Clostridium difficile diarrhea and colonization after treatment with abdominal infection regimens containing clindamycin or metronidazole

Clostridium difficile Diarrhea and Colonization after Treatment with Abdominal Infection Regimens Containing Clindamyein or Metronidazole Dale N. Gerding, MD, Mary M. Olson, RN, Stuart Johnson, MD, Lance R. Peterson, MD, J a m e s Z. L e e , Jr., MD, PhD, Minneapolis, Minnesota

One hundred fifty-six patients with presumed or documented abdominal infections were treated with amikacin/metronidazole/placebo (Group 1, 56 patients), amikacin/clindamycin/placebo (Group 2, 57 patients), or amikacin/clindamycin/ampicillin (Group 3, 43 patients) to determine both the therapeutic efficacy of the various regimens and the type of complications due to Clostridium difficile. C. difficile diarrhea occurred in 15 of 156 patients (9.6%), and C. difficile colonization occurred in 14 of 156 patients (9%) during treatment and 30 days of follow-up. The number of C. difficile diarrhea cases in Group 1 (3 of 56) was significantly lower than in Group 2 (9 of 57, p <0.05), but not in Group 3 (3 of 43). Exclusion of all patients who received other antibiotics in the 30-day poststudy period revealed no C. difficile diarrhea or colonization in Group 1 (0 of 13) and an acquisition rate of 31% (14 of 45) with the regimens containing clindamycin (p <0.02). Successful treatment outcomes (106 evaluable patients) were not statistically different among the three groups (Group 1, 64%; Group 2, 76%; and Group 3, 88%), but these data were difficult to interpret because, by chance, significantly more patients in Group 1 had bacteremia at entry (p <0.01), and patients in Group 3 had significantly more biliary tract infections (p <0.02) and significantly more favorable acute physiology scores (p <0.05). Use of metronidazole can reduce complications related to C. difficile, particularly if additional antimicrobials other than aminoglycosides are avoided.

From the Infectious Disease Section, Medical Service, and the Surgical Service, Veterans Administration Medical Center and the University of Minnesota Medical School, Minneapolis, Minnesota. Presented in part at the 26th Interscience Conference on Antimicrobial Agents and Chemotherapy, New Orleans, Louisiana, September 28-October 1, 1986. Supported by a grant from Searle Pharmaceuticals, Inc., Chicago, Illinois, and the Veterans Administration, Washington, D.C. Requests for reprints should be addressed to Dale N. Gerding, MD, Infectious Disease Section/111 F, MVAMC, One Veterans Drive, Minneapolis, Minnesota 55417. Manuscript submitted November 11, 1988, revised February 28, 1989, and accepted March 21, 1989.

212

THE AMERICAN JOURNAL OF SURGERY

mpiric antibiotic therapy for intra-abdominal infecE tion remains a subject of interest because of the complications and high mortality rate associated with many of these infections. The importance of directing antibiotic therapy against anaerobic and aerobic pathogens has been recognized, in addition to proper surgical technique and timing [1,2]. One standard therapeutic regimen against which new regimens are compared is an aminoglycoside in combination with clindamycin [3]. Metronidazole has been evaluated against elindamycin in combination therapy in seven prospective clinical trials and is reported to be of comparable efficacy [4-10]. The side effects and adverse reactions of these antibiotics have been reported, but the complication of antibioticassociated colitis/diarrhea has not been systematically addressed in this setting. Clostridium difficile colitis/ diarrhea has been a significant problem in the Minneapolis Veterans A d m i n i s t r a t i o n Medical Center (MVAMC), with an annual case rate of between 81 and 149 cases per year from 1982 through 1987. The high incidence has been shown to be due to nosocomial acquisition, which occurs most often as a result of patient-topatient transmission (DNG, unpublished data). Use of clindamycin has frequently been associated with this complication and was the only single antimicrobial implicated in a previous, case-controlled study at this institution [11]. We, therefore, undertook a prospective, randomized, double-blind study to compare the complication rate of C. difficile diarrhea and colonization, as well as the therapeutic efficacy of the following regimens in the treatment of abdominal infections: metronidazole plus amikacin versus clindamycin plus amikacin plus ampicillin (the standard regimen at MVAMC for intra-abdominal infections) versus clindamycin plus amikacin alone. Our major study hypothesis was that patients treated with metronidazole would have a lower incidence of C. difficile diarrhea and colonization than those treated with clindamycin. A second hypothesis was that the combined use of ampicillin and clindamycin would result in higher C. difficile complication rates than clindamycin alone, possibly by the elimination of fecal enterococci with ampicillin [12,13]. PATIENTS AND METHODS Between August 1983 and July 1985, patients at the MVAMC with suspected or documented serious intraabdominal infections were randomly assigned to one of the three antibiotic treatment regimens after verbal consent was obtained from the patients or their next of kin. The study was approved by the MVAMC human studies

VOLUME 159 FEBRUARY 1990

C L O S T R I D I U M D I F F I C I L E AND ABDOMINAL INFECTION

research subcommittee. Patients were excluded from this study if they were allergic to aminoglycosides, clindamycin, penicillin, or metronidazole; if they had been treated with one of the study drugs within the previous 14 days; or if they were pregnant. The study was designed to enroll a minimum of 150 evaluable patients, approximately 50 patients in each of the three treatment regimens. Enrollment numbers were based upon previously determined C. difficile colonization rates of 20% to 25% for MVAMC surgical patients with the duration of hospitalization expected in this study. All drugs were administered over 30 to 60 minutes by intravenous drip infusion and labeled as amikacin, study drug A (clindamycin or metronidazole), and study drug B (ampicillin or placebo). The three treatment regimens were: Group 1, amikacin 7.5 mg/kg intravenous every 12 hours (dosage was initially adjusted for age and renal function according to the nomogram of Sarubbi and Hull [14] and subsequently by peak and trough serum levels), metronidazole 1 g intravenous initially, followed by 500 mg intravenous every 6 hours, and placebo intravenous every 6 hours; Group 2, amikacin (as for Group 1), clindamycin 600 mg intravenous every 6 hours, and placebo intravenous every 6 hours; and Group 3, amikacin (as for Group 1), clindamycin (as for Group 2), and ampicillin 1 g intravenous every 6 hours. The diagnosis of intra-abdominal infection was made on the basis of clinical, laboratory, and other adjunctive diagnostic studies. Patients who were initially entered into the study but who were without bacteriologic or other confirmation of infection were eliminated from the study. Other indications for elimination from the study included patient request, additional antibiotic therapy given during the treatment period, and significant drug toxicity, adverse effects, or both. Before treatment was started, two paired blood culture specimens (aerobic and anaerobic) were obtained; culture of the abdominal focus, if available, was submitted in anaerobic transport media for anaerobic and aerobic cultures; urine and sputum cultures were obtained if appropriate; a rectal swab or a stool specimen was cultured for C. difficile [15]; and if diarrhea was present (more than four watery stools/day), stool specimens were sent for ova and parasite examination, culture for Salmonella, Shigella, Staphylococcus, and Campylobacter species, and C. difficile culture and toxin assay. At the beginning of the study and weekly thereafter, a laboratory evaluation consisting of a complete blood cell count, platelet count, albumin, serum glutamic oxaloacetic transaminase (SGOT), alkaline phosphatase, bilirubin, and urinalysis was performed. The serum creatinine level was measured initially and three times per week. During therapy, all positive culture sites were recultured between days 3 and 6 of treatment, and the abdominal site was cultured weekly if possible. After treatment, all sites that were initially positive were recultured, all pretreatment laboratory tests were repeated 1 to 2 days after treatment, and a stool specimen was submitted for C. difficile culture and toxin assay. Patients were followed for occurrence of diarrhea (more than four watery

40 ~176

~

50 20 "5

I0 0

0'-4

5-'9 10-'14 1 5 ' - 1 9 20'-24 Acute Physiology Score

25'-29

Figure 1. Percentage of patients in each treatment group with acute physiology score in ranges indicated: Group 1, amikaein plus metronidazole (broken line); Group 2, amikacin plus clindamycin (doffed line); Group 3, amikacin plus clindarnycin plus ampicillin (solid Une).

stools/day) for I month after treatment. Those with diarrhea were evaluated with sigmoidoscopy when possible; stool specimens were sent for ova and parasite examination, cultured for C. difficile and other diarrheal pathogens as mentioned earlier, and assayed for C. difficile cytotoxin. An acute physiology score was retrospectively compiled for each patient, using data obtained upon enrollment into the study. The criteria used were modified from the APACHE system developed by Knaus et al [16] (Figure 1), using the same scoring system for each parameter utilized. The following variables were used unchanged: temperature, serum creatinine, serum albumin, total bilirubin, alkaline phosphatase, SGOT, white blood cells, platelets, positive cerebrospinal fluid culture, positive blood culture, and positive fungal culture. The following variables were added (with the weighted score in parentheses): cardiovascular/shock: receiv!ng vasopressors within 48 hours (4); respiratory: on mechanical ventilation within 48 hours (4) or diagnosis of chronic obstructive p u l m o n a r y disease (1); hematologic: hemoglobin more than 20 g/dL (4), 19.9 to 16.7 g/dL (2), 16.6 to 15.3 g/dL (1), 15.2 to 10 g/dE (0), 9.9 to 6.7 g/dL (2), less than 6.7 g/dL (4); septic: positive abdominal culture or gross peritoneal contamination (4); underlying diseases: diabetes mellitus (1); alcoholism: cirrhosis (5), portal hypertension or varices (4), pancreatitis or hepatitis (3), diagnosis of alcoholism (2). Age was weighted (1 to 8) as in the APACHE II system [17]. The maximum possible score in this scheme was 67. Stool specimens were cultured for C. difficile using selective CCFA media (cycloserine, cefoxitin, fructosecontaining egg-yolk agar) and incubated anaerobically for 48 hours [18]. C. difficile isolates were identified by colony morphology, Gram stain, and gas-liquid chromatography. C. difficile cytotoxin assay was performed on filtered stool supernatants using HEp-2 cell monolayers and confirmed by Clostridium sordellii antitoxin neutralization as described previously [19]. Blood, stool, pus, and other specimens were cultured by appropriate aerobic and anaerobic methods [20]. Toxicity of the treatment regimens was evaluated in

THE AMERICAN JOURNAL OF SURGERY

VOLUME 159

FEBRUARY 1990

213

GERDING ET AL

TABLE

I

Reasons for Patient Removal from Study after 3 or Less Days of Treatment* Group 1

Group2

Group3

Protocol violation (antibiotic) No infection (prophylaxis only) Protocol violation (other) Antibiotic toxicity Patient death Patient request

14 10 2 0 0 0

6 3 1 1 2 0

12 6 1 1 1 1

Total

26

13

22

* Group 1 received amikacin plus metronidazole, Group 2 received amikacin plus clindarnycin, and Group 3 received amikacin plus clindamycin and ampicillin.

those patients who completed a minimum of 3 days of treatment. A minimum of 5 days was required for evaluation of efficacy. The endpoint of the treatment period was the day that the study drugs were discontinued or when an additional antibiotic was added. Successful completion or treatment failure was based on clinical response (abdominal examination, fever, wound drainage, white blood cells, radiologic studies) and bacteriologic response. Diarrhea and renal and hepatic toxicities were specifically evaluated. Patients who developed diarrhea during treatment or within 1 month after stopping treatment were evaluated as mentioned earlier, and all patients were evaluated for C. difficile acquisition at the termination of treatment as previously mentioned. Renal toxicity was defined as an increase in the serum creatinine level of more than 0.5 mg/dL if the creatinine level was less than 3 mg/dL initially, and as an increase of more than 1 rag/ dL if the initial creatinine level was more than 3 mg/dL. Hepatic toxicity was determined from increases in SGOT, alkaline phosphatase, or bilirubin levels during or after therapy. Results were analyzed by Student's t test, chi-square, and Fisher's exact methods. RESULTS Two hundred seventeen patients were initially enrolled in the study, with 61 patients removed prior to day 3 of treatment. Twenty-six, 13, and 22 patients were removed from the amikacin/metronidazole/placebo (Group 1), amikacin/clindamycin/placebo (Group 2), and amikacin/clindamycin/ampicillin (Group 3) treatment groups, respectively. The protocol violations that necessitated elimination from the study are listed in Table I. A change in the antibiotic regimen was the most frequent reason for elimination, followed by failure to document infection. Of the remaining 156 patients, 56 in Group 1, 57 in Group 2, and 43 in Group 3 were treated for more than 3 days and were followed up for complications related to C. difficile disease. Disease severity risk factors, including shock and respiratory failure during treatment, for the three patient groups are compared in Table II. Patients were similar in risk factors except that 11 of the 14 patients (79%) with positive blood cultures at the time of 214

THE AMERICAN JOURNAL OF SURGERY

TABLE II Risk Factors and Underlying Disease in Evaluable Patients* Group 1

Group2

(n = 56)

(n = 57)

Group3 (n = 43)

p Value

42 24 20 5 22 10 20 11 11

46 29 14 2 19 12 14 8 2

29 19 12 5 13 4 10 5 1

NS NS NS NS NS NS NS NS <0.O 1

Surgical procedure Age > 6 5 years Alcoholism Diabetes mellitus Malignancy Shock

Respiratory failure Albumin <2.5 g/dL Bacteremia at entry

" Group 1 received amikacin plus metronidazole, Group 2 received amikacin plus cliedamycin, and Group 3 received amikacin plus clindamycin and ampicillin.

entry into the study were randomly assigned to Group 1 (p <0.01). One hundred seventeen of the 156 patients (75%) required abdominal operations, with a single operation performed in 78% of those 117 patients (32 Group 1, 35 Group 2, and 24 Group 3). Thirty-one of these 156 patients had therapy stopped at 4 days for the following reasons: successful completion (18), protocol violation (5), antibiotic toxicity (4), antibiotic failure (2), and patient death (2), and their data were not included in the evaluation for efficacy. One hundred twenty-five patients completed 5 or more days of treatment with the study medications, the minimum required by protocol rules for efficacy evaluation. Nineteen of these patients did not have abdominal infections (abdominal malignancy mimicking infection or infections located outside the abdomen documented instead), leaving 106 patients (33 Group 1, 41 Group 2, and 32 Group 3) who were evaluated for efficacy. The median durations of treatment with study medications were 9, 9, and 8 days, respectively, for patients in Groups 1, 2, and 3. Patients in Groups 1 and 2 in the efficacy analysis had nearly identical mean acute physiology scores of 13.9 and 14, which were significantly higher than the Group 3 mean of 11.3 (p <0.05) (Figure 1). The acute physiology score also correlated positively with patient mortality, with a mean of 12.5 for the survivors and a mean of 17.8 f o r those who died (p <0.01). The specific infectious diagnoses or diagnostic categories and their distribution within the three efficacy groups are listed in Table III. Abdominal abscesses were the most frequent infections. Biliary tract infection was the next most frequent diagnosis. Forty-seven percent (15 of 32) of abdominal infections in Group 3 originated in the biliary tract, compared with 21% (7 of 33) in the Group 1 patients and 22% (9 of 41) in the Group 2 patients (p <0.02 Group 3 versus Groups 1 and 2). All of the biliary tract infections were lower-risk infections, either cholecystitis or choledocholithiasis, except for two cases of obstructed common bile duct (one Group 1 patient and one Group 2 patient) and one case of ascending cholangi-

VOLUME 159

FEBRUARY 1990

C L O S T R I D I U M D I F F I C I L E AND ABDOMINAL INFECTION

T A B L E III MOSt Frequent Infection Diagnoses In Patients Evaluated for Efficacy of Treatment *t

Abdominal abscess Billary tract infection~

Peritonitis Perforated appendix Perforated colon Diverticulitis Perforated stomach/duodenum Perforated small bowel

Group 1 (n = 33)

Group 2 (n = 41)

Group 3 (n = 32)

14 7 7 4 4 5 6 3

15 9 10 6 4 5 2 2

11 15 4 3 3 6 1 1

* Totals in columns exceed number of patients due to multiple diagnoses in some cases. r Group 1 received amikacin plus metronidazole, Group 2 reoeiver amikacin plus clindamycin, and Group 3 received amikacin plus clindamycin and ampk;illln. p <0.02 for Group 3 vs. Groups 1 and 2.

tis (one Group 2 patient). Cultures of the abdominal focus were obtained in 64 of the 106 patients assessable for efficacy. Pathogens were isolated from 50 of these 64 patients (16 Group 1, 19 Group 2, and 15 Group 3). Mixed aerobic-anaerobic infections were documented in 52% (26 of 50), aerobic bacteria only were found in 42% (21 of 50), and 6% (3 of 50) had only anaerobic bacteria cultured. Bacteroides species were the most frequently isolated bacteria, followed by Escherichia coli and viridans streptococci (Table IV). Eleven of the 156 patients (7%) were colonized with C. difficile on entry into the study (before receiving study antibiotics), and 4 eventually developed C. difficile-associated diarrhea, 3 while still receiving study antibiotics and 1 after antibiotics were stopped. An additional 14 patients (9%) who were initially culture-negative became asymptomatically colonized but did not develop diarrhea (Table V). Asymptomatic colonization occurred during study drug administration in 11 of these patients and after stopping study drugs in 3. Overall, 10% of the patients (15 of 156) developed diarrhea attributed to C. difficile (either culture-positive or cytotoxin-positive). Four of these patients were colonized on entry into the study, but 11 did not have positive C. difficile cultures until the time diarrhea occurred. Sigmoidoscopy was performed on 7 of the 15 patients with diarrhea but was diagnostic for pseudomembranous colitis in only 1 patient, despite culture- and cytotoxin-positive stools in all 7 patients. The C. difficile colonization rates before treatment and during the study were not significantly different among the three treatment groups (Table V). All four of the Group I patients who became colonized with C. difficile after study entry, however, received additional antibiotics before colonization~ the antibiotics included cefaz0lin (two patients), clindamycin (two patients), cephalexin (one patient), amoxicillin (one patient), and penicillin (one patient). None of the six Group 2 patients and two of the four Group 3 patients who became colonized received additional antibiotics before colonization (p <0.02 Group

TABLE

IV

Organisms Isolated from Abdominal Sites by Treatment Group* Group 1 Anaerobic bacteria Gram-negative bacilli Bacteroides species B. fragilis Other Gram-positive bacilli Anaerobic cocci Total Aerobic (facultative) bacteria Gram-negative bacilli Escherlchia coil Klebsiella Other Gram-positive bacteria Vii'idans streptococci Enterococcus Other Total Yeast

7 6 3 1 5 5

Group 2

Group 3

Total

29 23 2 6 14 4

21 18 4 3 5 5

57 47 9 10 24 14 95

14 7 3 4 9 6 2 1

25 9 4 12 9 5 3 1

19 11 3 5 8 3 2 3

58 27 10 21 26 14 7 5

2

5

0

84 7

* Group 1 received amikacin plus metronidazole, Group 2 received [ amikacin plus clindamycin, and Group 3 received amlkacin plus clindamycin and ampicillin.

TABLE V Clostrldlum difficlle Colonization and Diarrhea by Treatment Regimen for all Patients Treated 3 Days or L o n g e r * t Group 1 Group 2 Group 3 C. difficile asymptomatic colonization at study entry (four patients later developed diarrhea) Became colonized with C. difficile during treatment and 30-day follow-up (no diarrhea) Developed symptomatic C. difficile diarrhea during treatment and 30-
4/56

5/57

2/43

4/56

6/57

4/43

3/56

9/57

3/43

7/56

15/57

7/43

0/13

10/26

4/19

" Group 1 received amikacin plus metronidazole, Group 2 received amikacin plus clindamycin, and Group 3 received amikacin, clindamycin, and ampicillin. r p <0.05, Group 1 vs. Group 2; 11 patients developed diarrhea without evidence of prior C. difficile colonization. 4 patients who developed diarrhea were colonized on admission to the stu0y and were culturepositive at the time diarrhea occurred. $ p <0.01, Group 1 vs. Group 2; p <0.02, Group 1 vs. Groups 2 and 3.

2 versus Group 1). The C. difficile diarrhea rate was significantly less in tile Group 1-treated patients than in the Group 2-treated patients (p <0.05). Again, all three of the Group 1 patients With C. difficile diarrhea had received additional antibiotics in the follow-up period,

THE AMERICAN JOURNAL OFSURGERY

VOLUME 159

FEBRUARY 1990

215

GERDING ET AL

TABLE VI Clinical Outcome of Patients Evaluated for Efficacy**

Successful completion Clinical failure* Death on treatment Antimicrobial toxicity Protocol violation [antibiOtic) Protocol violation (other) Patient request Total

Group 1

Group 2

Group 3

21 (64) 6 (18) 2 (6) 2 (6) 2 (6) 0 0 33

31 (76) 3 (7) 0 2 (5) 1 (2) 1 (2) 3 (7) 41

28 I88) 1 (3) 0 2 (6) 0 1 (3) 0 32

~ Group 1 received amikacin and metronidazole, GroUp 2 received amikacin and clindamycin, and Group 3 received amikacin plus clindamycin and ampicillin. T Values in parentheses are percentages. * p = 0.05 for Group 1 vs. Group 2; all other differences insignificant.

before onset of symptoms. When both colonization and diarrhea rates were compared, the C. difficile acquisition rate was lower in the metr0nidazole group (7 of 56, 13%) than in the groups with clindamycin regimens (22 of 100, 22%) but was not statistically different (p >0.05). When patients who received other antibiotics in the 30-day follow-up period were excluded, the C. difficile acquisition rate was significantly lower (p <0.02) in the metronidazole group (0 of 13) than in the clindamycin-treated patients (14 of 45, 3I%) (Table V). Clinical outcome as determined on completion of treatment with the study drugs is shown in Table VI. Eighty patients (75%) had successful outcomes. Although more Group 3 patients (88%) had successful treatment completions than Group 2 (76%) or Group 1 (64%) patients, this was not significant (p >0.05). In 10 patients, therapy was deemed to have failed (Table VI); the difference between Group 1 (18%) and Group 3 (3%) was significant (p = 0.05). The overall mortality, evaluated after the 60-day posttreatment period in all 156 patients, was significantly higher in Groups 1 and 2 than in Group 3 (p <0.05, Group I versus Group 3; p <0.05, Group 2 versus Group 3). When the actual cause of death was considered, the mortality rate directly attributable to infection was not significantly different among the three groups (Table VII). The cause of death in 5 of the 11 Group 1 patients was considered the result of metastatic intra-abdominal malignancy, whereas the deaths of 2 other Group 1 patients were attributed to a massive intraabdominal/retroperitoneal hemorrhage and myocardial infarction, respectively. The three infection-related Group 1 deaths were due to peritonitis or bacteremia secondary to peritonitis. One patient in Group 2 died of pseudomembranous colitis 43 days after completion of the study. Elevations of SGOT were seen in 21% (33 of 156) of the patients during treatment and follow-up. Elevations of alkaline phosphatase and bilirubin were seen in 16% (25 of 156) and 13% (21 of 156), respectively. There were no significant differences in liver function abnormalities among the three treatment groups as measured by these enzyme and bilirubin analyses. Five percent of the pa216

T H E A M E R I C A N J O U R N A L OF S U R G E R Y

TABLE V l l

Mortality in Evaluable Patients**

Ovel'all (60--day follow-up) T During therapyw Within 30 days follow-up Infection-related

Group 1

Group 2

Group 3

11 (20) 4 (7) 6(11) 3 (5)

8 (14) 0 7 (12) 5 (9)

1 (2) 0 1 (2) 1 (2)

" Group 1 received amikacin plus metronidazole, Group 2 received amikacin plus clindamycin, and Group 3 received amikacin plus clindamycin and ampicillin. r Values in oarentheses are percentages. t p <0.05 Group 1 vs. Group 3, p = 0.04 Group 2 vs. Group 3. wp = not significant Group 1 versus Group 3.

tients (8 of 156) had a significant rise in serum creatinine: Group 1, 6 of 56; Group 2, 2 of 57, and Group 3, 0 of 43 (p <0.05, Group 1 versus Group 3). COMMENTS Antibiotics other than the study drugs were used extensively during the 30-day follow-up period, which complicated the assessment and was not anticipated in the original study design. The use of additional antibiotics in the follow-up period clearly influenced the C. difficile colonization and disease rates in Group 1 patients. Patients treated with amikacin plus metronidazole alone in this study did not acquire C. difficile infection or its attendant complications. This result was significantly different from the clindamycin-treated patients, who had a 31% acquisition rate. The overall C. difficile asymptomatic colonization rate of 16% (25 of 156) was similar to the carrier rate of 21% previously documented at this same institution the previous year as well as in other hospitals [11,21]. Although C. difficile diarrhea/colitis is a frequent complication of antibiotic treatment in our hospital, when recognized and treated promptly, this complication is usually not life-threatening. However, its recognition may be difficult, and its potential severity is evidenced in this study by the patient in Group 2 who died of pseudomembranous colitis. Considering the previous published experience and favorable therapeutic response to metronidazole [4-10], we expected to confirm this response, while comparing the rates of C. difficile acquisition and disease with those Of clindamycin-containing regimens. The superior outcome of Group 3 patients over Group 1 and Group 2 patients in terms of overall mortality at 60 days (Table VII) was quite unexpected. This outcome could be interpreted as showing improved efficacy of the Group 3 regimen, possibly due to broader coverage for gram-positive bacteria (i.e., enterococci) afforded by ampicillin. The bacteriologic findings, however, do not support a dominant role for enterococci. The more likely possibility is that despite randomization, Group 3 and Groups 1 and 2 were not comparable in terms of premorbid health status and disease severity. Group 3 was significantly different from both Groups 1 and 2 when measured by the modi-

VOLUME 159

FEBRUARY 1990

CLOSTRIDIUM DIFFICILE ANDABDOMINALINFECTION

fled acute physiology score system, a scoring system for which lower scores correlated with a lower risk of death. In addition, nearly half of the patients in Group 3 had uncomplicated biliary tract infections, compared to only one fifth of the patients in Groups 1 and 2, which again, likely biased the observed treatment response in favor of Group 3. Analysis of the cause of death indicated that only 3 of the 11 deaths in Group 1 were attributable t o infection, also suggesting that these patients had more serious underlying diseases and conditions at entry. It was also unexpected that 11 of the 14 patients who were bacteremic at study entry were randomly assigned to the Group 1 study arm. W h e n these 14 patients were excluded from the analysis, the outcome figures (successful completion, treatment failure, and death while receiving therapy) showed no significant differences among the three groups. The overall mortality, however, was still higher in Groups 1 (7 of 45) and 2 (8 of 55) than in Group 3 (1 of 42) (p = 0.03, Group 1 versus Group 3; p = 0.04, Group 2 versus Group 3). Metronidazole has been effective in the treatment of intra-abdominal infections, and our data indicate a lower predisposition to C. difficile sequelae than clindamycin, supporting our major study hypothesis. However, our data do not support our second hypothesis of a higher rate of C. difficile disease in association with regimens containing both clindamycin and ampicillin compared to only clindamycin plus amikacin. Neither asymptomatic colonization nor diarrhea was significantly increased in patients receiving the Group 3 regimen when compared with Group 2 patients (Table V), an observation that confirms our previous study of the relationship of prior antibiotic usage to C. difficile disease [11]. Based on our results, we consider metronidazole the preferred antianaerobic agent for treatment of abdominal infections in institutions with high endemic rates of C. difficile disease.

Complications of well-intentioned therapy are never welcome. This study places in seemingly honest perspectire both colonization by and clinically significant diarrhea caused by C. difficile. REFERENCES 1. Swenson RM, Lorber B, Michaelson TC, Spaulding EH. The bacteriology of intra-abdominal infections. Arch Surg 1974; 109: 398-9. 2. Gorbach SL, McGowan K. Comparative clinical trials in treatment of intra-abdominal sepsis. J Antimicrob Chemother 1981; 8(suppl D): 95-104. 3. Gorbach SL. Current experience with clindamycin in the treatment of abdominal and female pelvic infections. Scand J Infect Dis 1984; 43(suppl): 82-8.

4. Van der Auwera P, Collier J, Goris RJA, Saario I, Willis AT. A comparison of metronidazole and clindamycin for the treatment of intra-abdominal infection: a multicenter trial. J Antimicrob Chemother 1982; 10: 57-66. 5. Canadian Metronidazole-Clindamycin Study Group. Prospective, randomized comparison of metronidazole and clindamycin, each with gentamicin, for the treatment of serious intra-abdominal infection. Surgery 1983; 93: 221-9. 6. Smith JA. Skidmore AG, Forward AD, Clarke AM, Sutherland E. Prospective, randomized, double-blind comparison of metronidazole and tobramycin in the treatment of intra-abdominal sepsis. Ann Surg 1980; 192: 213-20. 7. Stone HH, Fabian TC. Clinical comparison of antibiotic combinations in the treatment of peritonitis and related mixed aerobicanaerobic surgical sepsis. World J Surg 1980; 4: 415-21. 8. Collier J, Colhoun EM. Hill PL. A multicentre comparison of clindamycin and metronidazole in the treatment of anaerobic infections. Stand J Infect Dis 1981; 26(suppl): 96-L100. 9, Kirkpatrick JR, Anderson BJ. Louie J J, Stiver HG. Doubleblind comparison of metronidazole plus gentamicin and clindamycin plus gentamicin in intra-abdominal infection. Surgery 1983; 93: 215-6. 10. Biron S; Brochu G, Beland L, etal. Short-term antibiotherapy for peritonitiS: prospective, randomized trial comparing cefotaxime-metronidazole and clindamycin-tobramycin. J Antimicrob Chemother 1984; 14(suppl B): 213-6. 11. Gerding DN, Olson MM, Peterson LR, et al. Clostridium difficile-asseciated diarrhea and colitis in adults. A prospective case-centrolled epidemiologic study. Arch Intern Med 1986; 146: 95-100. 12. Rolfe RD, Helebian S, Finegold SM. Bacterial interference between CIostridium difficile and normal fecal flora. J Infect Dis 1981; 143: 470-5. 13. Malamou-Ladas H, Tabaqchli S. Inhibition of Clostridium difficile by fecal streptococci. J Med Microbiol I982; 15: 569-74. 14. Sarubbi FA, Hull JH. Amikacin serum concentrations: prediction of levels and dosage guidelines.Ann Intern Med 1978; 89: 6128. 15. McFarland LV, Coyle MB, Kremer WH, Stamm WE. Rectal swab cultures for Clostridium difficile surveillance studies. J Clin Microbiol 1987; 25: 2241-2. 16. Knaus WA, Zimmerman JE, Wagner DP, Draper EA, Lawfence DE. APACHE-acute physiology and chronic health evaluation:a physiologically based classification system. Crit Care Med t981; 9: 59t-7. 17. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: A severity of disease classification system. Crit Care Med 1985; 13: 818-28. 18. George WL, Sutter VL, Citron D, Finegold SM. Selective and differential medium for isolation of Clostridium difficile. J Clin Microbiol 1979; 9: 214-9. 19. Shanholtzer CJ; Peterson LR, Olson MM, Gerding DN. Prospective study of Gram stain stool smears in diagnosis of Clostridium difficile. J Clin Microbiol 1983; 17: 906-8. 20, Lennette EH, Balows A, Hausler WJ, Truant JP, eds. Manual of clinical microbiology, 3rd ed. Washington, DC: American Society for Microbiology, 1980. 21. Burdon DW. Clostridiumdifficile: The epidemiology and prevention of hospital acquired infection. Infection 1982; I0: 203-4.

THE AMERICAN JOURNAL OF SURGERY VOLUME 159 FEBRUARY 1990 217