Of 533 children originally recruited for the cohorts, 368 children (201 [55%] boys) met our inclusion criteria. The 165 excluded children were comparable in age and sex distribution with study children; for 25 children, no stool specimen was submitted, while <6 months of surveillance was completed for the others. At the time of entry into the study, 256 children (70%) were <1 year, 45 (12%) were 1–4 years, and 67 (18%) were 5–11 years of age. The 368 children contributed a total of 889 child-years of surveillance data; nearly half the data were from children <2 years of age. Children were followed for a mean of 2.4 years and had an average of 1.95 episodes of diarrhea/child-year of follow-up. The highest incidence of diarrhea, 3.3 episodes/child-year, was recorded in children 12–23 months old; after the age of 3 years the incidence of diarrhea declined to <1 episode/child-year. The median duration of diarrheal episodes was 2 days (range 1–27). A total of 44,042 stool specimens were screened for coccidian parasites, a median of 124 stools/child (range 24–227); 897 (2%) of the stool specimens were collected during diarrheal episodes.

Children had an average of 0.20 episodes of cyclosporiasis/year and 0.22 episodes of cryptosporidiosis/year of follow-up (Table 1). Of the 368 children, 123 (33%) had at least one detected episode of Cyclospora infection, 30 children had two infections, and 10 children had >3 infections. A total of 143 children (39%) had at least one Cryptosporidium infection; 34 children had two infections, and 9 had >3 infections. Rates varied by age: the incidence of cryptosporidiosis peaked at 1 year and then fell sharply, but cyclosporiasis incidence remained fairly constant during the 1- to 9-year age period. For the 189 children who were enrolled in the study before the age of 3 months, the mean age at first infection was older for cyclosporiasis than for cryptosporidiosis (1.69 versus 1.36 years; p<0.01). After an initial episode of cyclosporiasis, the likelihood of diarrhea decreased significantly (p=0.049) with each subsequent infection (Table 2). For cryptosporidiosis, this trend was less consistent and did not reach statistical significance.

In a regression analysis in which data were controlled for concurrent cyclosporiasis, the expected mean duration of oocyst shedding was longer for cryptosporidiosis associated with diarrhea than for cryptosporidiosis not associated with diarrhea (expected mean 9.4 versus 4.8 days; p=0.0002). In the analogous regression analysis for concurrent cryptosporidiosis, a similar relationship was found for Cyclospora cayetanensis shedding with and without diarrhea (15.7 days versus 6.2 days; p=0.004). Diarrheal episodes associated with cryptosporidiosis, but not cyclosporiasis, lasted longer than diarrheal episodes not associated with coccidian parasites (expected mean duration 4.67 days for cryptosporidiosis versus 2.55 days with no coccidia; p<0001; mean 2.96 days for cyclosporiasis versus 2.55 days with no coccidia; p=0.35).

Both parasitic infections were more frequent during December to May than June to November, but the effect was more marked for cyclosporiasis (relative risk [RR] 3.3; p<0.0001) than for cryptosporidiosis (RR 1.9; p<0.0001). After data were adjusted for seasonality and age, the risk for cyclosporiasis or cryptosporidiosis did not differ by household water supply at the time of entry into the study (Table 3). Using a field rather than a toilet or latrine for defecation was associated with a higher risk of cryptosporidiosis but not cyclosporiasis. No association with exposure to animals could be demonstrated for cryptosporidiosis, but cyclosporiasis was more common in children in households with avians, guinea pigs, rabbits, or any other domestic animal.

Our analysis confirms that cryptosporidiosis and cyclosporiasis are common infections in this community, with distinct age-related patterns of occurrence. As noted (4,15), cyclosporiasis affected cohort children at later ages than cryptosporidiosis. The reasons for this epidemiologic pattern are not clear. One possible explanation might be that early infections afford less effective immunity for Cyclospora cayetanensis than for Cryptosporidium parvum. However, our data appear to contradict this hypothesis: first episodes of cyclosporiasis, but not cryptosporidiosis, protect against later symptomatic infection with the same organism. The development of better laboratory tools will be essential to elucidate the immune mechanisms involved. Another possibility is that the differences in age-specific incidence are related to predominant modes of exposure. This hypothesis is consistent with the assumption that Cyclospora cayetanensis is usually transmitted by exposure to contaminated environmental sources, from which young infants are usually relatively protected, while cryptosporidiosis can be transmitted by many routes, including from one toddler to another.

The ability of cryptosporidiosis to cause multiple symptomatic episodes may also be related to genetic heterogeneity. In a previous study of specimens from the same cohort, most cryptosporidiosis in this shantytown was caused by the Cryptosporidium parvum human genotype. However, children also had infections with C. parvum bovine and dog genotypes, C. meleagridis, and C. felis (8). Heterologous immunity may be less effective than homologous immunity. Although some polymorphism has been demonstrated in Cyclospora cayetanensis isolates (19), genetic studies are still in the early stages, and, to date, this parasite appears to be less heterogeneous than Cryptosporidium species. Nevertheless, immunity to both organisms occurs in this highly endemic setting, since immunocompetent adults rarely experience symptomatic infections (4).

For both coccidia, we found a high proportion of infections without diarrhea: 67% of cryptosporidiosis and 77% of cyclosporiasis episodes were not associated with diarrhea. These results differ from those reported in studies with different designs and reflect our methods, which aimed to detect as many coccidial infections as possible, independent of symptoms. Each child had a stool specimen screened nearly every week, so that 98% of the stool specimens were not collected during diarrheal episodes. Nevertheless, serologic data suggest that even intensive stool surveillance may miss a substantial proportion of cryptosporidiosis episodes (20). Data from the same Peruvian community showed that cryptosporidiosis without diarrhea had a substantial effect on childhood growth (21,22); because such infections may have long-term sequelae it is somewhat misleading to call them asymptomatic.

Cryptosporidium species infect a wide range of mammalian hosts and can be zoonotic infections (6), while Cyclospora cayetanensis has never been convincingly demonstrated to infect a nonhuman host (9–11). However, the animals most commonly associated with zoonotic C. parvum were rare in this urban setting: no families had calves, only one family had goats and lambs, and six families had adult sheep. This finding may explain why we could not show an association of cryptosporidiosis with animal exposure. Children in households with animals, especially birds, guinea pigs, and rabbits, appeared to be at higher risk of cyclosporiasis. The finding of an association with avians is consistent with results of a C. cayetanensis case-control study in Guatemala (14), but this association is still unexplained. Possibly the presence of domestic animals is a marker for some other unmeasured risk factor. Cyclosporiasis appears to be much more common in Lima than in the mountains of Peru (Asociación Benéfica PRISMA, Lima, Peru, unpub. data), and most residents of the study community migrated from the mountains to Lima. Raising domestic animals, especially poultry, may be more common among recent rural migrants with less exposure and therefore higher susceptibility to C. cayetanensis.

Our exposure data were collected at the beginning of longitudinal surveillance, and the time that elapsed between their collection and the occurrence of an infection could have been as long as several years, which may have decreased our ability to detect associations. Studies specifically designed to identify risk factors close to the time of an infection and use of molecular techniques to distinguish genotypes or strains may help clarify some of these issues.

As more longitudinal data become available for these organisms, we are gaining a clearer picture of the overall epidemiology of cyclosporiasis and cryptosporidiosis in endemic settings. Both organisms cause a spectrum of disease, from apparently asymptomatic infection to prolonged episodes of diarrhea that may have a profound effect on a child’s well-being. Our findings for cryptosporidiosis are consistent with the view that the organism infects children very early in life and has multiple routes of transmission, but mysteries remain concerning the cycle that maintains cyclosporiasis as an endemic infection. In-depth study of the transmission of C. cayetanensis will be key to designing effective strategies for intervention.