Study sites and settings
The primary location was the pediatric gastrointestinal infections ward at the hospital for tropical diseases (HTD) in Ho Chi Minh City in southern Vietnam. The HTD is a 500 bed tertiary referral hospital treating patients from the surrounding provinces and from the districts within Ho Chi Minh City. The secondary location was Dong Thap provincial (DTP) hospital in Dong Thap province, approximately 120 km from the HTD in Ho Chi Minh City.
Studies contributing data for analysis
Data from three independent studies were combined and compared. All patients enrolled in the three studies were treated as inpatients and there were no fatalities. The initial period (referred to as period A from here onwards) was a study performed at the pediatric ward at HTD from January 1995 to August 1996. The enrollment and clinical observations for this randomized controlled trial are as described previously . Briefly, children that were aged >3 months and < l4 years, admitted to HTD with fever and bloody diarrhea (bloody diarrhea defined >3 loose stools with obvious blood) for <5 days were entered into the study provided that their parents or guardian gave fully informed consent. Additional strains for microbiological assessment only (nine in total) were collected for comparison within the same period of the study duration from DTP. Overall 80 strains were isolated from enrolled children over this period; clinical data was available for analysis on 63 patients with culture confirmed shigellosis.
The secondary period (referred to as period B from here onwards) was conducted only at the HTD, between March 2000 and December 2002. This period was a clinical and microbiological investigation of the etiology of diarrhea in the pediatric population admitted to the HTD in Ho Chi Minh City. Whilst the treatment criteria for this descriptive study were not controlled (> 90% of patients received treatment with fluroquinolones (norfloxacin or ofloxacin)), the remainder of the criteria for admission to the study were comparable, children were eligible for enrollment to the study if consent was given and they were aged less than 14 years. The obvious variation in the enrollment for this study was that children were enrolled on the basis of having any diarrheal syndrome, rather than specifically targeting those with dysentery and suspected shigellosis. One hundred and fourteen Shigella isolates were recovered during this period; clinical data was available for analysis on 113 patients.
The final period (referred to as period C from here onwards) in which data was combined was a trial conducted at the HTD and at DTP between June 2006 and December 2008. This was a randomized controlled trial for comparing the treatment of dysentery with ciprofloxacin and gatifloxacin in Vietnamese children (controlled trials number ISRCTN55945881) (HV and SB, unpublished data). The inclusion criteria were as period A. One hundred and three isolates were collected during this period and clinical data on all admitted children was available for analysis.
All three studies were approved ethical assessment by the Scientific and Ethical Committee of the hospital for tropical diseases and Oxford University tropical ethics committee (OXTREC) number 010-06 (2006).
From all studies, stool samples were collected from patients andcultured directly on the day of sampling. Initial isolation was asbelow, however, all bacterial isolates were stored in glycerol at-80°C and re-serotyped and checked for consistency with the original antimicrobial susceptibility profile for the purposes of this work. All specimens were processed and checked in the microbiology laboratory of the HTD.
Samples were cultured overnight in selenite F broth (Oxoid, Basingstoke, UK) and onto MacConkey and XLD agar (Oxoid) at 37°C. Colonies suggestive of Salmonella or Shigella (non-lactose fermenting) were sub-cultured on to nutrient agar and were identified using a 'short set' of sugar fermentation reactions (Kliger iron agar, urea agar, citrate agar, SIM motility-indole media (Oxoid)). After incubation for 18 - 24 h at 37°C, the test media were read for characteristic Shigella reactions and API 20E test strips of biochemical reactions (Biomerieux, Paris, France) were used to confirm the identity of Shigella spp. Serologic identification was performed by slide agglutination with polyvalent somatic (O) antigen grouping sera, followed by testing with available monovalent antisera for specific serotype identification as per the manufacturers recommendations (Denka Seiken, Japan).
Antimicrobial susceptibility testing of all Shigella isolates against ampicillin (AMP), chloramphenicol (CHL), trimethoprim- sulfamethoxazole (SXT), tetracycline (TET), nalidixic acid (NAL), ofloxacin (OFX) and ceftriaxone (CRO) was performed by disk diffusion following standardized Clinical and Laboratory Standards Institute methods . The minimum inhibitory concentrations (MICs) were additionally calculated for all isolates by E-test, according to manufacturer's recommendations (AB Biodisk, Solna, Sweden) and were compared to control strain E. coli ATCC 25922 and an in house fully sensitive E. coli control.
Clinical observations and statistical analysis
Clinical data was recorded on specialized clinical report forms for all three studies by clinical staff involved in the studies. The data collected was related to basic details of the patient, age (months), sex, location of residence and weight (kg). A history from all patients was also recorded, including; duration of illness prior to admission to hospital (days), fever (defined as a prolonged temperature > 37.5°C), abdominal discomfort, vomiting, watery diarrhea (defined as three or more loose bowel movements during a 24-h period), bloody or mucoid diarrhea (defined as >3 loose stools with obvious blood or mucus), estimated number of episodes of diarrhea before attending hospital, convulsions believed to be related to fever and/or infection and if there was any known pretreatment with antimicrobials. A white blood cell count was performed on all patients and stools were examined by microscopy (HPF (× 400)) to identify white and red blood cells, these observations were scored on scale from zero to three, scale 0 = 0 cells/HPF, scale 1 = 1 to 10 cells/HPF, scale 2 = 11 to 20 cells/HPF and scale 3 = >20 cells/HPF. Time in hours (from initial investigation in hospital) to the ceasing of bloody/mucoid and watery diarrhea was recorded. Duration of hospital stay was recorded in days post admission; patients were only discharged when all clinical symptoms had resolved completely.
Data were double entered into Microsoft Excel for storage and manipulation. Mapping data was entered, analyzed and draw in MapInfo software (Pitney Bowes MapInfo Corporation, USA). For intergroup comparisons, Chi-square tests were used for comparison of categorical variables. For the analysis of continuous variables, Wilcoxon rank sum, and Kruskal-Wallis test were used for non-normally distributed data. A p-value of less than 0.05 (two-tailed) was considered significant. Statistical analysis was performed in R http://www.r-project.org/.