Beginning in 1990, an outbreak of C. difficile diarrhea was identified at the Veterans Affairs Medical Center in Tucson, Arizona. The epidemic was eventually terminated after identifying an association of C. difficile infection with clindamycin use and after instituting clindamycin restriction. In this report, we describe the outbreak, identify a predominant clindamycin-resistant strain of C. difficile by restriction endonuclease analysis, document the association of illness with clindamycin use, and show a decreased frequency of clinical cases and the epidemic strain after clindamycin restriction.

Methods

Top Methods Results Discussion Author & Article Info References

The Veterans Affairs Medical Center where the epidemic occurred is a Dean's Hospital affiliate of the University of Arizona. Hospital discharges and daily census figures showed no significant differences for the 5 years before and during the outbreak; they average 6619 ±384 patient admissions per year and a census of 168 patients per day. The infection control program is administered by two nurse specialists under the supervision of an infectious diseases physician. The frequency of C. difficile-associated diarrhea has been monitored since 1986 by continuous review of microbiology reports of stool specimens submitted for C. difficile culture or toxin assay. Antibiotic use is restricted for most staff physicians and housestaff to a hospital formulary, and nonformulary antimicrobial agents are only available with approval from an infectious diseases physician. For at least the past 15 years, the formulary has included clindamycin (Cleocin; Upjohn Company, Kalamazoo, Michigan) and has excluded oral vancomycin.

Stool specimens from patients with diarrhea who are suspected by their primary care physicians of having colitis were tested for C. difficile. All specimens were tested for cytotoxin by cell culture assay [5] before December 1991 and by latex agglutination [5, 6] thereafter. All specimens tested for cytotoxin and all specimens with a positive latex test result were cultured on cycloserine-mannitol blood agar to isolate C. difficile [7]. For purposes of case definition, all patients who had diarrhea in association with a positive assay for C. difficile-stool cytotoxin or for latex agglutination and who had recent antibiotic therapy (within 60 days) were considered to have C. difficile infection. Diarrhea was defined as 4 or more loose or unformed stools in a 24- to 36-hour period; all specimens submitted to the laboratory were tested.

Typing of C. difficile organisms was done on all available unique isolates by restriction of total organism DNA with Hind III restriction endonuclease followed by agarose gel electrophoresis, as previously described [8]. Organism groups and types were determined from comparison with a library of more than 250 types [8].

Clindamycin susceptibility was determined for selected isolates of C. difficile using a microdilution method in Wilkins-Chalgren broth. Clindamycin dilutions were 0.125, 1.0, and 4.0 µg/mL. Plates were incubated for 48 hours in an anaerobic glove box. The organism was considered resistant to clindamycin if the minimal inhibitory concentration (MIC) was greater than 4.0 µg/mL. Susceptibility to imipenem, penicillin G, amoxicillin-clavulanic acid, ticarcillin-clavulanic acid, metronidazole, and ciprofloxacin were also tested by the same microdilution method.

To test for the significance of differences, we used the Fisher exact test for categorical variables and the Mann-Whitney U test for continuous variables. To test for the significance of associations, we used the Dunnett multiple range test [9].

Results

Top Methods Results Discussion Author & Article Info References

Description of the Outbreak

In July 1990, 8 patients who had C. difficile infection (15.8 infections per 1000 discharges) were identified at our center (Figure 1). For the previous 33 months, the average monthly incidence (mean ±SD) had been 1.5 ±1.1 infections (2.6 infections per 1000 discharges) diagnosed from a mean of 15 stool specimens per month submitted for toxin assay (10% of submitted specimens positive per month). Through July 1991, the incidence of new patients was fivefold higher than this norm, averaging 7.7 infections per month (a total of 101 patients), with all but 1 of these months exceeding the 95% surveillance threshold (4 or more patients per month, normal mean + 2 SD). The mean number of unique patient specimens submitted increased to 36 per month, of which 28% were positive for toxin during this period. Of toxin-positive specimens, 71% and 65% were culture positive before and during the epidemic, respectively. Both the group of 439 isolates tested before the epidemic and the 334 isolates tested during the epidemic that were toxin negative yielded C. difficile in culture 5% of the time.

View larger version (26K): [in this window] [in a new window]   Figure 1. Number of cases of Clostridium difficile-associated diarrhea detected each month before and after clindamycin restriction. The epidemic period is shown in black bars and the pre- and postepidemic periods by white bars. The dotted horizontal line represents a historical threshold of 2 SD above the mean. The number of type J7 isolates (identified using restriction endonuclease analysis) compared with the number tested are shown below the bars.

The clinical symptoms for approximately 90% of the patients were abrupt onset of abdominal distension, cramping, and profuse diarrhea. Clostridium difficile infection resulted in intestinal infarction and eventual death of one patient. In most patients, diarrhea did not begin to resolve until after institution of specific treatment with either metronidazole or oral vancomycin.

The epidemic was recognized within 1 month of its inception, and several attempts at control were instituted, including hospital staff education, increased use of gloves by nursing staff when assisting patients, and improvement of environmental hygiene. Educational efforts began in early August 1990 and were carried out through the remaining calendar year. The medical center staff who participated in the educational process included attending physicians, housestaff, nursing, and housekeeping personnel. The objectives of the educational programs included a review of the epidemiology, transmission, and clinical presentation of C. difficile. Emphasis was placed on prevention of C. difficile transmission through hand washing, glove use, timely changing of gloves, and increased availability of gloves. In addition, routine environmental sanitation practices were reviewed for thoroughness, with emphasis placed on patient-related items, such as bedside commodes and bathrooms. None of these efforts appeared to alter the frequency of the new cases.

Organism typing by restriction endonuclease analysis was carried out on isolates of C. difficile from 10 patients diagnosed in November and December of 1990 and from 48 other patients diagnosed from January through July of 1991. Seventeen restriction endonuclease analysis patterns were identified, indicating a high degree of polymorphism. However, isolates from 34 (59%) of the patients had an identical restriction endonuclease analysis pattern, type J7 Figure 2, with only 1 to 5 isolates representing each of the other types.

View larger version (148K): [in this window] [in a new window]   Figure 2. Restriction endonuclease analysis patterns of Clostridium difficile strains isolated during and after the epidemic. The left-hand lane contains DNA sizing markers; the other lanes contain digested DNA from different patients. The type J7 pattern (identified using restriction endonuclease analysis and shown in lanes 2 to 4) was evident in 59% of the patients, whereas other patients showed 16 other types, including type B12 (lane 5), type N1p (lane 6), and type D1 (lane 7).

The susceptibility to clindamycin of six strains of C. difficile, isolated from patients from Tucson (Veterans Affairs Medical Center), is shown in Table 1. All six isolates were susceptible to penicillin G, imipenem, amoxicillin-clavulanate, ticarcillin-clavulanate, and metronidazole and were resistant to ciprofloxacin (data not shown). Both J7 type strains were resistant to clindamycin.

Analysis of Clindamycin Use Associated with Clostridium difficile Infections

Several lines of evidence linked clindamycin use with the epidemic. First, before the onset of the epidemic, clindamycin use markedly increased. For the years 1986 through 1989, the average amount of clindamycin dispensed was 315 grams per month. In 1990, clindamycin use was 489 grams per month, an increase of 41% compared with the previous year (P < 0.001). Second, in a survey conducted by the College of American Pathologists in 1989 of 220 teaching facilities, our center ranked in the 99th percentile for clindamycin use [10]. This level of use was higher than for any other antimicrobial agent on our formulary (Table 2). Third, administration of clindamycin was significantly more likely to be associated with C. difficile infection than was administration of other commonly used antimicrobial agents (Table 2). Infected patients were three to nine times as likely to have received clindamycin compared with another antimicrobial agent. The association of C. difficile infection with clindamycin was greater than for each of the other antimicrobial agents (all P < 0.05 using the Dunnett multiple range test [9] for clindamycin compared with various antimicrobial agents).

The association of increased clindamycin use with the increased frequency of C. difficile infection was examined further by analyzing the temporal relation (Figure 3). For the 19 months ending in July 1991, a positive relation was found between clindamycin use in either the current or previous month and the number of infected patients (P = 0.04 and P = 0.01, respectively). The stronger correlation obtained by comparison with previous months' data could be due to a time lag for development and detection of C. difficile infections.

View larger version (18K): [in this window] [in a new window]   Figure 3. Relation of monthly clindamycin use from pharmacy records to new cases of Clostridium difficile-associated diarrhea. The regression line bounded by its 95% CIs indicates a positive relation between either current (left panel; r = 0.53, P = 0.02) or previous (right panel; r = 0.663, P = 0.003) months' use and the number of infected patients. The stronger correlation obtained by comparison with previous months' data could be due to a time lag for development and detection of C. difficile infections.

In a final analysis, a case–control strategy was used to assess the amount of clindamycin used in C. difficile-infected and uninfected patients. For each of the 27 patients who developed C. difficile infection and had received clindamycin, 4 other patients were identified from the same ward and time period (within 180 days of the case patient) who also received clindamycin but did not develop infection. Comparing the amount of clindamycin received by these two groups, we found that infected patients received more clindamycin than those without infection (83 grams per infected patient compared with 64 grams per uninfected patient; P = 0.022).

Effect of Decreased Use of Clindamycin

Based on these findings, clindamycin was removed from the hospital formulary beginning late in July 1991. With restriction, clindamycin use decreased from an average of 429 grams to 71 grams per month, a decrease of 83%. During the 15-month period, the mean proportion of C. difficile infections decreased to 1.9 per month (P < 0.001). Within 3 months after the restriction was instituted, the monthly incidence of infections remained below threshold (see Figure 1) and has continued to stay below threshold. A small decrease in the number of unique patients having specimens submitted for toxin detection also occurred [31.5 stool specimens per month during the epidemic compared with 26 specimens per month subsequently; P = 0.02], possibly attributable to fewer hospitalized patients with diarrhea.

After clindamycin restriction, continued analysis of isolates for their endonuclease digest pattern showed a marked decrease in the proportion of J7 type strains. Only 2 (13%) of 15 patients had isolates of the J7 type, and in both cases the J7 type was isolated in a single month (October 1991). This frequency was less than during the epidemic (P < 0.001).

Discussion

Top Methods Results Discussion Author & Article Info References

Several features of this nosocomial outbreak of C. difficile are uniquely well documented and lead to new insights into the epidemiology of this hospital infection problem that is increasingly more frequent. First, the epidemic began abruptly. Second, it was recognized quickly but failed to respond to institution of various standard infection control measures. Third, a single restriction endonuclease analysis pattern of C. difficile, the J7 type, accounted for nearly 60% of new cases. Fourth, the hospital was a high user of clindamycin. Fifth, clindamycin use was strongly associated with cases of C. difficile diarrhea. Sixth, the predominant strain was resistant to clindamycin. Seventh, a successful decrease in clindamycin use was associated with a marked decrease in new cases. Finally, the restriction endonuclease analysis pattern associated with the epidemic diminished with resolution of the outbreak. Although these features do not exclude multiple factors from being important to the development of this outbreak, they do strongly implicate clindamycin use as crucial.

Association of C. difficile diarrhea with the use of clindamycin is not new. The first papers identifying pseudomembranous colitis as a clinical entity by Cohen and colleagues [11] in 1973 and by Tedesco and colleagues [12] in 1974 were both titled "Clindamycin- associated Colitis". Bartlett [13] subsequently isolated C. difficile and its toxins from stool specimens of seven of the Tedesco and colleagues' patients after storage of the specimens for 5 to 6 years. Others have subsequently identified clindamycin use as a risk factor for C. difficile diarrhea [3, 5, 14-17] and have identified ampicillin and cephalosporins as similarly associated with an increased risk [3]. In our study, the association of clindamycin use with increased risk for symptomatic C. difficile disease was documented in an institution that had much more clindamycin use than other teaching hospitals. In addition, a case–control study showed that patients with C. difficile infection had received substantially more clindamycin than control patients. Although possible drug interactions were not considered (for example, by excluding from consideration those patients who had received more than one of the antibiotics during the outbreak), the analysis shown in Table 2 was only one of several that implicated clindamycin as the primary drug responsible for this outbreak of antibiotic-associated diarrhea. Although it is still possible that clindamycin use at our institution was not causally responsible for the development of the C. difficile epidemic, this possibility seems the most likely, especially because of the decrease in the numbers of patients with the clindamycin-resistant J7 strain that occurred after clindamycin use was restricted.

The role of resistance to clindamycin by C. difficile as an added risk factor in patients who receive clindamycin is not clear. Although it would seem to be an added risk, we are aware of only one study [18] that correlated infection with clindamycin-resistant C. difficile and previous clindamycin therapy, a study that also identified a unique strain of C. difficile by restriction endonuclease analysis. In one previous outbreak [14], the frequency of clindamycin resistance (MIC greater than 4.0 µg/mL) increased from 42% before the outbreak to 67% during the outbreak and decreased to 43% afterward (P < 0.05 for both). In a study [19] of 98 isolates of C. difficile, 50% were susceptible to 4.0 µg/mL or less of clindamycin and 90% were susceptible to 128 µg/mL or less. Clindamycin concentrations in stool after oral administration of 600 to 1200 mg/d range from 125 to 448 µg/g [20]. In this study, the two epidemic J7 isolates (identified using restriction endonuclease analysis) that were tested had an MIC greater than 4.0 µg/mL, whereas only 1 of 4 other restriction enzyme analysis types that were tested had an MIC greater than 4.0 µg/mL. Even though the J7 pattern was distinct from others in this epidemic and from others reported previously [14], it is possible that some J7 isolates might be clindamycin sensitive if the genetic loci responsible for clindamycin resistance were small. Given the relatively short time frame of the epidemic, however, this possibility is unlikely.

Some previous studies have shown the presence of a predominant outbreak strain among nosocomial C. difficile isolates [14, 21-24], whereas others have shown many strains [25-27]. In a study [14] of one epidemic that began and ended abruptly, clindamycin resistance was correlated with an epidemic strain that was characterized initially by a unique plasmid profile and later by a unique restriction enzyme profile. However, in that study, patient antibiotic risk factors were not evaluated. In our study, a predominant restriction endonuclease analysis pattern (type J7) accounted for 59% of all of the typed outbreak isolates and was also clindamycin resistant. The frequency of isolation of this strain decreased markedly after successful curtailment of clindamycin use. Whether the epidemic strain was completely eradicated is unknown because several post-epidemic isolates were lost and could not be tested.

Perhaps the most striking feature of this study was the prompt cessation of the outbreak after restriction of clindamycin. Antimicrobial agent restriction of any kind is difficult to accomplish but was successfully used in this setting after we found that clindamycin use was linked to C. difficile cases. Others have used antimicrobial agent restriction of clindamycin as one of multiple infection control measures that were used simultaneously [14-16]. In the present outbreak, the initiation of educational measures, enforcement of barrier precautions, and increased attention to environmental cleaning did not have an effect on the case rate. Only after institution of clindamycin control measures did this outbreak resolve. New case rates decreased to below threshold within 3 months and have remained low (see Figure 1). In addition, the frequency of isolation of the J7 type of C. difficile also decreased after institution of clindamycin control [P < 0.001].

It is not clear how or when the J7 strain of C. difficile first appeared in this institution. Isolates were not available for typing from the period before the outbreak. A recent study [27] suggests that C. difficile strains are frequently introduced to the hospital by admitted patients who are asymptomatic carriers. We postulate that such a patient, symptomatic or asymptomatic, brought the organism into this permissive environment. The high rate of clindamycin use enabled the organism to persist and spread. The rapid resolution of the outbreak and the simultaneous disappearance of the type J7 C. difficile organism (identified using restriction endonuclease analysis) suggest that clindamycin use was an integral component of this nosocomial outbreak.

Analyses of antibiotic use, as were done in our study, may prove valuable in efforts to control other infectious epidemics. Such data are available from the hospital pharmacy database at many hospitals and should be readily accessible to infection-control workers. The relation of clindamycin use with C. difficile infection that emerged from using pharmacy data was an important factor in convincing physicians to cooperate with clindamycin restriction. Similar relations exist between the use of other antimicrobial agents and resistant nosocomial pathogens. Application of pharmacy data coupled with effective control measures for antimicrobial agents has been successful in controlling such outbreaks. Clostridium difficile diarrhea should now be added to the list of antimicrobial agent-resistant nosocomial pathogens for which antimicrobial agent control measures can be used successfully to control an outbreak.