NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

National Collaborating Centre for Women's and Children's Health (UK). Diarrhoea and Vomiting Caused by Gastroenteritis: Diagnosis, Assessment and Management in Children Younger than 5 Years. London: RCOG Press; 2009 Apr. (NICE Clinical Guidelines, No. 84.)

Cover of Diarrhoea and Vomiting Caused by Gastroenteritis

Diarrhoea and Vomiting Caused by Gastroenteritis: Diagnosis, Assessment and Management in Children Younger than 5 Years.

Show details

7Antibiotic therapy

Introduction

Gastroenteritis is most often caused by a viral enteric pathogen, and even in those with bacterial or protozoal infections the disease is generally self-limiting. Treatment has usually been recommended for dysenteric shigellosis and amoebiasis, cholera, pseudomembranous colitis, and for some other organisms in particular circumstances.140 The evidence relating to the use of antibiotics in young children with gastroenteritis was reviewed with a view to making recommendations on good practice.

Clinical question

What is the role of antibiotic therapy in children with gastroenteritis?

A search for randomised trials or systematic reviews comparing antibiotic treatment with placebo or no treatment of gastroenteritis in children was performed. Of 130 citations identified, 25 were retrieved in full copy. Of these, 11 were excluded, but 14 compared use of antibiotics with placebo and were included in this review.

7.1. Salmonella

Evidence overview

Four randomised trials141–144 evaluated the effectiveness of antibiotic treatment compared with placebo or no treatment for children presenting to hospital with acute diarrhoea. These trials were conducted in the USA,141 Taiwan,142 Canada143 and Colombia.144 Three trials had three treatment arms141–143 and one trial144 had two treatment arms but according to protocol changed the route of antibiotic administration from intramuscular (IM) to oral in the second year of the study.

The first RCT, from the USA,141 included children aged up to 8 years (n = 45) presenting at hospital with acute diarrhoea and with salmonella species subsequently isolated in rectal swab cultures. Children with a history of adverse drug reactions to penicillins, with another focus of infection or who were under 6 weeks of age were excluded. Participants were randomised to one of three treatment arms to compare the effects of ampicillin (100 mg/kg per day) (n = 15), amoxicillin (100 mg/kg per day) (n = 15) and placebo (n = 14) given in four equal doses daily for 5 days for the treatment of salmonella gastroenteritis. Computer-generated random number lists were used to assign the children to pre-coded drugs. Separate randomisation lists were used for children under and over 1 year of age. The main outcomes assessed were the mean days until diarrhoea stopped (diarrhoea cessation defined as the day of the first formed stool without mucus), the mean days until diarrhoea improved (defined as improved stool consistency and a decrease in the number of stools), the mean days until the first negative culture (defined as the first of at least two consecutive negative cultures), the mean days until last positive culture, bacteriological relapse and diarrhoea relapse. Outcome assessors were blind to the treatment the children received. [EL = 1−]

The groups were broadly comparable at baseline, except that children receiving amoxicillin were younger (mean age 7.7 ± 1.7 months) than those in the ampicillin (mean age 15.7 ± 5.7 months) and placebo groups (mean age 19.8 ± 7.4 months). There were no statistically significant differences between the ampicillin, amoxicillin and placebo groups for the mean number of days until diarrhoea stopped or the mean number of days until the first negative culture. Participants receiving ampicillin and amoxicillin did have a significantly reduced mean number of days until diarrhoea improved compared individually with placebo (WMD −1.20 days; 95% CI −1.65 to −0.75 days and WMD −1.00 days; 95% CI −1.45 to −0.55 days, respectively), but the difference across all three groups was not significant (Kruskal–Wallis nonparametric ANOVA P > 0.2). Excretion of salmonella continued for significantly longer in the ampicillin group compared with the placebo group (days until last positive culture WMD 20.40 days; 95% CI 13.49 to 27.31 days) and the amoxicillin group compared with placebo (days until last positive culture WMD 16.10 days; 95% CI 8.75 to 23.45 days), but differences across all three groups were not statistically significant (Kruskal–Wallis nonparametric ANOVA P > 0.5). Eight patients in each antibiotic treatment group developed bacteriological relapse (16/30), although there were no relapses in the placebo group. This difference was statistically significant (P = 0.003). All but one of the relapses occurred between days 4 and 20 after entering the study. Three patients with bacteriological relapse from each antibiotic group also suffered a diarrhoeal relapse (6/30). There were no diarrhoeal relapses in the placebo group (n = 14), but this difference was not statistically significant.

The second three-armed study, from Taiwan,142 recruited 42 children older than 6 months presenting to hospital with suspected salmonella enteritis (defined as blood and/or mucoid diarrhoea with or without fever). Confirmation and serotyping of salmonella was performed using stool culture. Children were excluded if they had a negative salmonella stool culture, a toxic appearance, were vomiting, had abdominal distension indicative of sepsis or ileus or if they had taken antibiotics in the 72 hours prior to admission. Participants were assigned to treatment with oral azithromycin 10 mg/kg per day, in one dose daily for 5 days (n = 14), cefixime 10 mg/kg per day, in two doses daily for 5 days (n = 14) or to no treatment (n = 14). A computer-generated random number program was used to assign the children to treatment groups. No details of blinding of outcome assessors or of a power calculation were given. Patients returned to clinic 7 days after completing the course of treatment, and every week thereafter for stool sampling and culture, until two consecutive normal stools were noted. The main outcomes assessed were duration of diarrhoea and fever after initiation of therapy, and clinical or bacteriological relapse. [EL = 1−]

The three groups were similar at the baseline for sex and clinical parameters but children receiving cefixime were younger those in the other two groups (P < 0.05). No statistically significant differences were found between azithromycin or cefixime compared with each other or to no treatment for the mean duration of diarrhoea post treatment or the proportion of patients with positive cultures at week 3 post treatment.

A third three-armed treatment trial, from Canada,143 included children aged between 10 months and 15 years who were admitted to hospital with a history of diarrhoea, fever for 3 days or more and/or mucus and blood from diarrhoeal stools. Only those with a subsequent positive culture for salmonella remained in the study. Children who had received antibiotics in the previous 5 days or who had renal or hepatic disease, blood dyscrasia, or salmonella bacteraemia were excluded. Participants were randomly assigned to treatment with 20 mg/kg per day trimethoprim plus 100 mg/kg per day sulfamethoxazole oral suspension four times per day for 7 days (n = 14) or ampicillin 100 mg/kg per day oral suspension or capsules four times per day for 7 days (n = 10) or no antibiotic treatment (n = 12) for the management of salmonella enteritis. No details regarding allocation concealment, sequence generation, blinding of outcome assessors or a power calculation were provided. Follow-up was by daily physical examination and culture of stool specimens, during treatment. After treatment had finished, two or three consecutive daily stool samples were taken for culture at 1 week, 8 weeks and 6 months. Family contacts also had stool cultures performed at admission and as for participants post therapy. [EL = 1−]

The treatment groups were comparable at baseline for age and clinical parameters. No statistically significant differences were noted between the trimethoprim/sulfamethoxazole, ampicillin, or no treatment groups for the mean duration of diarrhoea after start of therapy (2.8, 3.1 and 3 days, respectively), the mean duration of hospitalisation after start of therapy (5.3, 5 and 6 days, respectively) or the mean duration of fever after start of therapy (3.2, 1.6 and 2.6 days, respectively).

The fourth trial, from Colombia,144 examined the effect of ampicillin versus placebo on salmonella infection. One hundred and ten of 282 malnourished infants and children younger than 2 years admitted to hospital with diarrhoea as a major symptom had salmonella isolated from culture of stool specimens. Children were recruited into the study once culture confirmation of shigella, salmonella or E. coli was made from rectal swab and stool specimens taken 12–16 hours previously. One patient without recognised pathogens was entered into the study for every two patients with shigella, salmonella or E. coli. Treatments were given intramuscularly (IM ampicillin versus IM sterile fructose) in the first year of the trial, and orally in the second (oral suspension of 100 mg/kg per day ampicillin or placebo suspension in equally divided doses every 6 hours for 5 days). Fifty-seven participants received either IM or oral ampicillin and 53 received either IM fructose or oral placebo. A random numbers list was used to assign children to treatment groups. The treatments were prepared in foils for suspension in water before use and allocation was not known to investigators, patients or outcome assessors. [EL = 1+]

The treatment groups were similar at baseline for age, sex, race, diarrhoeal duration, dehydration status, clinical examinations and prior treatment with antimicrobial or antidiarrhoeal drugs. No statistically significant differences were reported between the ampicillin and placebo groups for the mean number of days until diarrhoea improved or ceased or for the mean number of days until the patient became afebrile or culture negative. No patient receiving IM ampicillin relapsed (reversion to positive cultures after a period of negative culture) after the 5 day course of therapy. Although details of relapse in the placebo group were not presented, the authors asserted that this finding was statistically significant (P = 0.02) and that fewer patients receiving IM antibiotics than those receiving either placebo (P = 0.01) or oral ampicillin (P = 0.04) became short-term salmonella carriers (culture positive any time after completion of therapy).

Evidence summary

The trials included to inform this section were generally of poor quality (except for one RCT) with small sample sizes, and there was variation in the specific antibiotics used. Nevertheless, there was consistent evidence from the trials suggesting that antibiotic treatment did not shorten the duration of diarrhoea or lead to an earlier resolution of clinical symptoms. The good-quality RCT reported that IM ampicillin protected children against relapse and carriage of salmonella infection significantly better than placebo or oral ampicillin, while another trial (with EL = 1−) reported an increase in the carrier rate with oral ampicillin and amoxicillin use compared with the placebo.

7.2. Campylobacter

Evidence overview

Three RCTs were identified that compared erythromycin treatment with placebo or no treatment for campylobacter enteritis, and they were conducted in South Africa, Canada and Peru.

The first RCT, from South Africa,145 examined the effect of erythromycin treatment for campylobacter-associated enteritis compared with placebo in infants aged 1–24 months. Children admitted to hospital with diarrhoea of less than 96 hours’ duration and who had not received any antimicrobial therapy for this illness were included in this study (n = 26). Confirmation of Campylobacter jejuni and any other infection was from microscopic and culture examination of stool samples. Results for those children who were infected only with Campylobacter jejuni (n = 8) are presented here. Exclusion criteria details were not provided. Participants were randomised to receive an oral suspension of either 40 mg/kg per day erythromycin (n = 4) or placebo (n = 4) in divided doses for 5 days. Treatments were supplied as granules for reconstitution in pre-coded containers such that patients, investigators and outcome assessors were blind to treatment allocation. Follow-up was by daily physical and stool examination for 1 week. [EL = 1−]

Treatment groups were similar at baseline for age, sex, duration of diarrhoea, dehydration severity and weight. Although the study was well conducted, causative organisms were identified retrospectively and only eight children with Campylobacter jejuni infection alone were included, reducing the power of the study for these results. No statistically significant differences were found between the erythromycin and placebo groups for the mean durations of abnormal stool frequency and consistency, vomiting, dehydration or fever.

One RCT146 conducted in Canada recruited children of up to 12 years of age (and their household contacts) following prospective identification of a positive, erythromycin-sensitive stool culture of campylobacter. Children with symptoms of enteritis were recalled to hospital and were allocated to no treatment (n = 12) or to treatment with 40 mg/kg per day erythromycin every 6 hours for 7 days (n = 15). Exclusion criteria were presence of other enteric pathogens in the stool, lack of symptoms and antibiotic therapy being given in the 2 weeks prior to recruitment. No details of the randomisation process were presented, and assessors were not blinded to treatment allocation. Children were followed up until the entire household had three consecutive negative (weekly) stool samples. [EL = 1−]

The trial reported no statistically significant differences in the mean duration of diarrhoea experienced by participants receiving erythromycin or no treatment. The range in number of days with diarrhoea was 1–6 days in the erythromycin group and 1–15 days in the group receiving no treatment. A statistically significant difference was found in the mean number of days until the first negative culture between those children receiving erythromycin (2.0 ± 1.3 days) and those receiving no treatment (16.8 ± 12.5 days) (P < 0.001).

The third RCT, from Peru,147 examined the effects of early erythromycin treatment for campylobacter-associated enteritis compared with placebo in children aged 3–60 months brought in as outpatients for treatment of acute diarrhoea (n = 24). Participants had five or more loose stools a day with gross blood or mucus for no longer than 5 days and had not received antibiotics for another illness. Children with clinical signs of dehydration or who were under the third percentile for weight/length ratio (US National Center for Health Statistics standard) or who had had a separate episode of diarrhoea in the previous fortnight were excluded. Participants were randomised to receive 50 mg/kg per day erythromycin oral suspension (n = 14) or placebo oral suspension (n = 10) in four doses for 5 days. Treatments were randomised and pre-coded by the manufacturers such that patients, investigators and outcome assessors were blind to treatment allocation. Allocation to treatment groups was prior to stool culture confirmation of campylobacter. Follow-up by stool culture and parental reporting of symptoms was performed for 5 days. [EL = 1+]

The mean duration of diarrhoea was significantly lower in the patients receiving erythromycin (2.4 ± 0.4 days) compared with placebo (4.2 ± 0.3 days), (WMD −1.80 days; 95% CI −2.08 to −1.52 days). However, the number of patients with normal stools at 5 days in the erythromycin group (13/14) was not significantly different from those receiving placebo (5/10) (RR 1.86; 95% CI 0.98 to 3.51). The mean number of days until last positive stool culture was significantly lower for those receiving erythromycin (0.5 ± 0.3 days) compared with the placebo group (2.2 ± 0.6 days) (WMD −1.70 days; 95% CI −2.10 to −1.30 days), but no statistically significant difference was found in the proportion of patients in each group with a positive stool culture at 5 days.

Evidence summary

Of the three available small RCTs, one trial had insufficient power to detect any statistically significant treatment differences in a small subgroup who received erythromycin or placebo for treatment of Campylobacter jejuni enteritis. There were conflicting results from the two remaining studies. One trial [EL = 1−] found no difference between the groups in mean duration of diarrhoea. The second trial [EL = 1+] found that the mean duration of diarrhoea was shorter with erythromycin treatment, although the ‘diarrhoeal cure’ rate at 5 days was similar between treatment groups. The difference in mean duration of diarrhoea might be explained by the second trial’s early recruitment of participants to treatment groups without awaiting stool culture confirmation of campylobacter. Thus patients would be less likely to have had long episodes of diarrhoea prior to treatment and might be more uniform in severity of illness during the study. Both studies found that erythromycin treatment demonstrated antibacterial efficacy by either reducing the mean number of days until first negative stool culture or the last positive culture. However, it could not be established in the EL = 1+ trial whether erythromycin treatment caused fewer patients to excrete campylobacter at day 5 compared with placebo.

7.3. Yersinia

Evidence overview

One RCT was identified. This study148 was conducted in Canada and examined the treatment of yersinia enteritis with trimethoprim/sulfamethoxazole compared with placebo (n = 45) Participants were children younger than 15 years with symptomatic enteritis. Prior to recruitment, stool samples from participants had been positively cultured for yersinia. Participants and their household contacts were followed until all had three consecutive negative (weekly) stool samples. Clinical symptoms were assessed and reported daily by a parent and stool specimens were obtained for the first 7 days, then weekly thereafter. There was about 25% loss to follow-up and results for only 34 children were presented. The mean age of children in the antibiotic group (n = 18) was 2.5 ± 3 years and was 3.6 ± 4.3 years in the treatment group (n = 16). There were no statistically significant differences between the antibiotic and placebo groups for the clinical outcomes (median duration of diarrhoea, the number of patients experiencing diarrhoea for less than 7 days, or recurrence of diarrhoea). However, statistically significant differences between the treatment groups were found for bacteriological parameters. The findings for the median number of days until ‘bacteriological cure’ (erythromycin median 5.5 days, range 2–53 days versus placebo, median 17.5 days, range 3–62 days; P < 0.005) and for the number of patients with positive stool cultures at the end of treatment (erythromycin 2/18 versus placebo 13/16; P < 0.001) both favoured antibiotic use. However, more participants taking antibiotics (7/18) had a bacteriological relapse compared with those taking placebo (0/16) (P < 0.05). [EL = 1−].

Evidence summary

Evidence from one RCT suggested that erythromycin treatment compared with placebo for yersinia gastroenteritis did not make a significant difference to clinical diarrhoeal outcomes. Although erythromycin did reduce the time taken for patients to stop excreting yersinia, its administration caused more patients to have bacteriological relapses compared with placebo.

7.4. Shigella

Evidence overview

One RCT was identified. This study144 compared the effects of ampicillin with that of placebo in infants and children younger than 2 years admitted to hospital in Colombia with diarrhoea as a major symptom (n = 282). Children were recruited once culture confirmation of shigella, salmonella or E. coli was made from rectal swab and stool specimens taken 12–16 hours previously. One patient without recognised pathogens was entered into the study for every two patients with shigella, salmonella or E. coli. Treatments were given intramuscularly (IM ampicillin versus IM sterile fructose) in the first year of the trial, and orally in the second (oral suspension of 100 mg/kg per day ampicillin or placebo suspension in equally divided doses every 6 hours for 5 days). Overall, 37 participants had shigella infection – 16 received ampicillin and 21 received placebo. No statistically significant differences were found between the treatment groups for the diarrhoeal outcomes (mean number of days until diarrhoea improved, 2.4 versus 4.6 days, and mean number of days until diarrhoea ceased, 4.4 versus 6.8 days). IM ampicillin and the combined results for IM and oral ampicillin significantly reduced the mean number of days until the patient became afebrile, compared with the placebo (less than 0.5 versus 2.6 days; P < 0.05, and less than 0.5 versus 1.6 days; P < 0.05, respectively). Similarly, IM ampicillin and the combined results for IM and oral ampicillin significantly reduced the mean number of days until the patient became stool culture negative (0.4 versus 1.8 days; P < 0.01, and 0.9 versus 2 days; P < 0.05, respectively). IM ampicillin was found to be more effective in this respect than oral administration (0.4 versus 1.5 days; P < 0.05). [EL = 1+]

Evidence summary

Although the trial was well conducted, the sample size of children evaluated for this relevant section was small. The results suggested that children treated with ampicillin for shigella took fewer days to become afebrile and to stop excreting the organism than those treated with placebo. IM ampicillin was more effective than oral ampicillin in reducing the time to becoming stool culture negative, but the use of ampicillin did not reduce the time to improvement or cessation of diarrhoea.

7.5. Escherichia coli

Evidence overview

The Colombian trial described above144 also investigated the effects of ampicillin compared with placebo on serology-confirmed enteropathogenic E. coli infection, as well as shigella and salmonella infections (total n = 282). In total, 35 of 282 infants and children younger than 2 years admitted to hospital with diarrhoea as a major symptom had E. coli isolated by stool culture. Of these, 18 received either IM or oral ampicillin (100 mg/kg per day in equally divided doses every 6 hours for 5 days) and 17 received either IM fructose or oral placebo in every 6 hours for 5 days. There were no statistically significant differences between either ampicillin groups, or between the ampicillin and placebo groups in the mean number of days until diarrhoea improved or ceased or in the mean number of days until the patient became afebrile or stool culture negative. [EL = 1+]

Evidence summary

Results from one trial showed no statistically significant differences between either ampicillin groups or between the ampicillin and placebo groups in the mean number of days until diarrhoea improved or ceased or in the mean number of days until the patient became afebrile or culture negative. Although this trial was of good quality, there were only 35 children evaluated to inform this question.

7.6. Cryptosporidium

Evidence overview

Two studies were identified for inclusion. One was a good-quality trial from Zambia but it included HIV-seropositive and malnourished children, while the other trial had methodological limitations and included both children and adults as the study population.

An RCT149 conducted in Zambia recruited children (n = 100) who were admitted to hospital with diarrhoea, who had Cryptosporidium parvum oocytes identified from a pre-enrolment stool sample and whose parents consented to the child having an HIV test. All children were stabilised with fluid therapy, antibiotics and mineral supplementation as required throughout the study. Exclusion criteria were age under 1 year and receipt of a drug with antiprotozoal activity within 2 weeks of enrolment to the study. Following stratification by HIV status (n = 50 in each group), children were randomly assigned using a computer-generated random number list to treatment with 20 g/l nitazoxanide (n = 25) or placebo (n = 25) oral suspension (5 ml twice daily for three consecutive days). The patients and outcome assessors were not aware of the treatment allocation. The children were followed up in hospital for 8 days. The main outcomes were the clinical response on day 7 (well or continuing illness), the parasitological response, the time from first treatment to last unformed stool, and mortality by day 8. Results are presented here for the HIV-seronegative subgroup only (n = 50). [EL 1+]

Twenty-five HIV-seronegative children were allocated to each treatment group, although three from the placebo group were subsequently excluded as no Cryptosporidium parvum oocytes were detected in their stool at baseline. The groups were similar for sex, age, weight, malnutrition status, laboratory abnormalities and stool frequency. Children receiving placebo had a shorter mean duration of diarrhoea prior to enrolment than the nitazoxanide group, but this reflected one child in the nitazoxanide group reporting 91 days of diarrhoea prior to enrolment. Eleven children in each group had physical signs of malnutrition. Oedema indicative of kwashiorkor or marasmic kwashiorkor was detected in 22 children (10/25 nitazoxanide, 12/22 placebo).

Fourteen of 25 children receiving nitazoxanide were ‘well’ (rather than having ‘continuing illness’) at day 7 compared with five of 22 of those receiving placebo. ‘Well’ was defined as having no symptoms, no watery stools and no more than two soft stools within the previous 48 hours. ‘Continuing illness’ was defined as not fulfilling the definition of ‘well’. This difference was statistically significant (P = 0.037). At day 7, significantly more children in the nitazoxanide group (13/25) had a parasitological response (defined as two negative stool examinations) compared with the placebo group (3/22) (P = 0.007). There were no deaths in the nitazoxanide group but four children in the placebo group died, three from persistent diarrhoea with septicaemia and one who additionally had congestive heart failure (re-feeding syndrome). This difference was statistically significant (P = 0.041).

A comparative trial150 conducted in Egypt was identified that examined the effect of nitazoxanide and co-trimoxazole compared with placebo for clinical and microbiological ‘cure’ of cryptosporidium infection. This trial was poorly reported, with no details regarding the methods used or the baseline comparability of the treatment groups. Consequently, it was considered to be potentially highly biased. [EL = 1−]

Of 1087 patients with diarrhoea examined in the outpatient clinic, 150 were found to have cryptosporidiosis. This was confirmed by two stool diagnostic tests (Ziehl–Neelsen stain and direct immunofluorescent technique). Of these 150 patients, 73 were children. Adults and children were divided into three treatment groups (nitazoxanide, co-trimoxazole or placebo) and were followed for up to 10 days. Clinical cure was not defined in the study. Microbiological cure was defined as two consecutive negative stool samples. Results of numbers of patients ‘cured’ were presented, although it was not clear which ‘cure’ was measured and when measurements were taken – microbiological cure by the 10th day was presumed. There was a statistically significant difference in the number of children cured (21/24) following administration of nitazoxanide compared with placebo (9/25) (RR 2.43; 95% CI 1.41 to 4.19; P = 0.001). A statistically significant difference was not demonstrated for the comparison of co-trimoxazole (8/24) versus placebo (RR 0.93; 95% CI 0.43 to 2.00; P = 0.84).

Evidence summary

A good-quality RCT conducted in a population of malnourished children showed that nitazoxanide was effective in achieving a clinical and parasitological response to treatment, and in preventing deaths. Results from the other, potentially biased, controlled trial suggested that nitazoxanide but not co-trimoxazole was effective in achieving a microbiological cure in children younger than 12 years with diarrhoea of cryptosporidium infection.

7.7. Treatment without prior identification of a pathogen

Evidence overview

Four RCTs were identified as relevant: two studies were conducted in South Africa151,152 and two in Mexico.153,154 Data from 378 children (aged 2 months to 7 years) suffering from gastroenteritis were collected across the four studies, Three trials had two treatment arms151–153 and one trial had three treatment arms.154 The proportion of patients randomised but lost to follow-up was reported in all the studies (less than 20%). None of the trials included a sample size power calculation.

A trial conducted in South Africa151 compared the effect of trimethoprim/sulphonamide with placebo for the treatment of gastroenteritis. Children aged 5–30 months admitted to hospital for gastroenteritis (n = 34) were randomised into two treatment arms. There were 18 participants in one group and 26 in the other, although the authors did not specify which group received which treatment. No details of the randomisation process were given, but the investigators, patients and outcome assessors were blinded to treatment allocation. At baseline, the treatment groups were similar for age. No statistically significant differences were found between the two groups for the mean duration (in days) of diarrhoea, vomiting, pyrexia or hospital stay. [EL = 1−]

The second trial from South Africa152 examined the effect of erythromycin compared with placebo for the treatment of non-specific gastroenteritis. Included in the study were children aged 1 month to 2 years who had been admitted to a hospital with a history of diarrhoea not exceeding 96 hours and who had received no antimicrobial therapy for the illness (n = 78). Participants were randomised into two treatment groups and received either erythromycin oral suspension 40 mg/kg per day in divided doses for 5 days (n = 39) or placebo oral suspension (n = 39). Follow-up was by daily examination for 7 days. Treatments for the trial were supplied by the manufacturer as granules for reconstitution in pre-randomised and coded containers. The patients and the outcome assessors were blind to treatment allocation. The comparability of the groups at study entry was adequate. The distribution of pathogens was similar between groups and results were presented for 32 antibiotic and 33 placebo recipients. There were no statistically significant differences between the two groups in the mean duration of abnormal stool frequency, vomiting, dehydration or fever. However, the erythromycin group had a statistically significantly shorter mean duration of abnormal stool consistency compared with the placebo group (WMD −0.80 days; 95% CI −1.46 to −0.14 days). [EL = 1+]

A third RCT, conducted in Mexico,154 with three treatment arms compared the effects of furazolidone, trimethoprim/sulfamethoxazole and no antibiotic treatment for acute invasive diarrhoea in children. Patients aged 2–59 months brought to hospital and seen in outpatients with three or more watery stools in the previous 24 hours, with up to 5 days of diarrhoea prior to admission, and with presence of polymorphonuclear (PMN) leucocytes and blood in stool (n = 125) were included in the study. Patients who had received antimicrobials or antidiarrhoeal drugs in the previous 48 hours, who had amoeba in their stools, who had any concomitant disease or who had allergy or intolerance to the study drugs were excluded. Following a complete physical examination and submission of a stool specimen, participants were randomised to receive 7.5 mg/kg per day furazolidone in four equal doses a day for 5 days (n = 42), 8 mg/kg per day trimethoprim + 40 mg/kg per day sulfamethoxazole in two equal doses a day for 5 days (n = 52) or no treatment (n = 24). Participants were followed up with daily visits as outpatients to hospital, clinical assessment at day 3 and stool samples taken at days 1 and 6. Treatment success for participants with an identified pathogen was defined as clinical cure (absence of diarrhoea and alleviation of all symptoms) at day 3 and bacteriological cure (negative stool culture) at day 6. For patients with negative culture, treatment success was defined as clinical cure (absence of diarrhoea and alleviation of symptoms) at day 3. The methods of randomisation were not reported and it was unclear as to whether patients or outcome assessors were blinded to treatment. [EL = 1−]

At baseline, patients were similar for age, sex, weight, height, body temperature and mean stools passed per day. However, those children receiving furazolidone had fewer days with diarrhoea prior to recruitment compared with patients receiving either trimethoprim/sulfamethoxazole or no antibiotic treatment (P < 0.02). A statistically significantly higher number of patients (who took antibiotics) were clinically cured by day 3 compared with the no antibiotic group (furazolidone RR 1.93; 95% CI 1.21 to 3.09, trimethoprim/sulfamethoxazole RR 1.82; 95% CI 1.13 to 2.92, and for both antibiotics together RR 1.87; 95% CI 1.18 to 2.98). Similar results were seen for clinical cure rate by day 6 (furazolidone RR 2.78; 95% CI 1.25 to 6.19, trimethoprim/sulfamethoxazole RR 3.05; 95% CI 1.38 to 6.72, and for both antibiotics together RR 2.92; 95% CI 1.33 to 6.39).

However, among those patients who had negative stool cultures, there were no statistically significant differences in the proportion of patients who had been clinically cured at day 3, for either furazolidone or trimethoprim/sulfamethoxazole individually or for both antibiotics together compared with no antibiotic treatment. For patients with positive stool cultures, bacteriological cure at day 6 was only statistically significantly different from placebo when data for antibiotics were combined (RR 2.33; 95% CI 1.04 to 5.25). No statistically significant differences were found for furazolidone (RR 1.76; 95% CI 0.76 to 4.12) or trimethoprim/sulfamethoxazole (RR 1.97; 95% CI 0.85 to 4.56) alone compared with placebo.

Another RCT from Mexico153 recruited children aged 3–84 months seen in hospital with diarrhoea into a treatment trial of trimethoprim/sulfamethoxazole compared with placebo. Participants had passed three or more unformed stools in the previous 24 hours, had less than 72 hours’ duration of diarrhoea, no antibiotic treatment in the previous 7 days and were not severely dehydrated (n = 141) and were randomised into two treatment groups to receive 10 mg/kg per day trimethoprim + 50 mg/kg per day sulfamethoxazole oral suspension in two divided doses a day for 5 days (n = 73) or placebo oral suspension (n = 68). Daily assessments were made throughout the duration of treatment and once more at 2 weeks. Although details of the randomisation process were not reported, patients and outcome assessors were blind to treatment allocation. The groups were similar at study entry for age, pre-study diarrhoea duration, mean stool passage at 24 hours pre-therapy, fever, dehydration status and faecal leucocytes. Fifty of 141 participants had body weight under the third percentile for age according to Mexican standard criteria. [EL = 1−]

The mean time to last diarrhoeal stool was statistically significantly shorter with antibiotic use compared with placebo in all patients (58.2 versus 75.5 hours; P = 0.021), those with fever (58.2 versus 75.5 hours; P = 0.021) and those with faecal leucocytes (>3/HPF) (57.7 versus 106.5 hours; P = 0.025). However, there were no statistically significant differences between the antibiotic and placebo groups for the mean number of unformed stools in a 5 day period and in week 1 and week 2 post treatment for all patients and the subgroups (those with fever and patients with >3/HPF).

Evidence summary

Four clinical trials performed in South Africa and Mexico examined the use of antibiotics in children presenting with diarrhoea in whom the results of stool microbiological investigations were not yet available to inform management. Only one trial was of good quality and its results suggested that erythromycin treatment reduced the mean duration of abnormal stool consistency. The other three trials had methodological limitations, evaluated different antibiotics and reported contrasting results. While one small trial reported that the use of trimethoprim/sulphonamide had no effect on duration of clinical symptoms (diarrhoea, vomiting or pyrexia) or on the length of hospital stay, another trial found evidence that this drug reduced the duration but not the severity of diarrhoea in the first 5 days of treatment. A fourth trial reported that furazolidone and trimethoprim/sulfamethoxazole achieved a ‘clinical cure’ in all patients within 3 days of starting treatment. This effect was not seen for patients who were subsequently found to have negative stool cultures. A protective effect of antibiotic administration was only seen for patients with positive stool cultures when the data for both antibiotics were combined.

7.8. Traveller’s diarrhoea

Evidence overview

No trials including children were identified but a Cochrane systematic review of antibiotic treatment for traveller’s diarrhoea in adults was found.155 [EL = 1+] The authors included all trials in any language in which travellers older than 5 years were randomly allocated to antibiotic treatment for acute non-bloody diarrhoea and where the causative organism was not known at the time of treatment allocation. Patients with acute bloody diarrhoea for longer than 14 days were excluded. Twelve trials were included in the Cochrane review in total, but only nine trials that compared antibiotic therapy with placebo were relevant to this guideline (n = 1174). Participants were students, soldiers, tourists, hotel guests or volunteers who had travelled to Mexico (five trials), Morocco (one trial), the Gambia (one trial), Belize (one trial) and unspecified developing countries (one trial). The antibiotics used in the trials were ofloxacin, bicozamycin, ciprofloxacin (two trials), trimethoprim and trimethoprim-sulfamethoxazole, norfloxacin (two trials), fleroxacin and aztreonam. Although all nine trials reported the mean duration of diarrhoea (assessed by time to last unformed stool), only three reported the mean and standard deviation and one trial reported the mean and P value from which a pooled standard deviation was derived.

Four comparisons of antibiotic (n = 199) with placebo (n = 264) were made in three trials and statistically significant reductions in the mean duration of diarrhoea were found in those receiving antibiotics (WMD −25.86 hours; 95% CI −32.58 to −19.14 hours). One study reported a mean duration of diarrhoea of 26 hours in the antibiotic group (n = 8) compared with 60 hours in the placebo group (n = 9) (pooled SD = 28 hours)

Six trials reported the number of patients cured at 72 hours. There were statistically significantly more in the antibiotic groups who were cured at 72 hours (330/391) compared with the placebo groups (154/306) (OR 5.90; 95% CI 4.06 to 8.57).

Change of severity of diarrhoea (number of unformed stools per 24 hour period) over 72 hours was reported by two trials. There was a small but statistically significant reduction for those receiving antibiotics (n = 117) compared with those receiving placebo (n = 106) sustained over 72 hours (0–24 hours WMD −1.59; 95% CI −2.66 to −0.52, 25–48 hours WMD −2.10; 95% CI −2.78 to −1.42, and 49–72 hours WMD −1.38; 95% CI −1.94 to −0.82).

Five studies reported side effects from treatment: 110/523 participants receiving antibiotics experienced a side effect compared with only 38/339 in the placebo groups (OR 2.37; 95% CI 1.50 to 3.75) although these were said not to have been clinically serious or to have resolved on withdrawal of the treatment.

Evidence summary

No evidence in children younger than 5 years was identified. However, results from a systematic review found that antibiotic treatment was effective in reducing the duration and severity of diarrhoea in adult patients with traveller’s diarrhoea, although there was an increased incidence of side effects.

7.9. Groups for whom antibiotic treatment may be indicated

Clinical question

Are there any particular circumstances where antibiotics should be given?

Evidence overview

Searches were conducted for observational studies and 203 references were returned. On the basis of the titles and abstracts, 33 were retrieved in full copy for further examination. Of these, four studies were included here.

7.9.1. E. coli O157:H7

Two relevant studies were identified with regard to this pathogen, which is the main cause of haemolytic uraemic syndrome (HUS).

One prospective cohort study156 conducted in the USA investigated whether antibiotic treatment affected the risk of HUS in children. [EL = 2+] In total, 71 children younger than 10 years who had diarrhoea caused by E. coli O157:H7 were recruited to the study. Stool culture was obtained within the first 7 days of illness. On confirmation of E. coli O157:H7, investigators sought consent and recruited the infected child to the study. A questionnaire was administered to caregivers to record previous and ongoing clinical signs and symptoms, prescription and other medications taken (for example antibiotics and antimotility drugs). Prescription of medications was at the discretion of each physician and was confirmed retrospectively from notes. Only the initial temperature readings and laboratory test results were used for analysis. Daily blood counts and renal function tests were performed until HUS developed and resolved or until diarrhoea resolved. Multivariate regression analysis was used to evaluate the risk of HUS after adjustment for two risk factors that had been previously reported to be associated with it – the initial white cell count and the day of illness on which initial stool cuture is taken. The two groups of children – those who received antibiotics (n = 9) and those who did not (n = 62) – were similar for demographic characteristics, clinical and laboratory sparameters at the baseline. Overall, 10/71 (14%) developed HUS. Of the nine children who received antibiotics, five (56%) subsequently developed HUS while of the 62 who did not receive antibiotics, five (8%) developed HUS. This difference was statistically significant (P = 0.001) and remained so after adjustment, although confidence intervals were wide and the lower estimate was close to unity. (Antibiotics given within the first 7 days after onset RR 17.3; 95% CI 2.2 to 137; P = 0.007, and within the first 3 days after onset RR 32.3; 95% CI 1.4 to 737; P = 0.03). A statistically significant linear trend was observed for initial white blood cell count and development of HUS (P = 0.005). This remained statistically significant after reanalysis as a continuous outcome and adjustment (adjusted RR 1.5; 95% CI 1.1 to 2.1; P = 0.02). A statistically significant linear trend was also observed for the difference in HUS development according to the day that stool culture was taken (P = 0.01). This remained statistically significant after adjustment (adjusted RR 0.3; 95% CI 0.1 to 0.7; P = 0.008).

A retrospective cohort study,157 also conducted in the USA, evaluated risk factors for progression of E. coli O157:H7 to the development of HUS. [EL = 2+] Participants were younger than 16 years, resided in Washington state and either had symptomatic, culture-proven E. coli O157:H7 infection or had developed HUS in January or February 1993 (during an E. coli O157:H7 outbreak from a ‘fast food outlet’ chain. Demographic, symptomatic and medication data were gathered from three sources: two telephone questionnaires administered to participants’ parents (within 2 weeks of illness onset and 2–4 months later) and from examination of patients’ medical records. Cases and controls were similar for age, sex and annual family income at baseline. The median age of participants was 6 years (range 0–15 years). In total, 33/278 patients developed complete HUS and 4/278 developed incomplete HUS. There were three fatalities.

Children who reportedly vomited (29/153) were statistically significantly more likely to develop HUS than those who had not vomited (8/125) (RR 3.0; 95% CI 1.4 to 6.2). Although more children who had bloody diarrhoea or fever developed HUS, these differences were not statistically significant (RR 2.0; 95% CI 0.5 to 7.7, and RR 1.8; 95% CI 0.8 to 4.1, respectively).

To investigate early predictors, the risk of HUS was evaluated according to clinical outcomes measured within the first 3 days of illness. Vomiting remained a statistically significant risk factor in this time interval (RR 1.9; 95% CI 1.0 to 3.5) and the association was modified by age. Vomiting in children younger than 5.5 years was strongly associated with HUS development (RR 3.5; 95% CI 1.4 to 9.4), but the association was not evident in children older than 5.5 years (RR 1.0; 95% CI 0.4 to 2.4).

The use of medications was also analysed. In total, 50 children received a range of antibiotics in the first 3 days of illness. These children were more likely to live in a household with annual income over $29,000 (RR 1.7; 95% CI 1.0 to 2.8). Eight of these children went on to develop HUS compared with 28/218 who did not receive antibiotics (P = 0.56). In total, 31 children received antimotility agents in the first 3 days of illness. Six went on to develop HUS, compared with 20/234 who received no antimotility treatments (P = 0.10). There was no statistically significant difference in the development of HUS in children who received adsorbant and antimotility drugs compared with those who did not (P = 0.26).

There were no statistically significant associations between HUS development and haematocrit, platelet count, blood urea nitrogen, segmented neutrophils count or band forms at presentation. However, children who had a white blood cell count of over 10 500 per microlitre were at increased risk of developing HUS (RR 5.2; 95% CI 1.6 to 17.0; P < 0.01), and for those with a white blood cell count of over 13 000 per microlitre this risk was larger (RR 7.2; 95% CI 2.8 to 18.5; P < 0.01).

Evidence summary

There was consistent evidence from two studies that a raised white blood cell count in children with E. coli O157:H7 was a risk factor for the subsequent development of HUS. In one study, vomiting in children younger than 5.5 years was strongly associated with the risk of developing HUS. However, there was conflicting evidence on the effect of antimicrobials. One study reported that antimicrobial treatment was an independent risk factor for HUS but the study lacked precision for this finding. The second study did not find treatment with antimicrobials or with antimotility agents (with or without adsorbant agents) was associated with increased risk of HUS.

7.9.2. Salmonella

One retrospective review158 conducted in Malaysia sought to characterise the incidence, risk factors and outcome of invasive non-typhoid salmonella gastroenteritis in children aged between 1 month and 14 years. [EL = 2+]. Participants were 131 children with positive stool cultures for salmonella species but no second enteropathogen, seen in an outpatient department. Of these, 67% of children were younger than 1 year. Demographic, clinical (diarrhoea, vomiting, fever, hydration status), blood and stool outcome measures were recorded from case notes and examined. Overall, 124 children were found to have non-invasive salmonellosis and seven had invasive complications (five bacteraemia, two meningitis). Three risk factors were identified for the development of invasive salmonellosis. In total, 45 (85%) of the 124 with non-invasive disease were younger than 6 months compared with six of the seven with invasive disease (P < 0.01). Only 53 of those in the non-invasive group had a temperature of over 38 °C, compared with all seven of the invasive group (P < 0.003). Dehydration was found in five of the seven with invasive complications, but in only 25 of the 124 with non-invasive disease. One infant with bacteraemia died while awaiting a blood culture result. The authors suggested that empirical antibiotic treatment should be given to infants younger than 6 months who are febrile and dehydrated.

Another retrospective review,159 conducted in Hong Kong, included 126 children with salmonella (n = 86), rotavirus (n = 55) or unspecified gastroenteritis (n = 126) who were admitted to hospital. [EL = 2+]. Demographic, clinical (dehydration, vomiting, fever, diarrhoea, abdominal pain), stool and medication outcomes were collected from case notes and analysed according to gastroenteritis type. Patients with salmonella were more likely to have bloody (OR 6.1; 95% CI 3.2 to 11.7; P < 0.0001) or mucoid stools (OR 4.8; 95% CI 2.6 to 8.9; P < 0.001) compared with the combined rotavirus/non-specified gastroenteritis groups. They were statistically significantly younger (median 7.1 months versus 14.6 months; P < 0.0001), had a longer stay in hospital (median 3.4 days versus 2 days; P < 0.0001), passed more stools per day (median 6.2 versus 4.2; P < 0.0001) and more of them experienced fever during their admission (OR 3.6; 95% CI 1.6 to 8.4; P = 0.001). Additionally, patients with salmonella were statistically significantly more likely to have been given antibiotics than children in the other two groups (OR 3.6; 95% CI 1.9 to 6.9; P < 0.0001), although administration of antibiotics was not dependent on age.

Evidence summary

Compared with other enteropathogens, salmonella gastroenteritis has repeatedly been shown to particularly affect younger children. A retrospective review from Malaysia found that 67% of all salmonella-infected children were younger than 1 year. Most children developing invasive salmonellosis (bacteraemia or meningitis) were younger than 6 months. Similarly, a study from Hong Kong found the median age to be 7.05 months (range 3.9 to 13.6 months). Fever was a significant characteristic in both studies, compared with other pathogens and as an indication of invasive salmonellosis.

Dehydration was statistically significantly associated with more severe disease in one study. Diarrhoea was more frequent in salmonella infection and stools were characteristically bloody and/or mucoid, although blood in stools was not found to be indicative of invasive salmonellosis. Vomiting was less frequent than with viral infection. Children with salmonella had longer hospital stays and were more likely to be treated with antibiotics regardless of age.

GDG translation from evidence to recommendations

The GDG recognised that gastroenteritis in children in the UK is usually a self-limiting illness that does not require antibiotic treatment. Most cases are due to viral enteric pathogens. Even with the more common non-viral pathogens, treatment is usually not indicated in the UK (see below). Antibiotic treatment is associated with a risk of adverse effects, and is a very common cause of diarrhoea.

In most cases of childhood gastroenteritis, the healthcare professional will not know the causative agent. Stool microbiological investigations are performed in selected cases only (see Chapter 3). If a bacterial pathogen is identified by stool culture, the result would not be available at the time of first presentation.

The recommendations below took into account the experience of the GDG and of the advisers, and the limitations of the studies identified (not UK-based and with few children in the study populations).

The GDG was aware of studies conducted in South Africa and Mexico several decades ago in which antibiotic therapy was administered while awaiting the results of stool microbiological investigations. There was some evidence that at that time in those studies empirical treatment had some apparent benefit. However, the GDG did not consider that those studies were relevant to the current UK setting.

The spectrum of pathogens commonly responsible for gastroenteritis in the UK is such that benefit from empirical antibiotic treatment would be highly unlikely.

Recommendations regarding the treatment of specific enteric pathogens were also considered.

A series of randomised controlled trials found little evidence of clinical benefit from antibiotic treatment for children with salmonella gastroenteritis. Indeed, there was evidence to suggest that treatment might increase the risk of salmonella carriage. The GDG therefore concluded that antibiotic treatment should not routinely be given in salmonella gastroenteritis. However, there are some individuals at increased risk of systemic sepsis. Young infants are at increased risk of developing salmonella gastroenteritis, and those younger than 6 months are at increased risk of systemic spread. Others likely to be at high risk of sepsis are those with immune deficiency states including HIV/AIDS and malnourished infants and children. The GDG concluded that in such cases antibiotic treatment should be recommended.

The efficacy of antibiotic therapy for patients with campylobacter spp. is somewhat uncertain. One randomised controlled trial in which treatment with erythromycin began while culture results were pending found that treatment was associated with a shortened mean duration of diarrhoea. There was no evidence to suggest that antibiotic therapy was beneficial in those whose treatment began after a positive culture was available. The GDG therefore concluded that antibiotics should only be used where septicaemia is suspected.

There was no evidence that antibiotic therapy was beneficial in the treatment of yersinia enteritis. The GDG considered that antibiotics should be reserved for those with suspected or confirmed yersinia septicaemia.

The GDG was aware that there was evidence to support the efficacy of antibiotic therapy in patients with dysenteric shigellosis in adults. One randomised controlled trial of antibiotic therapy for bacterial gastroenteritis in children found that, although there was no benefit in relation to diarrhoea, the duration of fever and the time to clearance of stool pathogens were reduced. The GDG therefore concluded that those with dysenteric shigellosis should receive antibiotic treatment.

The GDG was aware that antibiotic treatment was effective in adults with enterotoxigenic E. coli, a very common cause of traveller’s diarrhoea. The effect of antibiotic treatment for enteropathogenic and enteroinvasive E. coli is uncertain.

Two studies were available regarding antibiotic treatment for the protozoal pathogen Cryptosporidium parvum. Both evaluated the broad-spectrum antibiotic nitazoxanide that has activity against protozoa. Both reported some evidence of benefit, but these studies had methodological limitations. Routine treatment of gastroenteritis due to cryptosporidium was therefore not recommended.

Young children with E. coli O157:H7 appeared to have a risk of 10–15% of developing HUS. There was evidence to suggest that those with a leucocytosis at presentation are at higher risk. Vomiting, especially in children younger than 5 years, was also associated with an increased likelihood of progression to HUS. There was some evidence that antibiotic treatment might have been a risk factor for HUS, although this finding was not consistent between studies. It might be that this reported observation can be explained by the administration of antibiotics to those with more severe disease. Nevertheless, the GDG considered that there was insufficient evidence to recommend antibiotic treatment for E. coli O157:H7.

The GDG recognised that a number of other potential enteric pathogens exist that could cause gastroenteritis, but there were no available clinical trials on treatment in children. Clostridium difficile-associated pseudomembranous colitis is normally treated with antibiotics. The same is true of Vibrio cholerae. Protozoal infections – including Isospora belli, Cyclospora cayetanensis, Entamoeba histolytica and Giardia lamblia might all respond to antibiotic therapy, based on studies in adults.

There was no clinic trial evidence on the treatment of traveller’s diarrhoea in children, but the GDG considered that trials in adult patients were relevant, and these showed evidence of benefit from antibiotic treatment. It was therefore agreed that in such cases consideration should be given to seeking specialist advice regarding antibiotic treatment in children presenting with acute diarrhoea shortly after return from overseas travel.

Recommendations on antibiotic therapy

Do not routinely give antibiotics to children with gastroenteritis.

Give antibiotic treatment to all children:

  • with suspected or confirmed septicaemia
  • with extra-intestinal spread of bacterial infection
  • younger than 6 months with salmonella gastroenteritis
  • who are malnourished or immunocompromised with salmonella gastroenteritis
  • with Clostridium difficile-associated pseudomembranous enterocolitis, giardiasis, dysenteric shigellosis, dysenteric amoebiasis or cholera.

For children who have recently been abroad, seek specialist advice about antibiotic therapy.

Copyright © 2009, National Collaborating Centre for Women’s and Children’s Health.

No part of this publication may be reproduced, stored or transmitted in any form or by any means, without the prior written permission of the publisher or, in the case of reprographic reproduction, in accordance with the terms of licences issued by the Copyright Licensing Agency in the UK [www.cla.co.uk]. Enquiries concerning reproduction outside the terms stated here should be sent to the publisher at the UK address printed on this page.

The use of registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant laws and regulations and therefore for general use.

Bookshelf ID: NBK63849

Views

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...