Cryptosporidiosis: An Outbreak Associated with Drinking Water Despite State-of-the-Art Water Treatment

Methods

Case Ascertainment and Confirmation

All testing of specimens from county residents for Cryptosporidium is done by two laboratories in Clark County and one at the Nevada state health department; testing is done only at the request of a physician. We reviewed the records from laboratory A, which does more than 95% of this testing, and the surveillance records at the county health department to identify all cases of Cryptosporidium infection newly diagnosed during the 4-month study period (1 January through 30 April 1994). We also reviewed the death certificates of persons who had had laboratory-confirmed cryptosporidiosis to determine whether cryptosporidiosis or Cryptosporidium was listed as an immediate or contributing cause of death.

Case-Control Study of Adults with HIV Infection

A case-patient was defined as an HIV-infected adult (≥ to 18 years of age) living in Clark County in whom laboratory-confirmed cryptosporidiosis was diagnosed for the first time during the study period. Each case-patient was matched by primary physician or clinic with three HIV-infected adult controls from the county; one control in each of three CD4⁺ cell count categories (< 100 cells/mm³, 100 to 199 cells/mm³, and more than or equals to 200 cells/mm³) was selected. Stool specimens from controls were not tested for Cryptosporidium.

Telephone interviews were done using a standardized questionnaire in May and June 1994. The exposure period for case-patients was defined as the 4 weeks before the case-patients became ill; matched controls were asked about exposures during this period and about illness in the 4 months before their interviews. Study participants were asked whether they had been exposed to persons who may have been infected with Cryptosporidium (whether they lived in the same household with or visited or cared for a person who had diarrhea); whether they lived in a household in which someone attended or worked in a child-care setting or in which a child wore diapers; whether they had changed a child's diaper; and whether they had had any type of sexual activity or had engaged in high-risk sexual activity (anal-oral intercourse). Participants were asked about contact with newborn animals (< 4 months of age) and about visits to farms, pet stores, animal shows, animal pounds, petting zoos, and veterinarians; about restaurant patronage and consumption of uncooked and cold foods, unpasteurized dairy products, health foods, and dietary supplements; about exposure to recreational water (in a pool, whirlpool bath, hot tub, lake, or river); and about types of drinking water used at home and work (for example, tap water, tap water filtered at its point of use, bottled water, or well water). Study patients were asked whether they had immunosuppressive medical conditions other than HIV infection, whether they took immunosuppressive medications, and whether they had used nontraditional therapy for HIV infection; they were also asked about miscellaneous exposures, such as travel outside of Clark County and attendance at bars, clubs, and social functions.

Case-Control Study of Immunocompetent Children

In a case–control study of children that was similar to the study of adults described above, each case-patient who had laboratory-confirmed cryptosporidiosis was matched with three controls by age (± 3 years), primary physician or clinic, and week of medical evaluation; parents were interviewed. In contrast to the study of HIV-infected adults, no questions were asked about sexual activity and attendance at bars, and additional questions were included (for example, about attendance at child-care facilities, diaper use, and extracurricular activities).

Community Health Survey

A questionnaire was distributed in June 1994 to all employees at two Clark County agencies (agencies A and B) to determine whether they had had diarrheal illness during the study period and to identify their sources of drinking water. Employees were asked only about water drunk at home; however, the agencies did not provide bottled water, and tap water at the agencies was not filtered at its point of use.

Water Quality Analysis and Environmental Survey

Data on water quality for source water (Lake Mead) were reviewed for a 50-month period (1 March 1990 through 30 April 1994); data on water quality for finished (treated) water were reviewed for a 28-month period (1 January 1992 through 30 April 1994). The water treatment plant that serves all of Clark County was inspected. We reviewed the treatment procedures and the log of malfunctions and repairs at the plant and in the pipes that distribute water throughout the county.

Statistical Analysis

We used conditional logistic regression (SAS version 6.10 for Windows [PROC PHREG] [SAS Institute, Cary, North Carolina]) to calculate matched odds ratios for the case–control studies, and we used the chi-square test (Epi-Info version 5.1 [Centers for Disease Control and Prevention, Atlanta, Georgia, and the World Health Organization, Geneva, Switzerland]) to compare proportions for the community health survey. The Wilcoxon two-sample test was used to compare the ranked distributions of ordinal variables. We report two-tailed P values.

Results

Case Ascertainment and Confirmation

We identified 78 persons in whom laboratory-confirmed Cryptosporidium infection was diagnosed during the study period (Figure 1). No procedural or personnel changes had been made that affected diagnosis or reporting. At laboratory A, the mean percentage of stool specimens per month that tested positive for Cryptosporidium had increased from 4% in 1993 to 21% in the first quarter of 1994.

Sixty-one of the 78 persons with cryptosporidiosis (78.2%) were HIV-infected adults Figure 2, and more than 90% of the 61 had CD4⁺ cell counts less than 100 cells/mm³ (Table 1). Four of the 78 (5.1%) were adults without HIV infection; 1 of these was receiving corticosteroid therapy for renal transplantation, and another was receiving chemotherapy for testicular cancer. Two of the 78 (2.6%) were HIV-infected children, and 11 (14.1%) were immunocompetent children.

Persons who had laboratory-confirmed cryptosporidiosis lived throughout Clark County (Figure 3, top), in four of the five geographic areas served by the water treatment plant (all except Nellis Air Force Base [Figure 3, bottom]). The epidemic curves of the dates of onset of illness for the HIV-infected adults and the immunocompetent children are similar (Figure 4). These epidemic curves and that for the month of diagnosis of the cases reported to the health department Figure 1 show that the outbreak apparently began in December 1993 (when the first infected persons reported onset of illness) and extended through June 1994 (after which the number of reported cases notably decreased). The total number of laboratory-confirmed cases associated with the outbreak period was 103 (78 during the study period, 16 in May, and 9 in June).

Of the 78 persons in whom Cryptosporidium infection was diagnosed during the study period, 41 (52.6%) had died by 30 April 1995, 1 year after the end of the study period. One of the 2 HIV-infected children had died, as had 40 (65.6%) of the 61 HIV-infected adults (Table 2). Of these 40 adults, 32 (52.5% of 61) had died by 30 June 1994, 2 months after the end of the study period. At least 20 of these 32 (62.5%) had cryptosporidiosis listed as a cause of death on their death certificates (Table 2). The CD4⁺ cell counts were similar for HIV-infected adults with cryptosporidiosis who lived and those who died (Table 2).

Case-Control Study of HIV-Infected Adults

We interviewed 36 (81.8%) of the 44 case-patients (59.0% of the 61 HIV-infected adults) who were alive during the interview period (mid-May through mid-June) and 107 matched controls (Figure 2). The median ages of the case-patients and controls were 35 years (range, 27 to 61 years) and 36 years (range, 25 to 68 years), respectively; more than 90% were men. Most case-patients had CD4⁺ cell counts less than 100 cells/mm³ (Table 1).

All case-patients but only 31 controls (29.0%) reportedly had diarrhea (≥ to 3 loose or watery bowel movements in a 24-hour period) during the study period (odds ratio, undefined). Diarrhea lasted a median of 60 days (range, 7 to 146 days) for case-patients and a median of 13 days (range, 1 to 133 days) for controls (P < 0.001). However, 24 case-patients (66.7%) and 5 controls (16.1%) who had diarrhea still had it when they were interviewed. Nausea, vomiting, abdominal cramps, and weight loss (median, 13.6 kg) were reported by at least 80% of the case-patients; fever was reported by only 25.0%.

No association was found between illness and travel or exposure to potentially contaminated persons, animals, food, or recreational water. Case-patients were no more likely than controls to have an immunosuppressive medical condition other than HIV infection or to use immunosuppressive medications. In addition, no case-patients were known to be well acquainted with each other, and we could not identify any event that was attended by all case-patients or that could have resulted in sequential person-to-person exposures.

In contrast, drinking tap water was significantly associated with illness (Table 3); 91.7% of the case-patients (33 of 36) but only 71.0% of controls (76 of 107) reportedly drank at least some tap water (odds ratio, 4.22 [95% CI, 1.22 to 14.65]; P = 0.02). Bottled water was protective, but only if it was the sole source of drinking water. Only 8.3% of case-patients (3 of 36) but 28.0% of controls (30 of 107) drank bottled water exclusively (odds ratio, 0.26 [CI, 0.08 to 0.88]; P = 0.03). When the 13 controls who reportedly had had diarrhea for 5 or more days were excluded from the analysis, the results were similar. The odds ratios were 4.61 (CI, 1.33 to 16.03; P = 0.02) for drinking any tap water and 0.61 (CI, 0.06 to 0.75; P = 0.02) for drinking only bottled water. Eleven of these 13 controls drank at least some tap water. For the case-patients and controls who had CD4⁺ cell counts less than 100 cells/mm³, the associations with water were even stronger (Table 3). The odds ratios were 13.52 (CI, 1.78 to 102.92; P = 0.01) for drinking some tap water and 0.07 (CI, 0.01 to 0.56; P = 0.01) for drinking only bottled water. The etiologic fraction [7], or proportion of persons whose disease would have been prevented had they not drunk tap water, was 70% for all case-patients and 87% for those case-patients with CD4⁺ cell counts of fewer than 100 cells/mm³.

When matched for CD4⁺ cell count category, the proportions of case-patients and controls who had died by 1 year after the end of the study period (30 April 1995) did not differ significantly (17 of 36 [47.2%] compared with 12 of 36 [33.3%], respectively; P > 0.2). However, significantly more case-patients than controls had died by 2 months after the end of the study period (30 June 1994; early deaths) (11 of 36 [30.6%] compared with 0 of 36, respectively; P = 0.001). We reviewed death certificates for 10 of the 11 case-patients (90.9%) who had early deaths and for all 6 case-patients who died after 30 June 1994 (late deaths); similar proportions of those who died early and those who died late had cryptosporidiosis listed as a cause of death (50% compared with 33.3%; P > 0.2, Fisher two-tailed).

Case-Control Study of Immunocompetent Children

In the case–control study of immunocompetent children, we interviewed 10 of the 11 case-patients and 26 matched controls (Figure 2). Case-patients and controls were similar in median age (4 years [range, 8 months to 14 years] and 3 years [range, 1 to 13 years], respectively) and in sex distribution (70.0% of the case-patients and 69.2% of the controls were male). All case-patients but no controls reportedly had diarrhea during the study period. No statistically significant differences were found between the two groups with respect to exposure to potentially contaminated persons, animals, food, drink, or recreational water. Only 2 case-patients (20.0%) and 6 controls (23.1%) attended child care facilities (P > 0.2), and only 3 case-patients (30.0%) and 9 controls (34.6%) wore diapers (P > 0.2). We could not test for an association between tap water and illness because all children drank some tap water.

Community Health Survey

Questionnaires were completed by 52.4% (131 of 250) of the employees from agency A and 69.0% (69 of 100) of the employees from agency B (Figure 2). Overall, 45.5% (91 of 200; 45.8% for agency A and 44.9% for agency B) of the employees who returned questionnaires reportedly had a diarrheal illness (≥ to 3 loose or watery bowel movements in a 24-hour period) during the study period (Figure 4). The illness, which lasted for a median of 3 days (range, 1 to 90 days), was notable for the presence of abdominal cramps (76.9%) and because only 19.8% of the ill employees reportedly had a fever. To our knowledge, no ill employee was evaluated by a physician or had a stool specimen tested for parasites, bacteria, or viruses. Of the 34 zip code areas of Clark County, employees lived in 31 (91.2%) and ill persons in 26 (76.5%). Half (80 of 160) of the employees who drank any tap water reportedly had a diarrheal illness; only 28.9% (11 of 38) of the employees who did not drink tap water had such an illness (relative risk, 1.73 [CI, 1.03 to 2.91]; P = 0.03).

Water Quality Analysis and Environmental Survey

The water treatment plant is a state-of-the-art facility that is fully automated and computerized. Water from Lake Mead enters the plant through an intake tunnel located 130 feet below the surface of the lake (Figure 3, bottom) and is then chlorinated and filtered. Effluent water at each of 20 filter beds is monitored for turbidity and for the size of suspended particulate matter. After filtration, the water is disinfected again, to yield a free chlorine residual of 1.4 mg/L. The treated (finished) water is stored in an enclosed holding tank (clearwell) until it is pumped into the distribution system. No malfunctions had been noted and no major repair work had been done at the treatment plant or in the large pipelines (diameter more than equals to 8 in) of the distribution system.

From December 1993 through April 1994, the maximum recorded turbidity values for source and finished water were 0.26 nephelometric turbidity units (NTU) and 0.17 NTU, respectively. For both types of water, the median of all maximum daily values was 0.10 NTU. Current national standards for finished water require that 95% of all daily turbidity measurements in a month not exceed 0.5 NTU [8]. For this same 5-month period, the maximum recorded total coliform count for source water was 150 coliforms/mL (median of all maximum daily values, 0 coliforms/mL); no coliforms were found in finished water. Source water samples that tested positive for total coliforms were not tested for fecal coliforms.

Although testing water for Cryptosporidium was not mandatory, testing of the plant's source water (once a month) and finished water (twice a month) was begun in April 1993. Oocysts were not found in any of the water samples that were tested through April 1994 (the end of the study period). By chance, source water samples that tested positive for total coliforms were not tested for Cryptosporidium.

We identified some potential sites of source water contamination (Figure 3, bottom). A marina and a beach are located south of the water intake; neither functioned near capacity during the study period. Sewage from the bathroom facilities and from boats moored at the marina is pumped into oxidation ponds about 2 miles from the shore. No malfunctions or backups of the bathroom or sewage facilities had been documented. Treated wastewater from the country is discharged into the lake about 6 miles north of the water intake. No studies have specifically addressed this issue, but, according to plant management, the flow dynamics in the lake are such that wastewater would be unlikely to reach the water intake area (Monscvitz JT. Personal communication).

Discussion

Our epidemiologic data indicate that contaminated municipal drinking water was the most likely vehicle of transmission in this outbreak and that the outbreak was not limited to HIV-infected persons. First, the widespread geographic distribution of the residences of the case-patients and of the county employees who had diarrhea, and the broad age range of the ill persons, are consistent with waterborne transmission. Second, the case–control study that included HIV-infected adults strongly implicated tap water and showed that drinking bottled water was protective, but only if it was drunk exclusively. If any of the 13 controls who had chronic diarrhea should have been classified as case-patients, we biased our results toward the null by including them as controls, because 11 of the 13 drank some tap water. Third, illness was more common among employees who drank any tap water than among those who did not. Fourth, the dates of onset of illness for the immunocompetent children and the employees were similar to those for the HIV-infected adults (Figure 4). This suggests that the mode of transmission was the same for all groups and supports the hypothesis that the municipal water was responsible for infection in the general community.

Because of the widespread geographic distribution of the case-patients, contamination of the water had to have occurred before the water reached the treatment plant, in the treatment plant, or in the very proximal portion of the distribution system (before the system branched; Figure 3, bottom). Given that no malfunctions or repairs occurred that could explain contamination in the treatment plant or the proximal distribution system, we conclude that the water was probably contaminated before it reached the plant and that small numbers of oocysts passed through the plant's filtration system.

We found no direct evidence to show that lake water was contaminated, and we could not identify a specific source of contamination. The quality of the source water and the state-of-the-art treatment plant in Clark County (which, even during the outbreak period, was producing water of higher quality than that required by current standards) would be the envy of most cities in the United States. The maximum recorded turbidity value for finished water was only 0.17 NTU for this outbreak but reached 1.7 NTU during the outbreak in Milwaukee [4]. Because the water quality was so good, water-borne transmission was not suspected during the outbreak period, no boil-water advisory was issued, and persons continued to become infected for as long as 14 weeks after the first suspicion (in mid-March) that an outbreak was occurring (Figure 1).

Source water from Lake Mead is disinfected with chlorine, to which Cryptosporidium oocysts are highly resistant, and it is also filtered, which mechanically removes oocysts. The filters are composed of a loose layer of sand and carbon and do not inherently function as microstrainers. Rather, chemical pretreatment of water is required to aggregate small particles into larger ones that can be physically removed by the filters. During each cycle of operation, which lasts 5 to 6 days in the Clark County plant, the filters become increasingly efficient as a film of particles forms on top of the sand. A backwash process removes this film at the end of each cycle. Oocysts are most likely to breach the filters during the first 15 to 40 minutes of a cycle, before the film has fully developed. The Clark County plant has 20 filter beds operating in parallel, any of which could have intermittently been breached by small numbers of oocysts. An alternative hypothesis is that an as-yet unidentified cross-connection in the treatment plant allowed source and finished water to mix.

The occurrence of cases of cryptosporidiosis over an extended period in conjunction with the failure to detect oocysts in water samples collected during the study period suggests low-level, probably intermittent, contamination of the water. Such contamination could easily have been missed because testing for Cryptosporidium was infrequent and the amount of water sampled each month (100 to 200 gallons) was a minuscule fraction of the amount treated (7.5 billion gallons). Perhaps even more important is that current methods for obtaining and testing water samples for Cryptosporidium oocysts are insensitive, with mean rates of recovery of oocysts ranging from 3% to 22% for spiked samples ([9, 10]; Clancy JL. Personal communication).

After the study period, water was sampled more often (May 1994 through September 1995), and the samples, which were tested using standard methods [11], showed so-called presumptive oocysts on several occasions in lake water and in filter backwash, and in one finished water sample. Presumptive oocysts are thought to be the residual shells of once-intact oocysts; they have the same size, shape, and staining characteristics with fluorescein-tagged monoclonal antibodies as Cryptosporidium, but they lack the internal structures necessary for definitive identification. The occurrence of presumptive oocysts in lake water and filter backwash provides evidence of lake water contamination with oocysts and indicates that intact oocysts can enter the intake to the treatment plant (located 130 feet below the surface of the lake). The presence of presumptive oocysts in the finished water indicates that some particles the size of oocysts can pass through the plant's filtration system.

Three major factors largely account for the fact that this outbreak was recognized and documented. First, because cryptosporidiosis is a reportable disease in Nevada, we could compare the number of cases reported during the study period with the number reported before the study period and could show that the increase was impressive, even though the total number of cases during the study period was relatively low. The increase was also impressive when compared with the number of cases reported after the study period. Despite heightened surveillance because of outbreak-related publicity, a mean of only 1.6 cases per month (range, 0 to 4 cases per month) were reported to the county health department from 1 January to 30 September 1995. Second, laboratory A and the health department have a close working relationship that facilitates the accurate and timely reporting of cases of cryptosporidiosis to the health department. Third, there is a large population of HIV-infected residents in the community. In such persons, the symptoms of cryptosporidiosis are typically too severe and persistent to ignore. These persons are also more likely than immunocompetent persons to have physicians who consider the diagnosis of cryptosporidiosis (because of its association with the acquired immunodeficiency syndrome [AIDS]) and know to request an examination for ova and parasites that includes the special techniques required to find Cryptosporidium[12, 13]. On the basis of the estimate that 630 adults in Clark County had AIDS during the study period, and given that 58 of the 61 HIV-infected adults who had cryptosporidiosis had AIDS, the attack rate for laboratory-confirmed cryptosporidiosis among adults with AIDS was 9.2%.

This outbreak underscores the fact that cryptosporidiosis, for which no curative therapy is available, can be life threatening for severely immunosuppressed persons. The short-term mortality rate for the HIV-infected adults who had cryptosporidiosis was high, and two thirds of those who died during or shortly after the study period had cryptosporidiosis listed on their death certificates. These data are impressive even though we could not differentiate dying “of” from dying “with” cryptosporidiosis. However, comparison of the mortality rates for the case-patients and controls showed that early mortality but not 1-year mortality was higher for case-patients.

The outbreak highlights the need for better guidelines for preventing waterborne-Cryptosporidium infection in immunosuppressed persons [1, 14]. The major obstacles to the development of such guidelines are the absence of water quality indicators that can reliably predict the occurrence of an outbreak (if the treatment plant is operating within existing federal and state standards) [14], the absence of data about the frequency with which outbreaks such as the one we have described occur in the United States, and poor understanding of the importance of waterborne transmission (compared with other modes of transmission) of C. parvum among HIV-infected persons.

The potential for low-level, waterborne transmission may be greater than previously appreciated. Recent studies have shown that up to 97% of the surface waters serving water treatment plants and 54% of treated (filtered) waters contain low numbers of Cryptosporidium oocysts [15-17]. However, questions remain about the viability and infectivity of these oocysts and about the infectious dose required to produce illness. In a recent study in which Cryptosporidium was fed to human volunteers, the median infective dose was estimated to be 132 oocysts [18]. Other investigators have predicted, on the basis of mathematical models, that infection could occur in a small proportion of persons exposed to just one viable oocyst [19]. The usefulness of experimental and modeling data in assessing the true risk for waterborne-Cryptosporidium infection remains to be seen.

Although data suggest that person-to-person transmission is an important mode of spread of Cryptosporidium among HIV-infected persons [20-22], the relative importance of waterborne transmission is poorly defined. The risk probably varies among cities, depending on factors such as the quality of source water and of water treatment. As reported for 1991-1992 in the most recent surveillance summary for waterborne disease outbreaks in the United States [6], Cryptosporidium caused 3 of the 11 outbreaks (27.3%) associated with water intended for drinking for which an etiologic agent was identified. However, the importance of waterborne-Cryptosporidium infection as a cause of diarrhea in the United States in nonoutbreak settings is unknown.

In 1993, the year the outbreak in Clark County began, cryptosporidiosis was reportable in only 6 (including Nevada) of the 28 states for which information was available. In 1995, it was reportable in 24 of 50 states, and many other states are considering making cryptosporidiosis reportable. Enhanced surveillance will provide important information about the public health importance of the disease. However, because the diagnosis of Cryptosporidium infection is often not considered, and because most laboratories do not routinely test for Cryptosporidium, cryptosporidiosis will continue to be under-reported.

Despite the inadequacies of our knowledge about waterborne-Cryptosporidium infection, we should not assume that municipal drinking water is “risk free.” Alternative sources of drinking water (such as boiled, filtered, and bottled water) for immunocompromised persons have recently been reviewed [1, 14, 23]. However, the use of such sources requires changes in lifestyle and added expense, and, as shown by this investigation, it is protective only if no tap water is drunk. To determine the public health significance of waterborne-Cryptosporidium infection in the United States, we need more sensitive and rapid methods for detecting oocysts in water, workable surveillance systems able to detect cases associated with low-level transmission of Cryptosporidium, and epidemiologic studies specifically designed to address the risk for waterborne transmission of Cryptosporidium in nonoutbreak settings [14].

Drs. Juranek and Herwaldt and Mr. Hightower: Centers for Disease Control and Prevention, Division of Parasitic Diseases, Mailstop F-22, 4770 Buford Highway NE, Atlanta, GA 30341-3724.

Dr. Ravenholt, Ms. Martin, Ms. Mesnik, Mr. Griffiths, Ms. Bryant, and Mr. Reich: Clark County District Health Department, 625 Shadow Lane, Las Vegas, NV 89106.