The Northern Region of Ghana has in all parts savannah-type vegetation and climate with rains from May to October. Malaria is hyperendemic with little seasonal variation 9. Tamale, the regional capital, has a population of approximately 350,000 inhabitants but a rather rural character with hamlets and thatched, mud-wall huts scattered over a vast area.

Between November 2005 and January 2006, i.e., during the dry season, 243 children with acute diarrhoea were consecutively enrolled at the Bulpeila Health Centre. The Bulpeila Health Centre, a study site of the Institute of Tropical Medicine Berlin, is divided into a research section and an outpatient and maternity clinic. The facility is providing primary health care for the approximately 50.000 adults and children of Bulpeila subdistrict, the catchment area of the study. The population structure and the rural character of Bulpeila is representative for the entire Northern Region where the dominant ethnic group are the Dagomba who speak Dagbani and are mainly Muslims. As in the whole region the socio-economic standard of Bulpeila's population is low and subsistence farming and small scale trade are the prevailing sources of income.

Patients met the following inclusion criteria: acute diarrhoea (≥ 3 watery or loose stools within 24 hours for < 14 days 10), age ≤ 12 years, absence of diseases requiring referral to hospital, no anti-infective therapy 48 hours prior to recruitment, and informed written consent from the children's guardians. The study was designed with a broad target age of 0–12 years because an analysis in different age groups was intended and the actual age distribution of children with acute diarrhoea attending Bulpeila Health Centre was not known in advance. 445 children fulfilled the inclusion criteria andof those 243 (54.6%)wereenrolled into the study, because stool specimen could be obtained at the study site. In addition, 124 control children, in similar age groups as the patients and residing in the same part of Tamale were recruited at four nurseries and during routine Child Welfare Service at Bulpeila Health Centre. Controls were recruited if they had no acute or chronic diarrhoea, according to the WHO definitions 10, did not receive anti-infective therapy in the last 48 hours, and if verbal consent was obtained from their guardian. The age of patients and controls were documented on weighing cards of the Child Welfare Clinic. The study protocol was approved by the Ethics Committee, University for Development Studies, Tamale.

Patients were clinically examined by the study physician (FAS) and a medical history was obtained. Fever was defined as an axillary temperature > 37.4°C, and underweight and wasting by z-scores of < -2 for weight-for-age and weight-for-height, respectively 11. Epidemiological data were collected using a standardized questionnaire adapted from the Integrated Management of Childhood Illness initiative 12. Rehydration therapy and specific anti-infective treatment were carried out according to WHO guidelines 10, and all patients received zinc per os for 14 days 1314. Other diseases were treated according to Ghana Health Service guidelines 15. Patients were followed up in individual intervals until complete recovery (93.4 %) or loss to follow-up or withdrawal of consent (6.6 %).

Trained health workers collected faecal specimens from patients and controls. Specimens were immediately processed. Aliquots were frozen at -20°C and preserved using merthiolate-iodine formaldehyde (MIF) solution. Aliquots were shipped to Berlin, Germany, for further investigation, on dry ice in case of frozen samples. Blood samples from patients were collected into EDTA-coated plastic tubes.

Full blood counts were performed applying a KX-21N cell counter (Sysmex, Japan). Anaemia was defined as haemoglobin < 11 and < 12 g/dL in children below six and twelve years of age, respectively 16. Malaria parasites were counted microscopically per 500 white blood cells on Giemsa-stained thick blood films. Submicroscopic infections and Plasmodium species were determined by nested PCR assays 17 after DNA extraction (QIAmp, Qiagen, Germany). Rota-, astro-, noro-, and adenoviruses were detected by nested (RT-)PCR assays 1819. Intestinal parasites were diagnosed by light microscopical examination of native and MIF-concentrated stool specimens 20. Furthermore, for the detection of Giardia lamblia, Cryptosporidium sp. and Entamoeba histolytica, stool antigen assays (MeriFluor Crypto&Giardia, Meridian Bioscience, USA; TechLab E. histolytica II, TechLab, USA) were applied. Positive Cryptosporidium sp. and E. histolytica results were verified by Kinyoun modified acid-fast staining and PCR, respectively 2122. Microsporidia were excluded by calcofluor white and Ryan's modified trichrome stains 2324. Bacterial pathogens were identified by culturing methods. To detect Salmonella, Shigella, and Vibrio species, standard agar plates and enrichment media were used (MacConkey, XLD, TCBS agars, selenite broth, peptone water). In addition, selective media were inoculated for the detection of Yersinia, Aeromonas, and Campylobacter species (CIN, Ryan, Karmali agars; all Oxoid, Germany). Suspicious bacterial colonies were further isolated and differentiated using routine techniques (e.g., oxidase and motility for Campylobacter spp., agglutination for Salmonella, Shigella, and Vibrio) or the API system. All differentiated and all suspicious isolates were stored at -20°C using the cryobank system (Mast Diagnostica, Germany). Additionally, collective samples harvested from MacConkey agar plates were frozen (MacConkey cryobanks). All cryobanks were shipped on dry ice to Berlin, Germany, for further analyses. There, samples from MacConkey cryobanks were spread on routine agar plates for detection and differentiation of enteric pathogens as before. In case of discrepant results, single isolates from cryobanks were further studied.

Data analysis was done using StatView statistical software (SAS Institute Inc., USA). Proportions were compared by χ²-tests, and odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated. Age- and sex-adjusted ORs for diarrhoea comparing cases and controls were calculated by logistic regression models for each potential enteropathogen. A P value < 0.05 was considered statistically significant.

The characteristics of the 243 patients with acute diarrhoea are shown in Table 1. Despite a target age of 0–12 years, 98 % of patients enrolled were under five(mean age: 0.8 years; range 0–11 years); and both underweight and wasting were seen in one out of five children.

The household questionnaires revealed that the most frequently used water source was the public tap (43%); one third of the households obtained water from ponds, rivers, or streams. For approximately half of the patients' households, a public pipe ventilated covered latrine was the most frequently used toilet facility whereas one third stated to commonly defecate somewhere out of doors, e.g., in the field (Table 1).

The mean duration of diarrhoeal symptoms as reported by the patients' guardians was 3 days (range, 6 hours to 1 week) and the mean stool frequency in the last 24 hours was 5.5 (range, 3–15). Main reported symptoms were fever, abdominal pain, vomiting, and weakness. On examination, more than half of the patients were found to be febrile, and more than a quarter of patients showed pronounced weakness as a sign of dehydration. Anaemia was present in 90.9%. Moreover, lower respiratory tract infection and skin diseases were common concomitant diseases (Table 1).

Control children had a mean age of 1.5 years (range, 0–10 years). None of these children showed overt signs of clinical disease, and in particular, no signs of acute or chronic diarrhoea as defined before. Still, 65.3% of their stool samples were classified as loose or watery.

Potential enteropathogens were isolated from 186 patients (76.5%) and 66 controls (53.2 %; P < 0.0001). A single infectious agent was detected in 50.2% of the patients and multiple infections in 26.3%. Co-infections were significantly more frequent in patients than in controls (26.3%, 64/243 vs. 14.5%, 18/124; P = 0.01)Enteropathogenic viruses were found in the stool of 178 patients (73.3%) and of 57 controls (46.0%; P < 0.0001). Rotavirus was the most common enteropathogen, which was present in more than half of the patients. Adjusting for age and sex, children with acute diarrhoea had eight times increased odds of being infected with rotavirus as compared to control children (P < 0.0001; Table 2). In contrast, adenovirus, norovirus, and astrovirus did not differ significantly in frequency between patients and controls (Table 2).

Intestinal parasites and G. lamblia in particular, were observed more than twice as frequently in controls than in patients (Table 2), however, this difference was not statistically significant. Cryptosporidium sp., Hymenolepis nana, Strongyloides stercoralis, and Ancylostoma duodenale were found only sporadically in both patients and controls.

Bacterial enteropathogens were detected more frequently in patients than in controls, at borderline statistical significance (adjusted OR, 4.8; 95%CI, 0.9–27.0). Among patients, the following bacteria were identified: 2 × Campylobacter jejuni, 2 × Salmonella enterica Gr.2, 2 × Salmonella enterica serovar Typhimurium, Salmonella enterica serovar Colindale, Salmonella enterica serovar Typhi, 2 × Shigella boydii, 2 × Shigella flexneri. In controls, one Campylobacter coli and one Shigella boydii were isolated. Yersinia, Aeromonas, or Vibrio species were not detected in this study.

Malaria parasites (exclusively Plasmodium falciparum) were found microscopically in 24.8% of the patients (geometric mean parasite density, 6335/μL; range, 31–214,741/μL) and in 36.2% by PCR. Malaria, defined as fever plus microscopically confirmed parasitaemia, was seen in 14.8%. Plasmodium falciparum infection occurred at similar prevalence in patients without an identified cause of diarrhoea (40.4%, 23/57) and in patients with identified enteropathogens (34.9%, 65/186; P = 0.5). However, symptomatic malaria occurred significantly more often in children without enteropathogens (24.6%; 14/57) than in children with isolated enteropathogens (11.8%; 22/186; P = 0.018).

Rotavirus-infection was significantly associated with age: it was seen in 66.1% (84/127) of patients of less than one year of age and in 42.2% (49/116) of older children (P = 0.0002). Vice versa, adenovirus occurred more often in the older (37.1%, 43/116) than in the younger children (18.9%, 24/127; P = 0.002). Finally, rotavirus was the only pathogen, which showed associations with clinical presentation, i.e., infected patients revealed a higher prevalence of fever than uninfected patients (62.4%, 83/133 vs. 44.5%, 49/110; P = 0.005). This difference was more pronounced when excluding cases of symptomatic malaria (58.7%, 71/121 vs. 29.1%, 25/86; P < 0.001). In addition, rotavirus infected patients presented more frequently with watery stools (58.5%, 107/183 vs. 43.3%, 26/60; P = 0.041) than uninfected patients. No associations between parasitic and bacterial pathogens on the one hand and epidemiological and clinical characteristics on the other were observed (data not shown).