Reply to Cheng et al

To the Editor—We appreciate Cheng et al′s interest in our research [1] and offer the following comments and response to their comments [2].

First, concerning the possibility that confounding variables affected our analysis, we recognize that unmeasured potential confounders, such as the severity of previous Clostridium difficile infection (CDI), host immunity, and perception of risk of recurrent CDI due to an impending course of systemic antibiotics, could have affected our providers′ decision to prescribe oral vancomycin prophylaxis (OVP). However, although propensity scores in a retrospective study such as ours might have balanced observed baseline covariates between the 2 groups, they could not have balanced unmeasured characteristics or confounders [3]. Using univariate analysis, we compared 15 patient variables (including several commonly cited risk factors for CDI, such as older age, exposure to specific high-risk antibiotics or gastric acid suppressants, and duration of systemic antimicrobial therapy [4, 5]) and failed to find any significant differences between the 2 groups that would have explained lower CDI rates in the OVP group. Contrary to Cheng et al′s assertion, we did not suggest that “the OVP patients were more ill, and are likely to relapse.” Instead, we stated that we could not exclude the possibility that patients perceived to be at high risk of CDI might have been preferentially placed on OVP by their providers. If true, such perception could have served as an unmeasured potential confounder, which would have in turn likely negated, not overestimated, the measured efficacy of OVP. Whether the providers’ perception of recurrent CDI in our study actually reflected reality cannot be assumed, however.

As for the number of deaths in the 2 groups, we found no evidence that they had any impact on our study results. Specifically, there were 3 death in the OVP group and 6 deaths in the control group during the study period (4.2% vs 4.5%, respectively; P = 1.0). Recalculation of the data when all deaths were excluded from analysis resulted in 3 (4.4%) of 68 patients in the OVP group and 35 (27.8%) of 126 patients in the control group being diagnosed with recurrent CDI (odds ratio, 0.12; 95% confidence interval, .04–.41). These results are virtually identical to those reported in our article [1].

As for the possibility of “inadequate” duration of patient follow-up in the diagnosis of recurrent CDI, we chose an observation period of 4 weeks following discontinuation of systemic antimicrobial therapy because the highest incidence of CDI in hospitalized patients occurs 4 weeks after hospital discharge [6], as we stated. Since many patients were discharged from the hospital while receiving antibiotics, the duration of follow-up following discharge was often longer than 4 weeks. Although the risk of CDI may be increased for up to 90 days following cessation of antibiotic therapy, as stated by Cheng et al, it is equally important to stress that up to 60% of recurrences are due to the acquisition of new strains of C. difficile, often months following oral vancomycin therapy [7]. Furthermore, the patients in our OVP group completed their oral vancomycin course shortly (mean 0.8 days) after completion of their systemic antibiotics. Under these circumstances, extending the patient follow-up period beyond 4 weeks would have had questionable value since OVP would not have been expected to have an appreciable impact on CDIs caused by C. difficile strains acquired after completion of its course. To the best of our knowledge, Cheng et al′s contention that OVP may delay the time to relapse beyond 4 weeks lacks experimental and epidemiologic evidence and may not be very likely given the expected rapid drop in vancomycin stool concentrations soon after discontinuation of therapy [8, 9].

Concerning the potential for underutilization of testing for CDI in the OVP group, we suggest that the current era of heightened awareness of CDI among healthcare providers (and the public in general) is more likely to lend itself to overutilization [10] rather than underutilization of CDI testing. We also believe that this phenomenon may be particularly relevant in patients whose providers might have perceived them to be at high enough risk of recurrent CDI to warrant OVP. Indeed, our finding that 2 of 3 patients in the OVP group who tested positive for C. difficile were eventually considered to be asymptomatic and probably had little indication for C. difficile testing supports this position.

As suggested by Cheng et al [2] and stated in our article [1], further prospective trials would be ideal to evaluate the efficacy of OVP in patients with prior history of CDI but in need of systemic antimicrobial therapy. We also acknowledge that our study raises questions that merit further investigation. For example, although we included patients who had previous bouts of CDI up to 23 months prior to their placement on systemic antibiotics (mean 6 months), it is not clear if patients with longer intervals from their last CDI would also benefit from OVP. It is also not known if we should consider the severity of prior CDI when deciding which patients may benefit from OVP. Similarly, the ideal OVP regimen as relates to its dosage (eg, 125 mg vs 250 mg) or frequency (eg, twice daily vs less frequent) has not been determined. The impact of OVP on the gastrointestinal flora, particularly as relates to colonization and infection with multidrug-resistant organisms such as vancomycin-resistant enterococci, also warrants formal studies.

In conclusion, our work should be regarded only as a first step toward evaluating a possible role for OVP in preventing recurrent CDI in patients who require systemic antibiotic therapy. Nevertheless, until further studies are performed, OVP may offer a potential preventive strategy in reducing the risk of recurrent CDI, particularly in patients considered to be at high risk of complications.

Note

Potential conflict of interest. Both authors: No reported conflicts. Both authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References