Professional Documents
Culture Documents
Stauber 2014
Stauber 2014
a r t i c l e i n f o a b s t r a c t
Article history: Aims: Annually, more than 2 million diarrheal disease deaths can be attributed to the lack of access to water, san-
Received 3 December 2013 itation and hygiene. These deaths occur mostly in developing countries where water quality testing resources are
Received in revised form 13 February 2014 limited. Several tests are currently used to detect and quantify Escherichia coli and other fecal bacteria in drinking
Accepted 13 February 2014 water; however they can be expensive, complex, and technically demanding. There is a need for a simple, reli-
Available online 21 February 2014
able, low-cost water quality test that can be used in resource limited settings. Therefore, the purpose of this re-
search was to perform a rigorous evaluation of the recently developed compartment bag test for detection and
Keywords:
Environmental health
quantification of E. coli against the standard method of membrane filtration.
Drinking water quality Methods and results: A total of 270 water samples were collected from forty-five various naturally contaminated
Microbial water testing water sources around metro-Atlanta from August 2011 through April 2012. Samples were processed using the
Fecal indicator compartment bag test and membrane filtration with mI agar. Concentrations of E. coli were significantly correlat-
E. coli ed with a correlation coefficient of 0.904 (95% CI 0.859–0.950). Sensitivity and specificity were 94.9% and 96.6%,
Safe water respectively.
Conclusions: These results suggest that the compartment bag test produces results consistent with those pro-
duced by membrane filtration on mI agar. Based upon its performance, the compartment bag test has the poten-
tial to be used as a reliable, affordable drinking water quality test where other microbial water quality testing
resources are not readily available.
© 2014 Elsevier B.V. All rights reserved.
1. Introduction Health Organization has specified that zero E. coli per 100 ml of
water is the goal for all water supplies (WHO, 2006). There are sev-
Globally, roughly 783 million people lack access to improved drinking eral current tests used to detect and quantify E. coli and other fecal
water, and over two billion have no basic sanitation (WHO/GLAAS, 2012). bacteria in drinking water; however they can be expensive, complex,
There are approximately two billion cases of diarrheal disease globally and time consuming (Boubetra et al., 2011). Most of these tests require
every year, making it one of the leading causes of preventable deaths trained laboratory personnel and a laboratory setting that may not be
worldwide and the second leading cause of mortality and morbidity in available in remote areas or those with limited resources (NRC, 2004).
children under the age of 5 years (Liu et al., 2012). One of the major In particular, the lack of access to microbial water analysis kits or
causes of the more than 2 million deaths annually due to diarrheal disease laboratories is an issue for many communities in the developing
is the consumption of contaminated water (WHO/UNICEF, 2009). world (Sundram et al., 2000).
The microbial quality of water has a large impact on health in de- There is a need for simple, reliable, low-cost microbial tests that will
veloping countries where access to safe drinking water is limited be readily available to people and institutions in developing countries
(Fewtrell and Bartram, 2001). Contaminated drinking water may with limited access to water, sanitation, and hygiene in order to provide
contain unsafe levels of microorganisms that pose a risk to human data that facilitate the prevention of waterborne diarrheal disease
health (JMP, 2010). Fecal (also called thermotolerant) coliforms, (McMahan et al., 2011). To meet this testing need, the compartment
particularly Escherichia coli (E. coli), have been used as indicators bag test, developed by researchers at the University of North Carolina
of fecal contamination of drinking water (Horan, 2003). The World at Chapel Hill, is a self-contained, portable, less expensive and simple
to use test to detect and quantify the presence of E. coli in drinking
water based upon the most probable number principle of quantification.
⁎ Corresponding author. Tel.: +1 404 413 1128. This test can be used without extensive laboratory equipment that is
E-mail address: cstauber@gsu.edu (C. Stauber). normally needed for standard tests for drinking water quality. The
http://dx.doi.org/10.1016/j.mimet.2014.02.008
0167-7012/© 2014 Elsevier B.V. All rights reserved.
C. Stauber et al. / Journal of Microbiological Methods 99 (2014) 66–70 67
purpose of this study was to evaluate the compartment bag test for the
detection and quantification of E. coli in water against a standard meth-
od, membrane filtration using mI agar, for various natural water sources
in Atlanta, GA.
respectively. Using continuous and log10 transformed data, comparisons isolates from both the compartment bag test and membrane filtration
were made with mean comparison tests. A Wilcoxon rank sum test was not presumed to be E. coli based on colony color were also biochemically
also performed to compare the two methods. tested. Only 1 of these isolates was identified as E. coli. The biochemical
identities of these isolates included Enterobacter cloacae, Enterobacter
aerogenes, Serratia liquefaciens, Klebsiella pneumoniae, Enterobacter
3. Results amnigenus, Citrobacter freundii, and Entrobacter agglomerans. Based on
the biochemical analysis of presumptive E. coli isolates, 99% were iden-
A total of 270 naturally contaminated water samples from various tified as E. coli with one isolate (1%) from membrane filtration identified
water sources were tested and grouped by the type of water source as Enterobacter aerogenes.
(Table 1). The geometric and arithmetic mean E. coli counts were 1.52
(95% CI 1.39–1.64) MPN/100 ml and 212.4 (95% CI 170.5–254.2) MPN/
4. Discussion
100 ml, respectively for the compartment bag test, versus 1.63 (95% CI
1.51–1.76) CFU/100 ml and 232.5 (95% CI 190.7–274.3) CFU/100 ml for
In this study, the compartment bag test, a self-contained, compact,
membrane filtration. The Wilcoxon rank sum test for paired samples field portable method to detect and quantify E. coli in water, performed
revealed that the arithmetic median concentrations of E. coli were signif-
just as effectively as membrane filtration by the US EPA Method 1604 in
icantly different between methods (p = 0.0002). a variety of different water samples collected seasonally in metropolitan
As shown in Table 2, E. coli concentrations for the two methods were
Atlanta, GA. Documenting the reliability of this test is an important step
significantly correlated with a Spearman's rank correlation coefficient of to provide evidence that this simple, low cost water quality test is not
0.904 (95% CI 0.859–0.950). Furthermore, when compared on the basis
only reliable but also practical for use in limited resource settings.
of categorical E. coli concentrations, there was relatively high agree- This evaluation indicated that the compartment bag test had high
ment, especially at higher concentrations as shown in Table 3. These
specificity (96.6%), positive predictive value (N97%), and sensitivity
categories [b 1 CFU/100 ml (safe or low risk), 1–10 CFU/100 ml (inter- (94.9%), but the negative predictive value was somewhat lower
mediate risk), 11–100 CFU/100 ml (high risk), and N100 CFU/100 ml
(~ 70%). When tested against membrane filtration with mI agar, the
(very high risk)] are used to indicate fecal contamination and the possi- compartment bag test was able to correctly detect 94.9% of all E. coli
ble association with level of waterborne disease risk. As shown in positive results, and 96.6% of all E. coli negative results. As a result of pro-
Table 3, the extent of agreement in categorical concentrations of E. coli ducing few false positives (0.45%), a positive result observed with the
between the two methods was 70%, 71.4%, 73.5% and 92.8% for the con- compartment bag test is a good predictor of the presence of E. coli
centration categories of b1, 1–10, 11–100 and N100/100 ml, (PPV = 99.6%). However, the compartment bag test produced a moder-
respectively. ate number of false negatives (30%), suggesting that a negative result
Out of 263 samples tested with both methods, 12 were negative for does not always assure the absence of E. coli (NPV = 70%). These results
E. coli (b1 MPN/100 ml) when tested with the compartment bag test, were comparable to those found by McMahan et al. with sensitivity,
whereas the same samples were positive for E. coli (N1 CFU/100 ml) specificity, NPV, and PPV determined to be 73%, 100%, 68%, and 100%,
when tested with membrane filtration. These were considered to be respectively for the compartment bag test when applied to the detec-
false negatives. One (0.45%) sample was considered to be a false positive tion of hydrogen sulfide producing bacteria in surface water. In studies
as it was positive for E. coli when tested using the compartment bag test, of the performance of some other E. coli methods, results for these pa-
but negative for E. coli when tested using membrane filtration. For the
rameters have been similar. Brenner et al. (1993) found that the sensi-
samples analyzed, the compartment bag test produced a sensitivity of tivity and specificity for mI agar when used to detect E. coli was 69.6%
94.9% and a specificity of 96.6% when compared with membrane filtra-
and 95.7%, respectively.
tion on mI agar (Tables 4 and 5). When compared using a two sample t-test and Wilcoxon rank sum
By EnterotubeTM II biochemical test results it was found that 100%
test for paired samples, we found statistically significant differences in
(43 out of 43) of the presumptive E. coli isolates obtained from the com- concentrations for the compartment bag test compared to membrane
partment bag test were identified as E. coli (Table 6). Out of a total of 109
filtration with mI agar. The statistically significant difference in concen-
presumptive E. coli isolates from the compartment bag test (43 pre- tration of E. coli between the two methods is likely due to the difference
sumptive E. coli isolates) and membrane filtration (66 presumptive
in media formulation as well as in how each method enumerates E. coli.
E. coli isolates), a total of 108 isolates were identified as E. coli. Fourteen Similar to how IDEXX Colilert Quantitray works, the compartment bag
test uses a broth culture medium to promote the growth of E. coli. Due
Table 1
Descriptive characteristics of water samples collected between August 2011 and April
to differences in the format as well as media formulation, it will likely
2012. not detect metabolically the exact same organisms as membrane filtra-
tion. Differences in concentrations may also be due to the relatively limit-
Type of source Number of samples Season Total
ed most probable number combinations available in a five compartment
Creek 18 Summer 84 bag test compared to a wide range of colony count outcomes (0 to 100
22 Fall
colonies per test) conforming to the Poisson distribution in membrane fil-
24 Winter
20 Spring tration analysis. In their comparison of seven different E. coli enumeration
Lake 22 Summer 86 methods, Hörman and Hänninen (2006) documented similar results.
27 Fall They found that the correlation coefficient suggested high correlation be-
17 Winter tween the MPN method (Colilert) and membrane filtration but mean
20 Spring
Pond 4 Winter 9
concentrations were significantly different between the two methods
5 Spring (Horman and Hanninen, 2006).
Stream 1 Summer 24 Statistically significant correlation of E. coli concentrations (by
3 Fall Spearman's rank correlation) by the two methods illustrates that quan-
8 Winter
tification of E. coli was comparable for both. Hence, the compartment
12 Spring
Rainwater 6 Summer 64 bag test was generally comparable to membrane filtration with mI
42 Fall agar for the quantification of E. coli in source waters of Atlanta, GA. Fur-
8 Winter thermore, based upon the decimal categories of E. coli concentrations,
8 Spring the compartment bag test and membrane filtration methods were in
Other 3 Winter 3
general agreement for matched results in each category. Agreement
C. Stauber et al. / Journal of Microbiological Methods 99 (2014) 66–70 69
Table 2
Results from Spearman's rank correlation comparing E. coli concentrations of the compartment bag test and membrane filtration.
Membrane filtration (CFU/100 ml) Coefficient Standard error t P 95% CI Spearman's rho
for the two methods was 70%, 71.4%, 73.5%, and 92.8% for E. coli concen- of most probable number values are possible within the range of b 1 to
tration categories of b1, 1–10,11–100, andN100 MPN/100 ml or CFU/ nearly 100 MPN/100 ml (McMahan et al.).
100 ml, respectively. Hence, the compartment bag test is able to effec- Although some differences in E. coli presence were observed with
tively indicate categorical concentrations of fecal contamination in samples collected from rainwater sources when tested with the com-
water samples, possibly predicting the level of waterborne disease risk. partment bag test and membrane filtration, they were not statistically
Current methods for detection and quantification of fecal bacteria in significant (data not shown). However, membrane filtration on mI
water can be complex and technically demanding (Boubetra et al., agar gave higher concentrations of E. coli than did the compartment
2011). They require extraneous laboratory equipment, and are difficult bag test, and in some cases there were wide differences in the concen-
to perform in developing countries and during natural disasters. In addi- trations of E. coli detected by the two methods. These findings suggest
tion, other resources including a cold chain, electricity, and incubators the need for further investigation into why these differences were ob-
are too expensive and impractical for use in resource limited settings served. It is noteworthy that in a study of the compartment bag test
(Brown et al., 2011). These requirements present tremendous obstacles compared to the IDEXX Quanti-Tray method for E. coli detection in har-
for microbial water quality testing in developing countries (Chuang vested rainwater samples collected in rural households in Thailand,
et al., 2011). The compartment bag test is self-contained, compact, light- there was not a significant difference in E. coli detection (unpublished
weight and field portable, which is beneficial in situations where access results presented at the 2011 Annual General meeting of the American
to laboratory equipment and supporting materials like pipets, graduat- Society for Microbiology). Reasons for differences found in rainwater
ed cylinders, and membrane filtration apparatus is not feasible. It can be samples in our study versus the study performed in Thailand may be
used in settings where electricity may not readily be available with the due to difference in the estimation format (membrane filtration enu-
utilization of ambient temperature incubation. Despite the fact that the meration of colonies and defined substrate technology quantified by
water samples were processed in a laboratory setting for this analysis, Quantitray MPN) (Hornan and Hinnanen, 2006).
our results demonstrate that, the compartment bag test is likely to be The main limitations of the compartment bag test are the upper de-
capable of performing efficiently in remote settings where resources tection limit of 100/100 ml and potential for variation in reading the re-
are limited, such as in developing countries. Furthermore, the compart- sults of this test, which include color detection and interpretation of the
ment bag test is similar in cost to the other methods available. The cost variation in color production by hydrolysis of the chromogenic E. coli
per test is between $5 and $10, with the cost per test depending on how substrate in the bacteriological medium. Since we performed these ex-
many are purchased. When purchased in progressively larger quanti- periments, the method of color interpretation has been somewhat
ties, the price incrementally decreases (personal communication, Dr. more standardized as the test is being sold commercially and detailed
Mark Sobsey). Estimates for membrane filtration and Colilert MPN instructions on interpretation are now provided (see http://www.
methods range from $0.50 to $6.00 per sample (Bain et al., 2012). aquagenx.com/how-to-use-the-cbt/).
While the compartment bag test is at the higher end of this range, it One additional problem with the test was that the compartment
does not require additional equipment for processing, which can cost bags occasionally leaked which resulted in the need to repeat or the
between $2000 and $4000 depending on the method (Bain et al., loss of the sample. This happened in very few instances (b2%) but
2012), and the test includes its own decontamination system whereas with production and distribution of the test increasing, additional mea-
the others do not. sures of quality assurance and control have been implemented to re-
There are possible sources of analyst variation in reading the results duce this likelihood, according to the test developers and producers. In
of this test, which include color detection and interpretation of the var-
iation in color production by hydrolysis of the chromogenic E. coli sub- Table 4
strate in the bacteriological medium. There may be a small or light Comparison of results from the compartment bag test and membrane filtration based on
production of color change that may only be visible with close examina- sample positivity for E. coli.
tion, and observers may interpret the presence of the diagnostic blue or Membrane filtration
blue-green color differently. With only 5 sample volumes in the com-
CBT Negative Positive Total
partmentalized bag, there are a small number of likely positive bag com-
Negative 28 12 40
partment combinations for the test, and as a result only a small number
70% 30% 100%
Positive 1 222 223
0.45% 99.6% 100%
Total 29 234 263
Table 3 11% 89% 100%
Comparison of categorical concentrations of E. coli from the compartment bag test (CBT)
and membrane filtration.