Prospective study of acute health effects in relation to exposure to cyanobacteria

Science of the Total Environment 466–467 (2014) 397–403

Contents lists available at SciVerse ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Prospective study of acute health effects in relation to exposure to cyanobacteria Benoît Lévesque a,b,c,⁎, Marie-Christine Gervais b, Pierre Chevalier b, Denis Gauvin b, Elhadji Anassour-Laouan-Sidi c, Suzanne Gingras b, Nathalie Fortin d, Geneviève Brisson b, Charles Greer d, David Bird e a

Université Laval, Faculté de médecine, Département de médecine sociale et préventive, 945 Ave. Wolfe, Quebec, Quebec G1V 5B3, Canada Institut national de santé publique du Québec, 945 Ave. Wolfe, Quebec, Quebec G1V 5B3, Canada c Centre de recherche du CHU de Québec, Axe Santé des populations et pratiques optimales en santé, Edifice Delta 2, Bureau 600, 2875 Blvd. Laurier, Quebec, Quebec G1V 2M2, Canada d National Research Council Canada, Energy, Mining and Environment, 6100 Royalmount Avenue, Montreal, Quebec H4P 2R2, Canada e Université du Québec à Montreal, Faculté des sciences, Département des sciences biologiques, Case postale 8888, Succ Centre-ville, Montreal, Quebec H3C 3P8, Canada b

H I G H L I G H T S • • • • •

The relationship between exposure to cyanobacteria and health symptoms was examined. Limited water contact with cyanobacteria was linked to gastrointestinal symptoms. Drinking water contaminated by cyanobacteria was associated with health symptoms. The public should be informed of symptoms associated with exposure to cyanobacteria. A management plan is needed for plants treating cyanobacteria-contaminated water.

a r t i c l e

i n f o

Article history: Received 22 April 2013 Received in revised form 2 July 2013 Accepted 13 July 2013 Available online xxxx Editor: Simon Pollard Keywords: Cyanobacteria Microcystin Bathing Drinking water Recreational water

a b s t r a c t We conducted a study to investigate the relationship between exposure to cyanobacteria and microcystins and the incidence of symptoms in humans living in close proximity to lakes affected by cyanobacteria. The design was a prospective study of residents living around three lakes (Canada), one of which has a water treatment plant supplying potable water to local residents. Participants had to keep a daily journal of symptoms and record contact (full or limited) with the water body. Samples were collected to document cyanobacteria and microcystin concentrations. Symptoms potentially associated with cyanobacteria (gastrointestinal: 2 indices (GI1: diarrhea or abdominal pain or nausea or vomiting; GI2: diarrhea or vomiting or [nausea and fever] or [abdominal cramps and fever]); upper and lower respiratory tract; eye; ear; skin; muscle pain; headaches; mouth ulcers) were examined in relation with exposure to cyanobacteria and microcystin by using Poisson regression. Only gastrointestinal symptoms were associated with recreational contact. Globally, there was a significant increase in adjusted relative risk (RR) with higher cyanobacterial cell counts for GI2 (b20,000 cells/mL: RR = 1.52, 95% CI = 0.65–3.51; 20,000–100,000 cells/mL: RR = 2.71, 95% CI = 1.02–7.16; N100,000 cells/mL: RR = 3.28, 95% CI = 1.69–6.37, p-trend = 0.001). In participants who received their drinking water supply from a plant whose source was contaminated by cyanobacteria, an increase in muscle pain (RR = 5.16; 95% CI = 2.93– 9.07) and gastrointestinal (GI1: RR = 3.87; 95% CI = 1.62–9.21; GI2: RR = 2.84; 95% CI = 0.82–9.79), skin (RR = 2.65; 95% CI = 1.09–6.44) and ear symptoms (RR = 6.10; 95% CI = 2.48–15.03) was observed. The population should be made aware of the risks of gastrointestinal symptoms associated with contact (full or limited) with cyanobacteria. A risk management plan is needed for water treatment plants that draw their water from a source contaminated with cyanobacteria. © 2013 Elsevier B.V. All rights reserved.

⁎ Corresponding author at: Université Laval, Faculté de médecine, Département de médecine sociale et préventive, 945 Ave. Wolfe, Quebec, Quebec G1V 5B3, Canada. Tel.: +1 418 650 5115x5214; fax: +1 418 654 3144. E-mail addresses: [email protected] (B. Lévesque), [email protected] (M.-C. Gervais), [email protected] (P. Chevalier), [email protected] (D. Gauvin), [email protected] (E. Anassour-Laouan-Sidi), [email protected] (S. Gingras), [email protected] (N. Fortin), [email protected] (G. Brisson), [email protected] (C. Greer), [email protected] (D. Bird). 0048-9697/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.scitotenv.2013.07.045

398

B. Lévesque et al. / Science of the Total Environment 466–467 (2014) 397–403

1. Introduction

2. Material and methods

Cyanobacteria occur naturally in freshwater and marine ecosystems in all climates (AFSSA-AFSSET, 2006). These microorganisms are of interest to researchers and public health authorities because of the toxicity of the cyanotoxins they produce. Cyanotoxins are a large group of toxins that produce a variety of negative health impacts. Some cyanotoxins are hepatotoxins, including nodularins and microcystins, of which there are at least 80 variants (Humpage, 2008). Microcystins are the most frequently documented toxins in blooms (WHO, 1999) in human and animal poisonings (AFSSA-AFSSET, 2006). A cytotoxin, cylindrospermopsin, has also been documented, mainly in Australia but also in Europe, Asia, South America and North America (Falconer and Humpage, 2005; Hamilton et al., 2005). Some cyanotoxins are neurotoxins (AFSSA-AFSSET, 2006), including anatoxin, saxitoxin and its derivatives and β-N-methylaminoL-alanine (BMAA), which is suspected of being associated with neurodegenerative diseases (Pablo et al., 2009). Lastly, endotoxins, structural components of cyanobacterial cell walls, are suspected of having irritative or allergenic effects (AFSSA-AFSSET, 2006). In 1996, an unfortunate incident occurred in Caruaru, Brazil that highlighted the toxic potential of cyanotoxins. At a hemodialysis unit, one hundred people developed liver failure after undergoing treatment with water contaminated by microcystins (Carmichael et al., 2001). Acute effects have been associated with drinking water, and several case series have been reported following exceptional situations, such as the use of copper sulfate to treat blooms in drinking water reservoirs or the inappropriate treatment of massive blooms. The most common health problems were gastrointestinal or hepatic symptoms (Hawkins et al., 1985; Lippy and Erb, 1976; Teixeira Mda et al., 1993; Veldee, 1931). A case–control study also showed an increase in gastrointestinal and dermatological symptoms in people whose drinking water supply came from a river which was contaminated with an extensive bloom and was chlorinated but otherwise not treated (El Saadi et al., 1995). While the World Health Organization (WHO) and Health Canada have established guidelines for total microcystin-LR in drinking water of 1 μg/L (WHO, 2008) and 1.5 μg/L (HC, 2013) respectively, there is a need for data from epidemiological studies on the health effects of exposure to cyanobacteria to clarify the risk (Falconer and Humpage, 2005). Case series have also been published regarding recreational exposure, with the most common symptoms being gastrointestinal, cutaneous and respiratory (Dillenberg and Dehnel, 1960; Turner et al., 1990; Williamson and Corbett, 1993). As with drinking water, few epidemiological studies with comprehensive study designs have been conducted. Two prospective studies conducted on visitors exposed to cyanobacteria at different beaches showed associations with health effects (Pilotto et al., 1997; Stewart et al., 2006a). For recreational water, the WHO has estimated that there is a low probability of adverse health effects below 20,000 cells/mL, while there is a moderate risk of effects, in particular acute gastrointestinal or skin problems, at concentrations over 100,000 cells/mL because at this level, 20 μg/L of microcystins may be present (WHO, 1999). Similarly, Health Canada has established guidelines of 100,000 cells/mL and 20 μg/L for total cyanobacteria and total microcystins, respectively (HC, 2012). To further substantiate the limited data on the acute human effects following drinking water and recreational water exposure to waterborne cyanobacteria, we conducted a prospective study over an eight-week period on people living on the shores of three lakes in the province of Quebec (Canada). Our objective was to investigate the relationship between exposure to cyanobacteria and microcystins (the most common cyanotoxins) and the incidence of symptoms in human populations living in close proximity to the lakes affected by cyanobacteria.

On the basis of their history of cyanobacterial presence (from a list prepared by the Quebec Ministry of Sustainable Development, Environment and Parks), the presence of an active residents' association willing to collaborate in the project and proximity to the research team, three lakes were selected: Lake William (LW), Lake Roxton (LR) and Lake Champlain's Missisquoi Bay (MB). At the latter site, a drinking water treatment plant supplied a sector of MB residents. The inclusion factors for study participants were: five years of age or over, access to the lake and residing in the targeted residence for more than two weeks during the study period. No more than three participants per family were accepted. Approximately 400 addresses were randomly selected for each lake. Data collection took place from June 27 to August 21, 2009. The final toxin analyses were completed in December 2011. The protocol was approved by the research ethics committee of the Centre Hospitalier Universitaire de Québec and participants signed an informed consent form. Two tools were used for data collection. First, a family questionnaire was used to document different individual variables (sociodemographic characteristics, medical history, symptoms in the two weeks prior to data collection, medication, occupation, travel in the last month) and information about the household (drinking water supply [treatment plant drawing its water from MB, water at risk of fecal contamination (surface well, water taken directly from the lake), others], presence of pets at home) and a daily journal of symptoms to collect data from participants on symptoms (eye, ear, respiratory, gastrointestinal, skin, muscle pain, headaches, mouth ulcers) potentially associated with cyanobacteria (Pilotto et al., 1997; Stewart et al., 2006a) including medical consultations or hospitalizations. At the same time, participants recorded full contact (swimming, waterskiing, windsurfing, use of watercraft involving launching, accidental falls) and limited contact (fishing, use of watercraft not involving launching) with lake water, the duration (less than 1 h, 1 h to less than 3 h, 3 h and more) and location of contact as well as head immersion and ingestion of water or not during water contact. Lastly, swimming in other lakes or rivers, in swimming pools and consumption of fish from the lake were also recorded. Daily surface water samples (duplicates at a water depth of 0.3 m and duplicates from 1.2 m water depth (MENVIQ, 1986) comprised one composite sample) were collected at five littoral sampling stations at LW and MB, and at four stations at LR. Limnetic samples were also collected (3 surface water samples to form a composite sample) twice a week at one station at LR and at two stations at LW and MB. Once a week, samples (littoral: 1 where water depth was 0.3 m and 1 where water depth was 1.2 m to form a composite sample; limnetic: 1 sample) were also collected at a depth of 15 cm to assess for the presence of Escherichia coli (E. coli) at the same stations. For days without sampling and other missing data (technical problems), values were interpolated for each day based on the chronologically closest values measured. For cyanobacteria, microcystins and E. coli, 3%, 7% and 87%, of the littoral samples and 86%, 85% and 88% of the limnetic samples were interpolated, respectively. Cyanobacterial cells were counted and identified as previously described (Rolland et al., 2005). Dissolved microcystin was measured in filtered water samples, which were passed through glass fiber filters of 0.7 μm pores (Whatman GF/F) and kept frozen. Microcystin particulates were captured on the same filter using volumes ranging from 20 to 500 mL, depending on the density of planktonic particles. Particulate toxins were extracted by sonication of frozen filters in distilled water, followed by clarification by centrifugation at 13,000 g for 10 min. The final extracts were diluted 20 to 40 times in order to avoid interferences present in the raw water. After filtration, dissolved and particulate microcystins were determined by ELISA (Abraxis 96-well microtiter plates PN 520011, Abraxis, Warminster, PA). This test is sensitive to the functional moiety of these hepatotoxins, the ADDA group, but the

B. Lévesque et al. / Science of the Total Environment 466–467 (2014) 397–403

standard curve is based on microcystin-LR. Where necessary, the extracts were diluted up to 30,000 times to ensure they were in the linear region of the 4-parameter logistic standard curve. The sum of dissolved and particulate microcystins was considered as total microcystins. E. coli was identified by membrane filtration based on USEPA method 1604 (US-EPA, 2002). Participants' symptoms were examined individually and by category. For respiratory symptoms, two indices were created: upper respiratory tract (runny nose or sore throat) and lower respiratory tract (cough or wheezing) (Tager and Speizer, 1976). Two indices were also created for highly credible gastrointestinal symptoms: GI1 (diarrhea [3 loose motions/day] or abdominal pain or nausea or vomiting) and an index for more severe symptoms, GI2 (diarrhea or vomiting or [nausea and fever] or [abdominal cramps and fever]) (Wiedenmann et al., 2006). A six-day symptom-free interval between episodes was taken into account (Dale et al., 2009). Latency periods were determined. Since the health effects of cyanobacteria are supposed to be secondary to intoxication (i.e. without an incubation period) (Stewart et al., 2006B), we postulated that they occur within 24 h of exposure. However, for muscle pain and gastrointestinal effects, logically secondary to ingestion, the latency period may be longer, in the region of 48 h (Williamson and Corbett, 1993). Thus, after an episode, we verified for the presence of contact with the lake on the same day and in the three days prior to the episode for muscle pain and gastrointestinal symptoms and on the same day and on the day prior to the episode for other symptoms. Exposure to E. coli categorized into two strata (≤ and N100 CFU/ 100 mL) (Wiedenmann et al., 2006) and was analyzed in two ways. Firstly, by assigning a concentration to each contact by statistical interpolation of adjacent concentrations, then more conservatively, by assigning a high risk index to any contact at a site where concentrations were higher than 100 CFU/100 mL at least once during the summer. Poisson regression (Kleinbaum et al., 1988) was used to model the incidence of the different symptoms in relation to contact with water bodies, cyanobacterial counts, potentially toxic species documented in the literature (MDDEP, 2008) and microcystin concentrations (in tertiles and at a concentration of more than 1 μg/L) taking into account potential confounders found to be significant in univariate analysis (p ≤ 0.1). Repeated measures by individual and clustering of participants within households were taken into account using generalized linear models (Liang and Zeger, 1986). When appropriate, trend tests were calculated to verify the trend between the occurrence of symptoms and exposure to different concentrations of cyanobacteria using orthogonal contrasts. It was impossible to examine symptoms by type of contact (full or limited) in relation to concentrations due to the small sample size. Since the analyses of cyanobacterial counts by toxic potential added little information to the analyses with total counts, it was not relevant to present models for them. Given the relatively low concentrations of exposure to microcystins documented and the strong correlation with the cyanobacterial counts (polychoric correlation coefficient = 0.81) (Uebersax, 2013), it was also not relevant to present models concerning the exposure to microcystins in tertiles. Finally, to verify the behavior of citizens in relation to cyanobacterial blooms, we also used Poisson regression to model the effect that cyanobacterial concentrations had on water contact (full and limited), taking into account precipitation and air temperatures. Precipitations and air temperature were obtained from the Environment Canada data bank for the closest weather stations of the lakes included in the study (Environment Canada, 2013). 3. Results Of a total of 1220 families contacted, we were able to reach 911. Of these subjects, 80 were not eligible and 506 refused to participate. Of the remaining 325 families, 12 were excluded (questionnaires completed incorrectly) and 46 withdrew. A total of 267 families, comprising 501 subjects, participated in the study. Of these, 35 were subsequently excluded (symptoms journal not kept). Ultimately, 267 families

399

and 466 subjects were included in the study: LW, 95 families (155 participants); LR, 83 families (150 participants); MB, 89 families (161 participants). In all, there were 245 women (53%; LW: 57%, LR: 46%; MB: 53%) and 221 men (47%; LW: 43%; LR: 53%; MB: 47%). The age distribution was as follows: 0–20 years: 11%; 21–40 years: 12%; 41–60 years: 39%; and N60 years: 38%. Only 12% of the families included in the study used their dwelling near the lake as a secondary residence. For others it was their permanent residence. Excluding two missing values, a proportion of 50% (n = 133) of dwellings included in the study received their drinking water from a municipal or a private aqueduct. From these families, 20 were supplied by a water treatment plant collecting the water in MB. Artesian wells were used by 114 families (43%). Finally, 18 families (7%) used a source at risk of fecal contamination (surface well, 15; directly in the lake, 3) as their tap water. As a whole, 68% of the families (aqueduct, 73%, artesian wells 64%, surface wells 60%, directly in the lake 67%) used the tap water as their main source of drinking water. There were 3163 instances of contact with the lakes: 1560 at LW (full: 654; limited: 884; unknown: 22), 765 at LR (full: 314; limited: 444; unknown: 7) and 838 at MB (full: 580; limited: 226; unknown: 32). The geometric mean of the E. coli samples for each of the stations ranged from 8 to 36 CFU/100 mL at LW, from 2 to 66 CFU/100 mL at LR and from 0 to 145 CFU/100 mL at MB. Fourteen percent, 11% and 11% of the samples collected at LW, LR and MB, respectively, exceeded 100 CFU/100 mL. Fig. 1 shows the cyanobacterial counts and microcystin concentrations at each station. Exposure to cyanobacteria (in log) and microcystins was much lower for LW residents than for the other two water bodies. Lake William residents therefore constituted a less exposed group. In fact, the median values of the cyanobacterial counts for littoral stations and limnetic stations in LW were 1032 cells/mL (range: 7–29,780 cells/mL) and 11,238 cells/mL (range: 1407–49,106 cells/mL), respectively. In contrast, cyanobacterial counts were much higher in LR (littoral stations: median = 20,001 cells/mL, range: 54–N106 cells/mL; limnetic station: median = 69,050 cells/mL, range: 766–781,340 cells/mL) and MB (littoral stations: median = 21,485 cells/mL, range: 18 –N 106 cells/mL; limnetic stations: median = 67,912 cells/mL, range: 8909–N106 cells/mL). For the microcystins concentrations in LW, medians for littoral and limnetic stations were under the limit of detection (LD) and the maximum values were 0.63 μg/L and 0.02 μg/L, respectively. The concentrations in LR (littoral stations: median = 0.23 μg/L, range: 0.008 μg/L–108.8 μg/L, N 1 μg/L = 16%, N 20 μg/L = 2%; limnetic stations: median = 0.12 μg/L, range: 0.04 μg/L–1.12 μg/L, N1 μg/L = 9%) were generally low with some high values, but few higher than 20 μg/L, the value used by WHO as the rationale for the limit of reference for recreational water (WHO, 1999). In MB, medians of the data were of the same order of magnitude than in LR. However, there were more high values (littoral stations: median = 0.70 μg/L, range: under LD — 773 μg/L, N1 μg/L = 44%,N 20 μg/L = 10%; limnetic stations: median = 0.35 μg/L, range: 0.001 μg/L–125 μg/L, N1 μg/L = 29%, N20 μg/L = 4%). Episodes of symptoms reported by participants were divided as follows: respiratory, 322; headaches, 228; gastrointestinal GI1, 180 and GI2, 88; eye, 169; muscle pain, 130; skin, 124; ear, 55; and mouth ulcers, 39. No association was found between exposure to more than 100 E. coli/100 mL during contact with the lakes and symptoms after interpolation of missing values or after assigning a high-risk index. Table 1 shows the results of multivariate models describing the relationship between gastrointestinal symptoms and contact with the water bodies. Owing to the interaction between contact and the lake, the results of contact are presented by lake. At LR and MB, there was a relationship between contact and GI1 and GI2. In contrast, for LW participants, the relative risks (RRs) were less than 1.0. The RRs were high for families whose drinking water supply came from MB, especially for GI1, and for families living in houses with water sources at risk of fecal contamination.

400

B. Lévesque et al. / Science of the Total Environment 466–467 (2014) 397–403

Fig. 1. A: Number of cyanobacteria by lake. B: Number of cyanobacteria in relation to World Health Organization guideline values for recreational water. C: Microcystin concentrations by lake. D: Microcystin concentrations in relation to World Health Organization guideline values for drinking water and recreational water.

None of the symptoms other than gastrointestinal were associated with contact with the lakes. However, in regression modeling contact with the lakes, significant RRs were observed for muscle pain, skin symptoms (erythema, itching, rashes) and ear symptoms (discharge, pain) in relation to drinking water supply from MB (Table 2). For gastrointestinal symptoms, we examined the effect of full and limited contact with the lakes separately. There was no increase in adjusted RRs (gender, gastrointestinal symptoms reported in the two weeks prior to data collection, residence's source of drinking water) for full contact (GI1: RR = 0.92; 95% CI = 0.49–1.74; GI2: RR = 0.86; 95% CI = 0.36–2.05). Furthermore, no relationship was observed with duration of contact or head immersion. However, the RRs were significantly higher for limited contact (GI1: RR = 2.73; 95% CI = 1.57–4.74; GI2: RR = 3.08; 95% CI = 1.47–6.47). This inconsistency can be explained by the fact that participants avoid full contact with the water during severe cyanobacterial proliferations. We used Poisson regression to model the relationship between contact and air temperature, precipitation and cyanobacterial counts. Full contact decreased with increasing counts, while limited contact did not. Moreover, full and limited contacts were significantly more frequent at warmer temperatures and significantly less frequent when there was precipitation (1 or 5 mm.) (Table 3).

Table 4 shows the relationship between gastrointestinal symptoms and contact with the lakes at different concentrations of cyanobacteria. The other significant variables related to the symptoms were included in the models (the lake where the contact occurred, the water source of the dwelling, sex, gastrointestinal symptoms reported in the two weeks prior to data collection). For GI2, a significant increase in risk was observed with increasing cell counts (trend: p = 0.001). As a whole, microcystin concentrations during contact were relatively low (1st tertile: b0.0012 μg/L; 2nd tertile: 0.0012–0.2456 μg/L; 3rd tertile: N0.2456 μg/L). The maximum concentration at which an episode of gastrointestinal symptoms was reported was 7.65 μg/L, much lower than the reference limit of 20 μg/L stated by Health Canada (2012) and WHO (1999). For exposure to more than 1 μg/L, adjusted RRs (gender, gastrointestinal symptoms reported in the two weeks prior to data collection, residence's source of drinking water) for GI1 and GI2 were 1.06 (95% CI = 0.32–3.52) and 1.48 (95% CI = 0.41– 5.23), respectively. 4. Discussion Of all the symptoms studied, only gastrointestinal symptoms were associated with contact with the lakes. The symptoms did not require

B. Lévesque et al. / Science of the Total Environment 466–467 (2014) 397–403 Table 1 Multivariate models describing the relationship between gastrointestinal symptoms and contact (full and limited) with the lakes adjusted for significant variables. Variables

GI1 RR

Contact: Missisquoi Baya No Yes Contact: Lake Roxtonb No Yes Contact: Lake Williamc No Yes Water source:d Not at risk From MBe At riskf Sex: Male Female Symptoms:g No Yes

1.00 2.48 1.00 2.72

GI2 95% CI

RR

1.14–5.36

1.00 2.71

1.31–5.63

1.00 3.99

1.00 0.43

0.16–1.16

1.00 0.28

0.08–0.97

1.00 3.87 1.89

1.62–9.21 1.00–3.57

1.00 2.84 3.17

0.82–9.79 1.36–7.35

1.00 2.79

1.70–4.58

1.00 3.26

1.64–6.48

1.00 4.00

2.33–6.88

1.00 6.15

2.36–16.05

1.50–10.60

a

Participants from Missisquoi Bay. Participants from Lake Roxton. c Participants from Lake William. d Residence's water source. e 20 families received their tap water from a treatment plant using Missisquoi Bay water. f 18 families used a source for their tap water at risk of fecal contamination (surface well, 15; directly in the lake, 3). g Gastrointestinal symptoms reported in the two weeks prior to data collection. b

a medical consultation. A significant increase in the symptoms was observed in residents who had contact with the two lakes that had significant proliferations (MB and LR), but not in those from LW where cyanobacterial counts were low (Table 1). At LW, the significant RR of less than 1.0 for GI2 suggests an uncorrected selection bias. It is very possible that people in better health had more frequent contact with the lake, thereby resulting in an underestimation of RRs. If so, this potential bias transposed to the residents of the other two lakes would imply an underestimation of the RR documented at these two other sites. Contact aside, people who reported gastrointestinal symptoms in the two weeks prior to data collection were more at risk of digestive problems indicating that they were probably more prone to these ailments. Women were also more likely to experience episodes of gastrointestinal symptoms. At least two studies made in the Canadian population by questionnaire, one by face-to-face interviews

Table 2 Multivariate models describing the relationship between symptoms other than gastrointestinal and contact (full and limited) with lakes adjusted for significant variables. Variables

Contact with lakes: No Yes Water source:a Others From MBb Symptoms:c No Yes a

Muscle pain

Skin symptoms

Table 3 Multivariate models describing the relationship between contact (full and limited) with the lakes and the cell counts of cyanobacteria adjusted for air temperature and precipitation. Variables

95% CI

1.00–7.35

Ear symptoms

RR

95% CI

RR

95% CI

RR

95% CI

1.00 0.90

0.57–1.41

1.00 1.20

0.75–1.94

1.00 0.60

0.19–1.83

1.00 5.16

2.93–9.07

1.00 2.65

1.09–6.44

1.00 6.10

2.48–15.03

1.00 13.33

8.11–21.90

1.00 7.17

4.38–11.72

1.00 20.09

9.15–44.09

Residence's water source. b 20 families received their tap water from a treatment plant using Missisquoi Bay water. c Related symptoms reported in the two weeks prior to data collection.

401

Cyanobacteria (cells/mL): b20,000 20,000–100,000 N100,000 Air temperature: b15 °C 15–20 °C N20 °C Precipitation:a +1 mm +5 mm a

Full contact

Limited contact

RR

95% CI

RR

95% CI

1.00 0.68 0.62

0.54–0.85 0.46–0.84

1.00 0.81 1.09

0.64–1.02 0.86–1.38

1.00 2.05 2.63

1.64–2.57 2.01–3.45

1.00 1.55 1.71

1.32–1.81 1.43–2.05

0.992 0.960

0.987–0.997 0.935–0.985

0.989 0.944

0.984–0.993 0.921–0.968

Analyzed as a continuous variable.

(Stanghellini, 1999) and the other one by a cross-sectional telephone survey (Majowicz et al., 2004), have shown a higher prevalence of gastrointestinal symptoms among women than men. In addition of the gastrointestinal symptoms GI1 (Table 1), a significantly higher incidence of various other symptoms (muscle pain, skin symptoms, ear symptoms) was observed in participants whose water supply came from MB (Table 2). The symptoms in question appeared relatively benign, since they did not require a medical consultation. The water intake is located near sampling station Missisquoi Bay 5 (Fig. 1), a site where samples were regularly contaminated by cyanobacteria and microcystins. The treatment plant uses a full treatment process (coagulation, decantation, filtration, disinfection) (Robert et al., 2005), and by regulation, monitoring of E. coli is performed 8 times/month. During the period of this study (June, July and August 2009), all the samples (n = 24) were negative. However, a recent study conducted at this site from 2008 to 2010 showed that small amounts of cyanobacteria in the order of 2000 to 3000 cells/mL and microcystins (up to 2.47 μg/L) could nonetheless be found in the treated water (Zamayadi et al., 2012). A proportion of 77% of participants who received their drinking water from MB said they drank tap water. For some symptoms, the statistical relationship was very significant. While it is premature to speak of a causal association, it is plausible that exposure to cyanobacteria contributed to the presence of these symptoms and that the water treatment process reduced the importance of symptoms without preventing them entirely. No relationship was found between gastrointestinal symptoms and full contact with water. However, highly significant adjusted RRs were observed for limited contact. Table 3 provides a logical explanation for this contradiction. The participants avoided full contact but continued to have limited contact during blooms suggesting that many of them are already aware of the potential health implications of contact with cyanobacteria. Dorevitch et al. (2012) recently demonstrated a significant risk in the order of 1.5 for gastroenteritis even in water considered acceptable for swimming based on fecal contamination indicators. In the present study, E. coli concentrations were generally low and no relationship was found with this potentially confounding factor, which supports the hypothesis that the effects observed are due to exposure to cyanobacteria. People who participated in limited-contact activities may have had contact with cyanobacteria in numerous ways, for instance, when getting into their boats, by putting their hands in the water and then in their mouth, etc. Furthermore, exposure through aerosols over water bodies (Backer et al., 2008) is also plausible. We documented a relationship between contact (full and limited) and gastrointestinal symptoms in relation to cyanobacterial counts (Table 4). For the most severe symptoms, the adjusted RR increases with concentration suggested a dose–effect relationship (Table 4. p trend = 0.001). As indicated in the Material and methods section, we were

402

B. Lévesque et al. / Science of the Total Environment 466–467 (2014) 397–403

Table 4 Multivariate models describing the relationship between gastrointestinal symptoms and contact (full or limited) with different concentrations of cyanobacteria, adjusted for significant variables. Variables

Contact with cyanobacteria (cells/mL): No b20,000 20,000–100,000 N100,000 Contact: Lake William Lake Roxton Missisquoi Bay Water source:a Not at risk From MBb At riskc Sex: Male Female Symptoms:d No Yes

GI1

GI2

RR

95% CI

RR

95% CI

1.00 2.19 1.27 2.26

1.15–4.16 0.47–3.41 1.26–4.06

1.00⁎ 1.52 2.71 3.28

0.65–3.51 1.02–7.16 1.69–6.37

1.00 2.37 1.44

1.16–4.82 0.63–3.31

1.00 1.70 0.73

0.65–4.44 0.21–2.52

1.00 3.42 1.75

1.42–8.21 0.95–3.24

1.00 2.67 3.00

0.79–8.96 1.31–6.87

1.00 2.75

1.67–4.53

1.00 3.33

1.66–6.67

1.00 4.15

2.38–7.22

1.00 6.73

2.52–18.00

⁎ Significant trend, p = 0.001. a Residence's water source. b 20 families received their tap water from a treatment plant using Missisquoi Bay water. c 18 families used a source for their tap water at risk of fecal contamination (surface well, 15; directly in the lake, 3). d Gastrointestinal symptoms reported in the two weeks prior to data collection.

low in comparison with the reference limit of 20 μg/L (Health Canada, 2012; WHO, 1999). However, other cyanotoxins could be involved in physiopathological processes. Moreover, a recent study suggested that the bacterial community associated with cyanobacteria, particularly Aeromonas strains, could be responsible for gastrointestinal symptoms declared following recreational exposure to cyanobacterial bloom (Berg et al., 2011). This warrants verification. 5. Conclusions The statistical relationships with different symptoms in families who receive their water from a plant equipped with a full treatment system whose source is highly contaminated by cyanobacteria support recommendations concerning the importance of implementing a risk management plan for communities affected by potential contamination of drinking water by cyanobacteria (WHO, 1999). As in the study by Dorevitch et al. (2012), the gastrointestinal effects documented in relation to limited contact indicate that this mode of contact may also expose the population to substantial risk. A RR of 3.28 for the GI2 index for exposure by full or limited contact to concentrations higher than 100,000 cells/mL is the same order of magnitude as that of 3.55 for exposure when swimming at over 180 E. coli/100 mL (Wiedenmann et al., 2006). While the mechanisms of action must be investigated, the population should be made aware of the risks of gastrointestinal symptoms associated with exposure to cyanobacteria through full or limited contact with water. Acknowledgments

unable, due to sample size, to document this relationship by type of contact. However, we verified the interaction between the type of contact and the exposure to cell counts with fewer categories (no exposure, b20,000 cells/mL, ≥ 20,000 cells/mL) and we observed an increase in the adjusted RR for the two types of contact (data not shown). To our knowledge, this is the first study to demonstrate such a relationship with gastrointestinal symptoms, which are most often reported anecdotally in the literature. This study differs from other studies on the subject owing to its prospective design; it was conducted over several weeks with residents of two lakes contaminated by cyanobacteria that were compared with residents of another lake much less contaminated. Contamination by E. coli was documented; it was relatively low and was not associated with gastrointestinal effects in residents who had contact with the water bodies. Even though data for E. coli were interpolated (see the third paragraph of the Material and methods section), we analyzed the data in two ways, by interpolation of the missing data, but more conservatively, by assigning a high risk index to any contact at a site where concentrations were higher than 100 CFU/100 mL at least once during the summer. It is unlikely that it could have confounded the relationship documented with exposure to cyanobacteria. A weakness of the study is the 32% participation rate, which was largely due to the substantial data collection requirements for participants. Nevertheless, a selection bias cannot be ruled out. As mentioned previously, if there is a selection bias, it would likely result in an underestimation of the RR in relation to contact. Finally, as for E coli, we had to interpolate missing data for cyanobacteria and microcystin for littoral stations, but more for limnetic stations. Moreover, we did not verify the variation in concentrations during the day. There could be a potential bias of misclassification of exposure concerning the relations between the exposure to different counts of cyanobacteria and gastrointestinal symptoms. However, here again, the potential bias should be non-differential, resulting in a lowering of the RR. In the present study, microcystins do not appear to be associated with symptoms and the highest concentration (7.65 μg/L) at which an episode of symptoms was documented following contact was relatively

Sincere thanks to Drs. Michel Savard, Benoit Barbeau and Éric Dewailly for their constructive comments concerning the research protocol, and to Dr. Patrick Poulin for the technical help in Fig. 1. Many thanks to all the people who participated in data collection and to the staff of the laboratories of the Department of Biological Sciences of the Université du Québec à Montréal and of the National Research Council Canada, Energy, Mining and Environment in Montreal who participated in the analysis of samples. Lastly, we cannot overlook the collaboration of municipal leaders, watershed and citizen committees and the residents and families who generously agreed to participate in the project. References Agence française de sécurité sanitaire des aliments et Agence française de sécurité sanitaire de l'environnement et du travail (AFSSA-AFSSET). Risques sanitaires liés à la présence de cyanobactéries dans l'eau. Paris, France: Agence française de sécurité sanitaire des aliments et Agence française de sécurité sanitaire de l'environnement et du travail; 2006. Backer LC, Carmichael W, Kirkpatrick B, Williams C, Irvin M, Zhou Y, et al. Recreational exposure to low concentrations of microcystins during an algal bloom in a small lake. Mar Drugs 2008;6:389–406. Berg KA, Lyra C, Niemi RM, Heens B, Hoppu K, Erkomaa K, et al. Virulence genes of Aeromonas isolates, bacterial endotoxins and cyanobacterial toxins from recreational water samples associated with human health symptoms. J Water Health 2011;9:670–9. Carmichael WW, Azevedo SM, An JS, Molica RJ, Jochimsen EN, Lau S, et al. Human fatalities from cyanobacteria: chemical and biological evidence for cyanotoxins. Environ Health Perspect 2001;109:663–8. Dale K, Wolfe R, Sinclair M, Hellard M, Leder K. Sporadic gastroenteritis and recreational swimming in a longitudinal community cohort study in Melbourne, Australia. Am J Epidemiol 2009;170:1469–77. Dillenberg HO, Dehnel MK. Toxic waterbloom in Saskatchewan, 1959. Can Med Assoc J 1960;83:1151–4. Dorevitch S, Pratap P, Wroblewski M, Hryhorczuk DO, Li H, Liu LC, et al. Health risks of limited-contact water recreation. Environ Health Perspect 2012;120:192–7. El Saadi O, Esterman AJ, Cameron S, Roder D. Murray River water, raised cyanobacterial cell counts, and gastrointestinal and dermatological symptoms. Med J Aust 1995;162: 122–5. Environment Canada. National Climate Canada and Information Archive. http://www.climate.weatheroffice.gc.ca/climateData/canada_e.html, 2013. [Accessed June 18]. Falconer IR, Humpage AR. Health risk assessment of cyanobacterial (blue-green algal) toxins in drinking water. Int J Environ Res Public Health 2005;2:43–50.

B. Lévesque et al. / Science of the Total Environment 466–467 (2014) 397–403 Hamilton PB, Ley LM, Dean S, Pick FR. The occurrence of the cyanobacterium Cylindrospermopsis raciborskii in Constance Lake: an exotic cyanoprokaryote new to Canada. Phycologia 2005;44:17–25. Hawkins PR, Runnegar MTC, Jackson ARB, Falconer IR. Severe hepatotoxicity caused by tropical cyanobacterium (blue-green algae) Cylindrospermopsis raciborskii (Woloszynska) Seenaya and Subba Raju isolated from a domestic water supply reservoir. Appl Environ Microbiol 1985;50:1292–5. Health Canada (HC). Guidelines for Canadian recreational water quality. 3rd ed. Ottawa, Ontario: Health Canada; 2012. Health Canada (HC). Blue-green algae (cyanobacteria) and their toxins. http://www.hc-sc.gc. ca/ewh-semt/pubs/water-eau/cyanobacter-eng.php, 2013. [Accessed March 25]. Humpage A. Toxin types, toxicokinetics and toxicodynamics. Adv Exp Med Biol 2008;619: 383–415. Kleinbaum DG, Kupper LL, Muller KE. Applied regression analysis and other multivariable methods. Boston, MA: PWS-Kent; 1988. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986;73:13–22. Lippy EC, Erb J. Gastrointestinal illness at Sewickley, PA. JAWWA 1976;68:606–10. Majowicz SE, Doré K, Flint JA, Edge VL, Read S, Buffett MC, et al. Magnitude and distribution of acute, self-reported gastrointestinal illness in a Canadian community. Epidemiol Infect 2004;132:607–17. Ministère de l'environnement du Québec (MENVIQ). Procédure pour la verification de la qualité des eaux de baignade. Québec, Qc: Ministère de l'Environnment du Québec; 1986. Ministère du Développement durable, de l'Environnement et des Parcs (MDDEP). Fiche: Cyanobactéries à potentiel toxique. Québec, QC: Ministère du Développement durable, de l'Environnement et des Parcs; 2008. Pablo J, Banack SA, Cox PA, Johnson TE, Papatrepopoulos S, Bradley WG, et al. Cyanobacterial neurotoxin BMAA in ALS and Alzheimer's disease. Acta Neurol Scand 2009;120:216–25. Pilotto LS, Douglas RM, Burch MD, Cameron S, Beers M, Rouch GJ, et al. Health effects of exposure to cyanobacteria (blue-green algae) during recreational water-related activities. Aust N Z J Public Health 1997;21:562–6. Robert C, Tremblay H, DeBlois C. Cyanobactéries et cyanotoxines au Québec: suivi à six stations de production d'eau potable (2001–2003). Québec, QC: Ministère du Développement durable, de l'Environnement et des Parcs; 2005. Rolland A, Bird DF, Giani A. Seasonal changes in composition of the cyanobacterial community and the occurrence of hepatotoxic blooms in the eastern townships, Québec, Canada. J Plankton Res 2005;27:683–94.

403

Stanghellini V. Three-month prevalence rates of gastrointestinal symptoms and the influence of demographic factors: results from the domestic/international gastroenterology surveillance study (DIGEST). Scand J Gastroenterol 1999;231:20–8. Stewart I, Webb PM, Schluter PJ, Fleming LE, Burns Jr JW, Gantar M, et al. Epidemiology of recreational exposure to freshwater cyanobacteria—an international prospective cohort study. BMC Public Health 2006a;6:93. Stewart I, Schluter PJ, Shaw GR. Cyanobacterial lipopolysaccharides and human health—a review. BMC Public Health 2006b;5:7. Tager IB, Speizer FE. Surveillance techniques for respiratory illness. Arch Environ Health 1976;31:29–32. Teixeira Mda G, Costa Mda C, de Carvalho VL, Pereira Mdos S, Hage E. Gastroenteritis epidemic in the area of the Itaparica Dam, Bahia, Brazil. Bull Pan Am Health Organ 1993;27:244–53. Turner PC, Gammie AJ, Hollinrake K, Codd GA. Pneumonia associated with cyanobacteria. Br Med J 1990;300:1440–1. Uebersax JS. Introduction to the tetrachoric and polychoric correlation coefficients. http://www.john-uebersax.com/stat/tetra.htm, 2013. [Accessed March 25]. US Environmental Protection Agency (US-EPA). Method 1604: total coliforms and Escherichia coli in water by membrane filtration using a simultaneous detection technique (MI medium). Washington, DC: USEPA; 2002. Veldee MV. An epidemiological study of suspected water-borne gastroenteritis. Am J Public Health 1931;21:1227–35. Wiedenmann A, Krüger P, Dietz K, Lopez-Pila JM, Szewzyk R, Botzenhart K. A randomized controlled trial assessing infectious disease risks from bathing in fresh recreational waters in relation to the concentration of Escherichia coli, intestinal enterococci, Clostridium perfringens, and somatic coliphages. Environ Health Perspect 2006;114: 228–36. Williamson M, Corbett S. Investigating health risks from riverine blooms of blue green algae. NSW Public Health Bull 1993;4:27–9. World Health Organization (WHO). In: Chorus I, Bartram J, editors. Toxic cyanobacteria in water: a guide to their public health consequences, monitoring and management. London, UK: World Health Organization; 1999. World Health Organization (WHO). Guidelines for water quality, Vol. 1, 3rd edition incorporating 1st and 2nd addenda. Geneva, Switzerland: World Health Organization; 2008. Zamayadi A, MacLeod SL, Fan Y, McQuaid N, Dorner S, Sauvé S, et al. Toxic cyanobacterial breakthrough and accumulation in a drinking water plant: a monitoring and treatment challenge. Water Res 2012;46:1511–23.