Details of sample collection and microbiological analyses have been published (2). The protocol has been approved by the Institutional Review Board at the National Institute of Cholera and Enteric Disease. Briefly, fecal specimens were collected systematically from patients entering the hospital from November 2007 through February 2010. Of note, the previous study analyzed data through October 2009; however, the systematic sampling is still ongoing. The specimens were collected from every fifth patient with diarrhea on 2 randomly selected days each week. Only patients with diarrhea (defined by World Health Organization guidelines as passage of >3 loose or liquid stools per day or more frequently than is normal for the person) were eligible for inclusion in the study. Samples were collected from an average of 5.6% of eligible patients.

Each patient contributed 1 sample, and each sample was tested for all 26 common diarrheagenic pathogens. Standard microbiological techniques were used to examine the samples. Samples were collected in McCartney bottles (using sterile catheters or rectal swabs) containing Cary-Blair medium and examined (within 2 hours of collection) for bacterial, viral, and parasitic pathogens by a combination of conventional, immunologic, and molecular methods. The bacterial pathogens (V. cholerae, V. parahaemolyticus, V. fluvialis, Campylobacter jejuni, Campylobacter coli, Salmonella spp., Shigella spp., and diarrheagenic Escherichia coli) were isolated from appropriate selective media and identified by standard biochemical tests. Species and subtypes were confirmed by serotyping (for V. parahaemolyticus, Shigella spp., and Salmonella spp.) with commercially available antiserum (Denka Seiken, Tokyo, Japan; BioRad, Marnes-la-Coquette, France) and by PCR (for V. cholerae [4], V. fluvialis [5], enterotoxigenic E. coli [ETEC, including heat-labile and heat-stable enterotoxin producers], enteropathogenic E. coli [EPEC, typical and atypical], enteroaggregative E. coli [EAEC] [6], enteroinvasive E. coli, and Shiga toxin–producing E. coli [7]). Rotavirus was detected by polyacrylamide gel electrophoresis and silver staining (8). Noroviruses (groups I [NVG1] and II [NVG2]), sapovirus, and astrovirus were detected by reverse transcription–PCR with random primers for reverse transcription and specific primers for PCR (9). Adenoviruses were detected by the commercially available RotaAdeno VIKIA Kit (bioMérieux, Marcy l’Etoile, France). All samples were screened by using a highly sensitive antigen capture ELISA (TechLab, Inc., Blacksburg, VA, USA) of G. lamblia, Cryptosporidium parvum, E. histolytica, and Blastocystis hominis.

To test for possible associations, we used the Fisher exact test to compare pairs of pathogens (1, both, or neither) with an independent assortment based on the overall frequency with which pathogens were detected. To establish criteria for statistical significance, we calculated p values, odds ratios (ORs), and 95% confidence intervals (CIs). Additional covariates were collected and examined for confounding and interaction. These included patient age, gender, residence, and religion and season of infection. Seasons were defined as summer (March–June), monsoon (July–October), and winter (November–February). All analyses were conducted by using SAS version 9.2 (SAS Institute, Cary, NC, USA).

Fecal samples were submitted from 2,748 patients. Patient demographic characteristics are listed in Table 1. A large proportion (44%) of patients were 15–45 years of age, ≈13% were <1 year of age, 80% resided in urban areas, 74% were Hindu, and 25% were Muslim. The following pathogens were detected in at least 1 sample: adenovirus, Aeromonas spp., astrovirus, B. hominis, C. jejuni, C. parvum, EAEC, EPEC, ETEC, E. histolytica, G. lamblia, NVG1, NVG2, rotavirus, Salmonella spp., sapovirus, Shigella spp., V. cholerae O1, V. cholerae O139, V. cholerae non-O1, V. cholerae non-O139, V. parahaemolyticus, and V. fluvialis. No pathogens were detected in 766 (28%) of the 2,748 samples (Table 1), but test results were positive for the other 72%. One pathogen was found for 1,169 (43%) samples and multiple pathogens for 813 (29%) (Table 1). The 2 most commonly detected pathogens were V. cholerae O1 and rotavirus, which were found in 24% and 22% of samples, respectively.

Figure

Figure. Odds ratios (ORs) showing odds of A) Vibrio cholerae or B) rotavirus co-occurring with various other pathogens relative to the odds of V. cholerae or rotavirus co-occurring independently with various other...

V. cholerae O1 was detected in 661 samples. V. cholerae was the sole pathogen in 379 samples; however, it was isolated along with another diarrheagenic pathogen from 282 samples. The co-infection of V. cholerae and rotavirus was highly significant (p = 1.12 × 10^–33). Co-infection with V. cholerae and rotavirus was ≈5-fold less likely (OR 0.18, 95% CI 0.13–0.25; Figure, panel A) to occur among those with than among those without V. cholerae infection. A negative association might be expected if a case of severe diarrhea caused by any given pathogen excluded other pathogens. Consistent with this expectation, the presence of C. parvum, adenovirus, Shigella spp., ETEC, and V. parahaemolyticus was decreased significantly (p = 7.87 × 10^–5 to 1.32 × 10^–9) and was 12.5-fold (with V. parahaemolyticus, OR 0.1, 95% CI 0.02–0.33) to 2.44-fold (with C. parvum) less likely to occur among those with than among those without V. cholerae infection. However, antithetically, the rate of G. lamblia co-infection was significantly higher among V. cholerae O1–positive than among V. cholerae O1–negative fecal samples (OR 1.71, 95% CI 1.32–2.21). A significant difference in infection rates among those with and without V. cholerae O1 infection was not found for EAEC, C. jejuni, V. fluvialis, E. histolytica, astrovirus, NVGII, and EPEC. Tests for association were not performed for Salmonella spp., NVGI, Aeromonas spp., B. hominis, C. coli, sapovirus, V. cholerae non-O1, V. cholerae non-O139, and V. cholerae O139 because the low number of patients infected with those pathogens resulted in insufficient power.

Rotavirus was detected in 594 of the fecal samples and was the sole pathogen found in 253 of them. Rotavirus and at least 1 other gastrointestinal pathogen were found in 341 samples; 119 samples were co-infected with rotavirus and >2 other pathogens. When the effect of rotavirus co-infection with other pathogens was tested (Figure, panel B), Shigella spp. were significantly less likely to be found in samples with rotavirus than in samples without rotavirus (OR 0.30, 95% CI 0.17–0.52). In contrast, EAEC, Cryptosporidium spp., and adenovirus were significantly increased in samples with rotavirus (p = 6.15 × 10^–6 to 1.61 × 10^–26; ORs 2.14–5.80. A significant effect was not observed for G. lamblia, C. jejuni, EPEC, ETEC, V. parahaemolyticus, V. fluvialis, E. haemolyticus, astrovirus, and NVGII. Tests for association were not performed for Salmonella spp., NVG1, Aeromonas spp., B. hominis, C. coli, sapovirus, V. cholerae non-O1, V. cholerae non-O139, and V. cholerae O139 because the low number of patients infected with those pathogens resulted in insufficient power. Analysis of samples from patients infected simultaneously with G. lamblia, V. cholerae, and rotavirus (n = 41) revealed that the frequency of co-infection with G. lamblia was not significantly affected by co-infection with V. cholerae O1 and rotavirus (p = 0.08).

Analysis of covariates indicated that gender, religion, and residence largely had no effect on the associations between pathogens; however, in some instances, age and season were identified as confounders or effect modifiers (Table 2). To examine the effect of these covariates, we stratified the data by age and season and found that many associations remained significant (Table A1). Associations between rotavirus and adenovirus remained significant for all age and season strata except among children <1 year of age. After adjusting for age and season by using logistic regression models, we found that co-infection with rotavirus and Cryptosporidium spp. and co-infection with rotavirus and Shigella spp. remained significant (Table 2). The negative association between V. cholerae and adenovirus remained significant after adjustment by logistic regression for age and season (OR 0.36; 95% CI 0.21–0.64); associations between V. cholerae and many other pathogens remained significant within specific strata of age and season. Assessing the effect of covariates was limited in some instances because of small cell sizes. For this reason, we did not include stratified results for co-infection with V. cholerae and V. parahaemolyticus in the Table A1.

Our analyses revealed that co-occurrence of gastrointestinal pathogens in feces of patients with polymicrobial infections and severe diarrhea necessitating hospitalization was not in proportion to the pathogens’ presence in all patients with diarrhea. Tests for association were performed with V. cholerae O1 and rotavirus because they were the most commonly detected pathogens and, hence, had the greatest power to detect an association with the other pathogens. Some combinations of pathogens occurred less frequently than expected (e.g., V. cholerae and rotavirus [OR 0.18, 95% CI 0.13–0.25]); some combinations appeared more frequently than expected (e.g., rotavirus and adenovirus [OR = 5.8, 95% CI 4.20–7.99]), and some combinations occurred at the same frequency whether with or without V. cholerae or rotavirus. After adjustment for age and season, these variables often acted as confounders or effect modifiers, but in general the associations remained significant. However, many of the stratified analyses had small numbers for comparison.

V. cholerae O1 exhibited a positive association with only G. lamblia, suggesting that something may be unique about the co-occurrence of those 2 gastrointestinal pathogens. In support of that idea are 1) a report that co-infection with G. lamblia and V. cholerae results in G. lamblia being present in trophozoite form rather than in the cyst form found in feces of control patients (10), and 2) a previous finding that G. lamblia trophozoites can bind cholera toxin (11). Alternatively, each is a pathogen with substantial environmental reservoirs, and the positive association may simply represent acquisition of both pathogens from the same environmental source.

Phylogenetic relatedness alone does not explain the apparent competitive inhibition or negative association that we found between V. cholerae and other pathogens. For example, although the closely related V. cholerae O1 and V. parahaemolyticus exhibited a 10-fold negative association, V. fluvialis, which is phylogenetically only slightly farther from V. cholerae O1 than is V. parahaemolyticus, did not show any inhibition in the presence of V. cholerae. Also, although 2 members of the family Enterobacteriaceae (Shigella spp. and ETEC) were found less frequently than expected in combination with V. cholerae O1, 2 other members of that family (EPEC and EAEC) occurred in the expected proportion in samples from patients with mixed V. cholerae O1 infections.

Rotavirus had multiple strong positive associations (ORs >2) with the other gastrointestinal pathogens detected in the fecal samples, even after considering age and season. The positive association of rotavirus, an RNA virus that affects cells covered in microvilli, and adenovirus, a DNA virus that affects cells that are dividing to generate new cells with microvilli, may represent an interaction between pathogens to cause more severe diarrhea (in our sample all patients were hospitalized) in patients 15–45 years of age (OR 10.81, 95% CI 5.63–20.78) or a way to escape or circumvent immunity from previous exposures. In a previous study, Koh et al. (12) found that among children in Korea with virus-caused diarrhea, adenoviruses occurred preferentially in the presence of rotavirus and that rotavirus and norovirus, although most common, occurred in polymicrobial infections in proportion to their numbers in the samples. Both observations are consistent with the results of our study. Furthermore, Bilenko et al. (13) observed that among Bedouins, G. lamblia was frequently found in polymicrobial infections and, when present with rotavirus, produced less severe diarrhea than rotavirus alone. In addition, Souza et al. (14) found that among young children in São Paulo, those with rotaviral and bacterial co-infections were more likely to have severe diarrhea than were children infected with either pathogen alone. However, the study had insufficient power to examine differences among bacterial pathogens, as did the rest of the studies of rotavirus infections mentioned in a recent review (15).

The presence of multiple pathogens in one third of patients with diarrhea has potential implications for treatment and raises several questions. Do cases of diarrhea caused by V. cholerae or rotavirus and a second pathogen differ from those caused by V. cholerae or rotavirus alone? Does 1 pathogen lead the way for another to successfully infect a person? Do the pathogens behave synergistically to escape immunologic detection? Because the cross-sectional nature of our study did not enable us to investigate the temporal sequence of pathogen infection, future research is needed to provide more evidence concerning the causal pathway(s). Also, the clinical significance of our findings must be more rigorously evaluated by studies that include infected patients and controls. A more substantive investigation into how age and season might affect polymicrobial infections should also be conducted.

The results of our current study indicate that associations can occur between some pathogens affecting the human gastrointestinal tract. The observation of selective positive associations among some gastrointestinal pathogens raises the question of how they interact in vivo; e.g., is the critical factor a modification of gastrointestinal tract microflora? Understanding the association(s) among various co-infecting pathogens may help direct the development of treatment strategies.