Data Analysis
To calculate rates of agreement between the quantitative and P/A methods, we created 2 × 2 contingency tables for total coliforms and E. coli [33]. We also compared testing methods by examining the numbers (fractions) of positive and negative results using logistic regression and chi-square (X 2 ) tests [20]. We also performed the analysis on data disaggregated by institution (see Appendix A).

Results and Discussion
We conducted parallel tests on a total of 1048 water samples from the five urban water supply systems ( Table 2). Comparisons of the proportions of samples that tested positive or negative for total coliforms and E. coli showed that the results of the two diagnostic methods were similar. Sixteen percent of the samples (168/1046) tested positive for total coliforms according to the P/A method and 14% (150/1046) tested positive for total coliforms according to quantitative methods. Twelve percent of the samples (123/1048) tested positive for E. coli according to the P/A method and 10% (106/1048) tested positive for E. coli using quantitative methods (Figure 1). There was no significant difference between testing methods (P/A vs. quantitative) when comparing fractions of positive samples for total coliforms (16% vs. 14%, p = 0.29) or E. coli (12% vs. 10%, p = 0.23) (Tables 3 and 4). Combined results from these tests demonstrated that the agreement rates between the quantitative and P/A methods were 97.9% (1024/1046) for total coliforms and 97.8% (1025/1048) for E. coli (Table 2). For the five samples with a negative P/A and positive quantitative result (two samples for total coliforms and three for E. coli), all had <5 CFUs for the quantitative method.
We also disaggregated samples by institution (Tables 3 and 4). For total coliforms, there was no significant difference among testing methods when comparing the fractions of positive samples at the institution level (all p > 0.05) ( Table 3). However, at one institution (Site 2) there was a significant difference between the two methods in E. coli detection-11% of samples tested positive for E. coli according to quantitative assays and 18% of samples tested positive according to the P/A method (p = 0.03) ( Table 4).
The higher fraction of positive results for E. coli that were generated by the P/A method at Site 2 merits further discussion. While it is possible that some of these are false positives, the P/A method demonstrated a low false positive rate of only 1% for E. coli and 5.3% for total coliforms at 24 h in the U.S. EPA Alternative Test Procedure (ATP) approval study [34]. Among Site 2 samples that tested positive for E. coli using the P/A method and negative using the quantitative method, 5/18 were either positive control strains or raw water, which were likely to be true positives. The remaining samples were piped water samples, which likely contained chlorine-injured organisms. The P/A method was specifically designed to improve the detection of fecal contamination by resuscitating chlorine-damaged bacteria, which may have contributed to the higher rate of E. coli detection by the P/A method at Site 2 [35]. Site 1 samples, which used the same quantitative method as Site 2, however, did not reveal a significant difference in E. coli or total coliform recovery. These data suggest that the P/A method may be more accurate at detecting E. coli contamination for certain water sample types than the quantitative method, particularly when injured organisms are present. One study found that recovery performance of some methods may be affected by sample matrix differences, such as heterotrophic background growth, alkalinity, and pH, though Colitag™ produced high recovery rates regardless of sample type [36].
Furthermore, water testing results from the distribution network samples indicate low levels of contamination in the piped systems. According to the quantitative results, 96.3% (674/700) of distribution network samples tested negative for total coliforms and 99.1% (694/700) tested negative for E. coli. These low levels of contamination are comparable to our previous analysis of over 27,000 piped water samples in Africa, which found that 97% were free from fecal contamination [25]. Low levels of contamination are suitable for P/A tests when screening for fecal indicator bacteria.    Furthermore, water testing results from the distribution network samples indicate low levels of contamination in the piped systems. According to the quantitative results, 96.3% (674/700) of distribution network samples tested negative for total coliforms and 99.1% (694/700) tested negative for E. coli. These low levels of contamination are comparable to our previous analysis of over 27,000 piped water samples in Africa, which found that 97% were free from fecal contamination [25]. Low levels of contamination are suitable for P/A tests when screening for fecal indicator bacteria.
The strong correlations between the results of the P/A and quantitative assays for microbial water quality indicators and the low levels of contamination in the piped distribution networks indicate that simple P/A tests, whose performance has been approved by an independent agency, such as the U.S. EPA, are appropriate alternatives to laborious quantitative diagnostics for monitoring well-managed chlorinated piped distribution networks in Africa. As most water quality monitoring institutions in sub-Saharan Africa do not achieve testing levels specified in standards or guidelines, the simpler P/A test may, therefore, provide a solution to increasing sampling rates.  The strong correlations between the results of the P/A and quantitative assays for microbial water quality indicators and the low levels of contamination in the piped distribution networks indicate that simple P/A tests, whose performance has been approved by an independent agency, such as the U.S. EPA, are appropriate alternatives to laborious quantitative diagnostics for monitoring well-managed chlorinated piped distribution networks in Africa. As most water quality monitoring institutions in sub-Saharan Africa do not achieve testing levels specified in standards or guidelines, the simpler P/A test may, therefore, provide a solution to increasing sampling rates. African water sector authorities should consider these results in their designs of water quality monitoring strategies and regulations.
However, even with regulatory acceptance of P/A methods, their application by African water suppliers and surveillance agencies will likely depend on the costs of procuring validated P/A tests that are imported from Europe or the United States (however, it is worth noting that most quantitative supplies are also imported from North America and the U.S.). The U.S. EPA considered reduced costs for P/A testing in comparison to quantitative methods when it adopted the Total Coliform Rule in 1989 [22]. Our previous analysis of reported testing costs in sub-Saharan Africa (including equipment, consumables, and labor) found that the per-test cost of H 2 S presence/absence tests (8.3 USD) was lower than the average per-test cost for membrane filtration (12.5 ± 8.1 USD) and MPN (14.0 ± 12.4 USD) [37]. However, this was not the case when only examining consumable costs: H 2 S presence/absence consumable costs (5.5 USD) were higher than average consumable costs for membrane filtration (3.1 ± 1.8 USD) or MPN methods (4.3 ± 3.9 USD) [37] (it is important to note that H 2 S cost data was only from one institution and no data was available on E. coli P/A methods). Similarly, a study by Bain et al. [16] that obtained costs from websites, catalogues, and quotations from manufacturers and suppliers found that the cost per test for P/A testing was approximately equal to quantitative methods (ranging from 0.6-5.0 USD and 0.5-7.5 USD, respectively, though these did not include import, delivery, or distributer mark-up prices in Africa). However, both studies found that equipment costs were generally lower for P/A compared to quantitative methods [16,37]. Furthermore, there is a possibility for additional cost savings from reduced staff time required for P/A methods (<5 min for sample handling) compared to quantitative methods (up to 30 min) [16].
A limitation of our study is that we did not generate all of our positive control samples using known control strains. Instead, in cases when water suppliers did not have access to control strains, they drew samples from likely contaminated sources, such as untreated groundwater or untreated raw water. Additionally, we were unable to conduct confirmatory tests of the positive results because of the capacity of the site laboratories.
The economic perspective outlined above suggests that despite the performance of the P/A test in monitoring chlorinated, piped distribution networks in African cities, the low consumable costs of quantitative methods will continue to favor their use over internationally validated, ready-to-use P/A tests that have higher per-unit costs, particularly among monitoring agencies that have already invested in the equipment and staff needed to perform quantitative methods. Nevertheless, validated P/A tests may still prove cost-effective in settings with chlorinated water supplies that are not supported by established laboratories.