Each outbreak of verotoxigenic Escherichia coli infection renews interest in interventions to prevent the complication of hemolytic-uremic syndrome. Donald Farquhar reviewed in CMAJ1 a paper by Wong and colleagues that raises important concerns about risk factors for the progression to hemolytic-uremic syndrome.2 Although the authors' concerns about antibiotic use in this context may be valid, it is critical that their conclusions not dissuade investigators from performing prospective, controlled antibiotic studies.
Wong and colleagues concluded that the association between antibiotic use and progression to hemolytic- uremic syndrome was strong, but the 95% confidence interval of the adjusted relative risk is extremely wide for antibiotic use within the first 3 days after onset of illness (1.4–737).2 If the next 1 or 2 patients with hemolytic-uremic syndrome had not received an antibiotic, the significance level might not have been maintained in the multivariate analysis. Both youth and the use of antimotility agents did not prove to be risk factors, contrary to previous findings.3,4,5 In addition, the authors did not find an association between antibiotic use and progression to hemolytic-uremic syndrome in a previous study with much larger numbers of patients.5 These discordances could be a function of inadequate patient numbers in the latest study.2
Perhaps more importantly, I am concerned that the approach taken to categorize antibiotic use may not have been appropriate. Antibiotics were initially grouped together then later divided into the categories of cotrimoxazole and β-lactams. When Carter and colleagues concluded 14 years ago that antibiotics could be harmful, they also included all antibiotics regardless of class.6 Shortly thereafter, my colleagues and I published evidence that a specific group of agents recognized to be effective in shigellosis did not have a detrimental effect.7 Our subsequent studies indicated that prolonged use of similar agents was associated with a lesser risk of hemolytic-uremic syndrome.3,4
One obvious reason for the discrepancies between studies could be the categorization of antibiotics. We initially chose our stratification for shigellosis on the basis that there is considerable evidence that certain antimicrobials do not have any beneficial effect and perhaps have a detrimental effect (for example, antibiotics to which the bacterium is resistant).8 If shigellosis investigators had pooled all β-lactam agents they may never have seen a benefit attributable to ampicillin, because first-generation oral cephalosporins were ineffective.9 A wide spectrum of antibiotics may be used to treat patients infected with E. coli O157:H7, including erythromycin and metronidazole. Would it be logical to pool such antibiotics or even cephalosporins with cotrimoxazole or ampicillin in analyses, and would they ever be considered as trial agents in a prospective study? Is it logical to pool all medications as shown in Table 1 of the paper by Wong and colleagues?2 Furthermore, although patient recruitment was delayed, the study by Proulx and colleagues did not show a detrimental effect for cotrimoxazole.10
Some in vitro studies have found increased toxin liberation from verotoxigenic E. coli isolates exposed to antibiotics.11,12 While the latter studies continue to be used as arguments against antibiotic use, recent evidence has indicated that the results are considerably dependent on the methodology.13
Therefore, there is still a need for a prospective, randomized controlled study of ampicillin and placebo as suggested over a decade ago.7 This study should include sufficient numbers of patients, and patients should be recruited early in the course of the illness. The argument for choosing ampicillin over cotrimoxazole in such a trial rests with the unknown but theoretical risk of administering a moderately soluble sulfonamide during an evolving nephropathy. Ampicillin resistance among verotoxigenic E. coli remains reasonably low in most regions. Without such studies, the role of some antibiotics in protecting against or complicating verotoxigenic E. coli infections will continue to be uncertain.