Antimicrobial Therapy in Patients With Escherichia coli O157:H7 Infection

Shiga toxin–producing Escherichia coli O157:H7 is an important cause of enteric illness in developed countries.¹ Infection typically presents as a diarrheal illness, often with bloody stools. In approximately 8% of patients,² infection progresses to hemolytic uremic syndrome (HUS), a life-threatening condition characterized by microangiopathic hemolytic anemia, thrombocytopenia, and renal failure.

The use of antimicrobial agents in the management of patients with E coli O157:H7 infection is controversial. Most strains are susceptible to commonly used antimicrobial agents³ ; however, these agents have not been shown to reduce the duration of diarrhea.^{4 - 6} More important, researchers are currently debating whether antimicrobial therapy increases or decreases the risk of developing HUS. In the production of HUS, Shiga toxins are key virulence factors, and the synthesis of this toxin appears to be regulated through the induction of an integrated bacteriophage that encodes the toxin gene. Furthermore, phage production is linked to the bacterial SOS response, a ubiquitous response to DNA damage.^{7 - 8}

In vitro studies have shown that subinhibitory concentrations of antimicrobial agents can induce phage production and enhance Shiga toxin production.^{8 - 13} Fluoroquinolones and trimethoprim in particular have been associated with induction of Shiga toxin production in vitro, whereas protein synthesis and cell-wall inhibitors such as fosfomycin tromethamine have been less likely to induce toxin production.^{7 - 8} These effects may be tempered by strain differences, however. For example, in one study, fosfomycin markedly increased toxin production by 1 strain but had little or no effect on the other 2 strains.¹³

Studies in animal models have also demonstrated both beneficial and harmful effects of antimicrobial agents. In a study by Kurioka et al,¹⁴ E coli O157:H7–infected mice were treated with therapeutic levels of various agents, including trimethoprim-sulfamethoxazole, fosfomycin, and a fluoroquinolone on days 1 to 3 or days 3 to 5 after infection. All agents administered early and all except 1 agent administered on days 3 to 5 were associated with a decrease in the duration of pathogen excretion and in Shiga toxin levels in the stool and blood. However, mice given trimethoprim-sulfamethoxazole on days 3 to 5 had a higher toxin level in blood than did control mice; only these and the untreated mice died.¹⁴ The study by Zhang et al⁷ used another mouse model to compare outcomes after administering ciprofloxacin or fosfomycin in subtherapeutic levels (eg, 20% of the therapeutic dose of ciprofloxacin). Although both agents caused similar decreases in counts of E coli O157:H7 in mouse intestine, ciprofloxacin was associated with a marked increase in free fecal Shiga toxin and with death, whereas fosfomycin was not.⁷

As with in vitro experiments and animal models, the findings from clinical and epidemiological studies have been conflicting. In North America, most studies have shown no benefit from antimicrobial treatment and some have suggested an increased risk of HUS.¹ A recent prospective but nonrandomized study found an increased risk of HUS; 5 of 9 patients who were treated with antimicrobial agents developed HUS.¹⁵ Hence, the rate of HUS observed in this small group of patients treated with antimicrobial agents was much higher than that in larger observational studies.⁶ It is possible, in that study, that knowledge of the potential harm caused some clinicians to discontinue administering antimicrobial agents after a small number of doses, and this could have had an effect similar to subtherapeutic dosing, with enhanced Shiga toxin production. In contrast, a large retrospective study from Japan found that administration of fosfomycin within the first 2 days of illness was associated with a reduced risk of HUS, whereas use of fosfomycin on or after the third day did not significantly affect progression to HUS.¹⁶ The largest of the North American studies, a retrospective analysis involving 268 children infected with the same outbreak strain, found no association between use of antimicrobial agents and development of HUS.⁶ Also of interest, a study of 117 children with HUS found that those who had received antimicrobial therapy during their diarrheal illness had less severe HUS.¹⁷

All but 1 of the clinical studies of which we are aware were nonrandomized. The main methodological problem with these studies is that patients who are most likely to receive antimicrobial agents are probably those with the most severe illness, and severe illness may be independently associated with a higher risk of progression to HUS.⁶ In other words, a harmful association between antimicrobial therapy and HUS may not reflect causation but, rather, may be the result of the selection of particular patients for treatment. Conversely, patients who receive early antimicrobial treatment may receive other helpful therapy at the same time (eg, intravenous hydration), and a protective association between early antimicrobial therapy and HUS may also not reflect causation—it may simply reflect the results of better supportive care. Attempts to control for the effects of nonrandomization through statistical analyses and the evaluation of a few potential confounders do not necessarily account for the complexity of reasons why physicians decide to treat and why patients comply with a given treatment.¹⁸

In a meta-analysis published in this issue of THE JOURNAL, Safdar et al¹⁹ summarize 9 studies published between 1990 and 2000. Their analysis includes a total of 1111 patients; 16% (range among studies, 8%-35%) developed HUS. The pooled odds ratio for an association between antimicrobial agent use and HUS was 1.14 (95% confidence interval, 0.77-1.70). The authors note the major limitation of the meta-analysis: they were not able to analyze the risk of HUS according to choice of antimicrobial agent or timing and duration of therapy. As emphasized herein, in vitro studies and animal models suggest the importance of drug choice, drug timing, and infecting strain. Some studies indicate that early treatment with an appropriate dose of an appropriate antimicrobial agent may reduce the risk of HUS. Other studies indicate that antimicrobial agents may be detrimental, particularly if they cause a bacterial SOS response or if the dose is subtherapeutic. The authors appropriately conclude that the currently available data, including their meta-analysis, are insufficient to resolve this issue.

In the meantime, what is the clinician to do? Generally speaking, empirical antimicrobial therapy is not recommended for patients with diarrhea because of the self-limited nature of most illnesses, the cost of treatment, the potential for promoting antimicrobial resistance, and the possibility of an adverse drug reaction.²⁰ Concern that antimicrobial treatment could increase the risk of HUS in patients later found to have E coli O157:H7 infection provides yet another reason for clinicians to avoid empirical antimicrobial treatment. Nevertheless, empirical therapy of diarrhea with antimicrobial agents is appropriate for certain subsets of patients, such as those who are at high risk of invasive infections.²⁰ Empirical therapy is also often appropriate in selected groups of patients in developing countries, where infection with Shigella or Vibrio cholerae may be common and facilities for making microbiological diagnoses are limited.²¹ Clinicians need to carefully consider the setting and nature of illness when deciding whether the potential benefits of empirical antimicrobial treatment outweigh the potential harm.

What about patients known to be infected with E coli O157:H7? Such patients are likely to have been ill for several days. There is little evidence to suggest that beginning antimicrobial treatment at this point will be beneficial, and clinicians would seem well advised to not treat these patients except in a controlled clinical trial. Patients in whom E coli O157:H7 is diagnosed after initiation of a course of antimicrobial therapy pose a therapeutic dilemma. These patients may have already incurred many of the risks of antimicrobial treatment. Theoretically, early discontinuation could result in subtherapeutic antimicrobial levels and enhanced Shiga toxin production, thereby increasing the risk of HUS.

Until better data are available, reasonable clinicians may disagree whether the best course of action is to stop or complete therapy. Physicians need to exercise their best judgment based on the individual patient, the antimicrobial agent, and other factors. In all circumstances, careful attention to hydration, often intravenous, remains the cornerstone of management of patients with E coli O157:H7 infection.²²