OUP user menu

Detection of Cryptosporidium, Giardia and Enterocytozoon bieneusi in surface water, including recreational areas: a one-year prospective study

Stephane Coupe, Karine Delabre, Regis Pouillot, Stephanie Houdart, Maud Santillana-Hayat, Francis Derouin
DOI: http://dx.doi.org/10.1111/j.1574-695X.2006.00098.x 351-359 First published online: 1 August 2006


Accidental ingestion of natural waters while bathing carries a risk of infection by waterborne protozoa such as Cryptosporidium, Giardia and, possibly, microsporidia. In order to evaluate this risk, we conducted a one-year prospective study of two recreational lakes and three river sites located near Paris, where bathing and boating are frequent. Twenty-litre water samples were collected monthly from each site. Concentrated samples were submitted to immunomagnetic separation followed by immunofluorescence (IMS-IF) for Cryptosporidium and Giardia detection. PCR and PCR restriction fragment length polymorphism (PCR-RFLP) were used for the genetic characterization of Cryptosporidium species on IMS-IF-positive samples. PCR were systematically performed to detect Enterocytozoon bieneusi. Bacteria counts were also determined. IMS-IF revealed low counts of Giardia cysts and Cryptosporidium oocysts in the recreational lakes, with occasional peaks (max. 165 cysts/10 L and 9 oocysts/10 L). By contrast, the river sites were consistently and sometimes heavily contaminated throughout the year. Enterocytozoon bieneusi was found in only two river samples. PCR-RFLP genotyping showed the presence of C. hominis and C. parvum. No correlation was found between the presence or counts of parasites and bacteria, except between the presence of Giardia and high counts of Escherichia coli and enterococci. Based on a previously developed model for quantitative risk assessment of waterborne parasitic infections, we estimated that the mean risk of infection by Cryptosporidium and Giardia associated with swimming was <10−4 in the recreational lakes, and frequently higher at the river sites.

  • Cryptosporidium
  • Giardia
  • Enterocytozoon bieneusi
  • surface water
  • IMS-IF
  • quantitative risk assessment (QRA)


Waterborne outbreaks of parasitic diseases occur worldwide, and often involve Cryptosporidium and Giardia. Most cases have been linked to contamination of water resources and to deficient water-treatment procedures or water-distribution networks (Rose et al., 1991; Mac Kenzie, 1994; Dalle et al., 2003; Fayer, 2004). Other protozoa, such as Isospora belli, Cyclospora and Enterocytozoon bieneusi microsporidia, have been also implicated in waterborne outbreaks, although the presence of these parasites in water has rarely been documented (Sparfel et al., 1997; Dowd et al., 1998, 2003; Cotte et al., 1999; Miegeville et al., 2003). The frequent contamination of surface water by Cryptosporidium and Giardia is well established (Rose et al., 1991; Kramer et al., 1998; Fayer, 2004; Mathieu, 2004; Yoder et al., 2004).

Several limited outbreaks of cryptosporidiosis and giardiasis linked to recreational waters have been reported in the USA (Furtado et al., 1998; Kramer et al., 1998; Mathieu, 2004; Yoder et al., 2004), and swimming in fresh water has been identified as a significant risk factor for cryptosporidiosis and giardiasis (Stuart et al., 2003; Roy, 2004). In Europe, the quality of bathing water is governed by Directive 76/160/EEC (Council of the European Communities, 1976), which specifies two types of threshold values for microbiological and physicochemical parameters, corresponding to two bathing qualities (‘good’, and ‘average’). As parasites are not covered by this directive, they are not routinely monitored.

Therefore, we initiated a one-year prospective study of the contamination of five natural water sites (including two recreational areas), located in the Paris region. We used the French standardized procedure for identification and quantification of Cryptosporidium oocysts and Giardia cysts in water (Anonymous, 2001), and Cryptosporidium-positive samples were submitted to PCR-restriction fragment length polymorphism (PCR-RFLP) analysis for specific identification. Detection of microsporidia was systematically performed. Analysis of bacterial contamination was performed in order to evaluate potential links between bacterial and parasitic contaminations on the studied sites. From our results, we can estimate, based on quantitative risk assessment (QRA) models, the quantitative risk of infection during bathing for both Cryptosporidium and Giardia parasites.

Materials and methods

Site selection and sampling

Five sites of natural surface water were selected in the east of Paris, based on their use for swimming or nautical activities: Jablines (JAB) and Champs sur Marne (CSM) are public recreational lakes, supplied by alluvial water, with facilities for bathing and yachting. These two sites are regularly monitored for microbial contamination by the Departmental Hygiene Laboratory. The three other sites — Melun (MEL), Annet sur Marne (ASM), and Meaux (MEA) — are located along the river Marne; bathing and yachting are common, although bathing is normally forbidden in Melun.

From November 2002 to October 2003, each site was sampled on a monthly basis. One 20 L sample was collected at approximately 0.5 m depth in a sterile plastic tank for parasitological analyses, and an additional litre was collected in a sterile flask for bacterial culture. Samples were transported to the laboratory at 4°C within 6 h.

Sample filtration and concentration

The 20 L samples were treated according to the AFNOR NF T 90–455 procedure, which is officially recommended in France for the detection of Giardia and Cryptosporidium in water (Anonymous, 2001). Briefly, water was filtered at a flow rate of 2 L min−1 through an Envirochek® cartridge (1 µm pore size) (VWR, Val de Fontenay, France). The retained material was eluted from the cartridge and then centrifuged. The resulting pellet was resuspended in 300 mL of distilled water, and then separated into two 150 mL samples. One 150 mL sample was used for immunomagnetic separation and specific immunofluorescence (IMS-IF) detection of Cryptosporidium oocysts and Giardia cysts (see below). The other 150 mL sample was submitted to 4 successive filtration steps through filters of pore sizes 180 µm, 60 µm, 11 µm and then 0.4 µm (Millipore, Saint Quentin en Yvelines, France), according to Yoder (2004). Each filtration step was performed using a peristaltic pump (Millipore) at a flow rate of 20 mL min−1. Prior to sample processing, molecular-grade water was first passed through a 0.4 µm filter, for use as a negative control and to check for contamination of the glass bottles, plastic tubing and filter chambers used for sample processing. Each filter was kept at 4°C in a 2 mL tube until the end of the sampling campaign, and was further used for PCR detection of E. bieneusi, and Cryptosporidium in case of IMS-IF-positive results.

Detection of Cryptosporidium oocysts and Giardia cysts by IMS-IF

Immunocapture and immunofluorescence labelling of cysts and oocysts was performed according to the AFNOR NF T 90-455, procedure (Anonymous, 2001). Briefly, parasites were captured using antibody-coated magnetic beads (Chemunex, Ivry sur Seine, France). After elution from the beads, parasites were labelled with anti-Giardia and anti-Cryptosporidium monoclonal antibodies. After washes with phosphate-buffered saline (PBS) and then centrifugation, the whole pellet was examined in an automatic fluorimeter (Chem Scan, Cheminex, Ivry sur Seine, France) (De Roubin et al., 2002), allowing Cryptosporidium oocysts and Giardia cysts to be identified and counted. Results were expressed as oocyst and cyst counts per 10 L.

Bacterial analysis

For each site and sampling date, Escherichia coli and enterococci were detected and counted in the 1 L water sample by using, respectively, the NF EN ISO 9308-3 method (Anonymous, 1999a) and the NF EN ISO 7899-1 method (Anonymous, 1999b). Briefly, dilutions of the sample were inoculated on microplates containing appropriate culture media for E. coli and enterococci. Bacteria were counted using a most probable number (MPN) format.

Results were expressed as CFU 100 mL−1, and then classified into two groups according to the thresholds defining ‘good quality’ bathing water in Europe (Directive 76/160/EEC). Samples with E. coli counts >100 CFU 100 mL−1 or enterococcal counts >100 CFU 100 mL−1 were classified as ‘non conform’ (NC); samples with lower densities of both bacteria were classified as ‘conform’ (C).

DNA recovery

Each filter was overlaid with 300 µL of lysis buffer containing 2 N NaOH (VWR), 0.5% sodium dodecyl sulfate (SDS) and 0.5% polyvinylpyrrolidone 360 K (PVP) (Sigma, Saint Quentin Fallavier, France), and incubated in a hybridization oven at 70°C for 15 min. Then, 600 µL of guanidine buffer containing 2.5 M guanidine-HCl, 50 mM Tris-HCl, 20 mM EDTA, 2 M phosphate buffer (Sigma) and 37% absolute ethanol (VWR) was added.

The recovered sample volume (approximately 700 µL) was deposited on a silica column (Roche, Meylan, France) and centrifuged for 1 min at 16 000 g. The extraction procedure was followed by the addition of 600 µL of ‘inhibitor removal buffer’, as recommended by the manufacturer, except that the volume of the washing buffer was increased to 650 µL. DNA was eluted in 60 µL of prewarmed elution buffer and kept at –20°C until PCR experiments.

The presence of PCR inhibitors was tested on 10 µL of each eluted DNA sample, by PCR amplification of a limiting dilution of plasmids containing a fungal DNA fragment. M13 Fwd and M13 Rev universal primers were used. Amplicons were visualized on 2% agarose gel stained with ethidium bromide. DNA samples that contained inhibitors were incubated with 50 µL of 4% PVP for 20 min at room temperature and then submitted to a second column purification step as described above. The DNA elution volume was 50 µL.

PCR analysis and genotyping

All amplifications were performed with a Gene TC Digit Thermal Cycler (Techne, Fischer-Bioblock Scientific, France). PCR amplifications were performed in a 50 µL volume containing 10 µL of DNA template, 75 mM Tris-HCl, 20 mM (NH4)2SO4, 0.01% Tween 20, 0.6% BSA, 2.5 mM MgCl2, 0.15 mM each deoxynucleotide triphosphate (Amersham, Orsay, France), 0.2 µM each primer, and 1.25 U of Taq DNA polymerase (Eurogentec, Seraing, Belgium).

For E. bieneusi, we used a PCR targetting a 210-bp fragment of the internal transcribed spacer (ITS) region, as previously described (Liguory et al., 1997). For detection and identification of Cryptosporidium species, a nested PCR followed by RFLP was used (Council of the European Communities, 1976; Coupe et al., 2005): PCR amplified a 216-bp fragment of the 18S rDNA gene, which was visualized under UV light after 2% agarose gel electrophoresis and ethidium bromide staining; aliquots of amplified fragments were submitted to sequential enzymatic digestions for specific identification, as previously described (Coupe et al., 2005).

Assessment of PCR sensitivities after filtration processing

PCR/filtration sensitivities were assessed by serial dilutions of spores and oocysts in 500 mL of molecular-grade water (from 1 to 1000 per 500 mL), subjected to filtrations and subsequent DNA purification and PCR as described above. This allowed an estimate of the sensitivity of the procedure at 50 spores and 5 oocysts in 500 mL of spiked water.

We also checked that positive amplifications were solely the result of the presence of parasites, as filtration of water containing specific DNA did not result in positive PCR amplification.

Statistical analysis

Associations between the presence of Cryptosporidium, Giardia and bacteria (E. coli and enterococci) were analysed using the Chi-square test or Fisher's exact test (depending on the sample size) in a 2 × 2 design, using the results from the five sites. A relationship between variables was considered significant when the P-value was below 0.05.

Cryptosporidium and Giardia risk assessment

QRA models were used to estimate the risk of Cryptosporidium infection and illness per bathe and the risk of Giardia infection per bathe, according to a given IMS-IF value. Each model was composed of several modules: emission (based on observed oocyst or cyst counts in water, and taking into account the recovery rate of the counting methods), consumption (based on the volume of ingested water per bathe), exposure (as a function of the emission and consumption module outputs, and taking into account the proportion of viable oocysts and cysts) and effects (the dose–response relationships).

Models were adapted directly from a previous model developed by our team for QRA of waterborne cryptosporidiosis in tap water (Sparfel et al., 1997). Readers should refer to this paper for technical details and underlying assumptions.

The variability of the Cryptosporidium and Giardia recovery rates in the IMS-IF methods was estimated from personal data, using 11 duplicated assays of natural or potable water spiked with various amounts of Cryptosporidium oocysts (range: 78–331 oocysts) or Giardia cysts (range: 82–370 cysts). Beta-binomial adjustments of these data using the maximum-likelihood method led to a beta distribution (17.0, 25.8) for the Cryptosporidium recovery rate, and a beta distribution (39.6, 63.5) for the Giardia recovery rate (data not shown).

For both models, specific parameters had to be defined to estimate accidental ingestion of water during bathing. Parameters were adapted from the results of Dang (1996) on the length of exposure per bathe among non-professional swimmers in swimming pools and the amount of water ingested per hour of bathing. We assumed that bathes in recreational lakes lasted longer than bathes in swimming pools, and that the volume of ingested water per hour was more variable than suggested by Dang. The duration of exposure per bathe was finally modelled using a Pert distribution (Vose, 1998), with a minimum duration of 6 min, a most probable duration of 0.5 h, and a maximum duration of 2 h. Similarly, the amount of water ingested per hour of bathing was modelled through a Pert distribution, with a minimum value of 0.0 L h−1, a most probable value of 0.10 L h−1, and a maximum value of 0.30 L h–1. We assumed that the bathe duration and volume of water ingested per hour were independent parameters. The average water intake was estimated as 79.7 mL, which is in good agreement with the average of 50 mL per bathe found by Dang (1996). The distribution of oocysts or cysts was assumed to be homogenous, following a Poisson distribution

For Cryptosporidium, other parameters were characterized as described by Pouillot (2004).

For Giardia, the distribution of cyst viability was modelled though a beta distribution (2.93, 17.4) adjusted from data reported by Teunis (1997). Giardia infectious challenges of human volunteers reported by Rendtorff (1954) and Rose (1991) were used to estimate the parameter of an exponential dose-response model. A bootstrap method was used to estimate the uncertainty of this parameter, leading to an empirical distribution of r with mean r=2.00 × 10−2 (corresponding to a median infectious dose (N50=33 cysts) and a 95% confidence interval of [1.00 × 10−2; 3.75 × 10−2] (corresponding to a 95% confidence interval for N50 of [Hutin et al., 1998; 70 cysts]).

In both models, the risks of infection by Cryptosporidium and by Giardia per bathe were subsequently assessed using a second-order Monte Carlo simulation (Vose, 1998; Pouillot et al., 2004).


Analysis of parasitic contamination (Table 1)

View this table:
Table 1

Parasitic and bacterial contamination in two recreational lakes, Jablines (JAB) and Champs sur Marne (CSM); and in three rivers sites, Melun (MEL), Annet sur Marne (ASM), and Meaux (MEA)

IMS-IF (oocysts or cysts 10L−1)PCRi
SiteMonthCryptosporidiumGiardiaCryptosporidiumEnterocytozoon bieneusEscherichia coli (CFU 100 mL−1)Enterococci (CFU 100 mL−1)
Sept20C. hominis C. parvum38<38
Oct21C. hominis871119
Dec222C. parvum556119
Feb544C. hominis11958
ASMApr271C. hominis<581580
Sept919C. hominis60038
  • IMS-IF: immunomagnetic separation and specific immunofluorescence.

Twelve monthly samples were collected from four sites (JAB, ASM, MEA, MEL). Only nine samples could be collected from the CSM site (artificial beach), which was dry between April and June 2002. A total of 57 samples were thus collected. The sampling volumes were 20 L for four samples and <20 L for four samples (12, 12, 14 and 18 L).

Detection of E. bieneusi

In first-round PCR, we found that approximately two-thirds of the samples contained potent PCR inhibitors that were removed by incubation with PVP followed by an additional extraction step. When this procedure was applied to all water samples, contamination by E. bieneusi was observed only twice, in one lake sample (JAB) and one river sample (ASM).

Cryptosporidium and Giardia detection by IMS-IF

Fifteen samples were positive for Cryptosporidium. Each site was positive at least once. The two authorized bathing places (JAB and CSM) were occasionally contaminated (three out of 21 samples), with a maximum of 2 oocysts 10 L−1 (April 2002 at CSM). The contamination of river water was more consistent throughout the year, with a peak value of 9 oocysts 10 L−1 at ASM.

Contamination by Giardia was far more frequent at all five sites, with occasionally high parasite densities (up to 165 cysts 10 L−1). As with Cryptosporidium, the lowest levels of contamination were found at the two authorized bathing places (CSM and JAB). Nevertheless, a peak of contamination (3 and 108 cysts 10 L−1) was noted at the two sites, respectively.

Cryptosporidium PCR and genotyping results

PCR and then RFLP were performed on the second half of the water samples that were found to contain Cryptosporidium oocysts by IMS-IF. Six out of these 15 samples were found to be positive by PCR. The RFLP patterns corresponded to 5 C. hominis and 2 C. parvum, because one sample contained both species (Coupe et al., 2005).

Bacterial contamination (Table 1)

At the authorized bathing places, bacterial contamination was below the detection threshold in 9/21 and 16/21 samples for E. coli and enterococci, respectively. The threshold of 100 CFU 100 mL−1 was exceeded in 6/21 samples for E. coli and in 1/21 samples for enterococci.

At the river sites, marked variations in E. coli and enterococcal contamination were noted, with maximal values of 28 800 E. coli CFU 100 mL−1 at MEU, and 6220 enterococcal CFU 100 mL−1 at ASM. The threshold of 100 CFU 100 mL−1 was frequently exceeded (10 times at MEU, 11 times at MEA and 12 times at ASM), mainly as a result of E. coli contamination.

Relationships between parasitic and bacterial contamination (Table 2)

View this table:
Table 2

Statistical analysis of bacterial contamination as an indicator of parasite contamination

Cryptosporidium IMSCryptosporidium PCRGiardia IMS
Presence of GiardiaNSNS
Presence of Escherichia coliNSNSNS
Escherichia coli >100 CFU 100 mL−1NSNSP=6 × 10−4
Presence of EnterococcusP=3 × 10−3NSP=0.01
Enterococcus >100 CFU 100 mL−1NSNSP=8 × 10−4
  • NS: not significant

Possible relationships between the presence of Cryptosporidium, the presence of Giardia (detected by IMS-IF), and high bacterial counts (100 CFU 100 mL−1) were analysed with the Chi-square test or Fisher's exact test.

Statistical analysis showed no significant associations between the presence of Cryptosporidium and the presence of Giardia. Significant associations were found between the presence of Giardia and the presence of enterococci, and between high counts (>100 CFU 100 mL−1) of E. coli and enterococci.

Risk assessment models

Estimated risks of Cryptosporidium infection, cryptosporidiosis, and Giardia infection per bathe are shown in Table 3. These risks were estimated for observed IMS-IF values ‘on a given day’.

View this table:
Table 3

Estimated daily risk and confidence interval (CIinf; CIsup) determined by Monte Carlo simulations for immunocompetent individuals (per 10 000) older than 3 years of (a) infection by Cryptosporidium, (b) cryptosporidiosis, and (c) infection by Giardia

Estimated risk per 10 000, average and percentiles
No. oocystsAverage25507595
(a) Infection by Cryptosporidium
00.15 (0, 1.93)0.07 (0, 0.9)0.13 (0, 1.6)0.21 (0, 2.65)0.36 (0, 4.6)
10.58 (0.08, 3.21)0.27 (0.04, 1.5)0.48 (0.07, 2.67)0.79 (0.11, 4.41)1.37 (0.2, 7.66)
20.99 (0.17, 4.48)0.46 (0.08, 2.09)0.82 (0.14, 3.72)1.36 (0.24, 6.15)2.36 (0.42, 10.67)
52.37 (0.52, 9.06)1.1 (0.24, 4.23)1.96 (0.43, 7.52)3.25 (0.71, 12.43)5.64 (1.24, 21.58)
104.69 (1.08, 14.8)2.19 (0.51, 6.92)3.89 (0.9, 12.29)6.43 (1.49, 20.31)11.17 (2.59, 35.25)
(b) Cryptosporidiosis
00.06 (0, 0.9)0.03 (0, 0.42)0.05 (0,0.74)0.09 (0, 1.23)0.15 (0, 2.13)
10.25 (0.03, 1.57)0.12 (0.01, 0.73)0.21 (0.03, 1.3)0.35 (0.04, 2.15)0.6 (0.08, 3.73)
20.43 (0.07, 2.02)0.2 (0.03, 0.94)0.36 (0.06, 1.67)0.59 (0.09, 2.77)1.03 (0.16, 4.8)
51.02 (0.22, 4.23)0.48 (0.1, 1.98)0.85 (0.18, 3.51)1.4 (0.3, 5.8)2.44 (0.52, 10.07)
102.05 (0.43, 6.77)0.96 (0.2, 3.16)1.7 (0.36, 5.62)2.82 (0.59, 9.29)4.89 (1.02, 16.12)
(c) Infection by Giardia
00.17 (0, 1.87)0.08 (0, 0.88)0.14 (0, 1.55)0.23 (0, 2.57)0.41 (0, 4.46)
10.69 (0.09, 3.97)0.32 (0.04, 1.86)0.57 (0.08, 3.29)0.94 (0.13, 5.45)1.63 (0.22, 9.46)
21.23 (0.18, 5.83)0.57 (0.08, 2.72)1.02 (0.15, 4.84)1.68 (0.24, 8)2.92 (0.42, 13.89)
52.77 (0.52, 10.74)1.29 (0.24, 5.02)2.3 (0.43, 8.91)3.8 (0.72, 14.73)6.6 (1.25, 25.58)
105.62 (0.99, 18.83)2.63 (0.46, 8.8)4.67 (0.82, 15.63)7.72 (1.35, 25.84)13.4 (2.35, 44.84)

The estimated average risk of Cryptosporidium infection was greater than 10−4 for an oocyst density of >2 10 L−1. Such a density was found nine times in river water (MEU, ASM and MEA) but never in lake water (JAB and CSM). Similarly, the risk of clinical cryptosporidiosis was <10−4 for an oocyst density of <510 L−1, a level never observed at an authorized bathing place. The estimated average risk of cryptosporidiosis among immunocompromised bathers was 30 × 10−4 for an oocyst density of 1 10 L−1 (data not shown).

Regarding Giardia, the results obtained by the Monte Carlo simulation showed that the average risk of infection exceeded 10−4 for a density of >2 cysts 10 L−1. Such concentrations were frequent in unauthorized recreational waters, but were rare in authorized recreational waters (four times at JAB and once at CSM). The risk of clinical giardiasis could not be estimated, owing to the lack of a reliable model relating infection to disease.


Our study provides additional information on the contamination of fresh water by Cryptosporidium, Giardia and E. bieneusi, focusing on recreational water and estimating the risk for cryptosporidiosis during bathing.

Previous PCR studies have shown evidence of E. bieneusi DNA in surface water (Liguory et al., 1997; Sparfel et al., 1997; Dowd et al., 1998, 2003; Fournier et al., 2000). Using the same techniques, we found only two positive samples from the river Marne. Although the presence of PCR inhibitors might have resulted in an underestimation of this parasitic contamination, these results suggest that the risk of human infection by microsporidia during swimming at these sites is low.

In contrast, Cryptosporidium and Giardia were more frequently detected.

The two authorized recreational areas studied here (Jablines and Champs sur Marne), which are fed by alluvial water and are regularly monitored for bacterial contamination, were only occasionally contaminated, whereas all the river bathing places were frequently contaminated, Giardia being more often present than Cryptosporidium. We could not evidence a seasonality of contamination for either parasite.

From the 15 samples in which Cryptosporidium was evidenced by IMS-IF, only 7 could be successfully amplified by PCR and genotyped. This low rate of positive PCR may be related to the persistence of potent inhibitors or other interfering compounds in the DNA eluate (Jellison et al., 2002), or to the heterogenous distribution of the oocysts in the sample, resulting in possible differences of sensitivity between IMS and the corresponding PCR. Indeed, a higher volume of water sample would probably have led to better detection sensitivities, but water samples larger than 20 L often result in the saturation of the Envirochek® cartridge (1 µm pore size) used for filtration in the AFNOR NF T 90-455 procedure. RFLP yielded specific patterns of C. hominis and C. parvum, indicating a human and possible animal origin of the contamination. These results are consistent with the urban or peri-urban locations of the study sites, with human frequentation (whether or not bathing is allowed). A genotyping of Giardia isolates (not performed in this study) might have added further information regarding the possible origin of contamination.

The potential relevance of bacterial analysis for indirect assessment of parasitic contamination was examined by comparing the bacterial counts with the IMS-IF results for Cryptosporidium and Giardia. We estimated that CFU counts of >100 E. coli 100 mL−1 or >100 enterococci 100 mL−1 are associated with the presence of Giardia but not Cryptosporidium. Similar results were obtained when we used the thresholds proposed in the revised European directive (Council of the European Union) for microbiological water quality (E. coli <500 CFU 100 mL−1 and enterococci <200 CFU 100 mL−1) (data not shown). A similar lack of correlation was found by Villaginès (1996) for various water samples and, more recently, by Harwood et al., (2005) who reported that the monitoring of E. coli and/or enterococci was not informative on the presence of Cryptosporidium or Giardia. Thus, the search for parasites in water should be performed independently of bacterial analysis.

Our quantitative results allowed us to make a rough estimate of the risk of infection by Cryptosporidium and Giardia during swimming at the study sites. Given the lack of reliable data on water ingestion during bathing in natural waters, we estimated this quantity from data on swimming pools and from assumptions on the duration of bathing and on accidental water ingestion (Dang, 1996). We then adapted QRA models of the risk of Cryptosporidium and Giardia infection after tap-water consumption. These models imply full acceptance of several weak or strong hypotheses reflecting parameter variability and/or uncertainty (Pouillot et al., 2004). For example, the dose-response relationships for Cryptosporidium and Giardia had both been determined on a small sample of individuals, and only for Cryptosporidium parvum (DuPont et al., 1995; Teunis et al., 1997; Pouillot et al., 2004). Although we were able to genotype some of our Cryptosporidium isolates, it was not possible to quantify the oocysts of each species, as the IMS is genus-specific and the IF technique focuses on C. hominis and C. parvum species, although cross reactions with C. meleagridis, C. muris and C. baileyi species are possible. Thus the same dose-response relationship was used whatever the identified species, and this may have led to an over- or under-estimation of the risk, as no specific data on human infectivity are currently available for C. hominis.

Viability and infectivity of cysts or oocysts are also determinant factors for QRA. Because IMS-IF examination or PCR cannot reliably assess the viability of cysts/oocysts, the QRA model took into account this lack of information by assuming that the probability that any one recovered oocyst is viable and infective varies between 0 and 1, with a most probable value of 0.4. This variability was modelled with a Pert (0, 0.4, 1) distribution (Pouillot et al., 2004).

When applied to our data for Cryptosporidium, we found that a >10−4 risk of Cryptosporidium infection [US Environmental Protection Agency (USEPA)-determined threshold] was reached at a contamination level of >2 oocysts 10 L−1. Such a level was never found in the authorized recreational waters, but was exceeded on several occasions at the river sites. As expected, the risk was much higher for immunocompromised patients, supporting the recommendation that such patients should avoid bathing in fresh recreational waters (Kaplan et al., 2002).

The model used for Giardia was also informative, but it was probably less reliable than for Cryptosporidium, because of the greater uncertainty regarding both the infectivity for humans of all Giardia groups found in environmental water and the dose-response relationship, and the lack of data with which to model the relationship between infection and disease. As with Cryptosporidium, we roughly estimated that the risk was very high when bathing in river water, whereas it rarely exceeded >10−4 at the recreational lakes.

In conclusion, this epidemiological study, combined with QRA analysis, clearly confirms that bathing in surface water in the Paris region of France is associated with a significant risk of infection by Cryptosporidium or Giardia. This was especially the case for rivers not protected from human or animal fecal contamination. In authorized recreational waters, our genotyping study of Cryptosporidium isolates indicates that surveillance should focus on the prevention of contamination by human waste. Close microbiological surveillance should be maintained in other fresh waters, as occasional peaks of contamination can be observed, possibly owing to heavy rainfalls that go through fields to lakes or to high-contamination events that could occur upstream from the sampling river sites. Our results suggest that E. coli and enterococci are not relevant indicators of parasitic contamination, but that low levels of these bacteria are associated with the absence of detectable parasites. Parasitic surveillance of fresh recreational waters should thus be carried out independently of bacterial surveillance.


This work was supported by the Agence Française de Sécurité Environnementale (AFSSE) (grant RD-2003-010), and S.C. was supported by a grant from the Agence Nationale de la Recherche Technologique (ANRT).


  • Editor: Kai Man Kam


View Abstract