A prospective, 4-year cohort study of children born in an urban slum in northeastern Brazil was undertaken to elucidate the epidemiology of Cryptosporidium infection in an endemic setting, describe factors associated with Cryptosporidium-associated persistent diarrhea, and clarify the importance of copathogens in symptomatic cryptosporidiosis. A total of 1476 episodes of diarrhea, accounting for 7581 days of illness (5.25 episodes/child-year), were recorded: of these, 102 episodes (6.9%) were persistent. Cryptosporidium oocysts were identified in 7.4% of all stools, and they were found more frequently in children with persistent diarrhea (16.5%) than in those with acute (8.4%) or no (4.0%) diarrhea (P < .001). Low-birth-weight children and those living in densely crowded subdivisions were at greater risk for symptomatic infection. Disease course was highly variable and was not associated with the presence of copathogens. Recurrent Cryptosporidium infection and relapsing diarrhea associated with it were moderately common. In light of these data, the applicability of the current World Health Organization diarrheal definitions to Cryptosporidium-associated diarrheal episodes may need to be reconsidered.
Cryptosporidium parvum is a coccidian protozoan parasite that is now well recognized as a worldwide cause of diarrhea. In the United States and other developed nations, this parasite is a frequent cause of diarrhea among immunocompromised persons [1, 2], has been implicated in causing epidemic illness during waterborne and foodborne outbreaks [3–5], and has been linked to diarrheal outbreaks in day care settings . In the developing world, the association of Cryptosporidium with both acute and persistent diarrhea among immunocompetent children has been striking. Cross-sectional surveys of children with diarrhea suggest that cryptosporidiosis is endemic in the developing world, with the identification of this parasite in up to 26.9% of symptomatic children . Community-based longitudinal studies, although far fewer in number, have also consistently found an association between Cryptosporidium infection and diarrheal illness [8–10]. In addition to causing potentially severe diarrheal illness in young children in the developing world, both symptomatic and possibly asymptomatic early childhood infections with Cryptosporidium may be associated with a subsequent increase in diarrheal burden or a decrease in nutritional status (or both) [8, 11, 12].
Since long-term adverse outcomes may be associated with Cryptosporidium infection  and treatment of this parasite has been elusive [1, 13], it is important to clarify risk factors not only for disease acquisition but also for illness duration (i.e., risks for persistent diarrhea). One possibility is that Cryptosporidium species act in conjunction with other pathogens, particularly to cause persistent diarrhea. Most community studies, however, have not undertaken sufficiently complete microbiologic stool evaluations to clarify this issue [8, 9].
We therefore undertook a cohort study of children living in an urban slum (“favela”) in northeastern Brazil to (1) elucidate the epidemiology of Cryptosporidium infection in an endemic setting, (2) describe factors associated with Cryptosporidium-associated persistent diarrhea, and (3) clarify the importance of copathogens in symptomatic cryptosporidiosis.
We studied an urban slum community (Gonçalves Dias) of ∼2000 inhabitants near the medical campus of the Federal University of Ceará in Fortaleza, Brazil. Fortaleza (population of 2 million) is the state capital of Ceará (population of 6 million) in Brazil's poor northeastern region. The community was divided into 5 distinct subdivisions created by the presence of major streets, a swamp, a warehouse, and an adjacent neighborhood of relative affluence. One of these subdivisions was markedly different from the others in that houses were extremely small and connected by narrow alleyways with open sewers.
All pregnant women in the community were identified by study nurses and were offered enrollment in the project. Following informed consent, women choosing to participate completed a detailed demographic questionnaire with the help of a study nurse. In general, children were followed from birth, but in 16 cases, mostly at the beginning of the study, entry was delayed. Sixteen children entered the study at >2 weeks of age, the latest at 59 days.
Following a study child's birth, a study nurse visited the house three times a week (Monday, Wednesday, and Friday) to record diarrheal illnesses and dietary information. Mothers (or other guardians) of children with diarrhea were asked to provide detailed clinical information about the illness, including other symptoms (fever, vomiting, dehydration), stool consistency and character, and medication use. Children with diarrhea were visited daily by a study nurse until 48 h after the illness resolved. The surveillance period was from August 1989 through April 1993.
Definitions of diarrheal episodes
We used World Health Organization (WHO) criteria to determine diarrheal episodes . A total of 3 unformed stools in one 24-h period was determined to be a diarrheal day. Therefore, it was possible for a child to have a single liquid stool collected and recorded as a nondiarrheal stool if there was no more than one further liquid stool in that 24-h period. A minimum of 2 days without diarrhea was required to delineate distinct episodes of diarrhea. A single day without diarrhea between 2 diarrheal days was therefore counted as 1 day of a 3-day diarrheal illness. A diarrheal episode <2 weeks in duration was recorded as acute diarrhea, a 14- to 29-day episode was considered persistent diarrhea, and an episode lasting ⩾30 days was considered chronic diarrhea .
To assess the association between nutritional status and disease course, we obtained weight and height information for all children in the cohort at 3-month intervals. Weight was measured by use of a calibrated sling scale (0.1-kg scale), and height was measured by use of a supine measuring board (0.1-cm scale). Weight-for-age, height-for-age, and their weight-for-height Z scores were calculated by use of nutritional software (EPI-NUT, Epi Info 6.0; CDC, Atlanta, and WHO, Geneva). We defined children as stunted if their height-for-age Z score was <2 SD below the mean, as underweight if their weight-for-age Z score was <2 SD below the mean, and as malnourished if weight-for-height Z score was <2 SD below the mean . To compare nutritional status between children with Cryptosporidium-associated acute and persistent diarrhea, we used the latest measurements obtained <4 months prior to the diarrheal episode being analyzed.
We attempted to obtain stool samples from all children with an identified diarrheal episode. We also attempted to obtain surveillance stool samples (nondiarrheal samples a minimum of 2 weeks from any diarrheal episode) from all children in the cohort at 4-month intervals. Stools obtained within 3 days following the conclusion of a diarrheal episode were considered diarrhea-associated stools. Stool samples obtained >3 days but <2 weeks following a diarrheal episode or within the 2 weeks preceding a diarrheal episode were considered indeterminate.
Stool samples were collected by parents or guardians and placed into disposable plastic cups. A portion of each sample was subsequently placed by field staff into a tube containing 10% formalin. Fresh stools were transported on ice by field staff to the laboratory for processing and microbiologic evaluation.
Stool specimens were concentrated using formalin—ethyl acetate at a centrifuge speed of 800 g in a fecal parasite concentrator (Evergreen Scientific, Los Angeles), and two slides were made from the resulting pellet. Slides were stained with modified acid-fast and auramine stains and read at 1000× (oil immersion) and 400× (fluorescence microscope), respectively. A stool was considered Cryptosporidium-positive if typical oocysts 4—6 μm in diameter were visible on both acid-fast and auramine slides.
Specimens were examined for the presence of ova and parasites by use of Lugol's iodine stain of both fresh and concentrated stool at 400×. Stools were examined for the presence of fecal leukocytes, using both microscopy (Gram's stain) and a latex agglutination assay (LEUKO-TEST; Techlab, Blacksburg, VA) for the detection of lactoferrin . Stools from children with ongoing breast-feeding were not tested for fecal lactoferrin because of cross-reactivity with breast milk. A stool guaiac test (Hemoccult; SmithKline Diagnostics, Santa Cruz, CA) was used to detect the presence of occult blood.
Monoclonal ELISAs were used on fresh-frozen stool samples to detect rotavirus (Rotaclone; Cambridge Biotech, Worcester MA) and adenovirus types 40 and 41 (Adenoclone, Cambridge Biotech).
Fresh stool was plated onto appropriate agar medium (consisting of MacConkey, xylose-lysine-deoxycholate, thiosulfate-citrate-bile-sucrose, trypticase soy with 5% defibrinated sheep blood and ampicillin, or trypticase soy with 5% defibrinated sheep blood with Campylobacter jejuni supplement), to isolate Shigella species, Salmonella species, Vibrio species, Yersinia enterocolitica, and C. jejuni. Three lactose-fermenting colonies identified as Escherichia coli by the API biochemical system (bioMérieux, Hazelwood, MO) were isolated from each stool and subcultured to agar slants for storage at 4°C and subsequent characterization.
One E. coli isolate from each patient was examined, using established methods, for HEp-2 cell adherence and enterotoxigenicity . Adherence to HEp-2 cells was scored as aggregate, localized, or diffuse . Aggregate adherence was interpreted as the presence of enteroaggregative E. coli, and localized adherence was interpreted as the presence of enteropathogenic E. coli (EPEC), both of which have been implicated as pathogens in childhood diarrhea . The role of diffusely adherent E. coli in the etiology of diarrheal disease remains unclear .
One E. coli strain per stool was also evaluated, using established methods, by oligonucleotide probe hybridization for the heat-labile and heat-stable toxins at the University of Virginia or the University of Maryland  (Dupont NEN Research Products, Boston) and for eae (E coliattaching and effacing trait) , EAF (EPEC adherence factor) , DA (diffuse adherence) , IEC (enteroinvasive E.coli) , and EHEC(enterohemorrhagic E.coli)  at the University of Maryland, Baltimore.
Strains were classified into virulence categories of E. coli on the basis of one or more tests being positive. Therefore, E. coli strains were classified as EPEC if the eae probe, EAF probe, or HEp-2 assay for localized adherence was positive; as enterotoxigenic if either the heat-stable or heat-labile probe was positive; as diffusely adherent if either the HEp-2 assay or diffuse adherence probe was positive; as enteroinvasive if the IEC probe was positive; and as enterohemorrhagic if the EHEC probe was positive. Because no IEC or EHEC strains were identified in the first 47 stools, the remaining 24 stools were not tested.
Univariate analyses were performed using χ2 or Fisher's exact tests (for cross-tabulations with an expected value in any cell ⩽5) to compare proportions for categorical variables; t or Wilcoxon rank sum tests (for variables with nonnormal distributions) were used to compare continuous variables. Tests were considered significant when the two-sided P value was < .05.
To determine whether there were identifiable risk factors for symptomatic infection with Cryptosporidium species, we used Cox regression . We used age in days as the time variable and assessed risks for a first symptomatic episode of cryptosporidiosis. Children without symptomatic infection were censored as they left the study or on 30 April 1993, the end of the surveillance period. Because this was an open cohort, multiple entry and exit was permitted for each child prior to failure.
To account for possible secular trends, we stratified the analysis by year of birth. We included the following factors as potential covariates in regression models: birth weight (<2500 vs. ⩾2500 g), sex, breast-feeding practices (divided a priori into tertiles), family size, house type and location, water sources and usage, hygiene practices, presence of pets, and maternal education. Variables were considered independent predictors if the 95% confidence interval associated with the point estimate did not include 1. The initial model suggested that only low birth weight and subdivision of residence were independent predictors. The proportional hazards assumption was met for the independent predictors. Additional variables were included as confounders if they changed the hazard ratio of either independent predictor by >10%.
STATA software (release 5.0; STATA, College Station, TX) was used for all statistical analyses.
We enrolled 189 children during the study period. Children had a mean of 543 days (range, 9–1350) of follow-up, and the cumulative number of days of observation was 102,681. A total of 157 children (83%) were followed for at least 3 months. There were 1476 episodes of diarrhea identified in 153 children, accounting for 7581 days of illness; the rate of diarrhea was 5.25 episodes/child-year. The mean number of days per diarrheal episode was 5.1 (SD, 6.3); the median number of days was 3. Of these episodes, 371 (25.1%) represented single-day diarrheal illnesses, 987 (66.9%) represented acute diarrhea longer than 1 day in duration, 102 (6.9%) represented persistent diarrhea, and 16 (1.1%) represented chronic diarrhea. We obtained samples from 611 (41.4%) of diarrheal episodes. Overall, there were 1091 stool samples tested during the study period. Only 36 children (19%) experienced no episodes of diarrhea. Among the 157 children with >3 months of surveillance, 9 had no recorded episodes of diarrhea.
We identified Cryptosporidium oocysts in 58 (31.2%) of 189 cohort children and in 58 (36.9%) of 157 children followed for at least 3 months. Of 1054 specimens from unique diarrheal episodes or from nondiarrheal or indeterminate stools, we identified Cryptosporidium oocysts in 73 (6.9%; table 1). Cryptosporidium oocysts were found more frequently in stools from episodes of persistent (16.5%) and acute (8.4%) diarrhea than in nondiarrheal stools (4.0%; P < .0001, χ2, and test for trend). No symptomatic infections with Cryptosporidium were identified in children <3 or 130 months of age. The mean age at diagnosis of the first symptomatic Cryptosporidium episode was 11.9 months. Cryptosporidium infection in this community was highly seasonal (figure 1). The percentage of laboratory specimens positive for Cryptosporidium oocysts ranged from 16.4% during the rainiest month of the year (April) to 0% during 4 of the 5 driest months of the year (August–December).
Among the 12 asymptomatic children with a stool specimen positive for Cryptosporidium oocysts, 3 had short (1 or 2 days) episodes of diarrhea between 16 and 22 days prior to the positive nondiarrheal stool, and 3 had previous episodes of symptomatic cryptosporidiosis between 2 and 22 months earlier. The other 6 children had no diarrhea within a month of the Cryptosporidium-positive stool. In addition, there were 2 children with positive stool samples that were deemed indeterminate. One child had a short episode of diarrhea 6 days following the positive stool; the other had 2 days of diarrhea 9 days before the positive stool and then a 5-day episode 11 days later.
We found a total of 15 cases of recurrent Cryptosporidium infection among 13 children, which includes the 3 cases of secondary asymptomatic infection (table 2). In 6 cases (5 children), the 2 positive stool specimens were in close succession (<14-day gap between diarrheal episodes); in 4 cases (3 children), there was a moderate gap of 1–2.5 months; and among the remaining 5 children, there were gaps of 13–22 months. Two children who had persistent diarrhea at the time of presumptive primary infection had secondary episodes of acute diarrhea, whereas 2 children had a converse experience: acute diarrhea followed by persistent diarrhea during secondary infection.
Twelve children who were not diagnosed with recurrent Cryptosporidium infection nevertheless had a relapse of diarrhea within 7 days of the completion of the primary episode. These relapses occurred a median of 5 days later and lasted a median of 3 days. Two children had a second relapse within 1 week of the first relapse. In 2 of these 14 total relapses, a stool sample was obtained and was negative for Cryptosporidium oocysts; stool samples were not submitted for the other episodes.
Demographic and nutritional characteristics of the 59 children with symptomatic Cryptosporidium infection are shown in table 3. The mean age of all children was 13.7 months (range, 2.7–42.5); age did not differ between children with acute or persistent diarrhea. Mild malnutrition was common, and moderate or severe malnutrition (stunting, underweight, or wasting) was rare, regardless of the clinical course of Cryptosporidium infection. Exclusive breast-feeding averaged 6 weeks, and any breast-feeding averaged 48 weeks. Just over 50% of children had pets in the home, and just under 50% had running water. No demographic or nutritional characteristics were significantly different between those with acute or persistent diarrhea.
Cryptosporidium-associated diarrhea ranged from 1–31 days, and the median duration was lengthy (6 days). The range in the peak number of stools in a single 24-h period was also great (3–20); the associated means (in peak stool output) between those with acute diarrhea (5 stools) and persistent diarrhea (9 stools) differed significantly (table 4). The most common symptoms associated with cryptosporidial diarrhea were vomiting and tactile fever, although even these were found in <50% of cases. These symptoms did not differ between those with acute or persistent diarrhea. Mild dehydration was rare, and moderate to severe dehydration was not observed. Concomitant diarrhea in other family members was common. A significantly higher percentage of children with persistent diarrhea received antibiotics during their illness than did those with acute diarrhea (67% vs. 34%, P = .05).
Laboratory characteristics of 72 Cryptosporidium-positive stool samples are also shown in table 4. One-quarter of non-diarrheal stools had a liquid consistency upon laboratory examination. Although visible blood was observed in none of the specimens, occult blood was found in ∼18%. Fecal fat was common in the stools of symptomatic individuals but nonexistent in nondiarrheal stools; reducing substances were rare. Fecal leukocytes were seen by direct microscopy in approximately one-third of stools from symptomatic children but in no nondiarrheal stools. Fecal lactoferrin, a more sensitive marker for the presence of fecal leukocytes, was also detected only in symptomatic children. There was no significant association between fecal leukocytes or fecal lactoferrin and the presence of any copathogen (data not shown). No stool characteristics were significantly more common in persistent diarrheal stools than in acute diarrheal stools.
Potential copathogens detected in stools of cohort children with Cryptosporidium infection are shown in table 5. Enteroaggregative E. coli, enterotoxigenic, diffusely adherent E. coli, Giardia lamblia, and helminths were the most commonly identified potential copathogens. Entamoeba histolytica, rotavirus, adenovirus, IEC, and EHEC were not identified in any stool specimens. In addition, we found no Vibrio species, Y. enterocolitica, or C. jejuni in any Cryptosporidium-positive stool (data not shown). We found no difference in the prevalence of copathogens between acute and persistent diarrheal stools. No test for trends for stool copathogens among the 3 groups was statistically significant.
In a Cox proportional hazards model, we found that low birth weight and residence in the most crowded subdivision within the slum were associated with an increased risk of symptomatic Cryptosporidium infection, after adjusting for weeks of exclusive breast-feeding, sex, type of house, and previous childhood deaths in the home (table 6). Children weighing <2500 g at birth were at a >5-fold increased risk compared with all other children of having a symptomatic episode of Cryptosporidium infection detected.
These data demonstrate that Cryptosporidium infection is common among children <4 years of age in a Brazilian slum community, affecting nearly one-third of all children followed. Cryptosporidium was isolated significantly more frequently from children with diarrhea than from asymptomatic children; children with persistent diarrhea were also more likely to have Cryptosporidium oocysts identified in their stools than those with acute diarrhea. This association with persistent diarrhea is remarkably similar in magnitude to that found in Guinea Bissau, West Africa . In children with more than one documented episode of Cryptosporidium infection, persistent diarrhea was equally likely to be associated with either episode. Consistent with findings in other longitudinal studies in the developing world [26, 27], we found that cryptosporidiosis was a highly seasonal disease: throughout the 4-year study period, recovery of Cryptosporidium oocysts occurred almost exclusively in months in which there was significant rainfall. It is notable that no children >3 years of age were diagnosed with symptomatic Cryptosporidium infection in this cohort, although previous work in this community has demonstrated sporadic symptomatic secondary cases among older children and adults in households with an index case .
We found that both fecal leukocytes and fecal lactoferrin were common in the stools of children with symptomatic Cryptosporidium infection. This unexpected finding may be explained by recent data that indicate that Cryptosporidium infection of intestinal epithelial cells stimulates cellular production of interleukin-8. Production of interleukin-8 may lead to the mucosal recruitment of leukocytes and contribute to stimulation of secretion in Cryptosporidium infection .
In multivariate analysis, we found that children with low birth weight were significantly more likely to be diagnosed with cryptosporidiosis. Whether this is a result of ongoing immunologic immaturity or is related to persistently lower weight-for-age and height-for-age Z scores for these children prior to diagnosis, when compared with other children (data not shown), remains unclear. There did not appear to be any risk associated with self-reported water usage patterns, perhaps because scattered water sources in the community as well as the city water supply have been shown to be contaminated . In addition, we found that children in the most densely crowded subdivision (with open sewers) within the slum were indeed at greater risk for cryptosporidiosis. It is likely that the use of data from the full cohort and the use of multivariate analysis (Cox regression) unmasked geographic location as a risk factor in this analysis, compared with our previous univariate analysis of this same population .
These findings suggest that in developing world settings, there may be a blurring between the typical pattern of epidemic Cryptosporidium disease with a single-source exposure and endemic disease, in which person-to-person transmission has been presumed to play a dominant role. Both the current finding that crowding appears to be a risk factor and previous evidence that there is a high incidence of secondary infection among household members of index cases in this community  suggest endemic disease. However, we have previously demonstrated that various water sources in this community, including the city water supply, are contaminated with Cryptosporidium oocysts , suggesting that epidemic infection may also play a role. Experimental infections in healthy volunteers have demonstrated that the mean infective dose for Cryptosporidium is low (132 oocysts)  and that the quantity of oocysts excreted in the stools of infected individuals can be enormous (up to a billion oocysts per infection) , indicating the ease with which this infection might be amplified in crowded communities. In poor urban areas such as this, it may be that seasonal rains amplify ongoing low-level water source contamination to produce increased numbers of infections and that person-to-person transmission then further sustains endemic disease in the community until the return of the dry season, when the number of cases again becomes negligible.
We found that a pattern of intermittent diarrhea associated with Cryptosporidium infection was moderately common, consistent with data from the large waterborne outbreak in Milwaukee in 1993 [3, 33]. In that study, 39% of subjects reported the recurrence of diarrhea after ⩾2 days of normal stools. This pattern of recurrent diarrhea brings into question the WHO criteria for diarrheal episodes, in which the recurrence of diarrhea after 2 symptom-free days marks a new episode . This definition may need to be revised in areas with endemic cryptosporidiosis. We propose that the term “persistent relapsing diarrhea” may be a better description of disease pattern in children with diarrheal episodes separated by a week or less, particularly for children in whom Cryptosporidium infection has been diagnosed.
Our conclusions about risk associations with first Cryptosporidium episodes are limited by the fact that Cryptosporidium infection, unlike measles, for example, does not have a pathognomonic clinical presentation. Because the causes of diarrhea are multiple, stool examination is necessary for diagnosis of cryptosporidiosis. Although our sampling rate was quite good (41%), we are obviously missing many cases of symptomatic Cryptosporidium infection and many more asymptomatic infections. This fact is underscored by data from this same community showing that antibodies to Cryptosporidium are present in 95% of tested individuals and that infection, whether it is detected or not, is nearly ubiquitous . In addition, our sensitivity in the detection of pathogenic E. coli was reduced by our inability to test>1 strain per sample. Finally, our model of risks associated with symptomatic Cryptosporidium infection must be interpreted with caution. The small numbers of failures (symptomatic Cryptosporidium infections) are reflected in the wide confidence intervals for all covariates, even those found to be statistically significant.
Consistent with previous studies, we found that even in an endemic setting, infection with Cryptosporidium results in heterogeneous disease. Short-term adverse outcomes (e.g., dehydration, hospitalization) were rare, possibly because of our clinical interventions (e.g., oral rehydration solution, maternal education) or a muted clinical course in this highly endemic setting, compared with adverse outcomes in naive children exposed under outbreak conditions . The fact that we were unable to find relationships between illness duration and either infection with copathogens or demographic and nutritional characteristics suggests the existence of a yet-to-be-defined host-parasite interaction, even among children who are clinically immunocompetent. A recent case report of a child with chronic cryptosporidiosis and interferon-γ deficiency , as well as experimental evidence from animal models regarding inter- feron-g and other cytokines [36–38], suggests that these me- diators may play a role in the course of human disease.
In addition, there is recent evidence of genetic heterogeneity among C. parvum isolates, although the clinical significance of these differences in endemic disease has yet to be elucidated . Further work in characterizing the immunologic status of individuals with cryptosporidiosis and clarifying subtypes of C. parvum, and in attempting to correlate this information with clinical course, might begin to explain the relatively protean presentations of this disease.
In summary, we found that in this Brazilian slum, early childhood infection with Cryptosporidium oocysts was very common, highly seasonal, and strongly associated with persistent diarrhea. Low—birth-weight children and those living in densely crowded subdivisions appear to be at greater risk of symptomatic infection. Disease course was highly variable and was not associated with the presence of copathogens. Recurrent Cryptosporidium infection and relapsing diarrhea associated with it were common. In light of these data, the applicability of the current WHO diarrheal definitions to Cryptosporidium-associated diarrheal episodes may need to be reconsidered.
We thank Sayonara Alencar and Luzia Sousa de Melo for invaluable field assistance; Isabel McAuliffe, Jania Teixeira, Maria do Carmo Nunes de Pinho, Leah Barrett, and Conceição Nogueria Raulino for laboratory assistance; the people of Gonçalves Dias for their cooperation; and David K. Shay, Stephen Gloyd, and Robert L. Davis for thoughtful reviews of the manuscript.