Ciprofloxacin Resistance Does Not Affect Duration of Domestically Acquired Campylobacteriosis

    Cox Associates and University of Colorado, Denver
    Bayer HealthCare, Animal Health Division, Shawnee Mission, Kansas

    cIn their article "Prolonged Diarrhea Due to Ciprofloxacin-Resistant Campylobacter Infection" [1], Nelson et al. comment, "Although human infections with ciprofloxacin-resistant Campylobacter have become increasingly common, the human health consequences of such infections are not well described" (p. 1150). They then present analyses indicating that "persons with ciprofloxacin-resistant infection had a longer mean duration of diarrhea than did the persons with ciprofloxacin-susceptible infection (P = .01); this effect was independent of foreign travel" (p. 1150). They conclude, "Persons with ciprofloxacin-resistant Campylobacter infection have a longer duration of diarrhea than do persons with ciprofloxacin-susceptible Campylobacter infection. Additional efforts are needed to preserve the efficacy of fluoroquinolones" (p. 1150).

    We have a number of concerns with these analyses. The following comments address some of them. Reexamining the raw data raises questions about the validity and general applicability of the stated conclusions and interpretations. Table 1 shows the mean number of illness-days (and no. of cases) of Campylobacter infection that (1) are acquired via foreign travel versus domestically and (2) are ciprofloxacin resistant versus ciprofloxacin susceptible. This table shows the unadjusted data, without adjustment for the additional subset-selection and statistical-modeling steps used by Nelson et al. Although foreign travel is strongly associated with resistance to ciprofloxacin, and also with longer mean duration of diarrhea, resistance to ciprofloxacin is clearly not associated with longer mean duration of diarrhea among cases of domestically acquired campylobacteriosis.

    Any analysis that begins with crude data showing no apparent effect but that ends by concluding that there is a significant effect for some subsets of subjects must be especially diligent in the reporting of model diagnostics, as well as in correcting for potential model-selection bias, variable-selection and -coding biases, data subsetselection bias, and multiple-testing biases, all of which can threaten study validity by producing an excessive number of false-positives [2, 3]. Nelson et al. do not report such recommended diagnostics and corrections, which leaves open the possibility that the reported effects and conclusions result from modeling errors and biases. To remove this possibility, it would be desirable to (1) use nonparametric models (e.g., analysis of classification trees [4] instead of analysis of variance), (2) report the results of model-validation tests, and (3) correct the reported significance levels, to account for biases associated with the use of multiple tests [3]. The potential practical importance of such corrections is suggested by the fact that our own classification treebased analysis of the complete data set did not reveal any significant positive association between Campylobacter's resistance to ciprofloxacin and the duration of diarrhea.

    The policy implications of Nelson et al.'s article also deserve reexamination. Their stated premise that "human infections with ciprofloxacin-resistant Campylobacter have become increasingly common" appears to be incorrect, insofar as ciprofloxacin-resistant Campylobacter infections in humans are less common than they used to be [5]. In an article published during 2004, Kassenborg et al. estimated that 326,000 fluoroquinolone-resistant Campylobacter infections occurred in humans during 1998 [6]. On the basis of both 2002 data from the National Antimicrobial Resistance Monitoring System and updated data from FoodNet, it can be estimated that there are now 280,000 fluoroquinolone-resistant Campylobacter infections in humans per year [7, 8]; this represents a 14% reduction. The "human health consequences of such infections" (p. 1150) described in Nelson et al.'s articlethat is, more days of diarrheaappear to be associated specifically with foreign travel: it is not clear that they are in fact "consequences" of resistance in cases of domestically acquired Campylobacter infection. The conclusion that "persons with ciprofloxacin-resistant infection had a longer mean duration of diarrhea than did the persons with ciprofloxacin-susceptible infection (P = .01)" does not appear to hold for cases of domestically acquired Campylobacter infection overall (see table 1), and the statement that "this effect was independent of foreign travel" does not appear to be justified. Indeed, Nelson et al.'s report of a significantly (P = .01) decreased mean duration of hospitalization associated with ciprofloxacin-resistant Campylobacter infection (compared with ciprofloxacin-susceptible Campylobacter infection) is inconsistent with the conclusion that adverse consequences to human health derive from resistant infections.

    Finally, the recommendation "Additional efforts are needed to preserve the efficacy of fluoroquinolones" (p. 1150) does not follow from either the data or the analyses that Nelson et al. present; their data do not demonstrate a significantly increased duration of diarrhea in patients infected with Campylobacter strains that can be classified as fluoroquinolone resistant (P = .2), nor do they address the matter of ciprofloxacin's efficacy against resistant strains of Campylobacter. Their data show that, compared with cases of Campylobacter infection acquired domestically, cases acquired during foreign travel have both a greater likelihood of resistance to fluoroquinolone and more days of diarrhea; but this should not be confused with the assertion (implicit in Nelson et al.'s title) that excess days of diarrhea are caused by (or "Due to") resistance to ciprofloxacin, rather than by other foreign travelrelated factors.

    References

    1.  Nelson JM, Smith KE, Vugia DJ, et al. Prolonged diarrhea due to ciprofloxacin-resistant Campylobacter infection. J Infect Dis 2004; 190:11507. First citation in article

    2.  Greenland S. Modeling and variable selection in epidemiologic analysis. Am J Public Health 1989; 79:3409. First citation in article

    3.  Ottenbacher KJ. Quantitative evaluation of multiplicity in epidemiology and public health research. Am J Epidemiol 1998; 147:6159. First citation in article

    4.  Tong W, Hong H, Fang H, Xie Q, Perkins R. Decision forest: combining the predictions of multiple independent decision tree models. J Chem Inf Comput Sci 2003; 43:52531. First citation in article

    5.  Enrofloxacin for poultry: withdrawal of approval of new animal drug application NADA 140-828. FDA docket 00N-1571, hearing April 28, 2003. Available at: http://www.fda.gov/ohrms/dockets/dailys/03/Jul03/072403/00n-1571-tr00009-02-vol381.pdf. Accessed 17 March 2005. First citation in article

    6.  Kassenborg HD, Smith KE, Vugia DJ, et al. Fluoroquinolone-resistant Campylobacter infections: eating poultry outside of the home and foreign travel are risk factors. Clin Infect Dis 2004; 38(Suppl 3):S27984. First citation in article

    7.  Chiller TM, Stevenson JE, Barrett T, Angulo FJ, NARMS Working Group. National Antimicrobial Resistance Monitoring System (NARMS), 19962002: emerging multidrug and clinically important resistance in enteric bacteria. NARMS Scientific Meeting 2004 (Decatur, GA), 45 March 2004). Atlanta, GA: Centers for Disease Control and Prevention, 2004. First citation in article

    8.  Samuel MC, Vugia DJ, Shallow S, et al. Epidemiology of sporadic Campylobacter infection in the United States and declining trend in incidence, FoodNet 19961999. Clin Infect Dis 2004; 38(Suppl 3):S16574. First citation in article