No Clinical Benefit of Empirical Antimicrobial Therapy for Pediatric Diarrhea in a High-Usage, High-Resistance Setting

We identified high antimicrobial usage and high antimicrobial resistance in children hospitalized with bloody and/or mucoid bacteria-associated diarrhea. There was no clinical benefit of antimicrobial therapy in children with diarrhea; adequately powered randomized controlled trials are required.

diarrhea episodes in Vietnamese children found to be associated with a viral pathogen and in 60% of cases with unknown etiology [7]. The excessive use of antimicrobials in animals and humans in Southeast Asia has led to the current antimicrobial resistance (AMR) crisis in the region, with increasing resistance against many first-line antimicrobials, including fluoroquinolones and third-generation cephalosporins, in many gram-negative pathogens across the region [8]. Therefore, a better understanding of the bacterial agents of diarrhea, their corresponding AMR profile, the impact of antimicrobial treatment on clinical outcome, and the effects of empirical antimicrobials is required.
Little has been reported regarding the disease burden and clinical management of hospitalized pediatric diarrheal diseases in Vietnam. Although NTS, Campylobacter, and Shigella have been identified as major bacterial causes of diarrhea in Vietnamese children [7,[9][10][11], the epidemiology, AMR profiles, treatment, and the associated outcome of these bacteria in pediatric diarrhea have not been not well described. The Vietnamese healthcare system currently follows the World Health Organization (WHO) guidelines for treatment of pediatric diarrhea. These guidelines recommend the use of low-osmolarity oral rehydration solution, zinc, and ciprofloxacin or 1 of the 3 alternatives (pivmecillinam, azithromycin, or ceftriaxone) for all patients with bloody diarrhea, irrespective of age [12,13]. However, as the prevalence of AMR in enteric pathogens increases across the region, it is uncertain how existing guidelines correspond to circulating AMR profiles, antimicrobial treatment practices, and patient outcomes in children hospitalized with diarrhea. Therefore, we conducted a prospective multicenter crosssectional study in Ho Chi Minh City to improve the understanding of bacteria-associated diarrhea in Vietnamese children and to assess the duration of hospital stay in diarrheal patients infected with bacterial pathogens and receiving antimicrobials.

Ethics
Ethical approval for this study was provided by the ethics committees of all 3 participating local hospitals and the University of Oxford Tropical Research Ethics Committee (OxTREC No. 1045-13). Written consent from parents or legal guardians of all participants was obtained before enrollment.

Study Design and Enrollment
This study was a prospective, observational, multicenter crosssectional study to evaluate the etiology, epidemiology, and outcomes in children (aged <16 years) hospitalized for diarrhea. Study participants were recruited from 3 tertiary hospitals (Children's Hospital 1, Children's Hospital 2, and the Hospital for Tropical Diseases) in Ho Chi Minh City, Vietnam, from May 2014 to April 2016.
Children hospitalized with diarrhea, defined as ≥3 passages of loose stools within 24 hours [12] with ≥1 loose stool containing blood and/or mucus, were recruited into the study. Based on characteristics of the diarrheal stools and the duration of illness, participants were classified into 3 groups: acute nonbloody diarrhea (diarrhea with mucus, <14 days), acute bloody diarrhea (diarrhea with blood, <14 days), and persistent diarrhea (diarrhea with mucus and/or blood, ≥14 days). Children were not eligible if they had suspected or confirmed intussusception at the time of enrollment. After enrollment, a short questionnaire was completed, and a fecal sample was collected and processed within 24 hours. All enrolled patients were provided with the routine standard-of-care practices at each hospital. Treatment and proxy outcomes, including patient recovery status at 3 days after enrollment and duration of hospitalization, were recorded by clinical staff at study sites. Patient status was recorded as "recovered" if the patient had <3 passages of loose stools in the past 24 hours or "improved" if the patient had fewer episodes of diarrhea and/ or less mucus and/or blood compared with status at enrollment.

Microbiological Methods
Fecal specimens were inoculated onto MacConkey agar (MC agar; Oxoid) and xylose-lysine-deoxycholate agar (Oxoid) and into selenite broth (Oxoid) and incubated at 37°C for 18-24 hours. Salmonella and Shigella were detected based on their characteristic appearance on xylose-lysine-deoxycholate and MC agar and confirmed using matrix-assisted laser desorption/ionization timeof-flight mass spectrometry (Bruker) and API20E (bioMerieux), following the manufacturer's guidelines. Campylobacter was identified using Campylobacter selective agar (Oxoid) under microaerophilic conditions, followed by Gram staining and microscopy.
Antimicrobial susceptibility testing was performed using the Kirby-Bauer disc diffusion method on Mueller-Hinton agar (Oxoid) for Salmonella and Shigella, and on blood agar containing 5% sheep blood for Campylobacter and interpreted using Clinical and Laboratory Standards Institute guidelines [14] (Supplementary Table S1). Multidrug resistance (MDR) was defined as nonsusceptibility to ≥1 agent in ≥3 antimicrobial categories listed in Supplementary Table S1. Microbiology results were reported to the collaborating hospitals within 3 days of sampling.

Statistical Analysis
Data were analyzed using Stata (version 11; StataCorp) and R (versions 3.2.2; R Foundation for Statistical Computing) software. Figures were constructed with R software, using the ggplot2 [15] and prodlim packages. Descriptive comparisons between groups were conducted using nonparametric tests, including the Fisher exact test for categorical variables and the Mann-Whitney U test for continuous data. Statistical comparisons between >2 groups were conducted using the χ 2 test and the Kruskal-Wallis test for categorical and continuous variables, respectively. Kaplan-Meier curves for length of hospital stay were compared between groups, using log-rank tests. The growth status of participating patients was assessed using the WHO global database on growth and nutrition [16], the Institute for Clinical Systems Improvement's guidelines on preventing and managing obesity in children and adolescents [17], and the macro package for Stata software (version 11) developed by WHO.
An accelerated failure time (AFT) regression model (incorporating all study patients hospitalized for ≥1 day) was constructed in 3 steps. First, the best AFT distribution fit was identified for the dependent variable, that is, length of hospital stay. Second, 11 demographic variables (eg, sex and age), clinical symptoms of disease severity, treatment types, and MDR were tested by means of univariate analysis, using a log-normal distribution. The 11 variables were chosen because of their potential to affect duration of hospital stay. Third, a multivariate log-normal model was constructed using a stepwise backward elimination method, where variables were removed based on the likelihood ratio test (P < .05).

General Characteristics of Patients Hospitalized With Diarrhea
Between May 2014 and April 2016, a total of 3166 hospitalized children meeting the study criteria were recruited at the 3 study hospitals. The majority of patients were male (1945 of 3166; 61.4%), with ages ranging from 1 month to 15 years (median age, 10 months; interquartile range [IQR], 6.5-16.7 months). Patients were hospitalized for a median of 5 days (IQR, 3-7 days) with 88.7% (2808 of 3166) of patients showing improvement or resolving symptoms within 3 days of enrollment.
We cultured the 3 key bacterial diarrheal pathogens in Vietnam (ie, NTS, Campylobacter, and Shigella), isolating 816 pathogens from 804 patients (11 coinfections) and stratified clinical manifestations and treatment data by these organisms. At least 1 of these bacteria was isolated from the fecal specimens of 44.3% with bloody diarrhea (485 of 1096) and 17.9% (318 of 1775) with nonbloody diarrhea (P < .001) ( Table 1). Overall, NTS was the most frequently isolated of the 3 bacterial pathogens from the diarrheal children, accounting for 15.1% (478 of 3166) of all diarrheal cases, followed by Campylobacter and Shigella. Shigella infections were more common in older children (median age, 3 years) and were associated with more severe symptoms. However, children with Shigella infections recovered more rapidly than those infected with NTS or Campylobacter (Supplementary Table S2).

Antimicrobial Usage for Treatment of Hospitalized Diarrheal Diseases
We also recorded the treatment regimens of the enrolled patients, which included oral rehydration solution, intravenous rehydration, zinc, probiotics, and antimicrobials. More than 90% of patients were administered oral rehydration solution, and >80% were given zinc supplementation. The use of antimicrobials within this population was high, with 85.2% of patients (2697 of 3166) receiving empirical antimicrobial treatment after admission to the hospital and before a bacterial culture result was obtained. Fluoroquinolones were the most commonly used class of antimicrobials (1799 of 2697; 66.7%). Differences in standard-of-care treatment were observed between patients with bloody versus nonbloody diarrhea (Table 1); antimicrobials were more regularly administered to patients with bloody diarrhea than to those with nonbloody diarrhea (P < .001). Antimicrobials, specifically fluoroquinolones, were commonly (>70%) prescribed before an etiological diagnosis in those eventually found to be infected with Campylobacter, Salmonella, or Shigella (Supplementary Table S2).

Antimicrobial Susceptibility
Campylobacter, Salmonella, and Shigella isolates displayed a high prevalence of nonsusceptibility against many of the screened antimicrobials ( Figure 1

Diarrheal Disease Outcome
We assessed the effect of antimicrobial treatment on 2 proxy disease outcome measures, clinical outcome (ie, improved/ recovered) at 3 days after enrollment and the duration of hospital stay. More than 80% of patients showed improvement or had recovered at 3 days after enrollment, regardless of antimicrobial treatment (Supplementary Figure S1). However, those given an antimicrobial, specifically a fluoroquinolone, had a longer hospital stay than those not receiving an antimicrobial (P < .001 and P = .01, respectively) ( Figure 2). Notably, the duration of hospital stay did not differ significantly between those receiving and those not receiving an antimicrobial among patients with bloody diarrhea ( Figure 2C). However, antimicrobial treatment in those with nonbloody diarrhea was significantly associated with a longer hospital stay (median [IQR] hospital stay for antimicrobial vs no antimicrobial use, 5 [3][4][5][6][7] vs 4 [3][4][5] days; P < .001; (Figure 2C). Similarly, antimicrobial treatment in patients with low CRP levels (≤5 mg/L) was significantly associated with an increased hospital stay, compared with patients with high CRP levels (>5 mg/L) (P < .001; Figure 2D).
We then stratified all patients by antimicrobial treatment, and within those treated with antimicrobials we compared disease outcome (ie, duration of hospitalization) between those infected with MDR or non-MDR organisms and between those infected with fluoroquinolone-susceptible or nonsusceptible organisms. In patients empirically treated with any antimicrobial or a fluoroquinolone, >90% (526 of 564) and 70% (361 of 515) were infected with an MDR or a fluoroquinolone nonsusceptible organism, respectively. However, regardless of the MDR status and fluoroquinolone nonsusceptibility of the infecting organisms, no differences in the duration of hospitalization were observed among patients treated with antimicrobials ( Figure 3). Comparable findings were observed for recovery status at 3 days after enrollment (Supplementary Figure S1). All children with nonbloody diarrhea had mucus in stools. c P values represent comparisons between nonbloody and bloody diarrhea using Fisher exact test for categorical data or Mann-Whitney U test for continuous data. d Obese: weight for length z score >3 standard deviations [SDs] in children aged <24 months; body mass index (BMI) for age z score >3 SDs in children aged ≥24 months. Overweight: weight for length z score >2 SDs in children aged <24 months; BMI for age z score >2 SDs in children aged ≥24 months. Wasted: weight for length z score ≤2 SDs in children aged <24 months; BMI for age z score ≤2 SDs in children aged ≥24 months. Severely wasted: weight for length z score ≤3 SDs in children aged <24 months; BMI for age z score ≤3 SDs in children aged ≥24 months [18]. e Dehydration classified as described by Basaleem and Amin [19]. f Percentage of those receiving antimicrobials. Fluoroquinolones included ciprofloxacin and norfloxacin. g Condition was described as "recovered" if patient had <3 passages of loose stool in the past 24 hours or "improved" if patient had fewer episodes of diarrhea and less mucus and/or blood than at enrollment.
An AFT multiple regression model was considered to be an appropriate method for describing how each of the adjusted variables multiplicatively alters the duration of hospitalization. Therefore, an AFT was constructed to investigate the effect of antimicrobials on the duration of hospital stay, adjusting for age, disease severity, and other prescribed treatments. After adjustment for age, diarrhea presentation, and dehydration, antimicrobials were associated with a significant increase in the duration of hospital stay of diarrheal patients by a time ratio of 1.32 [1.24-1.41]. Finally, infection with an MDR organism was found to not significantly prolong hospitalization (P = .55; Table 2).

DISCUSSION
Moderate-to-severe diarrhea has a significant healthcare burden in Vietnamese children [20]. Although previous observational studies of diarrhea in Vietnam have described some of the epidemiological features, the bacterial causes, and their associated AMR profiles, these studies have focused chiefly on children with acute watery diarrhea [7]. Little is known about the epidemiology and clinical management of bloody and/or mucoid diarrhea in Vietnam. Therefore, we aimed to address this paucity of data by enrolling >3000 of children hospitalized with diarrheal. This large sample size not only enabled the isolation of >800 enteric pathogens, but it also provided data regarding antimicrobial usage in medical practice and outcome. Furthermore, this study investigated the clinical role of AMR in a relevant population empirically prescribed antimicrobials at presentation to the hospital.
AMR in pathogenic bacteria, including those associated with diarrhea, is a global public health problem [21,22]. Data from Vietnam highlight the increasing trend in diarrheagenic bacteria of AMR to the current first-line antimicrobials, such as fluoroquinolones and third-generation cephalosporins. Moreover, despite recent increases in fluoroquinolone resistance in Asia and beyond, current guidelines still endorse the use of this class of antimicrobials to treat bloody diarrhea [23][24][25]. Many of the organisms isolated during this investigation were also nonsusceptible to other (nonfluoroquinolone) antimicrobials, including some "last resort" choices, such as imipenem.
In comparison to estimates from other industrializing countries, we observed a similar or elevated prevalence of NTS exhibiting nonsusceptibility to third-generation cephalosporins and increased nonsusceptibility to both ciprofloxacin and azithromycin [26,27]. An extraordinarily high prevalence (~90%) of fluoroquinolone-resistant Campylobacter has been recently observed in other industrializing countries [28,29]. The situation in Vietnam seems to be exacerbated by the nonsusceptibility of Campylobacter isolates to macrolides. The Shigella isolated here also had a high MDR rate; emerging MDR Shigella isolates with resistance to fluoroquinolones and extended-spectrum cephalosporins are now commonly reported across Asia [10,30,31].
The treatment of diarrhea with antimicrobials is a complex issue. Apart from the limited capability of most LMICs to confirm etiological agents associated with disease and the current complication of increasing AMR, there are conflicting data regarding the clinical efficacy of antimicrobials in reducing symptoms [33]. In line with WHO recommendations, we observed that >85% of patients hospitalized with diarrhea containing mucus and/or blood were prescribed an antimicrobial, most commonly fluoroquinolones (ciprofloxacin/norfloxacin). Our current analysis found that antimicrobial use during hospitalized diarrhea did not add benefit to supportive therapy only (ie, rehydration and zinc supplementation).
The recovery of patients regardless of antimicrobial treatment or AMR status when treated with a first-line antimicrobial for bloody diarrhea treatment (eg, ciprofloxacin) may be explained by either the differing in vivo effects of antimicrobials or the self-limiting nature of the infections. In addition, in diarrhea with less pronounced inflammation (indicated by the absence of blood and/or a low CRP level), the use of antimicrobials was associated with a prolonged hospital stay. These observations support previous findings in studies of NTS and Campylobacter, where antimicrobial treatment did not provide a clinical advantage and sometimes even caused harm (compared with placebo) in decreasing the duration of symptoms [34,35]. The routine antimicrobial treatment may also affect the transmission of diarrhea-causing bacteria by increasing fecal carriage and consequently spreading AMR organisms [34]. The main limitation in the present study is that this investigation was observational, with patients receiving standard-of-care treatment, which made it difficult to assess whether our observations were biased by more severe cases being prescribed antimicrobials. Furthermore, the duration of hospital stay and recovery of patients at 3 days after enrollment were proxy measures of clinical    outcome, and we cannot discount that some children may have been discharged before the cessation of symptoms.
In conclusion, bacteria associated with pediatric diarrhea in southern Vietnam displayed an extensive AMR profile, thereby emphasizing the significance of etiological diagnosis of diarrhea in LMICs. Our key finding was that antimicrobial treatment was not associated with a reduction in diarrheal symptoms and even prolonged hospital stay in some groups. Therefore, we urge that adequately powered randomized controlled trials be conducted to better assess the potential benefits of antimicrobial therapy for treatment of diarrhea. These data will become essential for controlling antimicrobial usage during the present AMR crisis.

Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.