Correspondence

Risk of the Hemolytic–Uremic Syndrome after Antibiotic Treatment of Escherichia coli O157:H7 Infections

N Engl J Med 2000; 343:1271-1273October 26, 2000

Article

To the Editor:

Wong et al. (June 29 issue)1 report an association between antibiotic treatment in children with acute diarrhea caused by Escherichia coli O157:H7 and the development of the hemolytic–uremic syndrome. In Chile, shigella species cause 30 percent of cases of bloody diarrhea in children and enterohemorrhagic strains of E. coli cause 37 percent of such cases2 — a situation that may also be common in developing countries. For children with acute bloody diarrhea in these countries, the most widely accepted recommendation is to obtain a stool culture and initiate empirical antibiotic treatment for shigella, because appropriate treatment shortens the duration of the diarrhea, decreases the incidence of complications, and reduces the risk of transmission by shortening the duration of bacterial shedding.3

Given these points, the results presented by Wong et al. raise several questions. Knowing that E. coli O157:H7 and shigella species cannot be differentiated early in the clinical course on the basis of clinical findings or simple laboratory tests, does the potentially increased risk of the hemolytic–uremic syndrome associated with empirical antibiotic treatment of E. coli O157:H7 infections outweigh the risk of not treating shigella infections during the 72 hours needed to obtain culture results? In Chile, 70 percent of cases of the hemolytic–uremic syndrome are associated with serotypes of E. coli other than O157:H7.4 Can the increased risk of the hemolytic–uremic syndrome in children who receive antibiotics for E. coli O157:H7 infections be extrapolated to other strains of the organism?

We worry that the message of this study will be extended prematurely to the treatment of bloody diarrhea worldwide. In places where the prevalence of shigella and E. coli infections is similar and strains of E. coli other than O157:H7 are common, withholding antibiotic therapy until the cause of the diarrhea is known may have more risks than benefits.

Miguel O'Ryan, M.D.
Valeria Prado, M.D.
University of Chile, Santiago, Chile

4 References

1. 1

   Wong C, Jelacic S, Habeeb RL, Watkins SL, Tarr PI. The risk of the hemolytic-uremic syndrome after antibiotic treatment of Escherichia coli O157:H7 infections. N Engl J Med 2000;342:1930-1936
   Full Text | Web of Science | Medline

2. 2

   Lopez EL, Prado-Jimenez V, O'Ryan-Gallardo M, Contrini MM. Shigella and Shiga toxin-producing Escherichia coli causing bloody diarrhea in Latin America. Infect Dis Clin North Am 2000;14:41-65, viii
   CrossRef | Web of Science | Medline

3. 3

   Haltalin KC, Nelson JD, Ring R III, Sladoje M, Hinton LV. Double-blind treatment study of shigellosis comparing ampicillin, sulfadiazine, and placebo. J Pediatr 1967;70:970-981
   CrossRef | Web of Science | Medline

4. 4

   Prado JV, Martinez DJ, Arellano CC, Levine MM. Variación temporal de genotipos y serogrupos de E Coli enterohemorrágicos aislados en niños chilenos con infecciones intestinales o sindrome hemolítico-urémico. Rev Med Chil 1997;125:291-297
   Web of Science | Medline

To the Editor:

We question some of the epidemiologic concepts and analyses used by Wong et al. They justified the cohort study design because “the severity of illness might confound the association with antibiotic treatment.”

The severity of illness is not a potential confounder. To cause confounding bias, “a variable must be a risk factor for the disease among the nonexposed persons, must be associated with the exposure of interest in the population from which the cases derive, but must not be an intermediate in the causal pathway between the exposure and disease.”1 The severity of illness is probably a marker for an intermediate factor in the causal pathway or an early symptom or sign of the hemolytic–uremic syndrome. Its association with antibiotic treatment could result in what has been called “susceptibility” bias2 (thus requiring a stratified or multivariable analysis). However, antibiotic treatment was not associated with any indicator of severity, and therefore, this source of bias was unlikely. The results in Table 3 of the article confirmed this point: the univariate odds ratio of 14.3 does not meaningfully differ from the adjusted odds ratio of 17.3. The analysis can actually be reduced to a two-by-two table, which provides a more accurate and precise measure of risk (Table 1Table 1Number of Children Infected with Escherichia coli O157:H7 in Whom the Hemolytic–Uremic Syndrome Developed, According to Whether They Were Treated with Antibiotics.).

The risk of the hemolytic–uremic syndrome among the children who were given antibiotics was 56 percent (5 of 9 children), and the risk among the children who were not treated with antibiotics was 8 percent (5 of 62 children). The relative risk was 6.9 (95 percent confidence interval, 2.5 to 19.2). In cohort studies, one can directly calculate the relative risk.3,4 Instead, Wong et al. used the odds ratio to approximate the relative risk. Direct estimation is preferable, and when the incidence is high, as it was in this study, the odds ratio overestimates the relative risk.

Tomás Aragón, M.D., Dr.P.H.
Susan Fernyak, M.D., M.P.H.
Randy Reiter, Ph.D., M.P.H.
San Francisco Department of Public Health, San Francisco, CA 94102

4 References

1. 1

   Salas M, Hofman A, Stricker BH. Confounding by indication: an example of variation in the use of epidemiologic terminology. Am J Epidemiol 1999;149:981-983
   Web of Science | Medline

2. 2

   Feinstein AR, Horwitz RI. An algebraic analysis of biases due to exclusion, susceptibility, and protopathic prescription in case-control research. J Chronic Dis 1981;34:393-403
   CrossRef | Medline

3. 3

   Schwartz LM, Woloshin S, Welch HG. Misunderstandings about the effects of race and sex on physicians' referrals for cardiac catheterization. N Engl J Med 1999;341:279-283
   Full Text | Web of Science | Medline

4. 4

   Zhang J, Yu KF. What's the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA 1998;280:1690-1691
   CrossRef | Web of Science | Medline

Author/Editor Response

The authors reply:

To the Editor: O'Ryan and Prado are correct; we cannot directly extrapolate our findings to infections with Shiga-toxin–producing strains of E. coli other than O157:H7. However, the same cautious approach should apply, especially since antibiotics cause the release of Shiga toxin from non–O157:H7 strains of the organism.1,2 We also realize that E. coli and shigella can infect the same people and can have similar clinical manifestations.3 However, with the probable exception of infections with Shigella dysenteriae serotype 1 (which are rare in this country), we believe that the potential harm from antibiotic treatment of an infection with a Shiga-toxin–producing strain of E. coli exceeds the harm from delaying treatment of shigellosis until the culture results are available.

We agree with Aragón et al. that the relative risk can be directly calculated from our data. However, we were concerned that the simple comparison of the group of patients who received antibiotics with the group of patients who did not receive antibiotics might have been compromised by differences in the severity of illness between the two groups — an effect that is often termed “confounding by indication.”4

Because we considered a priori markers of the severity of illness as potential confounders, adjustments for these factors were important despite the lack of statistical significance.5 Although the association between the initial white-cell count and the use of antibiotics was not significant, physicians often use leukocytosis to justify the use of antibiotic therapy. Furthermore, the initial white-cell counts in our study were usually obtained before antibiotic therapy was given, making it unlikely that this marker of the severity of illness was an intermediate factor in the causal pathway of the hemolytic–uremic syndrome. Therefore, to address the question of interest, we found it necessary to adjust for the severity of illness in a multivariate logistic-regression analysis; the adjusted odds ratio can readily be estimated on the basis of this analysis.

Our data further support the inference of a causal relation between antibiotic therapy and the subsequent development of the hemolytic–uremic syndrome. The temporal sequence of events was appropriate; antibiotic therapy preceded the development of the hemolytic–uremic syndrome. To emphasize the prospective cohort design of our study, we thought it was appropriate to report our findings as estimates of the relative risk using the odds ratios. We acknowledge the limitations of the odds ratio to estimate the magnitude of the relative risk. However, for the purposes of our study, the analytic paradigm was sufficiently justifiable to answer the question of whether or not antibiotic therapy was associated with the development of the hemolytic–uremic syndrome after adjustment for the severity of illness.

Craig S. Wong, M.D., M.P.H.
Sandra L. Watkins, M.D.
Phillip I. Tarr, M.D.
Children's Hospital and Regional Medical Center, Seattle, WA 98105

5 References

1. 1

   Takahashi K, Narita K, Kato Y, et al. Low-level release of Shiga-like toxin (verocytotoxin) and endotoxin from enterohemorrhagic Escherichia coli treated with imipenem. Antimicrob Agents Chemother 1997;41:2295-2296
   Web of Science | Medline

2. 2

   Walterspiel JN, Ashkenazi S, Morrow AL, Cleary TG. Effect of subinhibitory concentrations of antibiotics on extracellular Shiga-like toxin I. Infection 1992;20:25-29
   CrossRef | Web of Science | Medline

3. 3

   Cunin P, Tedjouka E, Germani Y, et al. An epidemic of bloody diarrhea: Escherichia coli O157 emerging in Cameroon? Emerg Infect Dis 1999;5:285-290
   CrossRef | Web of Science | Medline

4. 4

   Weiss NS. Clinical epidemiology: the study of the outcome of an illness. 2nd ed. New York: Oxford University Press, 1996:81-96.
   

5. 5

   Clayton D, Hills M. Statistical models in epidemiology. New York: Oxford University Press, 1993:273.
   

Citing Articles (1)

Citing Articles

1. 1

   Piero Ruggenenti, Marina Noris, Giuseppe Remuzzi. (2001) Thrombotic microangiopathy, hemolytic uremic syndrome, and thrombotic thrombocytopenic purpura. Kidney International 60:3, 831-846
   CrossRef