Risk of Viral Acute Gastrointestinal Illness from Non-disinfected Drinking Water Distribution Systems

Posted in: Drinking Water News, Ground Water News, Water Contamination
Tags: , ,

 

Article Courtesy of Elisabetta Lambertini †, Mark A. Borchardt *‡, Burney A. Kieke , Jr.‡, Susan K. Spencer ‡, and Frank J. Loge *†/ July 27, 2012/Environmental Science & Technology/ Shared as educational materials
Department of Civil and Environmental Engineering, University of California Davis, One Shields Avenue, Davis, California 95616, United States
Marshfield Clinic Research Foundation, 1000 North Oak Avenue, Marshfield, Wisconsin 54449, United States
Environ. Sci. Technol., 2012, 46 (17), pp 9299–9307
DOI: 10.1021/es3015925
Publication Date (Web): July 27, 2012
Copyright © 2012 American Chemical Society
*Tel: +1 530 754 2297; fax: +1 530 752 7872; e-mail: fjloge@ucdavis.edu (F.J.L.). Tel: +1 715 387 4943; fax: +1 715 384 9157; e-mail: mark.borchardt@ars.usda.gov (M.A.B.).

distribution systems

Acute gastrointestinal illness (AGI) resulting from pathogens directly entering the piping of drinking water distribution systems is insufficiently understood. Here, we estimate AGI incidence from virus intrusions into the distribution systems of 14 nondisinfecting, groundwater-source, community water systems. Water samples for virus quantification were collected monthly at wells and households during four 12-week periods in 2006–2007. Ultraviolet (UV) disinfection was installed on the communities’ wellheads during one study year; UV was absent the other year. UV was intended to eliminate virus contributions from the wells and without residual disinfectant present in these systems, any increase in virus concentration downstream at household taps represented virus contributions from the distribution system (Approach 1). During no-UV periods, distribution system viruses were estimated by the difference between well water and household tap virus concentrations (Approach 2). For both approaches, a Monte Carlo risk assessment framework was used to estimate AGI risk from distribution systems using study-specific exposure–response relationships. Depending on the exposure–response relationship selected, AGI risk from the distribution systems was 0.0180–0.0661 and 0.001–0.1047 episodes/person-year estimated by Approaches 1 and 2, respectively. These values represented 0.1–4.9% of AGI risk from all exposure routes, and 1.6–67.8% of risk related to drinking water exposure. Virus intrusions into nondisinfected drinking water distribution systems can contribute to sporadic AGI.

Pathogenic microorganisms have been detected in municipal drinking water systems, even in countries with highly regulated water infrastructure.(1, 2) Pathogen occurrence can be due to contamination in the source water or to intrusions into the distribution system. Most microbial intrusions into distribution systems result from either major losses of physical integrity, such as main breaks, or negative pressure events.(3) Low or negative pressure transients can occur due to sudden shifts in water velocity, e.g. following breaks, valve operation, pump start-ups and shut-offs, or rapid demand shifts.(4, 5) As a result, contaminated water can be drawn into a pipe from the surrounding soil through leaks or faulty joint seals, or flow backward into the system from an unprotected cross-connection with nonpotable water.(6)
Microbial intrusions into distribution systems are considered a threat particularly when no residual disinfectant is applied to the finished water. In the United States, 147,330 public water systems (PWS) supply groundwater to more than 100 million people; of these, 95,631 systems serving 20 million people produce water without disinfection. Another 56.8 million people drink groundwater that, while treated, does not meet the goal of reducing viruses by 99.99%.(7) Applying a residual disinfectant does not guarantee a sanitary distribution system, as the disinfectant can become depleted or be ineffective against some microorganisms.(8, 9)
Pathogens contaminating drinking water as a result of distribution system deficiencies are a significant cause of disease outbreaks.(10-14) Between 1971 and 1998, 133 out of 619 AGI outbreaks in U.S. PWS (9–29% in each reporting period) were directly associated with distribution systems.(13) Recently, between 2001 and 2008, distribution system deficiencies in PWS were responsible for 15 out of 56 outbreaks (20–43% in each reporting period).(15-18) Many outbreaks occurred despite the application of a disinfectant residual.(8, 13) Viruses were responsible for the majority of illnesses related to drinking water.(15-17) Moreover, groundwater systems accounted for 76% of drinking water outbreaks in 1991–2002,(14) 87% in 2003–2006,(15, 17) and 95.2% in 2007–2008. (18)
In contrast to outbreak data, comparatively little is known about pathogen intrusions into distribution systems causing sporadic or endemic AGI. Some studies have suggested an association between AGI and distribution system performance. For instance, during a prospective epidemiological study in a PWS complying with current microbiological standards, Payment et al.(19, 20) observed higher AGI incidence in individuals that drank from household taps, compared to individuals that drank treated municipal water bottled at the treatment plant. Tinker et al.(21) observed a modest but significant association between hospital visits due to AGI and the residence time of drinking water in the distribution system. Similarly, Nygård et al.(22) found a significant association between rates of campylobacteriosis and average length of distribution system reaching a district. Conversely, Payment et al.(19) found no association between AGI rates and water residence time in the system. Some studies have also related AGI to specific distribution system events. For example, in 7 PWS in Norway Nygard et al.(23) observed increased AGI incidence in the population drinking water hydraulically downstream of main breaks or maintenance works. Also, Hunter et al.(24) observed a significant association between self-reported AGI incidence and low pressure episodes at household faucets in a U.K. PWS.
The study described herein was part of the Water And Health Trial for Enteric Risk (WAHTER) Study, which applied methods in epidemiology and quantitative microbial risk assessment (QMRA) to estimate AGI risk related to drinking water in 14 nondisinfecting, groundwater-source, community water systems. The WAHTER Study had four objectives: (1) assess the association between virus levels in tap water and AGI incidence in the study communities; these findings are reported in Borchardt et al.;(25) (2) estimate the fraction of AGI incidence due to virus contamination of the source groundwater; (3) accounting for virus contributions from the groundwater, estimate the fraction of AGI incidence due to virus contamination of the distribution systems; and (4) assess the association between virus levels in tap water and maintenance procedures or failure events in the communities’ distribution systems; these findings are reported in Lambertini et al.(26) The present report focuses on objective 3; specifically, our objective was to quantify human enteric viruses directly entering nondisinfected drinking water distribution systems, and estimate the fraction of sporadic or endemic AGI from distribution system contamination using a QMRA model.

2 Materials and Methods


Study Design

The WAHTER Study was a community-randomized trial with crossover intervention in 14 municipalities located in the state of Wisconsin, USA. The municipally owned PWS rely on groundwater extracted from sand/gravel or sandstone aquifers; groundwater is pumped and distributed to customers without any disinfection. The water systems are further described in Lambertini et al.(26) Intervention consisted of ultraviolet (UV) light disinfection (minimum dose = 50 mJ/cm2) installed on all active wellheads resulting in all drinking water in a community, regardless of tap location, being disinfected upon leaving the ground. Eight communities had UV reactors (WEDECO, Charlotte, NC) installed in the first study year, whereas the six remaining communities continued to use nondisinfected water. Crossover was implemented at the beginning of the second study year by transferring the UV reactors between the two groups; six communities then had UV-disinfected drinking water whereas the eight communities resumed using nondisinfected water. The difference in epidemiologically measured AGI incidence between control and UV intervention periods yielded the fraction of AGI attributable to contaminated groundwater (i.e., objective 2 of the WAHTER Study). Here, we use QMRA to estimate the fraction of AGI attributable to direct contamination of the distribution systems.
Water Sampling and Virus Enumeration

Water samples were collected monthly in each community, from all active well heads, immediately downstream from UV reactors during UV periods, and from 5–8 household taps. Distribution system maps provided by the utilities were used to select household sampling locations to obtain samples representative of different regions of the distribution system. Viruses were concentrated from water samples in the field (mean sample volume 868 L, total sample number 1452) using custom-made glass wool filters.(27) Filters were kept on ice and transported to the laboratory within 48 h. Sampling was carried out over four 12-week periods: April–June 2006, September–November 2006, March–May 2007, and September–November 2007. In the laboratory, the filters were eluted, nucleic acids were extracted, and viral targets were quantified by fluorescence-based reverse transcription quantitative polymerase chain reaction (RT-qPCR) or qPCR following procedures described in Borchardt et al.(25) Inhibition was quantified in every sample and mitigated accordingly by dilution. Primers, probes, and quality assurance parameters for standard curves are reported in the Supporting Information (SI). Six virus types were enumerated as genomic copies (gc)/L: enteroviruses, noroviruses GI and GII, adenoviruses, rotavirus, and hepatitis A virus. Another category, “all-viruses”, was defined as to include all six virus types; when a sample was positive for more than one virus type, the within-sample sum of virus numbers was divided by the sample volume and this concentration of all viruses was assigned to the sample. Samples with no detected viruses were assigned a zero value. Only virus types considered in the risk assessment analysis, enteroviruses, norovirus GI, and all-viruses, are discussed in this paper.
Risk Assessment

Risk of illness was calculated using exposure–response relationships derived as part of the WAHTER study.(25) Exposure was estimated as the arithmetic mean virus concentration in household tap water in each community over 12-week study periods, and related by Poisson regression to the epidemiologically measured AGI incidence during the same 12-week period in the same community. An AGI episode was defined as having three or more episodes of loose watery stools or one episode of vomiting in a 24-h period. AGI incidence was obtained from weekly health diaries completed by 621 households (1079 children ≤12 years old and 580 adults ≥19 years old at the beginning of the study) in the 14 communities, and expressed as number of episodes/person-year.(25)
Exposure–response relationships were expressed as(1)where βo and β are model coefficients, C is the 12-week mean virus concentration (gc/L), and ε is an error term N(0,σ2) with(2)In eq 1, when C = 0, AGI Incidence represents transmission routes other than drinking water (e.g., person-to-person, foodborne).
Seven exposure–response relationships were included in the risk assessment (Table 1). These were selected from among 34 models developed with arithmetic mean virus concentration as the independent variable (reported in Borchardt et al.(25)). Our intent was to encompass a range of model coefficients that differ by virus type, age group, and model type (fixed effect only or mixed) to create a range in estimates for AGI risk from distribution systems. For enterovirus and all-viruses, the models included community and study period as random effects(25) (not shown in eq 1). The norovirus GI models selected for the purpose of the present analysis did not include random effects.
Table 1. Coefficients of the Exposure–Response Models used in the Risk Assessment
exposure-response model by virus type and age group βo β var(βo)b var(β)b cov(βo, β)b p-valuec
all viruses – all agesa –5.4195 7.534 × 10–2 9.720 × 10–3 1.966 × 10–3 –9.646 × 10–4 0.0977
all viruses – adultsa –5.4539 1.715 × 10–1 1.593 × 10–2 3.999 × 10–3 –2.094 × 10–3 0.0101
enterovirus – adultsa –5.4080 2.920 × 10–1 2.294 × 10–2 1.665 × 10–2 –2.361 × 10–3 0.0296
norovirus GI – all ages –5.4271 1.723 × 10–1 1.543 × 10–3 2.250 × 10–3 –7.976 × 10–4 0.0006
norovirus GI – adults –5.4214 2.557 × 10–1 3.487 × 10–3 4.471 × 10–3 –1.770 × 10–3 0.0003
norovirus GI – children ≤12 years –5.4300 1.205 × 10–1 1.277 × 10–3 2.022 × 10–3 –6.680 × 10–4 0.0098
norovirus GI – children <5 years –4.9855 1.826 × 10–1 3.271 × 10–3 5.432 × 10–3 –1.826 × 10–3 0.0165
aModels adjusted for “community” and “study period,” considered as random effects.
bVar: variance. cov: covariance.
cp-values refer to the significance of the regression slope β.
Two distinct approaches were followed to characterize AGI risk associated with distribution systems during UV and no-UV periods. (Here we define risk as synonymous with AGI incidence rate in units of episodes/person-year.) Mathematically, the two approaches are analogous and can be expressed as:(3)(4)where risk(distribution system) represents the AGI risk from distribution system contamination, risk(tap) is the AGI risk estimate corresponding to tap water virus concentrations, and risk(well) is the risk estimate corresponding to virus concentrations measured at well heads. Wellpost-UV refers to the water sampling location immediately following UV disinfection, before water enters the distribution system. Risk(distribution system) is equivalent to an incidence rate difference.
Approach 1 was based on virus data collected during UV periods, capitalizing on the study design in which UV disinfection substantially reduced virus contributions from groundwater.(26) As a result, an increase in virus concentration between the point of UV disinfection at wellheads and downstream at household taps represents virus intrusions into the distribution systems, as UV light does not provide residual disinfection, no chlorine residual was present, and viruses cannot replicate in the environment. Real-time qPCR is a sensitive measure of virus inactivation by UV disinfection as the irradiation causes pyrimidine dimers in the virus genome, preventing the polymerase enzyme from creating target amplicon.(28-30) Unlike some bacteria, viruses lack the repair mechanisms for UV-damaged nucleic acids and, as such, require entry into a host cell and its enzymes before repair is possible.(29) De novo creation of qPCR-measurable virus genomes between UV disinfection and household taps was not possible.
As with Approach 1, Approach 2 (no-UV and no chlorine residual) relied on the measured difference in virus concentrations between wellhead and tap to represent distribution system intrusions, but without the benefit of UV reducing virus contributions from the groundwater source. With both approaches, although unlikely, virus measurements at some household taps could have resulted from plumbing flaws within the house itself and not the municipal distribution system.
Mathematically, the intercept in eq 1 represents AGI from sources other than viruses in tap water. Therefore, inputting into the exposure–response relationships the tap water virus concentrations during UV periods yields an estimate of risk(tapUV), the combined AGI risk from two sources: distribution system intrusions and other transmission routes, such as person-to-person (Figure 1a). Inputting the post-UV virus concentrations yields risk(wellpost-UV), the risk from other transmission routes including the negligible number of groundwater viruses still measurable after UV disinfection. The difference between the former and the latter estimates represents AGI risk from distribution systems (risk(distribution system)) (Figure 1a). Approach 2 is based on an analogous calculation, but the risk associated with groundwater contamination was subtracted only mathematically, instead of through UV disinfection (Figure 1b). A Monte Carlo resampling scheme was used to generate the frequency distribution of the risk outcomes (Figure 2), accounting for the frequency distributions of virus concentrations and the error in the exposure–response relationships.
figure

Figure 1. Conceptual portrayal of the approaches to estimate AGI risk from distribution systems. (a) Approach 1: UV periods when AGI risk from groundwater is reduced, where A = risk from distribution systems (DS) and other transmission routes (other, e.g. food), B = risk from other transmission routes, including risk from any groundwater viruses not inactivated by UV disinfection, and the difference AB = risk from distribution systems. (b) Approach 2: No-UV periods when AGI risk from groundwater is accounted for, where C = risk from distribution systems, groundwater (GW), and other transmission routes, D = risk from groundwater and other routes, and the difference CD = risk from distribution systems.

figure

Figure 2. Example of frequency distributions of AGI risk resulting from 100 000 Monte Carlo iterations using Approach 2 (no-UV) and the norovirus–children <5 years old exposure–response relationship. Main plot: blue histogram, risk(tapno-UV), obtained by inputting tap water virus concentrations into the exposure–response relationship to yield the combined risk from distribution systems, groundwater, and other transmission routes; and green histogram, risk(wellno-UV), obtained by inputting groundwater virus concentrations to yield the risk from groundwater and other routes. Inset: Frequency distribution of risk associated with the drinking water distribution systems, risk(distribution system). For each Monte Carlo iteration, one estimate of risk(wellno-UV) is subtracted from risk(tapno-UV) to derive an estimate of risk(distribution system).

Virus concentrations were pooled into four sets: concentrations in tap water and in groundwater post-UV during UV periods (Approach 1); in tap water and groundwater during no-UV periods (Approach 2). Samples excluded from the analysis, primarily those collected during short-term chlorination, are reported in the SI, as is also the effect of data exclusions on the risk analysis outcomes.
The following steps were performed for each iteration of the Monte Carlo simulation: (1)X concentration values were randomly selected from the pool of tap water samples (532 samples for Approach 1, 556 for Approach 2), where X was drawn from a uniform distribution spanning the number of samples collected at household taps in each community over a 12-week study period (X = 17–24).(2)The arithmetic mean of the X tap concentration values was calculated to obtain a simulated mean tap concentration, consistent with the level of time aggregation (12-week periods) used in deriving the relationships between virus concentrations and AGI incidence.(3)M concentration values were randomly drawn from the pool of post-UV samples (178 samples, Approach 1) or groundwater samples (186 samples, Approach 2), where M was drawn from a uniform distribution spanning the numbers of samples collected at post-UV or well locations in each community over a 12-week period (M = 3–12).

(4)The weighted arithmetic mean of the M post-UV or well concentration values was calculated to obtain a simulated 12-week mean post-UV or well concentration. Weights consisted of the total water volumes pumped from a well during the month when the sample was collected, using volume data recorded daily by water utilities to meet State reporting requirements;

(5)Simulated mean concentrations were input into the exposure–response relationship (eq 1) to derive risk(tapUV) and risk(wellpost-UV) for Approach 1 and risk(tapno-UV) and risk(wellno-UV) for Approach 2. The inputted simulated concentrations were limited to those that fell within the observed range of measured mean concentrations that were used to develop the exposure–response relationships; 0–3 gc/L for all-viruses and norovirus GI and 0–1.9 gc/L for enteroviruses.

(6)Distribution system risk was calculated as per eqs 3 and 4.

Steps 1–6 were repeated 100 000 times for each approach to obtain the frequency distributions of risk, expressed as AGI incidence, associated with drinking water distribution systems (Figure 2).
Attributable Risk Percent

The fraction of total AGI risk attributable to distribution systems was estimated by dividing the median risk(distribution system), obtained from either Approach 1 or 2, by the median risk(tapno-UV). Risk(tapno-UV) represents the total AGI risk from groundwater, distribution systems, and other nonwater-related exposures when the virus concentration in the exposure–response relationship is nonzero (eq 1, Figure 1b). Thus, median risk(distribution system)/median risk(tapno-UV) is equivalent to an incidence rate difference as a fraction of the total incidence under exposure.(31)
The fraction of drinking water AGI risk (source groundwater plus distribution systems) attributable only to distribution system contamination was estimated by dividing the median risk(distribution system), obtained from either Approach 1 or 2, by the median risk(drinking water). Risk(drinking water) was obtained at each iteration of the Monte Carlo algorithm as(5)where risk(other) represents AGI incidence corresponding to consuming tap water with no virus contamination, and was obtained by inputting a virus concentration of zero into the exposure–response relationships (eq 1 and Table 1). Risk(groundwater) is the difference between the median risk(drinking water) and the median risk(distribution system).

3 Results


Virus Occurrence

Viruses were detected in the groundwater and tap water of all study communities. During UV periods (Approach 1), UV disinfection reduced virus detection frequency and concentration prior to distribution; 10.1% of post-UV samples were positive for one or more virus types (all-viruses variable) (Table 2), and 95% of these positive samples had virus concentrations ≤1.1 gc/L (Figure 3a). In contrast, downstream from UV disinfection at household taps, 20.3% of samples were virus positive and 95% of positive samples had virus concentrations ≤8.0 gc/L (Figure 3b). This increase in virus detection frequency and concentration between UV disinfection and household taps indicates viruses were directly entering the distribution systems.(26) During no-UV periods (Approach 2), 32.3% of groundwater samples were virus-positive (Table 2), and 95% of positive samples had all-virus concentrations ≤23.6 gc/L (Figure 3c); a similar detection frequency, 27.5%, was observed downstream at household taps where 95% of positive samples had all-virus concentration ≤25.5 gc/L (Figure 3d).
figure

Figure 3. Frequency distribution of virus concentrations (all-viruses variable) in positive samples. UV periods (Approach 1) (a) well post-UV, 178 samples analyzed, 18 positive; (b) tap water, 532 samples analyzed, 108 positive. No-UV periods (Approach 2) (c) groundwater, 186 samples analyzed, 60 positive; (d) tap water, 556 samples analyzed, 153 positive. Histogram intervals are 0.4 genomic copies/L for plots (a) and (b), 5 genomic copies/L for plots (c) and (d).

Table 2. Virus Concentrations by Sampling Location During UV (Approach 1) and No-UV Periods (Approach 2)
virus group approach sampling location meana (gc/L) mediana (gc/L) SDVb (gc/L) maximum (gc/L) % positive samples
all-viruses 1 wellpost-UV 0.0340 0 0.17 1.8 10.1
tapUV 0.275 0 1.5 17 20.3
2 wellno-UV 3.05 0 21 220 32.3
tapno-UV 1.45 0 8.6 116 27.5
enterovirus 1 wellpost-UV 0.00690 0 0.088 1.2 1.1
tapUV 0.0950 0 0.84 13 6.4
2 wellno-UV 0.911 0 7.8 103 16.7
tapno-UV 0.180 0 1.5 31 12.6
norovirus GI 1 wellpost-UV 0 0 0 0 0
tapUV 0.111 0 1.0 16 2.1
2 wellno-UV 2.06 0 16 173 5.9
tapno-UV 1.18 0 8.3 116 5.6
a178 well samples and 532 tap samples were analyzed during UV periods (Approach 1); 186 well samples and 556 tap samples were analyzed during no-UV periods (Approach 2).
bSDV: standard deviation.
The virus types most commonly detected during the study were adenoviruses (13.5% and 12.8% of tap samples were adenovirus-positive in UV and no-UV periods, respectively) and enteroviruses, followed by norovirus GI (Table 2). Although adenoviruses were detected often, adenovirus concentrations were much lower than the other viruses and showed no association with AGI incidence,(25) and were therefore not included in the present analysis (except included with all other virus types in the “all-viruses” variable). Few samples were positive for rotavirus (1 positive sample at a tap during no-UV) or hepatitis A virus (1 positive sample in groundwater and 3 at taps during UV periods; 1 and 6 positive samples in groundwater and taps, respectively, during no-UV); therefore, exposure–response models were not developed for these viruses.(25) Adenoviruses and enteroviruses were detected throughout the study, while norovirus GI was detected predominantly in the first study year.
AGI Risk from Distribution Systems: Approach 1

AGI risk from distribution system contamination, for all exposure-response models considered, had median values greater than zero and ranged from 0.0180 to 0.0611 episodes/person-year (Table 3). For comparison, the epidemiologically measured AGI incidence for study participants of all ages over the four 12-week periods was 1.71 episodes/person-year.(25) Risk frequency distributions had high variance, as expected from the heterogeneity in virus occurrence and concentration. Fifty-five to 59% of the risk estimates exceeded the USEPA drinking water disease threshold of 1 infection per 10,000 individuals per year(32) (Table 3), assuming the USEPA threshold for infections is equivalent to an acceptable threshold for illnesses. This comparison is conservative because not all infections result in an illness. Note that if our median risk estimates had been at the USEPA acceptable level of 0.0001 episodes/person-year, by definition 50% of the risk frequency distribution would have exceeded the threshold.
Table 3. Summary Statistics for the Frequency Distributions of Risk(Distribution System).a
Approach 1a Approach 2a
exposure–response model median (AGI episodes/person-yr) standard deviation (AGI episodes/person-yr) percent of risk distribution >1:10 000 infection/yearb median (AGI episodes/person-yr) standard deviation (AGI episodes/person-yr) percent of risk distribution >1:10 000 infection/yearb
all-viruses, all-ages 0.0305 0.233 55.3 0.0315 0.282 54.7
all-viruses, adults 0.0661 0.304 58.9 0.0731 0.469 57.4
enterovirus, adults 0.0440 0.377 54.9 0.0011 0.499 50.1
norovirus GI, all-ages 0.0235 0.113 59.1 0.0559 0.313 63.5
norovirus GI, adults 0.0359 0.173 59.3 0.0846 0.516 63.6
norovirus GI, children ≤12 0.0180 0.094 58.0 0.0430 0.220 62.7
norovirus GI, children <5 0.0431 0.234 57.7 0.1047 0.577 62.7
aResults in the table refer to the intermediate outcome of three Monte Carlo simulations, each including 100 000 exposure events. Variations in the median risk outcome among the three repetitions were between 1.7% and 5.7% (Approach 1) and between 0.5 and 9.7% (Approach 2).
bPercent of the frequency distribution of risk(distribution system) that is greater than the USEPA acceptable level of risk in drinking water specified as less than 1 infection in 10 000 people per year,(32) assuming an infection is equivalent to an illness.
AGI Risk from Distribution Systems: Approach 2

Except for risk outcomes derived using the enterovirus-adults exposure–response model, Approach 2 yielded median AGI risks from distribution system contamination similar to those obtained using Approach 1: 0.0315 to 0.1047 episodes/person-year (Table 3). During no-UV periods, enterovirus concentrations in wells were very similar to concentrations in tap water, resulting in a distribution system risk much lower than for other viruses. The variance in risk outcomes was significantly higher using Approach 2, reflecting the greater variability in virus concentrations in source groundwater when no UV disinfection was applied. Similar to Approach 1, the majority of risk outcomes exceeded the USEPA drinking water disease threshold. In both approaches the risk frequency distributions were slightly positively skewed, resulting in mean values of AGI risk that were higher than the median values. Hence, medians were adopted as a more representative and conservative summary value.
Attributable Risk Percent

The fraction of total AGI risk attributed to distribution system contamination was between 1.1% and 3.8% (Approach 1) or between 0.1% and 4.9% (Approach 2), depending on the exposure–response model. Considering only AGI risk from drinking water consumption, the fraction contributed by distribution system contamination was between 27.5% and 61.8% (Approach 1) or between 1.6% and 67.8% (Approach 2). Results for Approach 1 are shown in Figure 4. The much lower end of the risk range for Approach 2 resulted from the enterovirus-adults exposure–response model for the reason mentioned earlier. Otherwise, Approaches 1 and 2 yielded similar attributable risk percent estimates.
figure

Figure 4. AGI attributable risk percent for distribution systems, where risk(distribution system) was estimated by Approach 1, in relation to (a) the total AGI risk (i.e., the combined risk from distribution systems, groundwater, and other transmission routes); and (b) the AGI risk from drinking water (i.e., the sum of risk contributed by distribution systems and source groundwater). Risk estimates were derived for the seven exposure–response models, using the medians of the risk frequency distributions. All = all-viruses, E = enterovirus, N = norovirus. Percentages refer to the distribution system risk in proportion to the entire bar.

4 Discussion


Results from this study corroborate previous evidence that contamination within drinking water distribution systems can be responsible for endemic AGI.(23, 24) Our estimates of the median AGI risk from distribution systems are consistent with a previous national estimate of 0.03 episodes/person-yr (95% confidence interval: 0.003–0.09 episodes/person-yr).(33) Assuming estimates from the present study can be generalized to the rest of the United States, pathogen intrusions into distribution systems could be responsible for a significant AGI burden. Like the study communities, most municipal water systems in the U.S. are in small communities that rely on groundwater.(7) As of 2006, 95,631 out of 147,330 (65%) U.S. PWS relying on groundwater did not disinfect, leaving roughly 20 million people potentially exposed to waterborne pathogens through either contaminated source water or distribution system intrusions.(7) Extrapolating our results to all nondisinfecting groundwater systems in the U.S. (based on lower and upper values of median risk for all ages from distribution systems, Table 3), AGI from distribution system intrusions would range from 470,000 to 1,100,000 episodes per year nationwide, which is 235–559 times greater than the USEPA acceptable risk threshold of 1 infection in 10,000 people per year. The actual national AGI burden from distribution systems is likely higher, as systems providing a disinfectant residual can still be vulnerable to pathogen intrusions.(8, 13)
AGI risk from contaminated groundwater plus distribution systems, risk(drinking water) (eq 5), as a fraction of total risk was in the range of 3.8% to 10.4%, depending on the exposure–response model. In a companion study, QMRA using the all viruses-all ages exposure–response model estimated the median fraction of AGI risk from nondisinfected drinking water was 7%;(25) this study, unlike the present study, did not exclude virus measurements in chlorinated samples and did not limit simulated virus concentrations to those that fell within the observed range used to develop the exposure–response relationships. Extrapolated to all nondisinfecting groundwater systems, these estimates would translate into 1.1–3.4 million episodes/yr. These values are consistent with nationwide estimates of AGI related to drinking water in all systems, ranging between 11 and 19.5 million episodes/yr, about 8.5% to 12% of AGI from all exposure routes.(14, 33, 34) In comparison, U.S. foodborne AGI is estimated to be 48 million episodes/yr.(35)
Risk estimates were the same order of magnitude across all seven considered exposure–response relationships, although differences existed between age groups and virus types. In particular, for the two all-viruses exposure–response relationships the adult age group was at higher risk than the overall population, while for the four norovirus models the age group with the highest risk was children <5 years old, followed by adults, all-ages, and last children ≤12 years old. The same trends were observed during both UV and no-UV periods, suggesting the choice of model had a greater bearing on the risk estimate than the Approach (1 or 2).
Virus occurrence in the distribution systems was intermittent and sporadic. No community had consistently clean or contaminated water. Positive samples were not visibly clustered in space or time. Concentration patterns observed over the course of the two years suggest that distribution system contamination events can range in magnitude and frequency between relatively more common low levels and rare high concentration spikes. This finding is consistent with other studies that have demonstrated intrusions into distribution systems can vary greatly in frequency, duration, and contaminant load.(36) For example, main breaks are rare events, but could introduce large volumes of contaminated soil or water into a distribution system over a short time period. Lambertini et al.(26) showed for the WAHTER Study communities that installing pipes into the distribution systems was significantly associated with virus contamination of tap water. Day-to-day events, such as negative pressure transients resulting from pump operation, could cause intrusions on a semicontinuous basis and can remain undetected for a long period of time.
The AGI risk frequency distributions exhibited long right-hand tails reaching very high AGI incidence values, demonstrating that infrequent but high-concentration virus spikes in a distribution system could result in considerably more AGI risk than that indicated by median estimates (Figure 2, inset). The right-hand tail is also affected by the shape of the exposure–response relationship. Because such relationships were modeled with exponential equations, applying them to concentration mean values beyond the range observed in the study communities could lead to unrealistically elevated risk estimates. To avoid such constructs, simulated concentration means were input into the exposure–response relationships only if they fell within the range of concentration means used to derive the relationships. This constraint resulted in a negligible number of Monte Carlo iterations being rejected in Approach 1 and 7–27% in Approach 2, which truncated a considerable portion of the frequency distribution. To evaluate this issue, risk estimates were also calculated extrapolating the exposure–response relationships beyond their observed concentration range, by inputting all simulated mean concentrations (SI Table S5). For Approach 1 median risk estimates were nearly identical to those calculated with the constraint in place. However, for Approach 2 median risk estimates increased 22% to 31% for six of the seven exposure–response relationships, and variances increased unrealistically by orders of magnitude. For the enterovirus-adults model, the median risk(distribution system) became negative because enterovirus concentrations were slightly higher in groundwater than in tap water. While a reasonable risk estimate is bracketed by the constrained and unconstrained estimates, the difference between the two highlights the importance of different assumptions about extrapolating exposure–response relationships.
Two strategies could be employed to control AGI associated with distribution system intrusions: (1) maintaining a disinfectant residual throughout the distribution system (secondary disinfection), and (2) ensuring that the distribution system is not subject to losses of physical or hydraulic integrity that could lead to intrusions. The Surface Water Treatment Rule mandates the maintenance of a residual disinfectant concentration of at least 0.2 mg/L at the distribution system inlet, and detectable levels throughout the system.(32) As a consequence, all surface water systems in the U.S. disinfect, although outbreaks from distribution system deficiencies still occur in disinfected systems.(14, 37) Groundwater systems are not required to disinfect. The recent USEPA Groundwater Rule relies on sanitary surveys and indicator monitoring to detect system deficiencies or contamination events, which can trigger corrective actions, including disinfection, changes in infrastructure or management, or further monitoring.(7) The USEPA Total Coliform Rule (TCR) is currently under revision, and new research on distribution system deficiencies has been considered in the revision process.(38) However, the proposed revised TCR still relies on infrequent indicator organism monitoring as the main tool to detect distribution system deficiencies and trigger a sanitary survey. Specifically, small systems, responsible for 94% of TCR violations,(39) could still monitor only on a quarterly or yearly basis. UV disinfection in the present study reduced virus contributions from groundwater, but, as expected, without residual disinfection viruses were still able to enter the distribution systems. Whether upgrades in distribution system infrastructure, such as replacing nonsealing joints and leaky pipes or preventing negative pressure events, could be sufficient in the absence of residual disinfection to reduce AGI risk from distribution system contamination below the USEPA acceptable threshold is an open question. As the U.S. water infrastructure ages, failures may become more frequent, and significant upgrades will be needed in the coming decades.(3) So far insufficient financial investments have been made to improve water infrastructure,(40) and small systems are particularly at risk for lack of funds and personnel. As most of the national water distribution infrastructure is reaching the end of its design life in the coming decades, the frequency and health impacts of distribution system deficiencies will likely worsen.
Supporting InformationAbbreviations and definitions; qPCR primers and probes; quality assurance parameters for the standard curves; model sensitivity analysis; risk results obtained using the entire data set, i.e. including samples collected during chlorination; and risk results obtained extrapolating the exposure–response relationships beyond the observed concentration ranges.This material is available free of charge via the Internet at http://pubs.acs.org.

The authors declare no competing financial interest.

Acknowledgment


The Wisconsin WAHTER Study (Water And Health Trial for Enteric Risks) was funded by USEPA STAR grant R831630. We thank the study communities for their participation and support. Technical assistance from Phillip Bertz, Carla Rottscheit, Sandy Strey, and Matt Volenec at the Marshfield Clinic Research Foundation is gratefully acknowledged. Halona Leung, UC-Davis, produced the visual abstract.

References


This article references 40 other publications.

  1. 1.

    He, X. Q.; Cheng, L.; Zhang, D. Y.; Li, W.; Xie, X. M.; Ma, M.; Wang, Z. J.First molecular detection of group A rotaviruses in drinking water sources in Beijing, China Bull. Environ. Contam. Toxicol. 2009, 83, 120– 124

  2. 2.

    Vivier, J. C.; Ehlers, M. M.; Grabow, W. O. K.Detection of enteroviruses in treated drinking water Water Res. 2004, 38, 2699– 2705

  3. 3.

    NRC(National Research Council). Drinking Water Distribution System Assessing and Reducing Risks; NAS: Washington, DC, 2006.
  4. 4.

    Gullick, R. W.; Lechevallier, M. W.; Svindland, R. C.; Friedman, M. J.Occurrence of transient low and negative pressures in distribution systems J. Am. Water Works Assoc. 2004, 96, 52– 66

    [CAS]
  5. 5.

    LeChevallier, M. W.; Gullick, R. W.; Karim, M. R.; Friedman, M.; Funk, J. E.The potential for health risk from intrusion of contaminants into distribution systems during pressure transients J. Water Health 2003, 1, 3– 14

    [PubMed], [CAS]
  6. 6.

    Karim, M. R.; Abbaszadegan, M.; LeChevallier, M.Potential for pathogen intrusion during pressure transients J. Am. Water Work Assoc. 2003, 95, 134– 146

    [CAS]
  7. 7.

    U.S. Environmental Protection Agency, National Primary Drinking Water Regulations. Ground Water Rule; Final Rule; Federal Register 71, no. 216, 2006; pp 65574–65660.
  8. 8.

    Payment, P.Poor efficacy of residual chlorine disinfectant in drinking water to inactivate waterborne pathogens in distribution systems Can. J. Microbiol. 1999, 45, 709– 715

  9. 9.

    Propato, M.; Uber, J. G.Vulnerability of water distribution systems to pathogen intrusion: How effective is a disinfectant residual? Environ. Sci. Technol. 2004, 38, 3713– 3722

  10. 10.

    Hunter, P. R. Waterborne Disease: Epidemiology and Ecology; Wiley: Chichester, UK, 1997.
  11. 11.

    Lahti, K.; Hiisvirta, L. In Causes of Waterborne Outbreaks in Community Water-Systems in Finland – 1980–1992; Pergamon-Elsevier Science Ltd, 1995; pp 33– 36.
  12. 12.

    Risebro, H. L.; Doria, M. F.; Andersson, Y.; Medema, G.; Osborn, K.; Schlosser, O.; Hunter, P. R.Fault tree analysis of the causes of waterborne outbreaks J. Water Health 2007, 5, 1– 18

  13. 13.

    Craun, G. E.; Calderon, R. L.Waterborne disease outbreaks caused by distribution system deficiencies J. Am. Water Works Assoc. 2001, 93, 64– 75

    [CAS]
  14. 14.

    Reynolds, K. A.; Mena, K. D.; Gerba, C. P.Risk of waterborne illness via drinking water in the United States Rev. Environ. Contam. Toxicol. 2008, 192, 117– 158

  15. 15.

    Yoder, J.; Roberts, V.; Craun, G. F.; Hill, V.; Hicks, L.; Alexander, N. T.; Radke, V.; Calderon, R. L.; Hlavsa, M. C.; Beach, M. J.; Roy, S. L.Surveillance for waterborne disease and outbreaks associated with drinking water and water not intended for drinking – United States, 2005–2006 MMWR Surveill. Summ. 2008, 57 ( No.SS-09) 39– 62

    [CAS]
  16. 16.

    Blackburn, B. G.; Craun, G. F.; Yoder, J. S.; Hill, V.; Calderon, R. L.; Chen, N.; Lee, S. H.; Levy, D. A.; Beach, M. J.Surveillance for waterborne-disease outbreaks associated with drinking water -United States, 2001–2002 MMWR Surveill Summ. 2004, 53 ( SS08) 23– 45

    [CAS]
  17. 17.

    Liang, J. L.; Dziuban, E. J.; Craun, G. F.; Hill, V.; Moore, M. R.; Gelting, R. J.; Calderon, R. L.; Beach, M. J.; Roy, S. L.Surveillance for waterborne disease and outbreaks associated with drinking water and water not intended for drinking-United States, 2003–2004 MMWR Surveill Summ. 2006, 55, 31– 65

    [PubMed], [CAS]
  18. 18.

    Brunkard, J. M.; Ailes, E.; Roberts, V. A.; Hill, V.; Hilborn, E. D.; Craun, G. F.; Rajasingham, A.; Kahler, A.; Garrison, L.; Hicks, L.; Carpenter, J.; Wade, T. J.; Beach, M. J.; Yoder, J. S.Surveillance for waterborne disease outbreaks associated with drinking water – United States, 2007–2008 MMWR Surveill. Summ. 2011, 60 ( No.SS-12) 38– 68

    [CAS]
  19. 19.

    Payment, P.; Siemiatycki, J.; Richardson, L.; Renaud, G.; Franco, E.; Prevost, M.A prospective epidemiological study of gastrointestinal health effects due to the consumption of drinking water Int. J. Environ. Health Res. 1997, 7, 5– 31

  20. 20.

    Payment, P.; Richardson, L.; Siemiatycki, J.; Dewar, R.; Edwardes, M.; Franco, E.A randomized trial to evaluate the risk of gastrointestinal-disease due to consumption of drinking-water meeting current microbiological standards Am. J. Public Health 1991, 81, 703– 708

  21. 21.

    Tinker, S. C.; Moe, C. L.; Klein, M.; Flanders, W. D.; Uber, J.; Amirtharajah, A.; Singer, P.; Tolbert, P. E.Drinking water residence time in distribution networks and emergency department visits for gastrointestinal illness in Metro Atlanta, Georgia J. Water Health 2009, 7, 332– 343

  22. 22.

    Nygård, K.; Andersson, Y.; Rottingen, J. A.; Svensson, A.; Lindback, J.; Kistemann, T.; Giesecke, J.Association between environmental risk factors and campylobacter infections in Sweden Epidemiol. Infect. 2004, 132, 317– 325

  23. 23.

    Nygård, K.; Wahl, E.; Krogh, T.; Tveit, O. A.; Bohleng, E.; Tverdal, A.; Aavitsland, P.Breaks and maintenance work in the water distribution systems and gastrointestinal illness: a cohort study Int. J. Epidemiol. 2007, 36, 873– 880

  24. 24.

    Hunter, Paul R.; Chalmers, Rachel M.; Hughes, S.; Syed, Q.Self-Reported Diarrhea in a Control Group: A Strong Association with Reporting of Low-Pressure Events in Tap Water Clin. Infect. Dis. 2005, 40, E32– E34

  25. 25.

    Borchardt, M. A.; Spencer, S. K.; Kieke, B. A. J.; Lambertini, E.; Loge, F. J.Viruses in non-disinfected drinking water from municipal wells and community incidence of acute gastrointestinal illness Environ. Health Perspect. 2012, http://dx.doi.org/10.1289/ehp.1104499
  26. 26.

    Lambertini, E.; Spencer, S. K.; Kieke, B. A.; Loge, F. J.; Borchardt, M. A.Virus contamination from operation and maintenance events in small drinking water distribution systems J. Water Health 2011, 9, 799– 812

    [CAS]
  27. 27.

    Lambertini, E.; Spencer, S. K.; Bertz, P. D.; Loge, F. J.; Kieke, B. A.; Borchardt, M. A.Concentration of enteroviruses, adenoviruses, and noroviruses from drinking water by use of glass wool filters Appl. Environ. Microbiol. 2008, 74, 2990– 2996

  28. 28.

    Simonet, J.; Gantzer, C.Inactivation of poliovirus 1 and F-specific RNA phages and degradation of their genomes by UV irradiation at 254 nm Appl. Environ. Microbiol. 2006, 72, 7671– 7677

  29. 29.

    Eischeid, A. C.; Meyer, J. N.; Linden, K. G.UV disinfection of adenoviruses: molecular indications of DNA damage efficiency Appl. Environ. Microbiol. 2009, 75, 23– 28

  30. 30.

    Süß, J.; Volz, S.; Obst, U.; Schwartz, T.Application of molecular biology concept for the detection of DNA damage and repair during UV disinfection Water Res. 2009, 43, 3705– 3716

  31. 31.

    Greenland, S; Rothman, K. J.; Lash, T. L.Measures of effect and measures of association. In Modern Epidemiology, 3rd ed.; Rothman, K. J.; Greenland, S; Lash, T. L., Eds.; Lippincott Williams & Wilkins: Philadelphia, 2008; pp 51– 70.
  32. 32.

    U.S. Environmental Protection Agency. National Primary Drinking Water Regulations; Filtration, Disinfection; Turbidity, Giardia lamblia, Viruses, Legionella, and Heterotrophic Bacteria; Final Rule. Federal Register, 54:124:27486, 1989.
  33. 33.

    Messner, M.; Shaw, S.; Regli, S.; Rotert, K.; Blank, V.; Soller, J.An approach for developing a national estimate of waterborne disease due to drinking water and a national estimate model application J. Water Health 2006, 4 ( suppl 2) 201– 240

  34. 34.

    Colford, J. M.; Roy, S.; Beach, M. J.; Hightower, A.; Shaw, S. E.; Wade, T. J.A review of household drinking water trials and an approach to the estimation of endemic waterborne gastroenteritis in the United States J. Water Health 2006, 4 ( Suppl 2) 71– 88

  35. 35.

    Scallan, E.; Griffin, P. M.; Angulo, F. J.; Tauxe, R. V.; Hoekstra, R. M.Foodborne Illness Acquired in the United States – Unspecified Agents Emerg. Infect. Dis. 2011, 17 ( 1) 16– 22

  36. 36.

    Westrell, T.; Bergstedt, O.; Stenstrom, T. A.; Ashbolt, N. J.A theoretical approach to assess microbial risks due to failures in drinking water systems Int. J. Environ. Health Res. 2003, 13, 181– 197

  37. 37.

    Craun, M. F.; Craun, G. F.; Calderon, R. L.; Beach, M. J.Waterborne outbreaks reported in the United States J. Water Health 2006, 4 ( Suppl 2) 19– 30

  38. 38.

    U.S. Environmental Protection Agency. Total Coliform Rule Revisions; 2010. http://water.epa.gov/lawsregs/rulesregs/sdwa/tcr/regulation_revisions.cfm.
  39. 39.

    U.S. Environmental Protection Agency. FACTOIDS: Drinking Water and Ground Water Statistics for 2009; EPA 816-K-09-004, 2009. http://www.epa.gov/ogwdw/databases/pdfs/data_factoids_2009.pdf.
  40. 40.

    ASCE (American Society of Civil Engineers). Report Card for America’s Infrastructure; 2009. www.infrastructurereportcard.org/sites/default/files/RC2009_full_report.pdf.

  41. Title: Risk of Viral Acute Gastrointestinal Illness from Nondisinfected Drinking Water Distribution Systems
    Author: Elisabetta Lambertini, Mark A. Borchardt, Burney A. Kieke, Jr., Susan K. Spencer, and Frank J. Loge
    Publication: Environmental Science & Technology
    Publisher: American Chemical Society
    Date: Sep 1, 2012
    Copyright © 2012, American Chemical Society
Want to Donate?
Please contact us for gifts in kind - Mail your check to: P.O. Box 545934, Surfside, Fl 33154