Harmful Algae 1 (2002) 157–168

Blue green algal (cyanobacterial) toxins, surface drinking water, and liver cancer in Florida Lora E. Fleming a,∗ , Carlos Rivero b , John Burns c , Chris Williams c,d , Judy A. Bean a , Kathleen A. Shea a , John Stinn a a

b

NIEHS Marine and Freshwater Biomedical Sciences Center, University of Miami Rosenstiel School of Marine and Atmospheric Studies, Miami, FL, USA GEOCORE Facility, University of Miami Rosenstiel School of Marine and Atmospheric Studies, Miami, FL, USA c CyanoLab, Palatka, FL, USA d St. Johns River Water Management District (SJRWMD), Palatka, FL, USA Accepted 20 June 2002

Abstract The blue green algae or cyanobacteria represent a diverse group of organisms that produce potent natural toxins. There have been case reports of severe morbidity and mortality in domestic animals through drinking water contaminated by these toxins. Microcystins, in particular, have been associated with acute liver damage and possibly liver cancer in laboratory animals. Although, there has been little epidemiologic research on toxin effects in humans, a study by Yu (1995) found an association between primary liver cancer and surface water. Surface water drinking supplies are particularly vulnerable to the growth of these organisms; current US drinking water treatment practices do not monitor or actively treat for blue green algal toxins including the microcystins. After a monitoring survey in Florida found organisms and microcystins (among other cyanobacterial toxins) in surface water drinking sources, a pilot ecological study was performed using a Geographic Information System (GIS) to evaluate the risk of primary hepatocellular carcinoma (HCC) and proximity to a surface water treatment plant at cancer diagnosis. The study linked all HCC cancers diagnosed in Florida from 1981 to 1998 with environmental databases. A significantly increased risk for HCC with residence within the service area of a surface water treatment plant was found compared to persons living in areas contiguous to the surface water treatment plants. However, this increased risk was not seen in comparison to persons living in randomly selected ground water treatment service areas or compared to the Florida cumulative incidence rate for the study period, using various comparison and GIS methodologies. Furthermore, these findings must be interpreted in light of significant issues of latency, high population mobility, and the lack of individual exposure information. Nevertheless, the issue of acute and chronic human health effects associated with the consumption of surface waters possibly contaminated by blue green algal toxins merits further investigation. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Hepatocellular carcinoma (HCC); Ecological studies; Geographic information systems (GIS); Cyanobacteria; Blue green algal toxins; Microcystins; Surface water; Drinking water

∗ Corresponding author. Present address: Department of Epidemiology and Public Health, University of Miami School of Medicine, 1801 NW 9th Avenue, Suite 200, Highland Park Building, Miami, FL, 33136, USA. Tel.: +1-305-243-5912; fax: +1-305-243-3384. E-mail address: [email protected] (L.E. Fleming).

1568-9883/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 1 5 6 8 - 9 8 8 3 ( 0 2 ) 0 0 0 2 6 - 4

158

L.E. Fleming et al. / Harmful Algae 1 (2002) 157–168

1. Introduction Primary hepatocellular carcinoma (HCC), a malignant epithelial tumor, is the most prevalent type of liver cancer in the world. It is one of the three leading causes of cancer mortality, accounting for 250,000–1,000,000 deaths annually world wide, with a ratio of mortality to incidence of 0.98:1 (London and McGlynn, 1996; Pisani et al., 1999). HCC is particularly a problem in the developing countries of the world where 81% of the world’s cases are found (Parkin et al., 1999a; Parkin et al., 1999b). For individuals living in developed countries versus developing countries, the age-standardized incidence rates are 7.6/100,000 and 17.9/100,000 for males, and 2.6/100,000 and 6.2/100,000 for females, respectively (Parkin et al., 1999a). Although, the incidence of HCC is relatively low in the US, it has been increasing (Martin, 1998; El-Serag and Mason, 1999). In Florida, since 1981–1998, there was a significant increase in the average annual HCC incidence rate (Shea et al., 2001)). In particular, the average Florida annual HCC incidence rates among male and female hispanics and blacks have been consistently twice the rate of white males and females as standardized rate ratios (SRRs; Shea et al., 2001). During this time period, the incidence rate in hispanic males (3.29/100,000) approached twice the rate reported in white males (1.82/100,000), while black males (3.86/100,000) had greater than twice the rate of white males; in females, the rates in both hispanic and black females (1.23/100,000 and 1.18/100,000, respectively) were approximately twice those of white females (0.6/100,000). The cyanobacteria or blue green algae are an ancient and ubiquitous family of organisms, many with photosynthetic abilities (Chorus and Bartram, 1999; Carmichael, 1994; Falconer, 1989; NHMRC, 1994). The cyanobacteria frequently are found growing in marine, brackish and fresh waters, including freshwater surface water drinking sources, such as lakes and drinking water reservoirs. Similar to marine algal blooms, such as red tides, cyanobacteria periodically exhibit significantly increased reproductive rates and total population biomass known as a “bloom.” The reasons for these cyanobacteria blooms are not completely understood, but most cases are related to the addition of nutrients via runoff (Philipp et al., 1991;

Carmichael and Falconer, 1993; Rapala et al., 1997). Large bloom events are classified as “harmful algal blooms” if they cause negative environmental impacts such as: (1) organismal mortality (fish, crabs, clams, etc.) due to a significant decrease in the oxygen content in the water via bloom respiration or degradation; (2) loss of submerged aquatic vegetation due to an increased absorption of sunlight that reduces light penetration necessary for maintenance/growth; (3) a decrease in ecosystem stability by interfering in foodweb dynamics by displacing normal phytoplankton species; and (4) the production of highly active natural compounds (cyanotoxins) that are known to be toxic or allergenic in nature (Chorus and Bartram, 1999; NHMRC, 1994; Carmichael and Falconer, 1993). There is a wide spectrum of cyanotoxins (blue green algal toxins) that predominantly affect the nervous, hepatic and dermatologic systems (i.e. neurotoxic, hepatotoxic, and dermatotoxic). The hepatotoxins are cyclic peptides, predominantly microcystins, nodularins, and cylindrospermopsin. Of note, these toxins are particularly toxic to the liver in part due to selective transport mechanisms that concentrate these toxins from the gut and blood into liver cells; they damage the liver by altering the cytoskeletal architecture of the hepatocytes (Chorus and Bartram, 1999; Carmichael, 1994; NHMRC, 1994; Elder et al., 1993; Humpage and Falconer, 1999; Ohtani et al., 1992; MacKintosh et al., 1990; Repavich et al., 1990). In experimental animals, Yu (1995) and others (Ito et al., 1997; Ueno et al., 1996) have shown that microcystins are promoters, with a possible synergistic effect between microcystins and alflatoxins, for HCC. Yu (1995) and others (Yu et al., 1989a,b; Junshi et al., 1990) have studied the possible relationship between the consumption of surface water (pond, ditch, river versus well water or deep well) and an increased risk for primary hepatic cancer in China. China has an extremely high rate of primary liver cancer, previously associated with hepatitis B and aflatoxin exposures (Yu, 1995; Chorus and Bartram, 1999). However, large epidemiologic studies found not only a significantly increased risk of primary liver cancer in areas of high surface water consumption (standardized incidence ratio, SIR = 2.6) compared with areas of non-surface water consumption (SIR = 0.34), but also a strong dose response relationship. Reportedly, changing from pond/ditch to deep well (at least 200 m) water lead to a

L.E. Fleming et al. / Harmful Algae 1 (2002) 157–168

subsequent decrease in the mortality rate from primary hepatic cancer in one Chinese province; in another province where there was no change in drinking water source, the liver cancer mortality rates continued to increase during the same time period. Monitoring studies using a sensitive enzyme linked immunosorbent assay (ELISA) for microcystins revealed detectable levels of these hepatotoxins, as well as the presence of blue green algae, in the surface waters as opposed to other drinking water sources (Ueno et al., 1996; Falconer and Humpage, 1996). Recent data suggest that another hepatotoxic cyanobacteria, cylindrospermopsin, may also be associated with tumorigenesis (Falconer and Humpage, 2001, Humpage et al., 2000). Although, the World Health Organization, (WHO, 1988) adopted a provisional guideline for microcystinLR in drinking water of 1.0 ␮g/l, blue green algae and their toxins are rarely monitored in US surface water drinking supplies. Furthermore, the removal of the organisms and their toxins is difficult and expensive; the use of activated carbon treatment during active blooms will decrease, but not necessarily eliminate, levels of cyanobacterial toxins in drinking water (Chorus and Bartram, 1999). At the present time, the Florida aquifer is the major source for drinking water in the state, however, with Florida’s ever increasing population in conjunction with the limitations of the Florida aquifer, the pressure to find alternate sources of high quality drinking water is increasing. Currently, over 10% of drinking water in Florida is from surface water sources with projected significant increases. Large freshwater surface water bodies are the most easily accessible and can provide substantial volumes of water, however, these water bodies are breeding grounds for toxic cyanobacteria, including the hepatotoxic microcystins and cylindrospermopsins. Recent monitoring studies in Florida (Williams et al., 2001) of recreational and surface water drinking supplies, found that 87/167 (54%) samples, from 75 individual water bodies, contained significant levels of potentially toxic blue green algae, including species that are known to produce microcystins. Ninety percent of these samples were positively identified as containing blue green algal toxins, with 80% lethality in mice. Among the 18 surface waters serving as water sources for drinking, 12 of the surface water drinking sources tested to date have been found to have toxic blue greens producing microcystins in samples prior

159

to drinking water treatment. Furthermore, samples taken after regular water treatment were found to contain microcystins in the post-treatment drinking water at 7 of the 16 plants, at levels greater than the WHO recommended provisional guideline. In summary, short term and long term health effects have not been thoroughly evaluated in persons with exposure to blue green algae and their toxins (NHMRC, 1994). Current drinking water treatment practices in the US do not regularly monitor or actively remove these toxins from the drinking water since this is a relatively new field of study and would involve extremely expensive measures (Falconer, 1989; Volterra, 1993; Falconer, 1999; Heinze, 1999). Even with treatment, low level chronic exposure to the carcinogenic hepatotoxins, such as the microcystins and possibly cylindrospermopsin, is possible in persons consuming drinking water derived from surface water treatment plants in Florida and other parts of the US. Therefore, a study was performed to evaluate the risk of HCC in persons living within service areas of surface water treatment plants in Florida. 2. Methods This pilot study was an ecological study linking incident cases of HCC diagnosed Florida from 1981 to 1998 by their place of residence at the time of cancer diagnosis with environmental and geographic data on drinking water plants and sources in a geographic information systems (GIS) model. 2.1. Human health data The Human Subjects Protocol was constructed and approved by the University of Miami School of Medicine Human Subjects Committee (Protocol #99/054) and by the Florida Department of Health. The study population was defined as all cases of HCC diagnosed in the state of Florida between 1981 and 1998 in Florida’s incident cancer registry, the Florida Cancer Data System (FCDS) database. Because FCDS has been independently estimated to capture more than 95% of all incident cancer cases in Florida, this project was considered to be population based. HCCs were defined as those cancer cases with the appropriate SEER ICD-0 diagnostic codes, as described below.

160

L.E. Fleming et al. / Harmful Algae 1 (2002) 157–168

Cancer incidence data are submitted to FCDS from all hospitals, laboratories, ambulatory surgical centers, and radiation therapy centers in Florida. These data are subjected to verification and validation procedures before entry into the FCDS database. For this study, HCC cases were defined as those cancer cases with the appropriate SEER ICD-0-2 diagnostic codes (8170, 8171, and 8180). The data set analyzed in this project was extracted using SQL queries of the FCDS database. All of these data were incorporated into the GIS database in tabular and ArcView shapefile formats. 2.2. Exposure data From Florida Department of Environmental Protection (DEP) and other websites, geo-referenced information (GPS) was acquired for the deep ground water wells, as well as the administrative addresses for their respective treatment plants. However, apparently due to data security issues, it was not possible to acquire the actual location of the deep ground water treatment plants in geographic coordinates, despite considerable efforts. Geocoded addresses of the actual surface water treatment plants, as well as accurate hardcopy maps of the water distribution service areas from the 18 major surface water treatment plants in Florida, were provided by St. Johns River Water Management District (SJRWMD). These maps were digitized using the UTM coordinate system (NAD 83) within ArcView GIS. These data formed the base data layers within the GIS database. 2.3. GIS methods The geocoded data (residence at time of diagnosis) of all cases of HCC from the FCDS database included the following variables: age; date of birth; gender; and race-ethnic information. In addition, racial, ethnic, and socioeconomic data from 1990 were acquired from the US Census Bureau. This information was mapped above the base layers described in the previous section. To compare surface water treatment service areas to ground water treatment service areas, the following GIS analyses were performed. The geographic center of the actual surface water distribution service area was used as the hypothetical location of the surface

water treatment plant in lieu of the actual location of the plant due to the significant distance between the treatment plant and its corresponding service area. The immediate service area surrounding the actual location of the surface water treatment plants was not an accurate representation of the population served by such plants. In other words, the treatment plants were far removed from their actual service areas (perhaps to avoid potential anthropogenic influence upon the surface water source). The average size plus two standard deviations of the surface water treatment service areas was determined and used to identify the standard area of service for the initial analyses. A circular buffer with an area equal to the standard area of service was created and applied to each hypothetical surface water treatment facility (Fig. 1). A hypothetical center for each ground water treatment plant was determined based on the average location of its corresponding wells. This was due to the absence of positional information for the groundwater treatment plants. This calculation was based on the assumption that the ground water wells can be located within their service area due to a minimal risk of anthropogenic influence (as opposed to surface water treatment plants). In addition, the service area boundaries for the treatment facilities were also unknown. Therefore, the same circular buffer used for the surface water treatment facilities was applied to each ground water treatment plant (Fig. 2). The geographic intersections between the 1990 census block group data and the standard area of service for each ground and surface water facility were determined. The demographic profile and population parameters for each service area were calculated from this intersection and used as approximate denominator data. In an attempt to control for confounding bias, 4 sets of 18 ground water treatment service areas (controls) were randomly selected from groups matching the following characteristics (Fig. 3): 1. random; 2. similar median income/rent to surface water treatment service areas; 3. similar race-ethnic distribution to surface water treatment service areas; 4. similar median income/rent and race-ethnic distribution to surface water treatment service areas.

L.E. Fleming et al. / Harmful Algae 1 (2002) 157–168

161

Fig. 1. Example of surface water treatment plant locations and actual service area boundaries with study surface water treatment plants and their service areas.

To compare populations contiguous to the surface water treatment service areas, a contiguous buffer analysis was performed (Fig. 4). The actual service area of each surface water treatment facility was used

to delineate the exposure domain. An unique contiguous actual buffer area was determined for each surface water treatment facility and applied within the geospatial information system to create contiguous

Fig. 2. Example of actual ground water wells with study ground water treatment plant and study service area.

162

L.E. Fleming et al. / Harmful Algae 1 (2002) 157–168

Fig. 3. Sites of 18 study surface water treatment service areas and 4 sets of 18 randomly selected matched control ground water treatment service areas.

Fig. 4. Example of actual surface water treatment service areas and their control contiguous buffer areas.

L.E. Fleming et al. / Harmful Algae 1 (2002) 157–168

buffer areas as control areas, approximating the actual surface water treatment service area in shape and area. For these analyses, it was assumed that individuals living within the surface water treatment service areas would be exposed to “contaminated” (i.e. with cyanobacterial toxins) water significantly more than those individuals living just outside of the water treatment service area. Since the service areas were relatively small as noted above, it was assumed that the areas immediately surrounding the water treatment service areas would be similar in demographic characteristics and lifestyle patterns. 2.4. Statistical methods Throughout the analyses, cumulative incidence rates for the study time period from 1981 to 1998 were calculated to allow for comparisons between the different geographic spaces modeled in the GIS. The HCC cases from 1981 through 1998 were spatially joined to their corresponding service area; initial cumulative incidence rates (and confidence intervals) were calculated for each of the surface water treatment service areas and for the randomly selected ground water treatment plants for the entire relevant population. In addition, the initial pooled hepatocellular cumulative cancer incidence rates for all study groups (surface and ground water treatment service areas) were also calculated. A similar process was repeated for the two contiguous buffered area analyses. Results were expressed as age-standardized incidence rates using the 1970 US standard million population and Florida population projections. Age-adjusted rates were calculated based on the 1990 census age groups. The standard errors for the age-adjusted rates were calculated after the method of Breslow and Day (Breslow and Day, 1982; Howe et al., 1996). For the individual cumulative incidence rates, the Mann–Whitney Rank Sum Test was used to compare the surface water treatment service areas to the different comparison groups of the ground water treatment service areas and to the contiguous buffered areas (Glantz, 1987). The incidence rates for the pooled Surface water treatment service areas were compared to the rates for the pooled ground water treatment service areas and the pooled contiguous buffered areas,

163

and expressed as SRRs with 95% confidence intervals (Breslow and Day, 1982; Howe et al., 1996). These SRRs can express the risk of a particular disease since SRRs compare the rates of the disease for one population relative to the rates of another population. All data were analyzed using proprietary programs of the FCDS, Microsoft Excel 97 software, and ESRI ArcView GIS Version 3.2 software.

3. Results There were 4741 incident cases of HCC from 1981 through 1998 in Florida, with an age-adjusted cumulative incidence rate for HCC of 1.41/100,000 (S.E. 0.0003). There were 18 surface water treatment service areas (see Fig. 3). 3.1. Surface versus ground water analyses There were 6614 ground water treatment service areas. For the first set of analyses comparing the surface water treatment service areas to the control ground water treatment service areas, there were a total of 461 cases of HCC residing at the time of cancer diagnosis in the 18 surface water treatment service areas combined. The individual surface water treatment service area cases ranged from 0 cases of HCC in three service areas to 113 cases, and the individual cumulative age-adjusted cancer rates for HCC ranged from 0 to 3.08/100,000 (S.E. 0.0188; the latter rate based on only five cases of cancer). There were 1249 total GIS-generated ground water treatment service areas after the GIS manipulations described above. Control 1 was simply a random sample of 18 service areas from a pool of 1249 ground water treatment service areas. Control 2 was a random sample of 18 service areas from a pool of 1176 ground water treatment service areas selected based on median income (US$ 14,750.36–132,934.8) and median monthly rent (US$ 241.58–619.68). Control 3 was a random sample of 18 service areas from a pool of 1246 ground water treatment service areas selected based on percent black (0–66.39%) and percent hispanic (0–43.68%). Control 4 was a random sample of 18 service areas from a pool of 1174 ground water treatment service areas selected based on the

164

L.E. Fleming et al. / Harmful Algae 1 (2002) 157–168

requirements for both Control 2 and 3. There was potential overlap between the different control groups. The Mann–Whitney Rank Sum Test as applied to the individual incidence rates for the 18 surface water treatment service areas in comparison with the incidence rates of the four different groups of randomly selected matched control ground water treatment service areas. There were no significant differences (at P < 0.05) between the incidence rates by the Rank Test. Therefore, in this analysis, the risk of having HCC was not statistically different between persons living in the surface water treatment service areas and persons living in the GIS-generated control ground water treatment service areas. The pooled surface water treatment service area cumulative age-adjusted cancer rate for HCC was 1.13 (S.E. 0.00075) for 1981–1998; the cumulative age-adjusted cancer rate of HCC for all of Florida for the same time period was greater at 1.41/100,000. The SRR of HCC for pooled surface water treatment service areas compared to the four different pooled control ground water treatment service areas ranged from 0.80 to 0.98 with statistically significant 95% confidence intervals (Table 1). The SRRs of HCC for pooled surface water treatment service areas compared to the general Florida population was 0.80 with statistically significant 95% confidence intervals (Table 1). Therefore, in this analysis, the risk of having HCC was statistically significantly less for all persons living in the surface water treatment service areas compared with all persons living in the four groups of GIS-generated control ground water treatment service areas and compared to the general Florida population.

3.2. Contiguous buffer analyses As stated above, for both the actual buffer analysis, it was assumed that the areas immediately surrounding the actual surface water treatment service areas would be similar in demographic characteristics and lifestyle patterns. Using 1990 census data, the unexposed contiguous buffer area were found to contain comparison populations with similar race ethnic and socioeconomic backgrounds (data not shown) to those within the actual surface water treatment service area. For the analyses comparing the actual surface water treatment service areas to their control contiguous buffered areas, there were a total of 335 cases of HCC residing at the time of cancer diagnosis in the 18 actual surface water treatment service areas combined. The individual surface water treatment service area cases ranged from 0 cases of HCC in four service areas to 103 cases, and the individual cumulative age-adjusted cancer rates for HCC ranged from 0 to 4.88/100,000 (the latter rate based on only three cases of cancer). For the actual service areas with their buffers, the Mann–Whitney Rank Sum Test as applied to the individual incidence rates for the 18 actual surface water treatment service areas in comparison with the incidence rates of the control contiguous buffer group. There was a significant difference (i.e. P < 0.02) between the incidence rates by the Rank Test. Therefore, in this analysis, the risk of having HCC was significantly higher for persons living in the actual surface water treatment service areas compared to persons living in control contiguous presumably unexposed buffer areas.

Table 1 Pooled age-adjusted incidence rates and standardized rate ratios (SRR) of HCC for pooled study surface water treatment service areas, pooled control ground water treatment service areas and the Florida population Age-adjusted rate

Surface Control Control Control Control a b

water 1 2 3 4

1.13 1.19 1.41 1.39 1.15

S.E.

0.00075 0.00099 0.00076 0.00066 0.00098

SRR Rates compared to control ground water service areasa

Rates compared to Florida populationb

0.95 (0.94–0.96) 0.80 (0.80–0.01) 0.81(0.80–0.82) 0.98 (0.98–0.00)

0.80 0.84 1.00 0.99 0.82

95% Confidence interval. State of Florida age-adjusted cumulative cancer rate (1981–1998) = 1.41/100,000 (S.E. 0.0003).

(0.79–0.81) (0.84–0.86) (0.99–1.00 (0.98–0.99 (0.81–0.83)

L.E. Fleming et al. / Harmful Algae 1 (2002) 157–168

165

Table 2 Pooled age-adjusted incidence rates and standardized rate ratios (SRR) of HCC for pooled surface water treatment service areas, pooled control contiguous buffered service areas, and the Florida population Age-adjusted rate

Actual surface water area Contiguous buffer control a b

1.15 0.83

S.E.

0.0009 0.0011

SRR Rates compared to control ground water service areasa

Rates compared to Florida populationb

1.39 (1.38–1.40)

0.82 (0.81–0.83) 0.59 (0.58–0.60)

95% Confidence interval. State of Florida age-adjusted cumulative cancer rate (1981–1998) = 1.41/100,000 (S.E. 0.0003).

The pooled actual surface water treatment service area cumulative age-adjusted cancer rate for HCC was 1.15 (S.E. 0.0009) for 1981–1998; the cumulative age-adjusted cancer rate of HCC for all of Florida for the same time period was greater at 1.41/100,000. The SRRs of HCC for pooled actual surface water treatment service areas compared to the pooled control contiguous buffer area was 1.39 with statistically significant 95% confidence intervals (Table 2). The SRRs of HCC for pooled actual surface water treatment service areas compared to the general Florida population was 0.82 with statistically significant 95% confidence intervals (Table 2). Therefore, in this analysis, the risk of having HCC was statistically significantly greater for all persons living in the surface water treatment service areas compared with all persons living in the contiguous buffered areas, but statistically significantly less compared to the general Florida population.

4. Discussion In this pilot ecological study using a GIS analysis, the risk of HCC in Florida was significantly associated with residence at the time of diagnosis in a surface water treatment service area when comparing actual service areas to GIS-generated control contiguous buffer areas. However, additional analytical comparisons with the general Florida population and with randomly selected GIS-generated ground water did not find an elevated association between risk of HCC and presumed exposure to surface water. The majority of the individual cumulative incidence rates and the cumulative pooled incidence rate for HCC for

surface water treatment service areas were less than the cumulative incidence rate for HCC of all of the selected ground water treatment service area comparison groups and for the Florida population as a whole during the same time period. 4.1. Study limitations This study was an ecological study of the possible association between HCC and the location of the service areas of surface water treatment plants. As an ecological study, this study can only be considered to be hypothesis-generating; it can not prove or disprove an etiological association. The Investigators note that the limitations of this study described below must be considered in its interpretation regardless of the results. With regards to exposure, this study assumed that the place of residence at the time of cancer diagnosis was the same place of residence at the time of cancer initiation; this was a major assumption given the significant mobility of the Florida population and the potential 15–25 year latency from exposure to disease diagnosis of HCC. Furthermore, the use of the exposure measure of proximity of residence at the time of cancer diagnosis to the service areas of drinking water treatment plants was only a surrogate for exposure. It is possible that persons could have lived within the service area of a drinking water treatment plant and not necessarily utilized the treatment plant as their major source of drinking water (e.g. individual well water and bottled water). Furthermore, it is not known if treated surface water contaminated with blue green algal toxins actually reached the tap consumer, nor if it was a constant or, more likely, an

166

L.E. Fleming et al. / Harmful Algae 1 (2002) 157–168

intermittent exposure. In addition, due to the lack of geocoded data for the ground water treatment plants, assumptions were made in the GIS modeling with regards to the hypothesized location of the ground water treatment facilities and their service areas. In future analyses, it would be important to obtain and incorporate the correct geographic position of the ground water treatment plants and their service areas. With regards to disease and population, the statistical analyses were limited by the relatively small number of cases for each individual treatment service area despite the long time period. The use of the 1990 census data for the population denominator assumed relative stability of the population over the 18 year period from 1981 to 1998. Furthermore, the use of 1990 census data even on the census block level did not completely eliminate possible confounding from race/ethnic or socioeconomic class or from other confounding variables. In addition to those limitations already mentioned, with regards to the comparison of GIS-generated surface and ground water treatment service areas, there were a number of assumptions and possible biases. The use of a large radius could have lead to misclassification of cancer cases to include unexposed as well as exposed cases for the surface water treatment service areas, biasing the risk towards the null. This was not as important an issue for the GIS-generated ground water treatment service areas since they were selected not to overlap with the surface water treatment service areas. Finally with regards to the contiguous buffer analyses, in addition to small numbers of cases, there were several limitations (as well as assumptions). It was assumed that populations living contiguous to, but outside of, surface water treatment service areas would not be exposed to surface water, but would be similar in terms of confounding variables. Examination of the actual control contiguous buffer areas using 1990 census data showed similar distributions of race ethnicity and median income/rent between the exposed and contiguous buffer surface water treatment service areas. It was difficult using GIS to create an area equal to the actual surface water treatment service area. However, the use of rates rather than cancer cases for comparison purposes should diminish this bias.

4.2. Recommendations This pilot study did show an association between possible exposure to surface water and the risk of HCC in Florida, although not in comparison to ground water populations or the general Florida population. Although, only a hypothesis-generating study, the exposures and possible health effects (both acute and chronic) of the cyanobacteria and their toxins need to be evaluated more extensively in Florida, particularly given the results of the St. John River Water Management Cyanobacterial Survey (Williams et al., 2001). Of note, a similar GIS study of the risk of colorectal cancer and exposure to surface water in Florida did not find any significant associations, despite a much larger sample size of cancer cases (Fleming et al., 2001). Future studies should focus on establishing the actual exposure at the tap of consumers by sampling of tap water supplied by a surface water treatment plant during a toxic bloom. In addition, more acute symptoms and health effects (such as liver enzyme elevations) should be evaluated in those persons with established exposure, and this cohort of exposed and possibly effected persons should be followed over time to evaluate exposure and chronic health effects. Other confounding causes of both liver damage and HCC should be evaluated, such as ethanol, solvent exposure, aflatoxins, and viral hepatitis B and C. With regards to the specific issue of HCC, although a cancer associated with a high mortality, a case control study of persons with HCC compared to control persons without this cancer to evaluate their lifetime history of drinking, occupational and recreational surface water exposure (as well as confounding information such as viral hepatitis and alcohol consumption) would be of interest, particularly among persons who have lived in potentially high risk areas such as Florida. Evaluation of tumor tissue and other body fluids from persons with HCC for microcystin and cylindrospermopsin exposures, as well as confounders such as aflatoxins and viral hepatitis, could also be performed (Fleming and Stephan, 2001). Finally, in addition to exposure monitoring and health surveillance for the blue green algae and their toxins, education and prevention of both the exposure and potential health effects needs to be a primary focus of Florida’s scientific and public health

L.E. Fleming et al. / Harmful Algae 1 (2002) 157–168

community involved in the study and prevention of the harmful algal blooms.

Acknowledgements The authors would like to acknowledge the funding of the Florida Harmful Algal Bloom Taskforce, as well as the support of the Florida Cancer Data System (FCDS), the National Institute of Environmental Health Sciences (NIEHS) Marine and Freshwater Biomedical Sciences Center at the University of Miami, and NIEHS. The authors received scientific support for this study from Alan Rowan DrPH MS and Steven Wiersma MD MPH, as well as members of the Harmful Algal Bloom Taskforce and the St. Johns River Water Management District. Ms. Gayl van de Bogart and Ms. Jill Tincher provided significant administrative support. These data were presented preliminarily at the 9 May 2000 Meeting of the HAB Taskforce at the Florida Marine Research Institute in St. Petersburg, Florida. In addition, these data were presented at the Annual Meeting of the North American Association of Central Cancer Registries (New Orleans, April 2000), the Annual Florida Epidemiology Meeting (Gainesville, FL in August 2000), the Annual Florida Cancer Data System Meeting (Melbourne, FL in August 2000), the Annual National Institute of Environmental Health Sciences Director’s Meeting (Detroit, MI in October 2000), NALMS Conferences (Miami, FL in November 2000 and Tallahassee, FL in May 2001), and the ISEA Annual Meeting (Charleston, SC in November 2001). References Breslow, N.E., Day, N.E., 1982. Statistical Methods in Cancer Research, Vol. 2. IARC Statistical Publications, Lyon, France, p. 59. Carmichael, W.W., 1994. The toxins of cyanobacteria. Scientific Am. 78–86. Carmichael, W.W., Falconer, I.R., 1993. Diseases related to freshwater blue green algal toxins, and control measures. In: Falconer, I.R. (Ed.) Algal Toxins in Seafood and Drinking Water. Academic Press, New York, pp. 187–209. Chorus, I., Bartram, J. (Eds.), 1999. Toxic Cyanobacteria in Water: A Guide to their Public Health Consequences, Monitoring and Management. E & FN Spon, London. Elder, G.H., Hunter, P.R., Codd, G.A., 1993. Hazardous freshwater cyanobacteria (blue green algae). Lancet 341, 1519–1520.

167

El-Serag, H.B., Mason, A.C., 1999. Rising incidence of hepatocellular carcinoma in the United States. N. Engl. J. Med. 340, 745–750. Falconer, I.R., 1989. Effects on human health of some toxic cyanobacteria (blue green algae) in reservoirs, lakes and rivers. Tox Assessment 4, 175–184. Falconer, I.R., 1999. An overview of problems caused by toxic blue green algae (cyanobacteria) in drinking and recreational water. Environ. Toxicol. 14, 5–12. Falconer, I.R., Humpage, A.R., 1996. Tumor promotion by cyanobacterial toxins. Phycologia 35 (6), 74–79. Falconer, I.R., Humpage, A.R., 2001. Preliminary evidence for in vivo tumor initiation by oral administration of extracts of the blue green alga cylindrospermosis raciborskii containing the toxin cylindrospermopsin. Environ. Toxicol. 16 (2), 192– 195. Fleming, L.E., Stephan, W., 2001. Blue green Algae, Their Toxins and Public Health Issues. The Florida Harmful Algal Bloom Taskforce, Florida St. Petersburg, Florida. Fleming, L.E., Rivero, C., Burns, J., Williams, C., 2001. Final Report: Blue green Algal Exposure, Drinking Water and Colorectal Cancer. The Florida Harmful Algal Bloom Taskforce, Florida St. Petersburg, Florida. Glantz, S.A., 1987. Primer of Biostatistics. McGraw-Hill, New York. Heinze, R., 1999. Toxicity of the cyanobacterial toxin microcystin-LR to rats after 28 days intake with the drinking water. Environ. Toxicol. 14, 57–60. Howe, H.L., Lehnherr, M., Derrick, L. (Eds.), 1996. Cancer incidence in North America, 1988–1992. Sacramento, CA: North American Association of Central Cancer Registries. Humpage, A.R., Falconer, I.R., 1999. Microcystin-LR and liver tumor promotion: effects on cytokinesis, ploidy, and apoptosis in cultured hepatocytes. Environ. Toxicol. 14, 61–75. Humpage, A.R., Fenech, M., Thomas, P., Falconer, I.R., 2000. Micronulceus induction and chromosome loss in transformed human white cells indicate clastogenic and aneugenic action of the cyanobacterial toxin, cylindrospermopsin. Mutat. Res. 472, 155–161. Ito, E., Kondo, F., Terao, K., Harada, K.-I., 1997. Neoplastic nodular formation in mouse liver induced by repeated intraperitoneal injections of microcystin LR. Toxicon 35 (9), 1453–1457. Junshi, C., Campbell, T.C., Junyao, L., Peto, R., 1990. Diet, Life Style and Mortality in China. Oxford University Press, Oxford, England. London, W.T., McGlynn, K.A., 1996. Liver cancer. In: Schottenfeld, D., Fraumeni, J.F. Cancer Epidemiology and Prevention, 5th Edition. Oxford University Press, New York, pp. 772–793. MacKintosh, C., Beattie, K.A., Klumpp, S., Cohen, P., Codd, G.A., 1990. Cyanobacterial microcystin LR is a potent and specific inhibitor of protein phosphatases 1 and 2A from both mammals and higher plants. Fed. Euro. Biochem. Soc. Lett. 264 (2), 187–192. Martin, P., 1998. Hepatocellular carcinoma: risk factors and natural history. Liver Transpl. Surg. 4 (1/5), S87–91.

168

L.E. Fleming et al. / Harmful Algae 1 (2002) 157–168

National Health and Medical Research Council (NHMRC). 1994. Health Effects of Toxic Cyanobacteria (Blue Green Algae). Australian Government Publishing Service, Canberra, Australia. Ohtani, L., Moore, R.E., Runnegar, M.T.C., 1992. Cylindrospermopsin: a potent hepatotoxin from the blue green alga Cylindrospermopsis raciborskii. J. Am. Chem. Soc. 114 (20), 7941–7942. Parkin, D.M., Pisani, P., Ferlay, J., 1999a. Estimates of the worldwide incidence of 25 major cancers in 1990. Int. J. Cancer 80, 827–841. Parkin, D.M., Pisani, P., Munoz, N., Ferlay, J., 1999b. The global health burden of infection associated cancers. Cancer Surv. 33, 5–33. Philipp, R., Rowland, M.G.M., Baxter, P.J., McKenzie, C., Bell, R.H., 1991. Health risks from exposure to algae. CDR (London: England Rev.) 1, 67–68. Pisani, P., Parkin, D.M., Bray, F., Ferlay, J., 1999. Estimates of the worldwide mortality from 25 major cancers in 1990. Int J. Cancer 83, 18–29. Rapala, J., Sivonen, K., Lyra, C., Niemela, S.I., 1997. Variation of microcystin, cyanobacterial hepatotoxins, in Anabaena spp. as a function of growth stimulation. App. Environ. Microbiol. 63, 2206–2212. Repavich, W., Sonzogni, W.C., Standridge, J.H., Wedepohl, R.E., Meisner, L.E., 1990. Cyanobacteria (blue green algae) in Wisconsin waters: acute and chronic toxicity. Water Res. 24 (2), 225–231.

Shea, K., Fleming, L.E., Wohler Torres, B., Mackinnon, J., Voti, L., 2001. Primary liver cancer in Florida. Cancer 91 (5), 1046– 1051. Ueno, Y., Nagat, S., Tsutsumi, T., et al., 1996. Detection of microcystins, a blue green algal hepatotoxin, in drinking water sampled in Hiamen and Fusui, endemic areas of primary liver cancer in China, by highly sensitive immunoassay. Carcinogen 17, 1317–1321. Volterra, L., 1993. Sanitary implications associated with the use of eutrophic freshwater. Ann. Inst. Superiore di Sanita 29, 327– 333. Williams, C.D., Burns, J., Chapman, A., Flewelling, L., Pawlowicz, M., Carmichael, W., 2001. Assessment of Cyanotoxins. In: Florida’s Lakes, Reservoirs, and Rivers. Cyanobacteria Survey Project, Harmful Algal Bloom Task Force, SJRWMD, Palatka, FL. World Health Organization (WHO), 1988. Guidelines for Drinking Water Quality. WHO, Geneva. Yu, S.-J., 1995. Primary prevention of hepatocellular carcinoma. J. Gastroenterol. Hepatol. 10, 674–682. Yu, S.-Z., Chen, Z.-Q., Liu, Y.-K., Huang, Z.-Y., Zhao, Y.-F,. 1989a. The aflatoxins and contaminated water in the etiological study of primary liver cancer. In: Natori, S., Hashimoto, K., Ueno, Y. (Eds.), Mycotoxins & Phycotoxins #88. Elsevier, Amsterdam, pp. 37–44. Yu, S.-Z., 1989b. Drinking water and primary liver cancer. In: Tang, Z.-Y., Wu, M.-C., Xia, S.-S. (Eds.), Primary Liver Cancer. Springer, Berlin, pp. 30–37.