Introduction

The diversity of North America’s hydrology, and by extension, hydrogeology, is a reflection of its bio-physical diversity including extremely variable physiography, climate, bedrock geology and quaternary geology. Accordingly, many regional hydrological issues exist related to microbial groundwater contamination [1].

Canada is comprised of nine hydrogeological regions, ranging from extensive areas of permafrost in the north to Cordillera, the Western Canadian sedimentary basin and the St Lawrence platform in the south [2]. Approximately 8.9 million Canadians (30.3% of the population) currently rely on groundwater for domestic use; with an estimated 3.8 million Canadians served by private supplies [3]–[6]. Hydrogeological units within the United States (US) are similarly varied with approximately 69.5% of groundwater in the US derived from unconsolidated aquifers (alluvial sand and gravel, coastal plain, glacial outwash and fluvial/eolian) [7]. Approximately 44% of the U.S. population depends on groundwater for its drinking water supply [8]. Private wells constitute the largest proportion of water wells in the United States – more than 13.2 million households have their own well, supplying 45 million people [8]. Typically, groundwater supplies in Canada and the United States are categorized into one of three supply types; large-scale, disinfected municipal systems supplying cities and municipalities with >1000 residents, community systems of varying scale with or without disinfection and typically supplying <1000 people, and largely untreated private domestic wells, often serving single households.

Groundwater contamination occurs as a direct result of human or animal waste ingress to groundwater resources (aquifers or wells). Enteric pathogens are prevalent in both developing and developed countries, including the United States and Canada [9]–[10], where septic tanks [11]–[13], sewage and municipal wastewater treatment [14]–[16], wildlife [17], grazing animals [18] and other agricultural activities [19]–[22] can be sources of groundwater contamination. Natural attenuation occurs by overlying subsoil strata; however, age deterioration, poor wellhead hygiene or design defects provide localized pollution pathways [13], [23]. Bedrock fracturing and faulting, unconsolidated bedrock materials and thin or absent overlying subsoil layers may also provide pathways for contamination transport. Where subsurface contamination does occur, adverse human health effects may result, in part due to inadequate treatment, particularly with private supplies [24], [25].

Waterborne illness is a public health issue in North America and around the world. Sensitive sub-populations including the young, the elderly, pregnant women, and the immune-compromised are particularly susceptible to enteric infections [26]. An estimated 45% of all waterborne outbreaks in Canada involve non-municipal systems, largely in rural or remote areas [3], [4], [27]. While private supplies serve about 13% of Canada's population, >20% of Canada's reported waterborne disease outbreaks were associated with private water supplies between 1974 and 1996 [28]. Consumption from private wells has been cited as a potentially significant risk factor for contracting campylobacteriosis and other enteric diseases in Canada [29]–[31]. A recent study from New Zealand predicts that consumers sourcing their drinking water supply from shallow groundwater in an intensive dairy region (Waikakahi catchment, Canterbury) had an estimated 60–75% risk of Campylobacter infection per irrigation season [22]. Consumption of contaminated groundwater is estimated to account for approximately 6.5 million illnesses per year in the US (34% of all waterborne infections) [32]. A recent review by Wallender et al. [16] found that 30.3% of 818 drinking water outbreaks reported to the United States CDC Waterborne Disease and Outbreak Surveillance Systems during 1971 to 2008 were attributable to untreated groundwater sources. Exposure to shallow groundwater sources contaminated with animal faeces has been identified as a risk factor for Escherichia coli O157:H7 infection in the US [33].

In their summation of a 2005 workshop entitled ‘Estimating Waterborne Disease Risks in the United States’ Craun et al. [34] state that significant data and method-based uncertainties remain in the improved estimation of enteric diseases attributed to drinking water, particularly with respect to risks associated with wells and groundwater. There is currently a paucity of integrated data pertaining to the presence and source of enteric pathogens in North American groundwater systems. Accordingly, the purpose of this work was to identify and review available literature relating to contamination of Canadian and US groundwater sources by enteric pathogens between 1990 and 2013. The overarching objective of this review was a comprehensive collation of the current state of knowledge relating to the presence of enteric pathogens in Canadian and US groundwater and an examination of related patterns (e.g. use of FIOs, enteric pathogen detection, study location, source types, etc.). To date, similar work has not been undertaken; it is considered that the results could be used by stakeholders to prioritize research and fill existing data gaps, thus supporting the effective management of groundwater supplies in Canada and the US.

Methods

2.1 Literature Identification and Screening

The research question that guided the review was:

What has been the prevalence of contamination by enteric pathogens and fecal indicator organisms in Canadian and US groundwater systems during the period 1990-present?

A review protocol was developed in order to capture existing, published references to contamination events in Canadian and US groundwater systems (Figure 1). The review protocol was adapted from previous studies [35]–[37]. Literature for inclusion was identified from four bibliographic databases (PubMed, MedLine Plus, Science Direct, and Scopus) with key search terms (Table 1). Overall, 21,300 article titles that included ≥2 key search terms were located within bibliographic databases; article titles were imported to an Excel spreadsheet and de-duplicated using a replacement macro, resulting in 3400 article titles (Figure 1a). All literature scans were undertaken using Boolean positional operators (“AND”, “OR”, “SAME”, “WITH”, “ADJ”) to focus literature searches. Study screening 1 (citation title) was subsequently undertaken; all articles published prior to 1990, relating to groundwater chemistry, non-enteric pathogens and/or groundwater systems outside of Canada and the US were excluded. This resulted in the inclusion of 1028 non-duplicate articles (Figure 1b).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Review protocol employed throughout the current study, including results of literature identification, literature scans and data extraction processes.

https://doi.org/10.1371/journal.pone.0093301.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Review primary terms, term descriptions and search terms used for literature identification, study screening and final database development.

https://doi.org/10.1371/journal.pone.0093301.t001

Study screening 2 (citation abstract) was based upon explicit inclusion/exclusion criteria (Table 2). No study types were excluded, however, at this stage all non-English articles were excluded, as were articles which examined surface, recreational or coastal waterbodies. Articles reporting groundwater contamination by non-enteric pathogens were also excluded during this screening phase. This screening phase was undertaken by one researcher, with 270 article included for further consideration at this stage (Figure 1c).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Inclusion and exclusion criteria for literature screening.

https://doi.org/10.1371/journal.pone.0093301.t002

Study screening 3 was completed by two review teams who independently assessed each publication (year, title, and abstract), based upon explicit eligibility criteria (Table 2). Articles that met all initial screening criteria, as defined by agreement between reviewers were included. Disagreements (approximately 15%) were resolved by discussion among reviewers and an additional round of screening. As shown (Figure 1d), 165 articles were included in the final database. Upon data extraction, 52 of 165 included articles were found to contain extractable homogenous data, with an additional three relevant articles located via article bibliographies added at this stage (Figure 1e).

Results

3.1 Descriptive Analysis of Literature

A total of 55 published studies included extractable data, of which 23.6% (n = 13) were Canadian studies, while the remaining 76.4% (n = 42) were from the United States (Appendix S1). Included studies spanned the review time period, with a peak (n = 7) noted in 2002. Overall, 45.4% of studies (n = 25) contained both fecal indicator organism (FIO) and enteric pathogen detection data, while a further 29.1% (n = 16) contained enteric pathogen data only. The remaining 25.4% of studies (n = 14) solely reported on FIO detection. Of those studies that included extractable data pertaining to enteric pathogen detection (n = 41), 80.5% (n = 33) were from the US, while the remaining studies (n = 8) were from Canada. From these 41 studies, 102 extractable records were found (Tables 3 & 4); most were derived from US studies (90/102). All extracted enteric pathogen records were included for pooled-analyses.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Enteric pathogen records from included US and Canadian studies (1990–2013) (n = 102); record number, percentage of total records, record number with specified pathogen present, percentage of records with specified pathogen present.

https://doi.org/10.1371/journal.pone.0093301.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Summary statistics: a. Extracted Studies (n = 55), b. Extracted Records (n = 102), c. Extracted records associated with confirmed enteric pathogen presence (n = 74), d. Sampled well numbers (n = 52) and analysed groundwater sample numbers (n = 39).

https://doi.org/10.1371/journal.pone.0093301.t004

3.2 Fecal Indicator Organisms

In total, 39 studies reported FIO data related to six organisms. Based on the definition of FIO as “non-pathogenic, exclusively intestinal in habit and therefore exclusively fecal in origin if found outside the intestine” [39], total coliforms are not considered a FIO, however their use as indicator organisms remains widespread and thus are included here. A median of 2 FIOs were reported per study, up to a maximum of 5. E. coli was the FIO analysed most frequently, while heterotrophic plate counts were employed least often (Table 4a). Two-tailed Spearman’s’ rho was used to examine correlations between individual FIOs and the presence of enteric pathogens, in studies where both data were reported (Table 5). The presence of total coliforms, E. coli, intestinal Enterococci and coliphage (Male-specific (F+) and Somatic) were all strongly correlated with each other (p<0.01). No significant correlation between the total percentage of sample wells with enteric pathogens present (regardless of pathogen type) and the percentage of wells with fecal indicators present (individually or combined) was found. Weak (−0.7< r <0.7) negative correlations were found between pathogen contamination percentage (% of sampled wells with enteric pathogens present, irrespective of pathogen category) and total coliforms (% of sampled wells with total coliforms present) (r = −0.289, p = 0.034) and intestinal enterococci (r = −0.435, p = 0.006). A weak positive correlation between bacterial pathogen presence and E. coli (r = 0.636, p = 0.02). Both protozoan (r = 0.866, p = 0.026) and viral pathogen presence (r = −0.494, p = 0.023) were correlated with the presence of intestinal enterococci.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Correlation matrix; fecal indictor organisms and total source contamination percentage (n = 25) (Clostridium perfringens and Heterotropic plate counts were excluded due to low record numbers).

https://doi.org/10.1371/journal.pone.0093301.t005

Discussion

A review of groundwater pathogen contamination studies in Canada and the US was conducted, locating 102 enteric pathogen records from 55 studies. The majority of studies were from the United States. While almost 24% of studies were Canadian, only 12% of enteric pathogen records were from Canadian studies. Typically, US studies were more likely to include “multi-pathogen” data (>1 enteric pathogen per study) and therefore >1 record, while Canadian studies were more likely to include FIO data with no accompanying enteric pathogen data. There is a paucity of studies focusing on direct pathogen detection in groundwater systems in both countries. Given concerns regarding the use of FIOs as surrogate organisms [40]–[42], the suitability of FIOs for use in the regulatory standards for protection of water quality is a particular concern. Overall, 38.5% of Canadian studies contained FIO data in the absence of enteric pathogen data, potentially resulting in significant under-estimates of enteric pathogen presence and/or magnitude. Future work should focus on direct detection of enteric pathogens; this data may be used to improve microbial risk assessment efforts via significantly improved exposure assessment.

Just over 42% of extracted records in this review were for private domestic wells. Currently 34–45% of groundwater users in Canada and approximately 33% of groundwater users in the US employ a private water source [5], [8]. There is a lack of enteric pathogen data from municipal supplies, reflecting a potentially critical knowledge gap. The inclusion of municipal groundwater supplies in future North American groundwater sampling studies would help inform this identified gap.

In Canada, the focus within scientific studies in the reviewed literature has been on protozoan detection in groundwater, while in the US, it has been on bacterial and viral detection, with viruses being the most commonly studied pathogen in this review. Pathogenic viruses are a particular concern in groundwater systems as their size enables them to evade natural attenuation processes and enter the aquifer [43], [44].

No correlation was found between viral presence and either total coliforms or E. coli, the two most frequently employed FIOs for groundwater system surveillance. Moreover no association between coliphage detection and viral presence was observed in this review. A 1-year study of 12 municipal wells in Quebec, Canada reported similar results [45]. As previously noted by Borchardt et al. [44], the majority of groundwater-associated municipal systems in the US produce water without disinfection, populations served by these systems are therefore potentially exposed to waterborne viruses and consequent health risks. The limited value of FIOs in predicting the presence of enteric viruses in groundwater highlights the need for direct enteric viral detection at groundwater sources used for drinking water [46]–[48]. The lack of significance between viral presence and source design, in concurrence with an association between viral presence and precipitation patterns, suggests a generalised mode of contamination associated with viral pathogens [13].

Geology was rarely recorded or presented as a potential contaminant pathway in Canadian studies, while 60% of US studies stated that local geological conditions were a potential cause for concern. Awareness of local geological and hydrogeological characteristics is critical in understanding likely contamination mechanisms or pathways, resource management and remediation and, by extension, human health risk assessment [49]–[50]. Future research relating to microbial contamination of North American groundwater systems, that adequately describes geological and hydrogeological characteristics of the groundwater source will be more useful for informing our understanding of public health risk. Future work related to the assessment of treatment requirements in geologically vulnerable regions based on the likelihood of pathogen presence would be particularly useful.

Bacterial pathogens accounted for 22.6% of records; all studies identified poorly designed or constructed well sources as a likely pathway. In contrast to the viral pathogen findings, this suggests localized, source-specific contamination as a predominant ingress mechanism e.g. groundwater contamination occurring at source due to poor wellhead protection as opposed to generalised contamination of all or part of the aquifer [13], [51]. This is further reflected by the lack of association between bacterial presence and local hydrogeological characteristics.

The increased likelihood of potential zoonotic pathogen presence in agricultural areas was anticipated; low travel times in the subsurface make it unlikely to find these pathogens in areas without animals (livestock or wildlife) [18], [22], [52]. Our review also found bacterial pathogen presence more likely than either viral or protozoan presence in rural areas with livestock present. Potentially zoonotic pathogens were also more likely to be present where the sampled well was located in a surface water catchment, suggesting the potential for direct (and potentially rapid) groundwater-surface water interaction as a contamination mechanism. Potentially zoonotic pathogens were significantly less likely to be present in urban areas. Studies outside North America have previously reported pathogens as more likely present in shallow hand-dug and spring sources; for example, Hynds et al. [13] found a higher prevalence of pathogen presence in shallow groundwater sources in the Republic of Ireland. Hand-dug wells and springs are typically associated with inferior design and construction, leading to an increased likelihood of direct pathogen ingress at the wellhead [51], [53]–[55]. It is important to note that both study design and reporting are likely to introduce bias in these results, for example, studies tended to focus on groundwater systems in rural/agricultural areas, while few examined urban systems. Accordingly, interpretation of these results may not adequately reflect hazard sources or contamination pathways, being are based solely upon available data.

Much of North America, particularly the United States, is characterised by unconsolidated aquifers [7]. Accordingly, unprotected shallow groundwater systems are regularly used, particularly for domestic supply. Wallender et al. [16] have reported that 26.2% of waterborne outbreaks from 1971 to 2008 were associated with karst limestone bedrock. Consumers and water managers in these areas would benefit from a clear understanding of the potential contamination risks, and consequent human health burden posed by using untreated shallow systems, and intervention options that are most effective. These findings may be useful for informing future groundwater surveillance/monitoring study design with respect to the most appropriate pathogen (category/species) to include for analysis, and where to focus the monitoring efforts. Moreover, where a human enteric disease outbreak is suspected, these criteria may be used to elucidate the potential infection etiology.

Based on this review, we note an inconsistency between studied groundwater pathogen (category) and reported groundwater-related human health burden in North America. A knowledge gap exists with respect to information on both prevalence and concentration of pathogen contamination of groundwater systems, particularly enteric bacteria. For example, it has been reported that 11% of waterborne outbreaks were caused by Escherichia coli O157 (VTEC) as a confirmed etiological agent [56]; however, only 4.9% of pathogen records in this review relate to VTEC (Table 3).

Little efficacy was found with respect to the predictive ability of FIOs and protozoan or viral contamination of groundwater systems. The lack of association between protozoan presence and local hydrogeological characteristics (as for bacterial pathogens) reinforces the importance of rapid by-pass mechanisms [13], with poorly designed wells as much as three times more likely to have protozoan pathogens present. Consequently, correctly implemented source management, including appropriate well location, design, construction and maintenance, in concurrence with periodic direct pathogen analysis is of paramount importance. This is particularly significant in light of the resistance of Cryptosporidium spp. to conventional chlorination [57]. It is considered that an increase in direct pathogen data analysis and reporting would prove useful for future risk management and communication strategies to inform intervention decisions and source water protection initiatives. For example, Morris & Foster [58] have noted that protozoan pathogens are commonly believed to be absent from “true” groundwater due to their size; however, with better data on groundwater contamination this assumption could be evaluated to further inform risk identification and mitigation strategies.

Results and conclusions presented here are based on the extracted data from published studies and are subject to the inherent biases of the original study objectives, designs and what was reported. Few Canadian studies referenced local hydrogeological characteristics as a potential contamination pathway when reporting on contamination events or groundwater-associated outbreaks, representing a limitation in study reporting. This lack of reporting may impact subsequent analyses of contamination pathways, for example, findings may reflect a lack of reporting or examination of these factors in a rigorous manner. Furthermore, only one included study occurred in winter, representing a potential limitation.

Study design differed significantly between US and Canadian studies. The majority of lab-based studies and outbreak investigations were in the US, while a higher proportion of repeat sampling records were from Canada. Repeat sampling studies integrated low well numbers with multiple samples from each well (mean – 6 samples). Outbreak investigations and “snapshot” studies comprised significantly higher well numbers, with wells typically sampled once. Outbreak investigations are an essential epidemiological tool; however, they are reactive and therefore not suitable for assessing endemic groundwater risks. Numerous studies report that multiple samples are necessary to positively determine contamination of a groundwater source [59]–[60]. Accordingly, repeat sampling studies are more likely to reflect contamination rates in Canadian and US groundwater systems; however few were identified in this review.

Enteric pathogens were present in 72.5% of records (74/102) and 16.1% of wells. Upon exclusion of outbreak investigation data, 77.1% of records (54/70) and 14.7% of wells (n = 1855/12,616; Std. Dev. 20.5%) were positive for at least one enteric pathogen. These figures may provide a useful foundation for endemic groundwater-associated exposure assessment, until more direct enteric pathogen data become available.

The lack of association between the incidence of outbreak investigations and municipal and private groundwater systems was not anticipated; however, as the majority of groundwater sampling was undertaken pre-treatment, this does not accurately reflect the public health burden posed by specific system types (e.g. many private sources remain untreated or poorly treated, thus raw water samples may be more reflective of drinking water quality). Large municipal systems are more likely to be adequately maintained and appropriately treated [24]–[25]. Smaller community systems were represented by a significantly higher proportion of outbreak investigations. A recent review of waterborne outbreaks by Hrudey [61] found that all, with the exception of one (Milwaukee, 1993), occurred in community systems. Furthermore, Craun et al. [62] have reported on ten cryptosporidiosis outbreaks in the US directly attributable to community water systems. Accordingly, the characteristics of a robust groundwater supply should include an understanding of the source, delivery of adequate treatment, adequate system monitoring and effective response to monitoring results when required, particularly in the case of private and community systems where the primary responsibility for source quality lies with owners and users [63].

Our review has highlighted a number of knowledge gaps and requirements relating to contamination of Canadian and US groundwater systems over the past two decades. Improved endemic risk assessment will be possible with increased direct enteric pathogen detection and reporting. Repeat-sampling studies represent a priority for the design of future studies for the collation of these data, as they are considered more representative with respect to human endemic risk. More data are required for groundwater-associated municipal systems, both pre- and post-treatment, as currently these systems are not adequately represented in the literature. Additionally, more data are required for bacterial pathogens, as these microorganisms are associated with a large public health burden. These data would also provide a foundation for improved risk management and communication via increased awareness and understanding among well owners, well users, system operators and policy makers.

Conclusions

• Within the literature the four study designs that were identified were of a proactive (pre-waterborne illness event) or reactive (post-waterborne illness event) nature, with the proactive designs being more suitable for background characterisation. Pooled analyses of reported data found that enteric pathogens were present in approximately 15% of sampled wells; this may be useful for future microbial risk assessments of human health relating to Canadian and US groundwater systems, although it should only be employed in the absence of area-specific data, and then, only with caution due to inherent limitations associated with the current study (e.g. reporting biases associated with amalgamation of differing study designs, lack of data relating to specific hydrogeological settings, lack of data associated with winter sampling and weather events, etc.). Accordingly, the authors recommend future research be focused on addressing this current lack of temporal and spatial data.
• It is critically important to account for study design when presenting and interpreting results of a groundwater study, given that groundwater is not similar to surface water and many confounding factors will influence the presence of pathogens in one system versus another, as well as the occurrence of an associated outbreak.
• Few studies have been published on pathogen presence in municipal groundwater systems. By addressing this knowledge gap, evidence would be developed to help support local decision-making around appropriate treatment or other interventions to further promote the safety of the drinking water supply. It is recommended that future health-related studies, particularly those associated with sporadic waterborne infection and quantitative health risk assessments, focus on direct enteric pathogen detection where possible, given the limited efficacy of fecal indicator organisms for enteric pathogen presence in groundwater systems. These data will aid more focused, pathogen-specific exposure assessment and hazard characterisation, therefore providing more data for evidence-based groundwater and human-health protection measures.
• Local (hydro)geology is an important risk factor in the likelihood of pathogen presence in a groundwater supply and was not reported in a substantial proportion of study records included in this review. The availability and reporting of these data could be used in combination with land-use, source type and seasonality to more accurately inform outbreak investigators, epidemiologists and sanitary engineers.
• Bacterial and protozoan contamination appear to be associated with localised contamination mechanisms (well-specific), lending credence to the concept of “consumer control” (education of private well users may directly affect contamination rates and health burden). Viral contamination may be more generalised (aquifer scale) but intervention at the source could be targeted to address all enteric pathogens (e.g. ultraviolet disinfection).
• Our findings further reflect the importance of frequent inspection and maintenance of both private and municipal groundwater supplies in Canada and the US. Additionally, it is important that potential system vulnerabilities including climate, local hydrogeological setting and adjacent hazard sources are considered during study design.

Supporting Information