The A, B, BI, and Cs ofClostridium difficile

The study by Belmares et al [1] in this issue of Clinical Infectious Diseases is the most comprehensive longitudinal study of the molecular epidemiology of Clostridium difficile infection (CDI) conducted to date. The study is impressive not only because restriction endonuclease analysis typing was performed on three-quarters of C. difficile isolates from 908 cases of CDI at the Minneapolis Veterans Affairs Medical Center (MVAMC) collected over a 10-year period but also because the investigators recognized the importance of conducting the study despite the 18-year time span between the collection of the C. difficile isolates and publication. The study confirms what other less systematic and/or less comprehensive studies have suggested: that the molecular epidemiology of C. difficile in health care facilities is frequently both diverse and dynamic [2–8]. The study by Belmares and colleagues must be interpreted in conjunction with the results of the other studies performed and in light of activities that took place during the same period at the MVAMC that may have affected C. difficile transmission [9–15]. Infection control policies, changes in health care provision, and interventions that took place over the 10-year period included the following: (1) placing patients with CDI in private rooms only if incontinent and unable to use a bedpan; (2) using a quaternary ammonium disinfectant for cleaning all hospital rooms; (3) authorizing nurses to send diarrheal stool samples for C. difficile testing without a physician order; (4) restricting the use of clindamycin starting in September 1985; (5) donning gloves as the standard for handling all body fluids starting in April 1988; and (6) moving all patients into a new building in June 1988. Although the study by Belmares and colleagues contributes significantly to our knowledge of C. difficile , it also highlights the need for more studies combining molecular epidemiology with CDI prevention efforts in order to understand the effectiveness of CDI prevention practices. Over the course of the study, 174 unique strains of C. difficile were identified. Three strains were associated with large clusters, 6 strains were associated with small clusters, and 1 strain was associated with 1 large and 1 small cluster. Of the 4 strains associated with large clusters, none were identified in >5 of the study years. Only 2 strains were identified in every year of the study. In contrast, 115 strains (66.4%) were sporadic and isolated in only a single study year, and no differences were found in the number of different strains identified during years with or without large clusters. This significant heterogeneity of strains over time appears to be different from the other 2 major health care-associated, multidrug-resistant, gram-positive organisms: methicillin-resistant Staphylococcus aureus and vancomycin-resistant enterococcus [16–18]. In addition, the identified molecular epidemiology of C. difficile makes it less clear whether any of the interventions to prevent CDI during the study period had an impact on the incidence of CDI and raise important questions about the prevention and future of C. difficile (Figure 1) [1, 12, 13]. Incidence of Clostridium difficile infection (CDI) and interventions that may have affected CDI incidence. Antimicrobial stewardship is believed to be one of the most effective methods to prevent CDI [19]. Restriction of single antimicrobials and more comprehensive antimicrobial stewardship programs have been associated with reductions in the incidence of CDI during both endemic and epidemic CDI periods [20–22]. In the study by Belmares and colleagues, 3 strains were associated with large clusters that had sufficient data to assess trends over time (B1 and B2 in 1982 and 1983 and B1 and F2 in 1985). The only intervention conducted in association with any of these clusters was clindamycin restriction in September 1985 (Figure 1) [1, 9, 10, 12]. The number of CDI cases associated with the B1 strain, which was clindamycin resistant, decreased from an average of 9 cases per month during the 3 months before clindamycin restriction to 1.2 cases per month after clindamycin restriction, and a significant reduction in CDI occurred in 1986 compared with 1985 (relative risk, 0.74; 95% confidence interval, 0.56–0.98). The B1 strain was not identified again after May 1986. However, clindamycin restriction in September 1985 does not explain why the B1 strain was identified only at very low levels in 1984 and in the first 5 months of 1985, nor does it explain what happened to the B2 or F2 strains in 1983 and 1985, respectively, or the increase in the incidence of CDI in 1991. A study by Clabots et al [15] conducted from September 1987 through June 1988 at the MVAMC provides important insights into the transmission of C. difficile at this health care facility and may help to explain the findings of the study by Belmares et al in this issue. Stool cultures or rectal swabs were obtained on admission and every 7 days thereafter for 94% of the patients admitted to the study ward, and cultures were obtained from the hospital environment at 4 time points. Sixty-five patients were colonized with C. difficile on admission to the ward, 54 patients acquired C. difficile , and the source of C. difficile acquisition could not be determined for 12 patients. Only 7 patients (5%) colonized with C. difficile were symptomatic (ie, had CDI). A total of 58 different C. difficile strains were identified. There were 19 strains that were associated with both being present on admission and acquisition. A sequence of introduction of one of these strains onto the ward by a colonized patient, followed by acquisition by another patient who was on the ward at the same time, was documented 16 times (84%). The carrier who introduced the strain onto the ward was asymptomatic for 15 of the 16 documented transmissions. In addition, only 1 of the 54 patients who acquired C. difficile shared a room with a patient colonized with the same strain, and 2 of the 15 patients admitted to a room with documented C. difficile contamination acquired the same strain as was in the room. The investigators concluded that patients asymptomatically colonized with C. difficile on admission to the hospital are an important source of transmission and that health care workers are the primary vector. No large clusters occurred during the study period, and the proportion of colonized patients with symptomatic CDI was significantly lower than in other studies [14, 15, 23]. The presence of a CDI outbreak or an increase in the proportion of patients with CDI may have produced different results [23, 24]. The study by Clabots and colleagues does help explain the diversity and high turnover of strains observed in the study by Belmares and colleagues. It also has important implications for infection prevention, because it indicates that identifying colonized patients on admission to the hospital may be necessary to prevent C. difficile transmission. Unfortunately, no validated methods for screening asymptomatic patients for C. difficile colonization are currently available. The most sensitive method to detect C. difficile is to perform stool cultures [25]. Culturing stool samples for C. difficile is relatively expensive, labor intensive, and slow. By the time stool culture results are finalized, many patients would be discharged. Although toxin assays have been used as a screening tool to identify asymptomatically colonized patients, no published data are available that validate the use of these assays for this purpose. The sensitivity and specificity of C. difficile toxin assays to detect asymptomatic colonization are unknown. Because fecal toxin levels are lower in patients with CDI who have milder symptoms, it is possible that the sensitivity of these assays to detect asymptomatic colonization will be even lower [25, 26]. At the very least, the lower prevalence of C. difficile colonization among asymptomatic patients will reduce the positive predictive value of these assays and increase the number of false-positive results. Use of screening tests with low sensitivity and low positive predictive values to detect C. difficile colonization could both miss patients actually colonized with C. difficile and unnecessarily increase the number of patients under contact precautions for C. difficile colonization because of false-positive results. Two notable events occurred in 1988 that may have affected the incidence of CDI—a new policy for wearing gloves when handling body fluids was implemented in April, and a new hospital was opened in June. The study by Johnson et al [13] that demonstrated the efficacy of glove use to prevent CDI was conducted at the MVAMC from July 1986 through August 1987 on 4 wards. Two wards were randomized to the glove intervention (glove use when handling body fluids), and a significant decrease was found in both the incidence of CDI (P=.015) and the proportion of patients colonized with C. difficile (P=.029) on the glove wards but not on the control wards. Unfortunately, insufficient time elapsed from the time glove use became a hospital-wide policy to the move to a new facility to know whether the sustained reduction in the incidence of CDI from 1988 until 1990 was due to enhanced glove use, the move to the new hospital, or both. Conversely, the reduction in CDI during this period may not have been related to either event, because there was already a significant downward trend in the incidence of CDI from 1982 to 1987 (P<.001, χ2test for trend) that continued through 1989 (Figure 1). What does the study by Belmares and colleagues say about the future of CDI? This is very difficult to assess because of the significant changes in CDI epidemiology that have occurred since the isolates were obtained for the study. Although some of the observed changes in CDI may be related to the epidemic BI strain, not all of the changes in CDI epidemiology can be explained by this strain. Not all CDI outbreaks since 2000 have been associated with the epidemic BI strain, and the presence of the epidemic BI strain has not always led to increases in the incidence or severity of CDI [3, 6, 8, 27]. Of the 908 cases of CDI at the MVAMC during the study period, the treatment failure rate was only 2%, and only 6% of patients had a documented relapse of CDI [12]. This finding is significantly different from the treatment failure rate of up to 38% and the relapse rate of up to 34% observed more recently [28–33]. The increase in treatment failure and relapse rates is not solely explained by the emergence of the epidemic BI strain because increases in treatment failure and relapse rates have been observed with non-BI strains as well [31, 34]. Only 7% of CDI cases at the MVAMC during the study period were not considered nosocomially acquired [12]. Nosocomial acquisition was defined as (1) onset of CDI >48 h after admission or (2) onset of CDI within 48 h of admission in patients who had been hospitalized at the MVAMC during the previous 2 weeks [11]. Patients with onset of CDI at an outside hospital or patients who were never previously hospitalized were not considered to have nosocomially acquired CDI [12]. Data from the Prevention Epicenters of the Centers for Disease Control and Prevention demonstrate that patients with community-onset, community-associated CDI and with onset of CDI at other health care facilities represent 21% of patients with CDI seen at the Prevention Epicenters' hospitals (26% if indeterminate cases are also included) [35, 36]. The increase in CDI on admission to the hospital may reflect shorter lengths of stay in the hospital (ie, patients are discharged after acquiring C. difficile but before the onset of CDI), increases in exposure to C. difficile in the community, increases in the transfer of patients with CDI from other health care facilities, increases in C. difficile virulence, or changes in patient risk factors (eg, medication exposure and underlying severity of illness) [37–40]. The Department of Health and Human Services [HSS] Action Plan to Prevent Healthcare-Associated Infections gives a 5-year prevention target to reduce health care-associated CDI by 30% but also notes the following: “Preventability of endemic CDI is unknown; therefore, the meeting attendee experts suggested that HHS revisit this target in 2 years as prevention research findings may become available” [41, p 14]. Several things must occur to have a better understanding of the preventability of CDI. First, we need a better understanding of the molecular epidemiology of C. difficile . Both hospital- and community-based surveillance systems are necessary to know from where specific strains of C. difficile originate and how changes in strain prevalence correlate with CDI incidence and outcomes. Second, we need to know more about what organism and host factors are associated with enhanced C. difficile virulence and transmission [40, 42]. Finally, we need properly designed, adequately powered studies of strategies to prevent CDI. Whenever possible, an assessment of the impact that interventions have on the molecular epidemiology of C. difficile should be included as well, as highlighted by the excellent work of Belmares and colleagues. I acknowledge Kimberly A. Reske, Margaret A. Olsen, and Victoria J. Fraser for editorial assistance. Potential conflicts of interest. E.R.D. has served as a consultant for Merck and Becton Dickinson and has received research funding from ViroPharma.