Case–Case Analysis Using 7 Years of Travelers' Diarrhea Surveillance Data: Preventive and Travel Medicine Applications in Cusco, Peru

In Cusco, Peru, and South America in general, there is a dearth of travelers' diarrhea (TD) data concerning the clinical features associated with enteropathogen-specific infections and destination-specific risk behaviors. Understanding these factors would allow travel medicine providers to tailor interventions to patients' risk profiles and travel destination. To characterize TD etiology, evaluate region-specific TD risk factors, and examine relationships between preventive recommendations and risk-taking behaviors among medium- to long-term travelers' from high-income countries, we conducted this case–case analysis using 7 years of prospective surveillance data from adult travelers' presenting with TD to a physician in Cusco. At the time of enrollment, participants provided a stool sample and answered survey questions about demographics, risk behaviors, and the clinical features of illness. Stool samples were tested for norovirus (NV), bacteria, and parasites using conventional methods. Data obtained were then analyzed using case–case methods. NV (14%), enterotoxigenic Escherichia coli (11%), and Campylobacter (9%), notably ciprofloxacin-resistant Campylobacter, were the most frequently identified pathogens among adults with TD. Coinfection with multiple enteropathogens occurred in 5% of cases. NV caused severe disease relative to other TD-associated pathogens identified, confining over 90% of infected individuals to bed. Destination-specific risk factors include consumption of the local beverage “chicha,” which was associated with Cryptosporidium infection. Preventive interventions, such as vaccines, directed against these pathogens could significantly reduce the burden of TD.


INTRODUCTION
Cusco, Peru, is a popular travel destination in the Andes mountain range that is famous for its archaeological remains, including Machu Picchu. Previous studies have identified Cusco as a destination associated with a high risk of acute travelers' diarrhea (TD), with U.S. travelers at higher risk than those from other countries (adjusted odds ratio [OR] = 1.28; 95% confidence interval [CI] = 1.09-1.50). 1 However, in Cusco, and South America in general, the specific etiologies of TD have not been well evaluated, and the clinical features associated with pathogen-specific infections, along with destination-specific risk behaviors, remain largely uncharacterized. Understanding these factors would allow travel medicine providers to tailor interventions to patients' risk profiles and travel destination. 2 Anticipatory guidance and interventions (e.g., immunizations and prophylaxis) aim to decrease the incidence of travel-related injury and illness. Previous work shows that 94% of travelers to Cusco seek some sort of pretravel preventive guidance. 3 However, North American travelers to Peru are less likely than western European travelers to seek pretravel preventive medical advice from a travel medicine practitioner (37% versus 45.8%, P < 0.01) and from a health-care professional (52.0% versus 67.1%, P < 0.01). 3,4 Further, current health-related guidelines and dietary precautions have little evidence of effect on the incidence of TD. [5][6][7] Although most patients with TD experience spontaneous symptom resolution, some may experience symptoms for weeks, 8 and up to half of travelers change travel plans as a result of symptoms. 5 Of note, immunocompromised patients from high-income countries are traveling at increasing rates, and their increased risk for developing disease highlights the importance of providing effective preventive recommendations and measures for such complex patients. 9 The development of subsequent irritable bowel syndrome has become an increasingly recognized risk in the past decade, 10 reported by up to 1/5 of returning travelers with TD. 5 Thus, the benefits of reducing TD incidence extend beyond the travel period.
Traditionally, TD has been difficult to characterize because most data are derived from sentinel surveillance systems, which often lack denominator data and asymptomatic controls for comparison. A case-case approach allows a comparison of symptom severity between pathogens identified in the stool, although it lacks mechanisms to question if a particular pathogen present in the stool is causally related to diarrheal disease. A case-case approach allows restricted and refined analysis of some unique exposures associated with different pathogens detected through a surveillance system. [11][12][13] Although the approach reduces selection and recall bias relative to case-control formats 14 by ensuring that both case and "control" subjects were all affected by similar disease, the design does remain vulnerable to other forms of selection bias, that is, greater selection of infected travelers with severe symptoms compared with milder symptoms. To describe the etiology of medically attended TD, evaluate regional risk factors, and examine relationships between preventive recommendations and risk-taking behaviors among medium-to long-term travelers from high-income countries, we conducted this case-case analysis of 7 years of prospective surveillance data from adults presenting to a physician in Cusco with TD.

MATERIALS AND METHODS
Ethics statement. This study was approved by the institutional review boards of Naval Medical Research Unit No. 6 and Universidad Peruana Cayetano Heredia. Participants provided written informed consent at the time of enrollment.
Study population. Between June 2003 and July 2010, we prospectively enrolled Spanish language students presenting consecutively with acute diarrhea to the Amauta Spanish School's physician. Male and female students, older than 18 years, from countries with low-to-moderate risk of TD, and seeking free medical attention were included. 15 We defined diarrhea as three or more loose or watery stools, or two loose or watery stools accompanied by nausea, vomiting, abdominal cramping/pain, or tenesmus, within a 24-hour period. 16,17 To reduce confounding by chronic or recently treated diarrhea, students who could not provide a stool sample at the time of presentation, had taken antimicrobial medications within the week prior to presentation, reported chronic diarrhea, or did not speak English were excluded.
Survey administration, sample collection, pathogen identification, and antimicrobial resistance testing. Using a standardized form, a trained health-care worker surveyed enrollees to determine age, gender, country of birth, country of residence, lodging, and length of stay in Cusco, and Peru, in addition to TD risk factors, such as history of previous episodes, comorbidities, preventive measures, location and type of foods ingested, and clinical features of the presenting episode. Nonrecommended foods were defined using standard lists of unsafe foods. 2 We collected survey responses from individual cases while waiting for laboratory results to further reduce information and recall bias. The collection of individual-level data permitted us to control for potential confounding.
Participants provided fresh stool samples for analysis. Upon collection, specimens were kept fresh in preservative-free containers and transported in Cary Blair media to a central reference laboratory (ServiSalud) in Cusco city for bacteria and parasite analyses. On arrival, stool specimens went through a microscopic examination looking for ova and parasites using saline wet preparations to detect protozoa and helminth infections. The Ritchie method 18 was used to identify Giardia spp. Cary Blair specimens were cultured for Campylobacter, Escherichia coli, Salmonella, Shigella, Aeromonas, and Plesiomonas using conventional microbiologic techniques. [18][19][20] We performed polymerase chain reaction (PCR) on five lactose-fermenting colonies morphologically resembling E. coli to identify heat labile and stable enterotoxigenic (ETEC), enteropathogenic (EPEC), and enteroaggregative (EAEC) E. coli. 21 Antibiotic susceptibility testing was performed using the disk diffusion method according to the Clinical and Laboratory Standards Institute (CLSI) guidelines. 22 The isolates were tested against the most commonly used antimicrobial agents: ampicillin, amoxicillin-clavulanic acid, cefazolin, ceftriaxone, erythromycin, azithromycin, nalidixic acid, ciprofloxacin, gentamicin, sulfamethoxazoletrimethoprim, and tetracycline. 22 In the case of azithromycin, because of the absence of an established breakpoint, minimal inhibitory concentration was also determined by the agar dilution method according to CLSI guidelines 22 on all isolates with an inhibitory halo < 15 mm. 23 We combined intermediate and resistant isolates for analysis purposes. To determine Enterobacteriaceae susceptibility given lack of breakpoints published in CSLI, our laboratory used Staphylococcus aureus for quality control, and then established a resistance cutoff of 16 mm 24 with a susceptible strain of E. coli 25922. 22,[25][26][27][28] After preparing fresh specimens for parasite identification, the remaining stool was frozen and sent to Naval Medical Research Unit No. 6 (NAMRU-6) in Lima, Peru, for additional parasite testing using microscopy and commercial enzymelinked immunosorbent assay (ELISA) kits for Entamoeba histolytica (TechLabs, Blacksburg, VA), Giardia lamblia (Bio-Tech Trading Partners), and Cryptosporidium (BioTech Trading Partners, Encinitas, CA). Cryptosporidium-positive results were confirmed by PCR. Samples positive by microscopy but negative by ELISA were considered positive; in any further cases of discrepancy between results of pathogen identification at ServiSalud and NAMRU-6, NAMRU-6 results were considered confirmatory. A suspension of each stored stool specimen (10% w/v) was prepared in phosphatebuffered saline and RNA extracted using the Qiagen QIAmp Viral RNA Kit (Valencia, CA) in accordance with the manufacturer's instructions. Viral RNA was tested for Norovirus (NV) GI and GII by real-time PCR. 29 A sample was considered positive if the negative control did not exhibit fluorescent curves. The threshold cycle for the sample was at least 37 for NV GI and 39 for GII.
Statistical analysis. Statistical analysis was performed using Stata version 13.0 (StataCorp LP, College Station, TX). By convention, P values < 0.05 were considered statistically significant. Exploratory analysis of participant demographics, diarrhea risk factors, and clinical features of illness was performed using the two-tailed Student's t test to compare continuous variable means for normally distributed variables, Wilcoxon rank-sum tests for nonparametric continuous variables, and Fisher's exact test to compare proportions. We used dichotomous variables to characterize any versus no exposure to individual diarrhea risk factors. After excluding individuals coinfected with more than one pathogen, we performed a case-case analysis to examine the association of single-pathogen infections with dietary risk factors, according to the methods of Wilson and others. 30 It is standard practice in a case-case analyses to take the most prevalent or most frequently notified cause as the referent or comparison group. 30 We selected ETEC as the case-case analysis reference group because it is the most commonly implicated pathogen worldwide, 2 has the added benefit of being the most frequently identified cause of TD in Latin America, and has well characterized risk factors and clinical feature profiles. 17 We considered age and gender to be confounders a priori, 13,14 and controlled for them using multiple logistic regression. Exploratory analysis did not identify additional potential confounding factors.
Approximately two-thirds (176/230) of study participants reported receiving advice to prevent diarrhea. These individuals reported following this guidance all (46%), some (51%), or none (3%) of the time. Among those who reported always following recommendations, 93% (84/90) reported consuming three or more of the abovementioned foods linked to TD. Among these three stratified risk groups, the majority consumed fruit juice (84%, 86%, and 83%), cheese (73%, 83%, and 83%), or raw green vegetables (58%, 81%, and 83%), and the majority (84%) of the group that reported always following recommendations also consumed at least three foods linked to TD. In addition, 55/230 (24%) study participants took medications to prevent diarrhea, at some time prior to 1 week before enrollment. Of these individuals, 50 provided information on the type of medication used: 32% had taken antibiotics, 4% took bismuth preparations, and 64% took "other" medications.
Pathogen identification. A pathogen was identified in 45% of 230 cases (Table 3). NV was identified in 14% of samples (6 GI, 26 GII), ETEC in 11%, Campylobacter in 9%, Shigella in 6%, EAEC in 4%, Giardia in 4%, and Cryptosporidium in 3%. No EPEC, E. histolytica, Salmonella, Aeromonas, or Plesiomonas were detected. Coinfections with more than one pathogen were identified in 12 (5%) participants. Coinfections were most frequent among cases infected with the parasites Giardia (44%, 4/9) and Cryptosporidium (43%, 3/7). Coinfections were also found in individuals infected with ETEC (28%, 7/25 ETEC cases) in combination with NV (19%, 6/32 NV cases), EAEC (11%, 1/9 EAEC cases), and Campylobacter (10%, 2/20 Campylobacter cases). NV and ETEC were most frequently identified together (N = 6 coinfected cases), followed by Cryptosporidium and Giardia (N = 2 coinfected cases). Coinfections with the following pathogens were each observed once: EAEC and NV, ETEC and Giardia, Campylobacter and Cryptosporidium, and Campylobacter and Giardia. Giardia was a coinfecting agent in the greatest number of pathogen combinations (three other pathogens). No Shigella cases were coinfected. Table 4 shows bacterial antimicrobial   resistance patterns, and Table 5 shows the relationship between pathogen prevalence and seasonality. TD clinical features and severity. Gastroenteritis episodes were most frequently associated with watery diarrhea (84%), abdominal cramping/pain (76%), loss of appetite (69%), fatigue (68%), and gas/bloating (63%) ( Table 2). Subjective fever was reported in 42% of TD cases, most frequently with Campylobacter (75%), Shigella (71%), and EAEC (67%). Thirty-four percent of cases reported vomiting. Fifty-six percent of those with EAEC and 50% of those with NV (13/26 GII, 3/6 GI) vomited. Ten percent reported bloody diarrhea, most often with Shigella (36%), Giardia (22%), Campylobacter (20%), ETEC (16%), and NV GII (9%); no NV GI case had bloody diarrhea. These statistics include two individuals who were coinfected with NV GII and ETEC, along with one individual coinfected with Campylobacter and Giardia. Of the 230 cases, two were hospitalized, 62% stayed in bed, 52% saw a physician, 23% missed a scheduled tour, 13% made a change in travel itinerary, and 18% reported no effect on daily activities. One hospitalized case had no pathogen identified; one was positive for Cryptosporidium. Both were females aged 23 who presented in the winter and reported consuming raw green vegetables, fruit juice, and cheese. Neither had chronic gastritis. The case with no pathogen was from Holland, stayed in Cusco for 2 weeks in the winter, and reported she always followed preventive recommendations. The Cryptosporidium case was from Belgium, stayed in Cusco for 2 months in the winter, gave no answer regarding following recommendations, and reported consuming chicha. Ninety-one percent of norovirus-associated diarrhea cases were confined to bed. ETEC and Shigella cases had the largest proportion of doctors' visits, 60% and 79%, respectively.
Case-case analysis. In the unadjusted case-case analysis of diarrhea risk factors, individuals with and without ETEC ate at similar locations and had similar food exposures during the week before illness. However, individuals infected with NV GII were significantly less likely to have consumed reheated buffet food (OR = 0.10, 95% CI = 0.01-0.91), and individuals infected with NV overall were significantly less likely to have consumed food from street vendors (OR = 0.11, 95% CI = 0.01-0.88) relative to those with ETEC, although the buffet food relationship did not persist when the NV strains were combined. Chicha consumption was significantly associated with Cryptosporidium (OR = 5.44, 95% CI = 1.1-27.3) and EAEC (OR = 7.33, 95% CI = 1.17-46.1) infections relative to infection with ETEC. Individuals with no pathogen detected were significantly less likely to have consumed ice cream in the past week relative to ETEC (OR = 0.40, 95% CI = 0.17-0.94). Being a permanent resident of the United States or staying in Cusco longer than 1 week were not significantly associated with any pathogen relative to ETEC (data not shown). After adjusting for the potential confounders of age and gender, the statistical significance, OR magnitude, and CI size of these associations remained unchanged (Table 6).

DISCUSSION
This study is one of the first to evaluate TD among mediumto long-term adult travelers to South America. The 28-day median length of stay reported by participants exceeds the  Campylobacter coli  3  33  100  0  33  33  100  33  33  100  0  100  Campylobacter jejuni  16  56  100  38  13  88  100  25  25  100  0  44  Enteroaggregative Escherichia coli  9  56  89  67  89  89  67  100  100  100  78 78 Enterotoxigenic E. coli*    5-day median stay in Cusco reported previously by travelers 1 and also the 17-day median stay in a study of global travelers with TD. 31 Thus, these results may be generalizable to backpackers, international students, individuals involved in volunteer/service programs, and military populations deployed to this region.
We most frequently identified NV (14%), ETEC (11%), and Campylobacter (9%) in participants with TD, which differs from the results of a recent review identifying ETEC (33%), NV (15%), and EAEC (13%) as the top three TD pathogens in Latin America. 17 Although the lack of reported CIs in this comprehensive review does not allow us to comment on the level of significance of this observation, this suggestion of a lower prevalence of ETEC could be due to the inclusion of more severe diarrhea cases among longer-term travelers, along with the absence of NV detection methods in previous TD studies. While cholera vaccination provides some protection against ETEC, 32 we did not systematically assess this exposure. Of two participants who volunteered vaccination status, ETEC was missing for one and negative for the other, which would not appreciably impact ETEC prevalence. The study methodology was limited by a lack of capability to perform molecular testing for the full range of viral pathogens. For example, we might expect to detect Sapovirus among adult travelers to Cusco, 33 and data on Rotavirus and Astrovirus could also provide new information if included in future analyses. A recent systematic review of TD from 1973 to 2009 only included one study evaluating NV among travelers to Latin America, with a prevalence of 17% among students traveling to Mexico. 17 Globally, NV is the most common cause of all varieties of diarrhea, accounting for double the number of cases as ETEC, 34 and diarrheagenic E. coli and NV are commonly identified among Latin American travelers with TD. 35 Coinfection in this study was common, with half of coinfected TD cases positive for both NV and ETEC, and almost half of coinfected cases testing positive for Giardia. NV and ETEC were implicated together in a 2008 gastroenteritis outbreak among U.S. Navy personnel visiting Lima, 36 and both generally indicate a high level of fecal contamination in the environment. In coinfected travelers, NV is believed to contribute to disease, although detection may be attributable to prolonged viral shedding following a previous symptomatic or asymptomatic NV infection. 17,37,38 Giardia was also frequently detected in coinfected cases in our study, making it worthwhile to note that a recent debate in the literature emphasizes the commensal gut bacteria role that Giardia may play for citizens of countries where it is endemic. [39][40][41] Excluding participants with chronic diarrhea could have underestimated the contribution of parasites to diarrhea, while excluding those with recent antibiotic use could have impacted the prevalence of antimicrobial resistant pathogens. The exclusion of individuals who could not provide a stool sample at the time of presentation may have biased this study toward the evaluation of cases with particularly severe disease, which assists in the identification of pathogens against which preventive measures might be most impactful. Our population experienced a high rate (62%) of bed confinement, similar to rates reported in military members with TD (52%). 31 Compared with 4% bed confinement rates among adult Peruvians with NV-associated diarrhea, 42 NV-infected travelers with TD reported 91% bed confinement, suggesting that preventive measures, such as vaccines, directed against this pathogen could significantly reduce the burden of disease. Given TD's mission aborting potential for military troops, 43 in whom incapacitation could pose a further threat to life, 33,42 recommendations to military members operating in this geographic area should take NV-associated risks into account.
This study has several travel medicine implications. First, the prevalence of ciprofloxacin-resistant Campylobacter in our study reinforces the existing recommendations to empirically treat severe TD with azithromycin. 44 Our data also suggest that individuals who reported always following preventive recommendations took relatively fewer risks; however, these individuals still engaged in risky behaviors, indicating a possible gap in traveler receipt or understanding of preventive recommendations, or reflecting the difficulty of full compliance with all travel recommendations. In addition to supporting the use of pretravel interventions, such as vaccines, this may indicate that pretravel guidance is not sufficiently region or country specific, as noted by other authors. 31 *List-wise deletion was used to handle missing variables, and the asterisk (***) indicates that exposure to a risk factor was a perfect predictor of the outcome. †Adjusted for age and gender. ‡The number of cases of both ETEC and the respective pathogen (or cases without pathogen identified, as relevant) included in each case-case variable. §"Unknown" calculated with denominator of number of observations for each outcome variable.
Case-case analysis identified ice cream consumption as a risk factor for ETEC compared with cases with no pathogen identified, which, depending on method of milk preparation, could be consistent with the literature showing that ETEC is associated with unpasteurized dairy products. 45 For instance, the "queso fresco" and other cheeses typical of the Andean diet are often unpasteurized. Case-case analysis identified buffet food and street vendor food as risk factors for ETEC compared with NV GII, as well as an association between chicha with both Cryptosporidium and EAEC, which, despite wide CIs, should be considered for future travel medicine recommendations.
Given the case-case methodology, study limitations include the binomial outcome requirement for statistical analysis and the conservative estimates of association it provides given that most TD pathogens share risk factors. 30 In conclusion, NV was the most frequently identified pathogen among medium-to long-term adult travelers from high-income countries presenting to a physician in Cusco with TD. ETEC and Campylobacter, notably ciprofloxacinresistant Campylobacter, were also frequently identified, and coinfection with multiple enteropathogens was common. NV caused severe disease relative to other TD-associated pathogens identified, confining over 90% of infected individuals to bed. Destination-specific risk factors include consumption of the local beverage chicha, which was associated with Cryptosporidium infection.