To better understand the association between antibiotic therapy for Escherichia coli (E. coli) 0157:H7 enteritis and the risk of haemolytic uremic syndrome (HUS), and to determine whether antibiotic therapy increases the risk of HUS.

MEDLINE and PubMed were searched for studies published in any language using the keywords 'hemolytic uremic syndrome', 'antibiotic', 'risk factor' and 'Escherichia coli 0157:H7'. The search was limited to reports on infections in human that were published between January 1983 (the year when Shiga toxin-producing E. coli was first found to be associated with HUS) and February 2002. The reference lists of recent publications, the Cochrane Controlled Trials Register, and the NIH website listings of ongoing trials were also reviewed. In addition, 12 authorities in the field were contacted for unpublished studies.

Study designs of evaluations included in the review

Studies of any design where a control group had been used, and which provided adequate data delineating the relationship between antibiotic therapy and the occurrence of HUS, were eligible for inclusion in the review. The studies also had to evaluate risk factors for HUS to be included.

Specific interventions included in the review

Administrations of antibiotic therapy for severe, infectious enteritis or dysentery caused by E.coli 0157:H7 infection were eligible for inclusion in the review. The antibiotics included were: trimethoprim, ampicillin, cephalosporins, metronidazole, fosfomycin, sulfamethoxazole, erythromycin, gentamicin sulfate, tetracycline, ciprofloxacin, sulfonamides and trimethoprim-sulfamethoxazole.

Participants included in the review

Critically ill children or adults with a confirmed diagnosis of severe, infectious enteritis or dysentery that had been caused by E. coli 0157:H7, including patients who developed HUS, were eligible for inclusion in the review. The age of the participants in the review aged from less than 1 year to 94 years. Studies of infections with E. coli serotypes other than 0157:H7 were excluded.

Outcomes assessed in the review

The incidence of HUS was assessed in the review. Clear definitions of HUS had to be given in the study.

How were decisions on the relevance of primary studies made?

Two of the review authors independently reviewed each report identified by the searches, and independently recorded predetermined information relevant to the inclusion criteria.

The authors do not state that they assessed validity.

The authors do not state how the data, other than those related to the inclusion criteria, were extracted for the review, or how many of the reviewers performed the data extraction.

Data were extracted on: the study design, the age range and numbers of the participants; the numbers of participants developing HUS; the type of antibiotic used for treatment; and the interval between the onset of acute diarrhoea and the introduction of antibiotic therapy. Crude odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using the data provided in the studies. Data on both antibiotic use and HUS were abstracted as dichotomous variables. If primary data were not reported, the OR and 95% CI reported in the study were used.

How were the studies combined?

The studies were pooled both statistically in a meta-analysis and combined narratively. Pooled estimates of the OR and 95% CIs were obtained using the fixed-effect model of Mantel and Haenszel (see Other Publications of Related Interest no.1). Publication bias was assessed by a funnel plot. No attempt was made to calculate a pooled adjusted OR as three of the studies did not adjust for severity of illness, and there were substantial differences in the methods used to adjust for confounding among the remaining studies.

How were differences between studies investigated?

Heterogeneity was tested using the test of Breslow and Day (see Other Publications of Related Interest no. 2).

Ten studies (9 publications; n=1,121) were included in the review: 1 prospective randomised trial, 1 randomised trial, 2 prospective cohort, 1 retrospective cohort, and 5 retrospective case-controlled studies.

The pooled OR was 1.15 (95% CI: 0.79, 1.68), showing no association between HUS and antibiotic use. The test for heterogeneity was highly significant (p<0.001) with one retrospective and one prospective study accounting for this. When these two studies were removed, the OR was 0.83 (95% CI: 0.54, 1.26). The funnel plot showed no substantial publication bias.

The authors stated that their meta-analysis did not show a higher risk of HUS associated with antibiotic administration. However, they also stated that an adequately powered, nationwide randomised trial, in which multiple distinct strains of E. coli 0157:H7 infection are used to permit early randomisation, is required before it can be concluded unequivocally that administration of antibiotic therapy to critically ill children or adults with severe, presumably infectious enteritis, especially dysentery that might represent E. coli 0157:H7 infection, is deleterious.

The review question and the study selection criteria were stated clearly. The literature search was limited to MEDLINE and PubMed, and although there were no language restrictions and attempts were made to identify unpublished literature, some relevant material may have been missed owing to the restrictions of the electronic searches. The method used to assess the validity of the primary studies was not reported. The statistical tests undertaken seem to have been appropriate, and the findings were presented and discussed adequately.

The authors' conclusions seem appropriate in the light of the data they present.

Practice: The authors did not state any implications for practice.

Research: The authors state that better data are required. Ideally, this would be gained through an adequately powered, nationwide randomised trial, in which multiple distinct strains of E. coli 0157:H7 infection are used to permit early randomisation, before it can be concluded unequivocally that administration of antibiotic therapy to critically ill children or adults with severe, presumably infectious enteritis, especially dysentery that might represent E. coli 0157:H7 infection, is deleterious.

Oscar Rennehohm Foundation (unrestricted gift).

1. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959;29:719-48. 2. Breslow NE, Day NE. Statistical methods in cancer research. Volume I. The analysis of case-control studies. Lyon: IARC Scientific Publications; 1980.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.