Summary of findings
The administration of antibiotics is known to be associated with the development of diarrhea in patients. Erythromycin and the clavulanate in amoxicillin-clavulanate cause diarrhea by accelerating gastrointestinal motility (Bartlett 2002). Still other antibiotics are thought to cause diarrhea by reducing the number of fecal anaerobes, thereby decreasing carbohydrate digestion and absorption, leading to an osmotic diarrhea (Bartlett 2002). Nearly 15% of hospitalized patients receiving beta-lactam antibiotics develop diarrhea (McFarland 1995). Clostridium difficile (C. difficile) infection of the colon is another cause of antibiotic-associated diarrhea that is thought to be caused by an overgrowth of native or newly acquired C. difficile (Fekety 1993). C. difficile is implicated in 20 to 30% of patients with antibiotic-associated diarrhea, in 50 to 70% of those with antibiotic-associated colitis and in more than 90% of those with antibiotic-associated pseudomembranous colitis (Bartlett 1980; Bartlett 1990; George 1982; Kelly 1994). Approximately 5% of healthy adults asymptomatically carry low concentrations of C. difficile in their colon, and the growth of these bacteria has been shown in vitro to be held in check by normal gut flora (Fekety 1993). The exact mechanism by which C. difficile overgrowth occurs is still unclear, but it can occur with the administration of oral, parenteral or topical antibiotics (Fekety 1993, Thomas 2003). The incidence of C. difficile-associated diarrhea (CDAD) ranges from 1 in 100 to 1 in 1,000 hospital discharges depending on the antibiotic prescribing habits of the hospital (Ho 1996; Lai 1997; Manian 1995). Additionally, the incidence may change over time at the same hospital as it did in one study from approximately 1 in 300 to 1 in 100 hospital discharges (Olson 1994).
The consequences of CDAD can be severe. Seventy nine deaths due to CDAD were reported in Quebec, Canada in 2003 (Pindera 2004). At one academic medical center over a three-year period, 21 of 710 cases (3%) of C. difficile colitis required intensive care unit admission or died as a result of their infection (Rubin 1995). These deaths were associated with co-morbid conditions such as malignancy, COPD or renal failure and therapies such as immunosuppressive drugs, antiperistaltic medications or the prior administration of clindamycin (McFarland 1995). At another university hospital the morbidity of C. difficile infection was higher, and 24 of 157 patients (15.3%) with C. difficile colitis died from their infection (Morris 2002). The economic consequences of C. difficile infection can also be severe, with one report finding a mean cost of $10,970 (US$) per patient for the treatment of the infection and its complications (McFarland 1999). The reports from Quebec are the most worrisome, and subsequent reports of mortality rates in that province are much higher than previous reports (Pepin 2005). This is related to the appearance of a new variant of C. difficile, which is capable of secreting much higher amounts of toxin A & B and is more resistant to standard antibiotic therapy (Louie 2005). This new variant results in a higher incidence of CDAD among hospitalized patients (McDonald 2005), a greater need for urgent colectomy for toxic colitis and a high mortality rate (e.g. 37%; Pepin 2005). This variant, ribotype 027, has been reported in Quebec, the USA, and Europe (Louie 2005; Laurence 2006). The emergence of this highly virulent bacterium adds urgency to the identification of effective therapy.
It is important to explain the terms used in conjunction with C. difficile infection, as the studies analyzed in this review involve various descriptions of C. difficile disease. CDAD occurs in a patient with diarrhea that has tested positive for C. difficile toxin and/or positive stool culture of C. difficile. C. difficile colitis involves a stool test positive for the organism and signs of mucosal inflammation seen on endoscopy. Pseudomembranous colitis refers to the actual presence of pseudomembranes seen on endoscopy. These terms are extensively overlapping, but not necessarily synonymous, as many patients with antibiotic-associated diarrhea do not undergo the endoscopy and biopsy necessary for the diagnosis of colitis or the endoscopy and visualization of pseudomembranes. This systematic review will use the term CDAD for all forms of symptomatic C. difficile infection for its analysis.
The aim of this review was to investigate the efficacy of antibiotic therapy for CDAD, to identify the most effective antibiotic treatment for CDAD in adults and to determine the need for stopping the causative antibiotic during therapy.
Criteria for considering studies for this review
Types of studies
Randomized, controlled trials assessing antibiotic treatment for CDAD.
Types of participants
The types of participants included:
(i) patients with diarrhea: various definitions exist usually describing the consistency of stool, number of bowel movements per day, duration of symptoms, and rarely stool volume;
(ii) patients with C. difficile in stool identified by stool culture positive for C. difficile and or by stool positive for C. difficile cytotoxin;
(iii) patients who had received prior antibiotic therapy for an infection other than C. difficile; and
(iv) patients 18 years of age of older.
Patients were excluded if they did not have diarrhea or if there was no evidence of C. difficile infection.
Types of interventions
Only studies involving antibiotic therapy for C. difficile-associated diarrhea were included. These studies may include comparisons among different antibiotics, between different doses of the same antibiotic, or between antibiotic therapy and placebo.
Studies were excluded if they compared antibiotic to non-antibiotic therapy (such as resins, vaccines, probiotics or biologicals) or if even in the face of antibiotic treatment, the intervention being tested was of a non-antibiotic such as those listed above. In addition interventions used for the prevention of CDAD are not included in this review.
Types of outcome measures
The following outcomes were sought:
a. initial resolution of diarrhea;
b. initial conversion of stool to C. difficile cytotoxin and/or stool culture negative;
c. recurrence of diarrhea;
d. recurrence of fecal C. difficile cytotoxin and/or positive stool culture;
e. patient response to cessation of prior antibiotic therapy;
f. emergent surgery: fecal diversion or colectomy; and
Search methods for identification of studies
MEDLINE (1966 through March 24, 2010), EMBASE (1980 - March 24, 2010), the Cochrane Central Database of Controlled Trials and the Cochrane IBD Review Group Specialized Trials Register were searched using the following text search terms: "pseudomembranous colitis and randomized trial"; "Clostridium difficile and randomized trial"; "antibiotic associated diarrhea and randomized trial". These search terms were chosen for their general applicability to our study, so as not to miss any possible relevant trials given the variety of names given to C. difficile disease. Studies published only as abstracts were not explicitly excluded.
Data collection and analysis
At least three (or more in this update: 3 for each study) authors examined all the citations and abstracts derived from the electronic search strategy and independently selected trials to be included. Full reports of potentially relevant trials were retrieved to assess eligibility. Reviewers were not blind to the names of trials' authors, institutions or journals. Any disagreements about trial inclusion were resolved by group discussion.
The methodological quality of eligible trials in the initial review was assessed independently by two reviewers in terms of method of random allocation, allocation concealment, number of withdrawals and dropouts, whether data were analyzed on an intention-to-treat basis, and whether the participants and the outcome assessors were blind to the treatment provided. Since the quality of allocation concealment may particularly affect the results of studies, each of the two authors determined whether the quality of allocation concealment and the blinding of assessor and participants were adequate, inadequate or unclear. Where the method of allocation concealment was not clearly reported, attempts were made to contact the trials' authors for clarification. All disagreements were resolved by group discussion. For the current update the ten authors assessed study quality using the above criteria and the Cochrane Risk of Bias tool (Higgins 2008). Studies were excluded if they were not randomized controlled trials or if they made comparisons other than those specified. These studies are listed in the Characteristics of excluded studies table.
Data extraction was performed independently by at least two authors. Results were compared between reviewers and all studies were presented for group discussion. Where data may have been collected but not reported, further information was sought from the individual trials' authors.
Data were analyzed using Review Manager 5. For dichotomous outcomes, relative risks (RR) and 95% confidence intervals (CI) were derived from each study. When appropriate, the results of included studies were combined for each outcome. For dichotomous outcomes, pooled RR and 95% CI were calculated using a fixed effect model, except where significant heterogeneity was detected, at which time the random effects model was used. Data heterogeneity was assessed using a chi square test with significance set at P < 0.10. For the current update, the denominators were the total number of patients randomized and missing values were considered to be treatment failures.
|Figure 1. Methodological quality graph: review authors' judgements about each methodological quality item presented as percentages across all included studies.|
|Figure 2. Methodological quality summary: review authors' judgements about each methodological quality item for each included study.|
Description of studies
In all twenty one studies were identified, by the search strategy, and six were subsequently excluded because patients were stool positive for C. difficile, but did not have diarrhea (Johnson 1992), duplicate publication (Noren 2006), a non-antibiotic arm (Louie 2006; Lowy 2010; Mattila 2008), or because the study was not a randomized controlled trial (McFarland 2002). McFarland 2002 is an uncontrolled observation of treatment failures from randomized, probiotic studies. The search strategy did not identify any studies published solely as abstracts. In the original review there were nine included studies. The first update included three additional studies (Lagrotteria 2006; Musher 2006; Wullt 2004). The current update includes three additional studies (Louie 2009; Musher 2009; Zar 2007).
Of particular interest is one of the excluded studies (Johnson 1992). Johnson 1992 is a randomized, placebo-controlled, non-blinded trial of oral vancomycin 125 mg four times a day for 10 days versus oral metronidazole 500 mg two times a day for 10 days versus placebo for the treatment of asymptomatic carriers of C. difficile. Asymptomatic C. difficile excretors were identified by rectal swab during infection control surveillance of medical and surgical inpatients at a US Veterans Affairs hospital. Patients were included if they did not have diarrhea or abdominal symptoms and were not allergic to vancomycin or metronidazole. C. difficile infection was identified by stool culture and cytotoxin assay. The main outcome was presence or absence of C. difficile in the stool during treatment and two months after treatment. A total of 30 patients were equally divided into the three treatment arms. C. difficile organisms were not detected during and immediately after treatment in 9/10 patients treated with vancomycin compared with 3/10 patients treated with metronidazole and 2/10 patients in the placebo group. In contrast, by the end of follow-up-ranging from 40 to over 90 days C. difficile organisms were not detected in 3/10 patients treated with vancomycin compared with 7/10 patients treated with metronidazole and 8/10 patients in the placebo group. The authors concluded that asymptomatic fecal excretion of C. difficile was transient in most patients and treatment with metronidazole was not effective in eliminating C. difficile from stool. Although treatment with vancomycin was temporarily effective, it was associated with a significantly higher rate of C. difficile carriage two months after treatment and therefore was not recommended for asymptomatic carriers of C. difficile. The results of this small study should be interpreted with caution. Athough these results are not statistically significant, they are of concern and should stimulate further investigation of strategies to prevent the long term carrier state of C. difficile, especially in hospitalized patients.
The fifteen included studies involve patients with diarrhea who have recently received antibiotics for an infection other than C. difficile. The definition of diarrhea ranged from at least two loose stools per day with an associated symptom such as rectal temperature > 38 ºC (Anonymous 1994), to at least six loose stools in 36 hours (Teasley 1983; Louie 2009). All of the studies included a "loose" or "liquid" description of the stool in their definitions of diarrhea. In terms of exclusion criteria, patients with HIV were only explicitly excluded from one study (Anonymous 1994). Pregnant women were excluded from two studies (Anonymous 1994 and Zar 2007). The other studies had varying and vague exclusion criteria ranging from "other obvious cause of diarrhea" (Teasley 1983) to "anatomic abnormality" (Young 1985). Both Louie 2009 and Musher 2009 excluded patients deemed clinically unstable and ''not expected to survive the study period''. Zar 2007 excluded patients with suspected or proven life-threatening intra-abdominal complications.
The studies used different methods for detecting C. difficile infection. Teasley 1983 defined C. difficile disease as stool being positive for C. difficile by culture or cytotoxin assay, or by the presence of pseudomembranes on endoscopy. Stool was tested for other enteric pathogens, but the authors did not comment on the presence or absence of these pathogens. Young 1985 and Zar 2007 defined C. difficile disease in the same manner as Teasley 1983. Young 1985 tested for the presence of other enteric pathogens and excluded patients if their stool tested positive for any other pathogen. Dudley 1986 and Fekety 1989 and Louie 2009 defined C. difficile disease by both culture or cytotoxin assay. Dudley 1986 and Fekety 1989 did not test for the presence of other enteric pathogens. de Lalla 1992 defined C. difficile disease as stool being positive for cytotoxin, culture and evidence of colitis on endoscopy. de Lalla 1992 excluded patients if their stool tested positive for any other pathogen. Wenisch 1996 defined C. difficile disease as stool positive for cytotoxin or endoscopic evidence of colitis and fecal leukocytes. Wenisch 1996 did not perform stool culture on any of the study patients. The Swedish CDAD study defined C. difficile disease as being positive for cytotoxin and/or culture and excluded patients if their stool tested positive for any other pathogen (Anonymous 1994). Keighley 1978 defined C. difficile disease as being positive for culture or positive for culture and cytotoxin. Keighley 1978 did not test stool for other enteric pathogens. Boero 1990 tested stool for cytotoxin only. Four studies used toxin assay and symptomatic diarrhea to define C. difficile disease (Lagrotteria 2006; Musher 2006; Musher 2009; Wullt 2004).
Ten of the fifteen included studies compared different antibiotics for the treatment of C. difficile-associated diarrhea. Teasley 1983 compared oral vancomycin 500 mg four times a day for 10 days with oral metronidazole 250 mg four times a day for 10 days. Young 1985 compared oral vancomycin 125 mg four times a day for 7 days with oral bacitracin 20,000 units four times a day for 7 days. Dudley 1986 compared oral vancomycin 500 mg four times a day for 10 days with oral bacitracin 25,000 units four times a day for 10 days. de Lalla 1992 compared oral vancomycin 500 mg four times a day for 10 days with oral teicoplanin 100 mg two times a day for 10 days. Wenisch 1996 compared four antibiotics: oral vancomycin 500 mg three times a day for 10 days with oral metronidazole 500 mg three times a day for 10 days with oral fusidic acid 500 mg three times a day for 10 days with oral teicoplanin 400 mg two times a day for 10 days. Boero 1990 compared vancomycin 500 mg two times a day for 10 days with oral rifaximin 200 mg three times a day for 10 days. Wullt 2004 compared 250 mg of fusidic acid three times per day to 400 mg of metronidazole three times per day for 7 days each. Musher 2006 compared 250 mg of metronidazole four times per day for 10 days to 500 mg of nitazoxanide twice a day given for either 7 or 10 days. Musher 2009 compared 125mg of vancomycin four times per day for 10 days to 500 mg of nitazoxanide twice a day for 10 days. Zar 2007 compared 125 mg of vancomycin four times per day with 250 mg of metronidazole four times per day for 10 days.
The Swedish CDAD study was a dose finding study which compared oral teicoplanin 100 mg twice a day for 7 days with oral teicoplanin 50 mg twice a day for 3 days followed by 100 mg twice a day for 4 days (Anonymous 1994). Fekety 1989 performed a dosage-finding study for oral vancomycin by comparing 125 mg four times a day with 500 mg four times a day for an average of 10 days. Louie 2009 compared fidaxomicin (OPT-80) given at either 50 mg, 100 mg or 200 mg doses twice daily for ten days. Only one study was placebo controlled. Keighley 1978 compared oral vancomycin 125 mg four times a day for 5 days with placebo four times a day for 5 days.
In addition to treating CDAD with antibiotics, the above studies also dealt with the initial, offending antibiotic in different ways. Only three studies explicitly stated that all initial antibiotics were stopped (Wenisch 1996; Young 1985; Zar 2007). Three of the other studies made attempts to stop the initial antibiotics, but this antibiotic was continued if it was deemed essential to the patient's clinical treatment (Anonymous 1994; Dudley 1986; Teasley 1983). Louie 2009 excluded any patients requiring concurrent antibiotic therapy, but made no reference to the initial antibiotics used in the included participants. The remaining eight studies either stated that the initial antibiotic was either stopped or changed (Fekety 1989) or did not make any reference to the initial antibiotic at all (Boero 1990; de Lalla 1992; Keighley 1978; Lagrotteria 2006; Musher 2006; Musher 2009; Wullt 2004).
Zar 2007 defined cure as a combination of resolution of diarrhea and conversion of stool to toxin negative, whereas other authors reported these outcomes separately. Zar 2007 is the first study to stratify patients (after randomization) by disease severity when assessing the relative efficacy of metronidazole and vancomycin. This is of importance because this general concept of efficacy assessment has been adopted in one subsequent study (Musher 2009) and three major guideline publications including the UK Health Protection Agency (HPA 2008), the European Society of Clinical Microbiology and Infectious Diseases (ESCMID) (Bauer 2009) and the combined Infectious Diseases Society of America (IDSA), and the Society for Healthcare Epidemiology of America (SHEA), which is reported on the Centers for Disease Control (CDC) web site (Cohen 2010).
All included studies assessed only orally delivered antibiotics, i.e., patients unable to tolerate oral medications were excluded from each study. None of the included studies stratified the randomization of patients based on cessation or continuation of the antibiotic that caused the CDAD.
Risk of bias in included studies
The results of the risk of bias analysis are presented in Figure 1 and Figure 2. Quality concerns include unclear allocation concealment, unclear generation of the allocation sequence and blinding of researchers and participants. Only six of the 15 studies reported adequate allocation concealment (Dudley 1986; Keighley 1978; Lagrotteria 2006; Musher 2009; Wullt 2004; Zar 2007). Only six studies (Dudley 1986; Fekety 1989; Lagrotteria 2006; Wenisch 1996; Wullt 2004; Zar 2007) adequately explained the generation of the allocation sequence. Only five studies (Musher 2006; Musher 2009; Wenisch 1996; Wullt 2004; Zar 2007) clearly stated that the researchers and the subjects were both blinded. Two studies stated that patients were blinded and it was unclear if outcome assessors were blinded (Dudley 1986; Young 1985). Outcome assessors were not blinded in the Fekety 1989 study. Patients were not blinded in the Lagrotteria 2006 study. Louie 2009 was a dose finding study and was not blinded. The Wenisch 1996, de Lalla 1992 and Boero 1990 studies were not blinded. Attempts were made to contact the authors to clarify these issues and as of the submission of this review correspondence was not established.
The included studies have additional quality issues. One diagnostic concern is that 11 of the 15 included studies used C. difficile cytotoxin stool assay as the means of identifying infected patients (Anonymous 1994; Boero 1990; Dudley 1986; Musher 2009; Teasley 1983; Wenisch 1996; Young 1985; Zar 2007). Two other studies (de Lalla 1992; Fekety 1989) required stool positive for cytotoxin and growth of C. difficile on culture to establish the presence of CDAD. Keighley 1978 separated out those patients who had stool positive for cytotoxin from those with C. difficile growth on culture. However the most important quality issue encountered in this review is that no study in which there were dropouts included those dropouts in their analyses, so effectively all participants were analyzed as treated. In some studies this had a major impact, since randomization occurred before disease confirmation, and participants were eliminated after randomization when found not to have CDAD. In some cases this was as infrequent as 2% of those randomized (Musher 2009), however in several studies the dropout rate was roughly 50% due to this reason (Anonymous 1994; Dudley 1986; Keighley 1978). For all studies in this review dropouts were treated as treatment failures.
The best means of identifying C. difficile is somewhat controversial (Thomas 2003). Gerding 1995 in their position paper for the SHEA recommend either a combination of stool culture for C. difficile and cytotoxin assay or stool cytotoxin testing alone. The paper goes on to state that most strains of C. difficile produce either both toxin A and toxin B, or neither (Gerding 1995). Additionally, treatment of patients with diarrhea who are toxin negative, but culture positive is controversial in that non toxigenic strains of C. difficile are not considered pathogenic (Gerding 1995). The American College of Gastroenterology practice parameters on the diagnosis and treatment of CDAD state that C. difficile can be identified by either positive culture or toxin assay and that if both tests are negative and the diarrhea persists, then additional stool studies with the same or different tests should be ordered (Fekety 1997). In light of these two practice guidelines, papers identifying C. difficile by either culture or toxin assay were included in our analysis. The most thorough discussion of the diagnosis of CDAD is in the most recent SHEA IDSA publication (Cohen 2010).
For the initial diagnosis of CDAD, most included studies did not exclude the presence of other pathogens in the stool as the cause of diarrhea (Boero 1990; Dudley 1986; Fekety 1989; Keighley 1978; Lagrotteria 2006; Musher 2006; Musher 2009; Teasley 1983; Wenisch 1996; Wullt 2004). It is possible that a patient could be infected with both C. difficile and another pathogen. The studies do not consistently state whether the initial, offending antibiotic was stopped and if it was stopped, whether a different antibiotic was started in its place. Although Zar 2007 did screen stool samples for Shigella, Salmonella and Campylobacter, in this study all patients were negative for these microbes. Additionally, only Young 1985 controlled for patients whose diarrhea rapidly resolved on its own with the removal of the initial antibiotic by including only those patients with persistent symptoms. For these patients the diarrhea may not have been caused by C. difficile, although it was present in the stool, or the bacterium was successfully purged by the patient. Recent studies relied on positive toxin assay and diarrhea alone to make the diagnosis. Some of the older studies required endoscopic confirmation of pseudomembranes, although this was not uniformly done even within those studies. Zar 2007 did include pre-existing pseudomembrane checks, although no patients had pre-existing pseudomembranous colitis in this study.
Each specific trial also had its own unique quality issues as well. In the Teasley 1983 study which compared vancomycin to metronidazole, the offending antibiotics believed to have brought about the CDAD were said to be discontinued unless they were "essential for treatment." The authors do not indicate how many patients did not have their offending antibiotic discontinued, which may confound the interpretation of results as the original antibiotic may have caused diarrhea by a different mechanism. Dudley 1986 and Zar 2007 did not have test results for C. difficile in the stool until after randomization. As a result, all of the patients with diarrhea that were C. difficile negative (50% of the total patients that were initially randomized in (Dudley 1986)) were dropped from the study. Wenisch 1996 had a total of seven dropouts from all of the treatment arms combined. However, Wenisch 1996 did not indicate the treatment allocation of these patients, therefore precluding an accurate intention to treat analysis of the results. In the Swedish CDAD study there was an intermediate measure of cure called "improvement," but the criteria for this "improved" classification of outcome were never described (Anonymous 1994). For the purposes of this systematic review, a conservative approach was taken and the "improved" patients were not considered to have their clinical symptoms resolved. Fekety 1989 had a dropout rate of 18% and the duration of the antibiotic therapy was determined by each individual physician treating patients. The average duration of therapy was 10 days with a range of five to 15 days (Fekety 1989). Boero 1990 had an unclear definition of inclusion criteria and symptomatic cure. Patients were described as having diarrhea, but there were no quantitative measures for the number of stools per day and cure was simply described as "normalization" of stools (Boero 1990). Dropouts were 13% and 23% respectively in the Wullt 2004 and Musher 2006 studies, making an intention to treat analysis difficult. However, both studies reported dropouts by allocation group. There were no dropouts in the smaller (Lagrotteria 2006) study. Zar 2007 had a dropout rate of 12.8% (22/172), however, the study still achieved predetermined power. Zar also excluded "suspected or proven life-threatening intra-abdominal complications" as well as prescription of either vancomycin or metronidazole in the 14 days prior to enrolment. Zar 2007 retested stool for recurrence at 21 days after cure (defined as symptom-free and toxin-free stool at day 10). This protocol was not based on evidence, and it should be noted that current guidelines for management of asymptomatic carriers of C. difficile are conservative. Although Zar screened for prior antibiotic use, no data were reported on which antibiotics were used. Although Musher 2009 was a multi-centre randomized trial, 37 of 50 patients came from the principal investigator's hospital. The study excluded patients undergoing enteral tube feeding. Another exclusion criterion was use of "any drug with anti-C. difficile activity - with the exception of ≤3 oral doses of metronidazole or vancomycin - within 1 week" of entry. It should be noted that Musher 2009 had a dropout rate of 18.4% (9/49) and was terminated early due to "slower-than-expected recruitment". It also fell far short of its predetermined sample size of 350.
Louie 2009 was a phase 2a trial and was an open label study for mild to moderate C. difficile infection and excluded all patients with severe disease and those who were "not expected to survive the study period". While patients kept a symptom diary, they were only formally observed at study entry and at the end of treatment or withdrawal.
Of particular interest among the included studies is the sole placebo-controlled study by Keighley 1978. This randomized, controlled trial was performed in England in 1978. It assessed the efficacy of oral vancomycin 125 mg taken four times a day for five days (a dose lower and shorter than other studies) compared to placebo for the treatment of pseudomembranous colitis. All patients were recruited from a surgical ward and blinded during the five day treatment period. It is unclear from the study report whether the outcome assessor was blinded or not. Broad inclusion criteria were used, and any patient with postoperative diarrhea was recruited. This led to the randomization of patients with postoperative diarrhea not related to C. difficile. The only exclusion criterion was a previous history of pseudomembranous colitis. The study not only divided patients by their treatment regimen, but also divided patients into three groups by results of their stool analysis. Group one had stool that was culture positive for C. difficile and positive for C. difficile cytotoxin. Group two had stool that was culture positive for C. difficile only. Group three had stool that was culture negative and negative for cytotoxin.
Keighley 1978 reported that only 21 of 44 randomized patients had some evidence of C. difficile infection. Of these patients 16 were toxin positive and 5 were culture positive. Twelve of these patients were in the vancomycin group and nine were in the placebo group. Keighley 1978 did not specifically define cure or relapse, but defined clinical response measured by stool frequency as resolution or improvement of diarrhea. Clinical response was higher in patients who received vancomycin in groups one and two, but response was not affected by treatment in group three. Keighley 1978 describes the numbers of patients taking vancomycin or placebo in whom stool converted to negative for C. difficile strain, negative for fecal cytotoxin and negative for histological evidence of pseudomembranous colitis. There are several major quality concerns with the Keighley 1978 study. The number of patients was small, and follow-up was inadequate. During the convalescence period, there were specimens that were either not taken or mislaid. Two-thirds of the patients in the vancomycin group fell into this category. Follow-up conclusions were based on only four patients from the total of 12 patients with evidence of C difficile infection that received vancomycin (Keighley 1978).
As mentioned above, the study by (Zar 2007) has had a great influence because of its stratification in outcome assessment by disease severity. Because of this, potential quality issues bear special attention. Despite the publication date, recruitment for this study ended in 2002, before the 027 strain of C. difficile reached the Chicago area, and the severity of disease seen in this mutation had not yet been witnessed, so the severity criteria would by that standard be fairly mild. Yet, in spite of excluding the sickest patients with CDAD from the study, 8 participants died within 5 days of the onset of therapy and were excluded from further analysis rather than treated as treatment failures. Randomization also occurred before the diagnosis of C. difficile was established, and 14 randomized subjects were subsequently found to be C. difficile negative and excluded from analysis. Randomization prior to stratification may not sufficiently validate this concept of severity stratification in antibiotic choice as described by (Zar 2007). What was done is in fact a post hoc subgroup analysis by disease severity, as was also done in (Musher 2009).
Effects of interventions
See: Summary of findings for the main comparison Metronidazole versus Vancomycin for Clostridium difficile-associated diarrhea in adults; Summary of findings 2 Teicoplanin versus Vancomycin for Clostridium difficile-associated diarrhea in adults; Summary of findings 3 Metronidazole versus Teicoplanin for Clostridium difficile-associated diarrhea in adults
In terms of symptomatic cure no statistically significant differences were found between metronidazole, bacitracin, rifaximin, nitazoxanide, fusidic acid, teicoplanin and vancomycin.
Metronidazole versus vancomycin
Symptomatic cure was achieved in 79% of patients who received vancomycin compared to 71% of patients who received metronidazole (3 studies; 335 patients; RR 0.91; 95% CI 0.81 to 1.03; P = 0.14).
Bacitracin versus vancomycin
Symptomatic cure was achieved in 46% of patients who received vancomycin compared to 27% of patients who received bacitracin (2 studies; 104 patients; RR 0.58; 95% CI 0.34 to 0.99; P = 0.05).
Rifaximin versus vancomycin
Symptomatic cure was achieved in 100% of patients who received vancomycin compared to 90% of patients who received rifaximin (1 study; 20 patients; RR 0.90; 95% CI 0.69 to 1.18; P = 0.46).
Nitazoxanide versus vancomycin
Symptomatic cure was achieved in 74% of patients who received vancomycin compared to 77% of patients who received nitazoxanide (1 study; 49 patients; RR 1.04; 95% CI 0.76 to 1.43; P = 0.79).
Fusidic acid versus vancomycin
Symptomatic cure was achieved in 77% of patients who received vancomycin compared to 66% of patients who received fusidic acid (1 study; 60 patients; RR 0.85; 95% CI 0.61 to 1.17; P = 0.31).
Teicoplanin versus vancomycin
Symptomatic cure was achieved in 73% of patients who received vancomycin compared to 87% of patients who received teicoplanin (2 studies; 110 patients; RR 1.21; 95% CI 1.00 to 1.46; P = 0.06).
In terms of initial symptomatic resolution, vancomycin was significantly more effective than placebo. Initial symptomatic resolution was achieved in 41% of vancomycin patients compared to 4% of placebo patients (1 study; 44 patients; RR 9.00; 95% CI 1.24 to 65.16). This result should be interpreted with caution because of the small number of patients for this comparison and the poor methodological quality of the trial. OPT-80 was significantly more effective when given as 200 mg twice daily (94% clinical cure rate) compared to lower doses of 100 mg or 50 mg twice daily (1 study; 48 patients; RR 0.80; 95% CI 0.65 to 0.98; P = 0.03). This result should be interpreted with caution due to the small number of patients and because there is no evidence of its efficacy compared to placebo or other antibiotics.
For bacteriologic cure, teicoplanin was more effective than vancomycin. Bacteriologic cure was achieved in 82% of patients who received teicoplanin compared to 45% of patients who received vancomycin (1 study, 59 patients; RR 1.82; 95% CI 1.19 to 2.78; P = 0.006). There was no statistically significant difference in bacteriologic cure between metronidazole and vancomycin. Bacteriologic cure was achieved in 45% of patients who received metronidazole compared to 53% of patients who received vancomycin (2 studies, 163 patients; RR 0.85; 95% CI 0.62 to 1.17; P = 0.33). There was no statistically significant difference in bacteriologic cure between vancomycin and fusidic acid. Bacteriologic cure was achieved in 69% of patients who received vancomycin compared to 48% of patients who received fusidic acid (1 study; 60 patients; RR 0.68 95% CI 0.44 to 1.06). For initial bacteriologic resolution, vancomycin was significantly more effective than placebo. Initial bacteriologic resolution was achieved in 45% of vancomycin patients compared to 4% of placebo patients (1 study; 44 patients; RR; 10.00 95% CI 1.40 to 71.62). Teicoplanin was significantly more effective than vancomycin for initial bacteriologic resolution. Initial bacteriologic resolution was achieved in 62% of vancomycin patients compared to 87% of teicoplanin patients (2 studies; 110 patients; RR 1.43; 95% CI 1.14 to 1.81; P = 0.002). No statistically significant difference in initial bacteriologic resolution was found between vancomycin and metronidazole (1 study; 62 patients; RR 0.96; 95% CI 0.70 to 1.30; P = 0.78). No statistically significant difference in initial bacteriologic resolution was found between vancomycin and fusidic acid (2 studies; 191 patients; RR 0.96; 95% CI 0.78 to 1.18; P = 0.68). No statistically significant difference in initial bacteriologic resolution was found between vancomycin and rifaximin (1 study; 20 patients; RR 0.90; 95% CI 0.69 to 1.18; P = 0.46). Vancomycin was significantly more effective than bacitracin for initial bacteriologic resolution (2 studies; 104 patients; RR 0.52; 95% CI 0.31 to 0.86; P = 0.01).
Risk of relapse
No statistically significant differences in relapse were found between bacitracin and vancomycin (2 studies; 104 patients; RR 1.11; 95% CI 0.50 to 2.48; P = 0.80), teicoplanin and vancomycin (2 studies; 110 patients; RR 0.44; 95% CI 0.14 to 1.36; P =0.15), fusidic acid and vancomycin (2 studies; 191 patients; RR 1.09; 95% CI 0.64 to 1.86; P = 0.76), and nitazoxanide and vancomycin (1 study; 49 patients; RR 0.61; 95% CI 0.06 to 6.33). No statistically significant difference in relapse was found between metronidazole and nitazoxanide (1 study; 142 patients; 95% CI 0.69; 95% CI 0.24 to 1.98; P = 0.49). No statistically significant difference in relapse was found between metronidazole and metronidazole plus rifampin (1 study; 39 patients; RR 0.95; 95% CI 0.33 to 2.77; P = 0.93). No statistically significant difference in relapse was found between metronidazole and teicoplanin (1 study; 59 patients; RR 2.26; 95% CI 0.48 to 10.73; P = 0.31). No statistically significant difference in relapse was found between metronidazole and fusidic acid (2 studies; 190 patients; RR 0.91; 95% CI 0.53 to 1.55; P = 0.73). No statistically significant difference in relapse was found between teicoplanin and fusidic acid (1 study; 57 patients; RR 0.26; 95% CI 0.06 to 1.11; P = 0.07).
The patients with relapse of diarrhea after treatment were re-treated in various ways. Dudley 1986 re-treated with the same antibiotic and Young 1985 re-treated with the antibiotic from the other arm of the study. de Lalla 1992 stated that the patients with relapse were re-treated, but analyzed the results in aggregate, making it impossible to compare the antibiotics for efficacy. Fekety 1989's vancomycin dose-finding study re-treated all nine relapses with vancomycin and subsequently cured these patients. Wenisch 1996 treated recurrent disease with teicoplanin with a reported a 100% cure rate. Louie 2009 re-treated with either metronidazole or vancomycin, but did not report on the success of this treatment. The remaining studies did not report on the re-treatment of relapses.
The secondary outcomes measured in this review (surgery and death) occurred infrequently in all of the studies. Dudley 1986 reported one death and one partial colectomy after failed treatment of CDAD with bacitracin. Both patients were crossed-over to vancomycin before their adverse outcome, with the first patient improving prior to his death, and the second patient continuing to have CDAD. Fekety 1989 reported one death in each of the two vancomycin treatment arms. The Swedish CDAD study reported one death in the two times per day teicoplanin group, but the authors did not specify if it was part of the CDAD sub-section of the study or part of the general dosage-finding portion of the study (Anonymous 1994). Teasley 1983 had three deaths in the whole study, including two deaths in the vancomycin treatment arm. The authors attributed all three deaths to underlying illness and not to CDAD or the antibiotic therapy, a common claim in other studies as well, though this would seem to be a difficult judgement. de Lalla 1992 reported two deaths and Musher 2009 reported one death. Both studies did not specify the treatment group and attributed the deaths to the patients' underlying disease. Zar 2007 reported eight deaths, four in each treatment group, but did not comment on the cause of mortality. There was a total of 18 deaths among 1152 participants in this systematic review.
Two studies quantified the cost of treatment (Musher 2009; Wenisch 1996). Wenisch 1996 compared four different antibiotics. The cost data for a course of treatment were: Metronidazole, $35; Fusidic acid, $287; Vancomycin, $2030; and Teicoplanin, $3430. Musher 2009 compared two antibiotics. The cost data are as follows: Vancomycin, $618; Nitazoxanide, $316.
The studies presented in this review are disparate in design and small in number and content. Based upon the paucity of evidence, and significant quality issues, especially in the placebo controlled trial, it remains an open question whether mild C. difficile-associated diarrhea needs to be treated with antibiotics versus simple supportive care. This represents a significant departure from current practice, so much so that several of the investigators we contacted were unwilling to undertake a placebo controlled trial. Perhaps patients with mild CDAD might resolve their symptoms as quickly without antibiotics, and with a lower recurrence rate. The Keighley 1978 study, which is the only placebo-controlled study, suggests vancomycin's superior efficacy over placebo. However, this result should be interpreted with caution, given the small numbers of patients in the trial and the poor methodological quality of the trial. The Johnson 1992 study of asymptomatic carriers suggests that placebo is better than vancomycin or metronidazole for eliminating C. difficile in stool during follow-up. A placebo controlled trial will be even more difficult to perform in light of the appearance of the highly virulent ribotype 027 toxigenic strain of C. difficile. The very high mortality rate and increasing use of surgical therapy for this disease emphasize the still inadequate development of an ideal antibiotic therapy.
If one does decide to treat, it should be with two goals: improvement of the patient's clinical condition and prevention of spread of C. difficile infection to other patients. Given these two considerations, one should choose the antibiotic that brings both symptomatic cure and bacteriologic cure. In this regard, teicoplanin may be a good choice, because it was significantly better than vancomycin for bacteriologic cure and has borderline superior effectiveness in terms of symptomatic cure. Teicoplanin is not readily available in the United States and remains expensive elsewhere, which must be taken into account (Hopkins 2009).
Cost effectiveness cannot be calculated when efficacy is still not established. Although the costs presented in this review are informative, they do not play a definitive role in antibiotic choice until efficacy is confirmed. A 10-day course of vancomycin 500 mg four times a day costs $7,358 (US$) (Mosby 2003). This regimen was used in the de Lalla 1992; Fekety 1989 and Teasley 1983 studies. If given at the lower dose of 125 mg four times a day, a ten day course of vancomycin costs $618 (Musher 2009). This regimen was used in Musher 2009 and Zar 2007. A 10 day course of metronidazole 500 mg three times a day as used in the Wenisch 1996 study costs $765 (US$) (Mosby 2003). A 10 day course of bacitracin 25,000 units four times a day as used in the Wenisch 1996 study costs $200 (US$). Teicoplanin and fusidic acid are not available in the United States; therefore, using the price listing from the British National Formulary and the exchange rate as of 1/29/04, we calculated the price of these drugs in US dollars. A 10 day course of Teicoplanin 400 mg two times a day costs $1,374 ($US) and a 10 day course of fusidic acid 500 mg three times a day costs $706 ($US) (BNF 2004). A 10 day course of Nitazoxanide 500 mg twice a day costs $316 (Musher 2009). Rifaximin has recently become available in the United States and has been available in Italy since 1988. It must be kept in mind that true costs are difficult to calculate. The numbers above are charges made for each antibiotic, as each hospital or health system negotiates different prices for medication with pharmaceutical companies.
Zar 2007 reported that vancomycin was significantly better than metronidazole for treating severe C. difficile diarrhea (P = 0.02). However 22 participants were excluded from the analysis (12.8%) and an intention to treat analysis was not reported. Bishara 2007 found no statistically significant difference in symptomatic cure between metronidazole and vancomycin in a pooled analysis of 3 studies (RR 0.93; 95% CI 0.80 to 1.09) treating exclusions, deaths and relapses as treatment failures. This result is similar to the pooled analysis reported in this review.
The severity scale designed by Zar 2007 and used in Musher 2009 is important as, for the first time, it attempts to stratify those affected by CDAD into mild versus severe disease. Such a scale would be extremely valuable, not only for the standardisation of future studies, but for use in clinical practice, particularly as the study by Johnson 1992, for example, suggests that mild disease may be successfully treated without antibiotics. Currently, the severity scale designed by Zar 2007 uses any two of the following to indicate severe disease: pyrexia (≥38.3°C), serum albumin of < 2.5 mg/dl, leukocytosis of 15,000 cells/mm
These recommendations should be treated with caution because there is poor evidence supporting recommendations for the treatment of severe disease due to the exclusion of such patients in the included studies. Patients in all included studies were treated with oral antibiotics alone, those too ill to tolerate oral medications were excluded. In addition the criteria for defining the most severely ill patients in the Zar 2007 study were not at that time clearly associated with prognosis. Most importantly the stratification occurred after randomization and after a number of exclusions, constituting a post hoc subgroup analysis by disease severity. The results of such analyses have been described as best used for exploratory purposes and should not affect trial conclusions (Assmann 2000). No statistical tests of interaction were performed. Yet one trial (Louie 2009) and three recent published guidelines (HPA 2010; Bauer 2009; Cohen 2010) have adopted this principal of stratification and the guidelines endorse the efficacy of vancomycin for severe disease.
Older guidelines discuss treatment of unresponsive disease with an increased dose of vancomycin 500 mg four times daily, intravenous metronidazole 500 mg three times daily and rifampicin 300 mg twice daily. However, this regimen was shown to have no clinical benefit in primary CDAD (Lagrotteria 2006). The criteria defining severity stratification have differed in each publication that is cited in this review that assesses or recommends treatment by disease severity (Musher 2009; Zar 2007; Bauer 2009; HPA 2010; Cohen 2010). Lactate > 5 is certainly a valid marker of extreme severity, but with an associated 75% operative mortality, an earlier marker and proven therapy would be of more use (Cohen 2010). This badly needs clarification, and the evidence does not exist in available studies to clarify the criteria for severity stratification and whether different antibiotic regimens should be applied to each severity stratum.
The placebo-controlled study by Keighley 1978, although flawed, raises some interesting points. The study suggests that vancomycin is not a very good therapy for non-specific diarrhea or antibiotic-associated diarrhea that arises from a mechanism other than C. difficile overgrowth. Whilst vancomycin did seem to be effective among C. difficile positive individuals for the early control of diarrhea, it was not effective for long term bacteriologic control. The Johnson 1992 study provides complementary information, being the only other placebo controlled trial (see description of excluded studies - above). Johnson 1992 suggests that the best means of accomplishing the second goal of treatment - elimination of the C. difficile bacterium from the stool of hospitalized patients to prevent spread of infection, is best accomplished with no antibiotic at all. While vancomycin has high in vitro activity against C. difficile, it is still not certain that CDAD is best treated with this or any other antibiotic (Bartlett 1980). Antibiotic synergy has only been investigated in one study (Lagrotteria 2006), and no clinical benefit was seen by adding rifampin to metronidazole. All of the included studies use oral antibiotics and, as yet, there has been no research into the use of intravenous antibiotics as an alternative.
Wenisch 1996 is an important study because it directly compares several antibiotics. This avoids making unrealistic assumptions about relative efficacy made across several randomized trials (Baker 2002). The increasing virulence of C. difficile and failure to demonstrate that any single antibiotic is superior to others, with the possible exception of teicoplanin underlines the need to consider new treatments. Data on two new antibiotics: nitazoxanide (Musher 2006; Musher 2009) and OPT-80 (Louie 2009) are included in this update. The concept of toxin binding has been investigated and deserves further attention (Louie 2006). A C. difficile vaccine has also been developed and may provide another alternative to antibiotic treatment (Sougioultzis 2005).
The most recent data show a drop in CDAD mortality in 2008, after a rapid rise in mortality up to 2004 (Karas 2010; Redelings 2007). This may be due to a greater than 50% drop in incidence in the UK that is the result of effective prevention and isolation measures (HPA 2010), and improving awareness of causative antibiotics (Privitera 1991; Thomas 2003; Novell 2010). Novel therapies, some long forgotten, such as fecal repopulation (Khoruts 2010; Bowden 1981) are reappearing.
The three studies included in this update do not alter the overall conclusion of this review, but provide a clearer picture of the further research that should be carried out in the future.
Implications for practice
Although antibiotic therapy for CDAD is universally prescribed, this review cannot establish the efficacy of antibiotic therapy for CDAD, as the only placebo-controlled study is inadequate. This review cannot define the efficacy of cessation of the initial, offending antibiotic because the above studies did not apply a strict or consistent protocol in regard to its continuation. As a corollary, this review cannot establish the proper amount of time between cessation of the initial antibiotic and beginning antibiotic therapy for CDAD (to see if the offending antibiotic might be causing diarrhea by a mechanism other than C. difficile overgrowth), as the above studies did not address this research question. Finally, this review cannot definitively make a specific antibiotic recommendation for the treatment of CDAD, as the results suggest that several antibiotics are more or less equally effective at achieving early symptomatic cure as well as other secondary outcomes. Current guidelines generally recommend metronidazole or vancomycin depending on the clinical situation, but there is no strong evidence base to support these recommendations. Teicoplanin may be a more effective alternative to vancomycin. Although its use is currently limited by cost and availability, it may be a possibility in the future.
Implications for research
Given the uncertainty involved in antibiotic therapy for CDAD, adequately powered, placebo controlled, randomized trials that assess (1) cessation of the initial antibiotic, (2) timing and duration of CDAD therapy and (3) efficacy of several antibiotic treatments is recommended. Such trials would require the randomization of a large number of patients. Vancomycin should be included, as it is the only antibiotic that has previously been compared to placebo. Metronidazole should be included because it is an inexpensive and generally well tolerated alternative to vancomycin. Teicoplanin should be included because it may be superior to both of these antibiotics. Nitazozanide should also be included, as it may be a cheaper possible alternative to vancomycin. These comparisons may help to determine which antibiotic is most suitable for treating CDAD. There should be strict criteria for the diagnosis of CDAD: prior exposure to antibiotics, persistent diarrhea that does not respond to removal of the offending antibiotic and supportive therapy, and C. difficile that is identified by toxin assay. The primary outcomes should be symptomatic cure, defined as cessation of symptoms during the time of treatment and for 30 or 60 days after, and evidence of bacteriologic cure. The secondary outcomes should be initial symptom resolution, initial bacteriologic resolution, surgery and death. Given the broad use of antibiotics, hopefully this proposed study will elucidate the treatment of choice for CDAD and improve the lives of many of those suffering from a major cause of antibiotic-associated diarrhea. If disease severity is to be a factor in antibiotic choice, this should be assessed prior to randomization. However, in light of recent developments in C. difficile virulence and prevalence it is uncertain if a placebo controlled trial will be undertaken.
It would be useful to develop a universal system for grading the severity of CDAD. The increasing severity of the disease also highlights the need to investigate antibiotic treatment in severely ill patients with CDAD, as data on this group are currently lacking. Other issues that need to be addressed in future trials include: the treatment of recurrent symptoms, the use of intravenous compared to oral antibiotics, antibiotic resistance, synergistic antibiotic combinations and other types of interventions (e.g. the use tolevamer combined with antibiotics).
Funding for the IBD/FBD Review Group (September 1, 2010 - August 31, 2015) has been provided by the Canadian Institutes of Health Research (CIHR) Knowledge Translation Branch (CON - 105529) and the CIHR Institutes of Nutrition, Metabolism and Diabetes (INMD); and Infection and Immunity (III).
Miss Ila Stewart has provided support for the IBD/FBD Review Group through the Olive Stewart Fund.
Data and analyses
- Top of page
- Summary of findings [Explanations]
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Last assessed as up-to-date: 12 October 2010.
Contributions of authors
Dr Rick Nelson prepared the update for Issue 3, 2007.
E Bricker, R Garg, R Nelson, J Hansen, A Loza, and T Novak all contributed to the original version of the review published in Issue 1, 2005 of the Cochrane Library.
This update was prepared by R.Nelson with the collaboration of the other named authors, students at the Unverisity of Sheffield School of Medicine as part of the Master Class Symposium
Declarations of interest
Sources of support
- none, UK.
- none, UK.
Differences between protocol and review
Medical Subject Headings (MeSH)
*Clostridium difficile; Anti-Bacterial Agents [adverse effects; *therapeutic use]; Diarrhea [chemically induced; *drug therapy; microbiology]; Enterocolitis, Pseudomembranous [complications; *drug therapy]; Randomized Controlled Trials as Topic; Treatment Outcome
MeSH check words