Epidemiologic studies of Cyclospora cayetanensis in Guatemala.

In 1996 and 1997, cyclosporiasis outbreaks in North America were linked to eating Guatemalan raspberries. We conducted a study in health-care facilities and among raspberry farm workers, as well as a case-control study, to assess risk factors for the disease in Guatemala. From April 6, 1997, to March 19, 1998, 126 (2.3%) of 5, 552 surveillance specimens tested positive for Cyclospora; prevalence peaked in June (6.7%). Infection was most common among children 1.5 to 9 years old and among persons with gastroenteritis. Among 182 raspberry farm workers and family members monitored from April 6 to May 29, six had Cyclospora infection. In the case-control analysis, 62 (91%) of 68 persons with Cyclospora infection reported drinking untreated water in the 2 weeks before illness, compared with 88 (73%) of 120 controls (odds ratio [OR] 3.8, 95% confidence interval [CI] 1.4, 10.8 by univariate analysis). Other risk factors included water source, type of sewage drainage, ownership of chickens or other fowl, and contact with soil (among children younger than 2 years).

Before 1995, the coccidian parasite Cyclospora cayetanensis was primarily reported as a cause of gastroenteritis among children living in poor sanitary conditions (1,2) and adults from industrialized countries who lived or traveled in developing countries (3)(4)(5). From 1990 to 1995, three small Cyclospora outbreaks were reported in North America, at least one of them epidemiologically associated with drinking water (6)(7)(8). However, in the spring and early summer of 1996, the largest ever reported outbreak of cyclosporiasis, affecting more than 1,400 persons in North America (9), was associated with eating fresh raspberries; traceback data indicated that the source of the berries was Guatemala (9). A year later, despite improvements in water quality and sanitary practices instituted on berry farms after the 1996 outbreak, Guatemalan raspberries were again implicated in cyclosporiasis outbreaks (10).
The source of contamination of the implicated raspberries has not been established. Most likely, fecal contamination of water used for spraying fungicides and other substances directly on fruit was the source (9). Another possibility is that berries were contaminated with sporulated organisms during hand-sorting and packing; however, because Cyclospora may require 1 to 2 weeks in the environment to become infectious (2), direct fecal contamination from the hands of an infected worker is less likely. The possibility that berries might have been contaminated by bird droppings before being picked has also been raised (11), although no natural (12) or experimental infection with C. cayetanensis has been confirmed in any animal (M. Eberhard, pers. comm.). Studies of Cyclospora have been limited by this lack of an animal model; by dependence on microscopy for detection; by the lack of sensitivity of the available methods in water, fruit, and other environmental specimens; and by inability to culture the organism.
Cyclospora in Guatemala was first reported in 1994 (13). Since then, the organism has been

Epidemiologic Studies of Cyclospora cayetanensis in Guatemala
In 1996 and 1997, cyclosporiasis outbreaks in North America were linked to eating Guatemalan raspberries. We conducted a study in health-care facilities and among raspberry farm workers, as well as a case-control study, to assess risk factors for the disease in Guatemala. From April 6, 1997, to March 19, 1998, 126 (2.3%) of 5,552 surveillance specimens tested positive for Cyclospora; prevalence peaked in June (6.7%). Infection was most common among children 1.5 to 9 years old and among persons with gastroenteritis. Among 182 raspberry farm workers and family members monitored from April 6 to May 29, six had Cyclospora infection. In the case-control analysis, 62 (91%) of 68 persons with Cyclospora infection reported drinking untreated water in the 2 weeks before illness, compared with 88 (73%) of 120 controls (odds ratio [OR] 3.8, 95% confidence interval [CI] 1.4, 10.8 by univariate analysis). Other risk factors included water source, type of sewage drainage, ownership of chickens or other fowl, and contact with soil (among children younger than 2 years).

Resear Resear Resear
Resear Research ch ch ch ch reported among AIDS patients (14) and in several case series in hospital outpatient departments in Guatemala City (15)(16)(17). However, diagnosis of Cyclospora infection is not part of routine stool examination in health-care facilities in Guatemala.
In 1997, we initiated a study whose objectives were to provide the first comprehensive description of the epidemiology of Cyclospora in Guatemala and explore potential environmental sources of contamination. The study included surveillance among patients in outpatient departments and among raspberry farm workers, a case-control study to identify risk factors for Cyclospora infection, and screening of surface water specimens for fecal contamination and for Cyclospora oocysts.

Outpatient Surveillance
In two government health centers (both in the Department of Guatemala) and three hospital outpatient departments (two in the Department of Guatemala and one in the Department of Sacatepequez), we interviewed outpatients whose health-care providers had ordered stool specimens for bacterial pathogen testing or screening for helminth eggs. In each center, we screened up to 25 specimens submitted each day. The study was explained to potential participants and written consent was requested. For each person who agreed to participate, part of the stool specimen was collected, and information was recorded about age, sex, and current diarrhea (defined as 3 or more loose or liquid stools in 24 hours) or other gastrointestinal symptoms. For data analysis, we defined gastroenteritis as diarrhea, other gastrointestinal symptoms, or both. Persons with symptomatic Cyclospora infection were treated with trimethoprim-sulfamethoxazole unless they had a history of allergy to sulfa drugs. Asymptomatic persons were not treated, as the risk for side effects was thought greater than the benefits of treating a self-limited asymptomatic infection.

Raspberry Farm Cohorts
On three raspberry farms (two in the Department of Guatemala and one in the Department of Chimaltenango), cohorts of workers and their family members were recruited in early April 1997. All participating farms, except one, had been implicated in the 1996 outbreak and were still exporting raspberries. The farms included at least one from each of the two main areas where raspberries are cultivated. The study was described in detail to participants, and written consent was solicited. A stool specimen was collected every other week from workers (less regularly from family members) and each time a participant reported having gastroenteritis. Persons found to have Cyclospora infection were treated with trimethoprim-sulfamethoxazole unless they had a history of allergy to sulfa drugs; they were monitored until their stool specimens were negative. Because of concerns about potential contamination of berries, all workers and their family members with Cyclospora infection, whether symptomatic or asymptomatic, were treated. Surveillance was intended to continue throughout the spring and fall berry harvest seasons.

Case-Control Study
From May 7 to September 3, 1997, we recruited participants for the case-control study. Because surveillance on raspberry farms ended earlier than anticipated, only four case-control participants were from the farms; all others were recruited from health-care facilities. All persons with symptomatic Cyclospora infection detected through the surveillance system were sought in their homes; however, some could not be located because of incorrect addresses. The purpose of the visit was to offer antibiotic treatment; at the same time, we solicited participation in the casecontrol study. We also requested stool specimens from family members of case-patients; however, family members were not included in the casecontrol study, and no rigorous attempt was made to include all family members.
We recruited controls from persons who had a stool specimen screened in any of the surveillance system's health centers or hospitals during the period in which we were enrolling cases. Participants were eligible to be controls if they reported no gastrointestinal symptoms and had negative stool results. We chose the controls to permit a final age distribution similar to that seen for Cyclospora cases in the surveillance system and a roughly equal distribution by surveillance site; individual cases and controls were not matched. Thus, each week we rotated the surveillance site from which we recruited controls, and we attempted to enroll 10 controls per week: one to two children younger than 1.5 years of age, four children 1.5 to 4 years of age, three 5 to 14 years of age, and one to two persons older than 14 years of age.
We administered a questionnaire that focused on potential risk factors for Cyclospora, including drinking water source and handling, household sanitation, presence of animals, socioeconomic variables, and consumption of specific fresh fruits, vegetables, and herbs. Several key variables were defined: drinking untreated water was defined as having drunk other than commercially bottled water that had not been boiled, chlorinated, or filtered in the household; water source was divided into high risk (public standpipe, well, spring, or water bought from a vendor) and low risk (in-house piped municipal water or commercially bottled water). Among children <5 years of age, we defined breastfed as having been breastfed at all, whether exclusively or partially, at the time the child provided the specimen.

Laboratory Methods
Stool specimens were processed by using a standard formalin-ethyl acetate concentration procedure and screened by two methods: lightmicroscopy examination of a modified acid-fast stained smear and UV epifluorescence examination of a wet mount (18). To confirm the identity of the parasite as C. cayetanensis, three positive specimens stored in 2.5% dichromate at ambient temperature (approximately 23 o C) were examined at regular intervals over a 2-week period, starting from the time the specimen was provided (18). We observed characteristic sporulation in all three.

Analysis of Water Samples
From May 13 to July 17, 1997, we collected a weekly 10-liter water sample from each of three rivers (Villa Lobos, Los Verdes, and Guacalate). Two sampling sites were located in the Department of Guatemala and one in the Department of Sacatepequez; these rivers all contain raw sewage outfalls. Standard methods were used to quantify the level of fecal contamination with Escherichia coli (19). The specimens were concentrated by the flocculation method developed for detection of Cryptosporidium in water (20), with the following modifications for logistical reasons: the specimens were left to flocculate overnight; the centrifugation speed was 600 g, the maximum for our centrifuge, rather than 3,000 g; and we used only one washing with Tween. The concentrated sediments were examined by bright-field and UV epifluorescence microscopy for Cyclospora oocysts.

Data Analysis
All data were entered in Epi-Info 6.04b. Surveillance data were analyzed in Epi-Info and in Excel spreadsheets. The case-control data were analyzed in SAS 6.12. Potential risk factors were assessed in a series of univariate analyses; unless otherwise specified, the significance of differences was tested by Mantel-Haenszel chisquare test. We evaluated potential interactions among variables through a series of stratified analyses before constructing models that included interaction terms. Factors that were significant at p<0.10 level were then examined in stepwise regression models. The final multivariable model was chosen from alternative models resulting from the stepwise regressions, including interaction terms when indicated.

Hospital and Health-Center Surveillance
From April 6, 1997, to March 19, 1998, we screened 5,552 specimens from health-care facilities. Overall, Cyclospora was detected in 126 (2.3%) specimens. The rate of Cyclospora detected began to rise (3.8%) in May and peaked at 6.7% in June (Figure). This period corresponds to the first 2 months of the rainy season in Guatemala City. The prevalence remained above 3% in July and August and fell to undetectable levels after mid-November. In January and February, Cyclospora was detected in fewer than 1% of specimens.
Cyclospora infection was more frequent among children and among persons with gastroenteritis ( Table 1). The prevalence of Cyclospora was 3.5% in specimens from children 1.5 to 4 years of age and 3.8% among children 5 to 9 years of age, irrespective of symptom status. Children in these age groups were five times more likely than adults screened in the surveillance system to have Cyclospora infection. Persons with gastroenteritis were two to three times more likely than asymptomatic persons of the same age to have Cyclospora. Although we did not collect data concerning HIV status, one hospital submitted specimens for screening from a clinic that served the HIVinfected population in Guatemala City. Of 32 specimens from adult patients attending this clinic, 9 (28%) were positive for Cyclospora, a much higher rate than that detected among other adults at the same hospital (2 [0.4%] of 448; RR 87, 95% CI 16.2, 847). Seasonality among these patients was the same as that observed for other participants in the surveillance system: eight of the nine Cyclospora detections occurred during May to August. These patients were not included in the case-control study.

Raspberry Farm Cohort
The raspberry farm cohort comprised 164 workers and 18 family members of workers who submitted specimens from April 6 to May 29, 1997. All 176 workers on the three farms agreed to participate; however, not all submitted specimens. We collected 269 specimens from workers and 31 from their family members, a median of one specimen per person (1 to 4). Six of 182 persons had an episode of infection with Cyclospora. Of the six infections, two occurred in family members, ages 1 and 10 years; both had gastroenteritis. The other four Cyclospora infections occurred in farm workers; all but one were asymptomatic. On May 29, 1997, the Guatemalan Berry Exporters Commission required us to suspend surveillance.

Case-Control Study
During the case-control study period, we recruited 69 persons with Cyclospora infection and 125 controls. Five controls were excluded because they had gastrointestinal symptoms when their specimens were submitted, leaving 69 cases and 120 controls for the analysis. Of 69 cases, 33 (48%) were in patients < 5 years of age and 26 (38%) were in male patients; among 120 controls, 56 (47%) were <5 years of age and 56 (47%) were male (p = 0.23).
In addition to specimens from the casecontrol study, we screened stool specimens from 182 family members belonging to 56 families of case-patients. Of these, 14 persons from 11 families tested positive for Cyclospora. The detection rate was highest among children 1.5 to 9 years (10 [11.6%] of 86). There were no cases among 14 family members <18 months of age and few among adults >14 years of age (2 [4%] of 57). Figure. Surveillance for Cyclospora cayetanensis infection in stool specimens from three hospital outpatient departments and two health centers in Guatemala, April 6, 1997, to March 19, 1998. From the bottom, the three graphs demonstrate the Cyclospora detection rate, mean temperature in centigrade, and rainfall in mm by month. Median number of specimens per month 444 (324-638).
The rate among family members ages 1.5 to 9 years did not differ from that among children of the same age in surveillance data from the same period (64 [9%] of 714; p = 0.42).
Data were available on characteristics of illness in 62 patients in the case-control study ( Table 2). The median duration of illness before diagnosis was 15 days in those <5 years of age and 10 days in those >5 years of age. Young children also tended to have a higher median number of stools per day than older patients and were significantly more likely to have fever and mucoid stools. Of the 62 patients, 40 (65%) reported predominantly watery or liquid stools, but 33 (53%) reported that diarrhea was intermittent. Of the 62 for whom data were available, 37 (60%) had been treated with an antibiotic, most often metronidazole, before Cyclospora was diagnosed.
A number of variables were associated with risk for Cyclospora infection in univariate analyses; all were related to water, sanitation, or presence of animals in the household (Table 3). Persons with Cyclospora infection were significantly more likely than controls to report having drunk untreated water in the 2 weeks before illness, having obtained drinking water from a high-risk source, or having swum in or drunk water from a river or spring. In addition, having a septic tank rather than municipal sewage drainage and having had direct contact with soil were associated with an elevated risk for infection. Persons with Cyclospora infection were twice as likely as controls to own dogs, chickens, or other fowl; other animals such as cats and pigs were not associated with increased risk for infection. We asked about eating 16 kinds of fresh, uncooked produce in the 2 weeks before illness, including raspberries, blackberries, lettuce, cabbage, mint, and cilantro; none was associated with an elevated risk for Cyclospora infection.
Among children <5 years of age, maternal education was protective: 10 of 33 children with  In stepwise regression analyses, the variables that remained significant were drinking untreated water and, among young children, contact with soil (Table 4). Drinking untreated water was associated with a fourfold increase in risk. For contact with soil, there was a significant interaction with age, carrying a 20-fold increased risk for Cyclospora infection among children <2 years of age but not among those 2 years of age or older (p for interaction term = 0.03). The presence of chickens in the household was associated with a twofold increased risk but did not reach statistical significance in the multivariable model (p = 0.054). The presence of chickens, poultry, or other birds in the household could be substituted for chickens in the model with similar results (OR 1.7; p = 0.09). The results of the model were unchanged when data were adjusted for socioeconomic variables, including educational level (for adult participants and of mother for children <15 years of age), housing type, and ownership of items such as radio, television, bicycle, or motor vehicle.

River Water Analysis
All three rivers had evidence of heavy fecal contamination throughout the period of study, with mean E. coli counts of 4.7x10 5 , 1.9x10 7 , and 4.5x10 6 colonies per 100 cc in the Department of Guatemala sampling sites and the Department of Sacatepequez site, respectively. We detected both sporulated and unsporulated Cyclospora oocysts in two of the 30 specimens, one specimen from each of the two rivers in the Department of Guatemala; both positive specimens were collected on May 26, 1997. The number of oocysts seen was small (2 oocysts and 3 oocysts, respectively). However, calculating from the amount of sediment obtained in each specimen (40 and 20 ml), the amount of sediment examined (20 and 40 µl), and the recovery rate of oocysts (5% to 10%; estimated from studies of seeded specimens, M. Alvarez de Mejia, M.B. Lopez,

Conclusions
This study advances our understanding both of the epidemiology of C. cayetanensis in a disease-endemic area and of the context in which repeated outbreaks of foodborne cyclosporiasis have occurred in North America. We now know that Cyclospora is commonly associated with pediatric gastroenteritis in Guatemala, as it is in Peru, Nepal, and other developing countries (1,2,5,21). Moreover, as in Peru and Nepal, the organism displays a marked seasonality. In Guatemala, the prevalence of Cyclospora in specimens submitted to health-care facilities increased around the time of the spring raspberry harvest in April-June (9). In June, we recorded weekly detection rates as high as 10%. In our surveillance data from May through June, 19% of cases of pediatric gastroenteritis were associated with Cyclospora. We were also able to detect oocysts in surface water specimens, despite the fact that the available methods for detection in water have very low sensitivity (22). By October, the detection rate in human surveillance specimens had dropped below 1%. These data suggest that the association of Cyclospora outbreaks with raspberries from the spring but not the fall harvest may be related to the inherent seasonality of the organism.
A major limitation of our study was our inability to test for other diarrheal pathogens. We cannot be certain that every gastroenteritis episode during which we detected Cyclospora was due to Cyclospora. However, in the aggregate data, we observed a consistent epidemiologic and clinical pattern associated with detections of Cyclospora.
Although 1 year of surveillance is insufficient to say with certainty what the pattern of Cyclospora infections will be from year to year, the data so far suggest that the seasonality of Cyclospora in Guatemala City is similar to that in Kathmandu, another subtropical city at approximately the same altitude (1,500 m) above sea level (3,23). In Guatemala as in Kathmandu, the rainy season starts in May or June, with warmer temperatures in the preceding months; in both places, Cyclospora appears around this time. Nevertheless, the organism's seasonality cannot be explained by rainfall alone, since in Lima an equally marked pattern is seen in the near absence of rain (21). Several years of surveillance in multiple sites with varying climates, as well as a better understanding of the organism's biology in the environment, will be necessary to explain why Cyclospora infections apparently fluctuate with the seasons.
Cyclospora has a number of other remarkable characteristics. It tends to cause prolonged diarrhea (2,3,24). Because all our patients were treated once the diagnosis was made, we could not estimate a total duration of illness; nevertheless, among children <5 years of age, the median length of illness before diagnosis was 15 days. Among children with diarrhea in developing countries, the proportion of persistent episodes (>14 days long) is 3% to 23% depending on the study cited. These episodes, however, carry a higher risk for malnutrition and death than shorter episodes (25)(26)(27). Perhaps because of the long duration of illness, persons with Cyclospora gastroenteritis are likely to receive various antibiotics before the correct diagnosis is made: in our study, 70% of children <5 years of age had received at least one antibiotic, and a third of all our patients had received metronidazole. Prompt diagnosis would avert such inappropriate use of antibiotics.
In North America, patients have been infected with Cyclospora through eating fresh raspberries, pesto dishes, and mesclun lettuce (9,(28)(29)(30). In Guatemala, the main vehicle of infection appears to be untreated water. That water truly is an important vehicle for Cyclospora in Guatemala is underlined by the other water-related factors that were associated with infection. Our results support the findings in other studies that contaminated water is a likely source of infection (1,3,31). However, multiple routes of transmission for Cyclospora in Guatemala almost certainly exist. Among very young children, soil contact was a strong risk factor; an outbreak investigation in the United States also raised the possibility that soil might be a potential source of infective oocysts (6). Family members of patients had a rate of infection similar to that of persons in the same age group screened in our surveillance system at the same time of year, a finding consistent with the postulated lack of direct person-to-person transmission.
The fact that the presence of chickens or other domestic fowl was a significant risk factor comm.). Although the association remains when the data are adjusted for socioeconomic status, we cannot rule out the possibility that ownership of fowl is a marker for some other unidentified factor. The role of birds in Cyclospora transmission merits further investigation. C. cayetanensis is a pathogen commonly associated with pediatric gastroenteritis in Guatemala, especially from May through August. The seasonality of Cyclospora in Guatemala follows a pattern similar to that seen in Kathmandu. Our case-control analysis suggests that contaminated water and, for young children, soil, are likely vehicles of transmission. More sensitive diagnostic tools are urgently needed to help establish how fresh produce becomes contaminated.