Clinical Gastroenterology and Hepatology 2014;-:-–-
Risk of Clostridium difficile Infection in Hospitalized Patients Receiving Metronidazole for a Non–C difficile Infection Sandra Rodriguez,* Marlow B. Hernandez,* Giorgio Tarchini,‡ Megan Zaleski,* Marjon Vantanchi,* Lyssette Cardona,‡ Fernando Castro–Pavia,§ and Alison Schneider§ *Department of Internal Medicine, ‡Department of Infectious Disease, and §Department of Gastroenterology and Hepatology, Cleveland Clinic Florida, Weston, Florida BACKGROUND & AIMS:
Antibiotics often are given to prevent infections but also constitute a risk factor for Clostridium difficile infection (CDI). Metronidazole is an effective treatment for CDI. We investigated whether prophylactic administration of metronidazole to patients before they receive other antibiotics reduces the risk of CDI.
METHODS:
We performed a retrospective cohort analysis of data collected from 12,026 high-risk patients admitted to Cleveland Clinic Foundation Hospitals from 2008 through 2012. High-risk patients were defined as age 55 or older who received a broad-spectrum antibiotic (piperacillin-tazobactam or ciprofloxacin) and a gastric acid suppressant (a proton pump inhibitor or a histamine-2 receptor blocker) during their hospitalization. Development of CDI was compared between patients who received metronidazole for non-CDI indications before broad-spectrum antibiotics (n [ 811) and those who did not (n [ 11,215). Logistic regression was used to control for patient demographics and comorbidities.
RESULTS:
The rate of CDI was 1.4% (n [ 11) among the patients who received metronidazole for non-CDI indications and 6.5% (n [ 728) among those who did not. This was observed to be an 80% reduction in CDI among patients who received metronidazole (odds ratio, 0.21; 95% confidence interval, 0.11–0.38; P < .001), adjusted for age, sex, and comorbidities.
CONCLUSIONS:
Based on a retrospective analysis, metronidazole might be used to prevent CDI in certain high-risk patients. Prospective controlled trials are necessary before making further recommendations.
Keywords: Diarrhea; Antibiotic; Prevention; Prophylaxis.
lostridium difficile infection (CDI) accounts for 15% to 30% of antibiotic-related diarrhea and is currently the leading cause of infectious diarrhea in the health care setting.1 In recent years, CDI has become increasingly prevalent.2–4 Incidences of up to 8.4 cases per 1000 admissions have been reported in the acute care setting in North America, as compared with 3.4 per 1000 admissions in previous years.5 This is a major economic health care burden with estimated costs ranging from $2871 to $4846 per case for primary CDI infection and from $13,655 to $18,067 per case for recurrent CDI infection. Mortality in this population has been reported to be as high as 5.7%.4–8 Important known risk factors for the development of CDI include advanced age (>60 y), prolonged hospitalization, and exposure to antibiotics.9 Classes of antibiotics that are associated most commonly with emergence of disease are penicillins, cephalosporins, sulfa antibiotics, and fluoroquinolones.3–7,10–12 The use of acid-suppressant medications, such as histamine-2 receptor blocker and
C
proton pump inhibitors, has also been associated with an increased risk for the development of CDI.13–17 The proposed mechanism suggests that acid suppression allows Clostridium difficile spores to thrive, but the exact mechanism of an increased risk remains controversial.18 More recently, CDI incidence has been found increasingly in other populations not previously considered at risk. Young healthy patients receiving a 1-time dose of antibiotic prophylaxis before surgery or even those being treated for upper respiratory infections may be at risk.3,7,19,20 An increase in the incidence also has been found in special populations, such as in patients with inflammatory bowel disease. These patients experience higher rates of hospitalization, surgery, and
Abbreviation used in this paper: CDI, Clostridium difficile infection. © 2014 by the AGA Institute 1542-3565/$36.00 http://dx.doi.org/10.1016/j.cgh.2014.02.040
2
Rodriguez et al
mortality with CDI.21,22 These risks are most likely a result of new community-acquired strains of C difficile, namely NAP1/BI/027, a hypervirulent strain initially described in North America.12 Multiple strategies have been studied to reduce the transmission of CDI. These strategies include implementation of contact precautions and dedicated patient equipment, such as disposable stethoscopes and singleuse rectal thermometers.7 Use of chlorine-based disinfectants, proper hand washing, avoiding unnecessary use of antibiotics, and reducing the use of acid-suppressant medications, together have proven effective.9,16,23,24 There are little published data on the use of prophylaxis for CDI. The available literature mostly emphasizes the use of probiotics as a method of prophylaxis. Their mechanism of action has been linked to restoration of host flora by immunomodulation and suppressing pathogenic bacteria colonization.25–28 However, concerns arise from lack of product standardization and lack of qualitycontrol regulations.29 Formulations with smaller bacterial counts appear to have little impact in the prevention of disease, while those with higher bacterial counts may be associated with higher rates of blood stream infections, namely fungemia.30,31 Given the increased incidence of CDI and high rates of resistant strains, research to find new effective prophylactic measures is warranted. The aim of this study was to determine whether metronidazole has a role in prophylaxis in selected patients at risk for CDI.
Methods This was a retrospective cohort analysis of patients aged 55 to 103 years admitted to Cleveland Clinic Foundation Hospitals between January 2008 and December 2012. This hospital system includes an academic tertiary care medical center as well as 8 community and affiliate hospitals located in 2 states. Patient cohorts were generated using the Explorys data platform (Cloudera, Inc, Palo Alto, CA), which provides access to anonymous patient information and can create a large database of electronic medical records, allowing for large-scale population studies. The study protocol was approved by the Institutional Review Board of the Cleveland Clinic Foundation Hospitals. A cohort of 12,026 patients was generated that was defined as high-risk: age 55 or older, receiving treatment with broad-spectrum antibiotics (piperacillin-tazobactam or ciprofloxacin), and concurrent treatment with a gastric acid suppressant (either a proton pump inhibitor or a histamine-2 receptor blocker). This population then was divided into 2 subgroups: high-risk patients who received metronidazole before broad-spectrum antibiotics during their hospitalization and high-risk patients who did not receive metronidazole during their hospitalization (Figure 1). The length of treatment or the dose of antibiotic was not determined because the focus of the study
Clinical Gastroenterology and Hepatology Vol.
-,
No.
-
Figure 1. Flow diagram of patient population evaluated for CDI.
was on whether patients received these antibiotics categorically. In the metronidazole group, the antibiotic was initiated 1 to 3 days before starting piperacillin-tazobactam or ciprofloxacin. This time interval was chosen to ensure that patients were not being placed on metronidazole empirically for suspected CDI and also to determine a role for CDI prophylaxis. The primary outcome was the rate of development of CDI for all patients at any time from day 1 to day 7 after the initiation of broad-spectrum antibiotics. This time interval was chosen because most cases of CDI develop within 14 days of antibiotic initiation.32 Longer time intervals would require outpatient follow-up evaluation, and our study was not designed to perform this type of follow-up evaluation. CDI cases diagnosed before or concomitantly to the initiation of broadspectrum antibiotics were excluded. Explorys generated the cohorts that met our specific inclusion specifications. Piperacillin-tazobactam and ciprofloxacin were chosen on the basis of hospital formulary availability; these were the commonly prescribed antibiotics in the hospitals to treat respiratory, genitourinary, and gastrointestinal infections. As for metronidazole, the most common diagnoses related to its use were intra-abdominal infections (ie, abscesses, empiric treatment for diarrhea). Patients with a prior history of CDI were included as well as patients with repeated admissions if they met the study criteria. The development of CDI was determined from hospital diagnosis codes during admission. In the Cleveland Clinic Foundation Hospitals, a diagnosis of CDI is made by stool collection. More specifically, before March 2012, diagnosis was made by detection of toxin A and/or toxin B via enzyme-linked fluorescent immunoassay on 3 separate stool samples. Protocol for testing then changed to include detection of toxin A and/or toxin B via enzyme-linked fluorescent immunoassay performed one time in addition to glutamate dehydrogenase antigen detection. This was followed by polymerase chain reaction, if testing was indeterminate. Patient demographics including age, sex, and race; comorbidities; and clinical data including a medication list and clinical outcomes were retrieved from the Explorys database. A t test was used to compare groups with relation to continuous variables (such as age).
-
2014
CDI Antibiotic Prophylaxis
Chi-square analysis was used to compare groups with regards to categoric variables (including incidence of CDI), and to produce univariate predictors of CDI. Linear regression was used to estimate CDI prevalence among age groups. Logistic regression was used to control for patient demographics and comorbidities. Odds ratios were reported along with the corresponding 95% confidence interval. A Kaplan–Meier curve was produced to compare the age-dependent incidence of CDI as a function of a cohort, particularly in those patients who received metronidazole and those who did not. A P value of less .05 was considered statistically significant. All statistical calculations were performed using Medcalc 2011 software (version 13.1.2; Ostend, Belgium).
3
Table 2. Comparison of Demographics and Clinical Characteristics Based on Diagnosis of CDI
Mean age, y Male sex Caucasian DM HTN CHF CKD COPD Pneumonia UTI Ciprofloxacin Piperacillin-tazobactam
No CDI n ¼ 11287
CDI n ¼ 739
P value
72.02 8.45 51% 80% 35% 60% 13% 13% 10% 8% 17% 82% 19%
73.177 8.88 47% 78% 49% 74% 19% 20% 16% 29% 30% 88% 12%
.0003 .0114 .1965
Risk of Clostridium difficile Infection in Hospitalized Patients Receiving Metronidazole for a Non–C difficile Infection Sandra Rodriguez,* Marlow B. Hernandez,* Giorgio Tarchini,‡ Megan Zaleski,* Marjon Vantanchi,* Lyssette Cardona,‡ Fernando Castro–Pavia,§ and Alison Schneider§ *Department of Internal Medicine, ‡Department of Infectious Disease, and §Department of Gastroenterology and Hepatology, Cleveland Clinic Florida, Weston, Florida BACKGROUND & AIMS:
Antibiotics often are given to prevent infections but also constitute a risk factor for Clostridium difficile infection (CDI). Metronidazole is an effective treatment for CDI. We investigated whether prophylactic administration of metronidazole to patients before they receive other antibiotics reduces the risk of CDI.
METHODS:
We performed a retrospective cohort analysis of data collected from 12,026 high-risk patients admitted to Cleveland Clinic Foundation Hospitals from 2008 through 2012. High-risk patients were defined as age 55 or older who received a broad-spectrum antibiotic (piperacillin-tazobactam or ciprofloxacin) and a gastric acid suppressant (a proton pump inhibitor or a histamine-2 receptor blocker) during their hospitalization. Development of CDI was compared between patients who received metronidazole for non-CDI indications before broad-spectrum antibiotics (n [ 811) and those who did not (n [ 11,215). Logistic regression was used to control for patient demographics and comorbidities.
RESULTS:
The rate of CDI was 1.4% (n [ 11) among the patients who received metronidazole for non-CDI indications and 6.5% (n [ 728) among those who did not. This was observed to be an 80% reduction in CDI among patients who received metronidazole (odds ratio, 0.21; 95% confidence interval, 0.11–0.38; P < .001), adjusted for age, sex, and comorbidities.
CONCLUSIONS:
Based on a retrospective analysis, metronidazole might be used to prevent CDI in certain high-risk patients. Prospective controlled trials are necessary before making further recommendations.
Keywords: Diarrhea; Antibiotic; Prevention; Prophylaxis.
lostridium difficile infection (CDI) accounts for 15% to 30% of antibiotic-related diarrhea and is currently the leading cause of infectious diarrhea in the health care setting.1 In recent years, CDI has become increasingly prevalent.2–4 Incidences of up to 8.4 cases per 1000 admissions have been reported in the acute care setting in North America, as compared with 3.4 per 1000 admissions in previous years.5 This is a major economic health care burden with estimated costs ranging from $2871 to $4846 per case for primary CDI infection and from $13,655 to $18,067 per case for recurrent CDI infection. Mortality in this population has been reported to be as high as 5.7%.4–8 Important known risk factors for the development of CDI include advanced age (>60 y), prolonged hospitalization, and exposure to antibiotics.9 Classes of antibiotics that are associated most commonly with emergence of disease are penicillins, cephalosporins, sulfa antibiotics, and fluoroquinolones.3–7,10–12 The use of acid-suppressant medications, such as histamine-2 receptor blocker and
C
proton pump inhibitors, has also been associated with an increased risk for the development of CDI.13–17 The proposed mechanism suggests that acid suppression allows Clostridium difficile spores to thrive, but the exact mechanism of an increased risk remains controversial.18 More recently, CDI incidence has been found increasingly in other populations not previously considered at risk. Young healthy patients receiving a 1-time dose of antibiotic prophylaxis before surgery or even those being treated for upper respiratory infections may be at risk.3,7,19,20 An increase in the incidence also has been found in special populations, such as in patients with inflammatory bowel disease. These patients experience higher rates of hospitalization, surgery, and
Abbreviation used in this paper: CDI, Clostridium difficile infection. © 2014 by the AGA Institute 1542-3565/$36.00 http://dx.doi.org/10.1016/j.cgh.2014.02.040
2
Rodriguez et al
mortality with CDI.21,22 These risks are most likely a result of new community-acquired strains of C difficile, namely NAP1/BI/027, a hypervirulent strain initially described in North America.12 Multiple strategies have been studied to reduce the transmission of CDI. These strategies include implementation of contact precautions and dedicated patient equipment, such as disposable stethoscopes and singleuse rectal thermometers.7 Use of chlorine-based disinfectants, proper hand washing, avoiding unnecessary use of antibiotics, and reducing the use of acid-suppressant medications, together have proven effective.9,16,23,24 There are little published data on the use of prophylaxis for CDI. The available literature mostly emphasizes the use of probiotics as a method of prophylaxis. Their mechanism of action has been linked to restoration of host flora by immunomodulation and suppressing pathogenic bacteria colonization.25–28 However, concerns arise from lack of product standardization and lack of qualitycontrol regulations.29 Formulations with smaller bacterial counts appear to have little impact in the prevention of disease, while those with higher bacterial counts may be associated with higher rates of blood stream infections, namely fungemia.30,31 Given the increased incidence of CDI and high rates of resistant strains, research to find new effective prophylactic measures is warranted. The aim of this study was to determine whether metronidazole has a role in prophylaxis in selected patients at risk for CDI.
Methods This was a retrospective cohort analysis of patients aged 55 to 103 years admitted to Cleveland Clinic Foundation Hospitals between January 2008 and December 2012. This hospital system includes an academic tertiary care medical center as well as 8 community and affiliate hospitals located in 2 states. Patient cohorts were generated using the Explorys data platform (Cloudera, Inc, Palo Alto, CA), which provides access to anonymous patient information and can create a large database of electronic medical records, allowing for large-scale population studies. The study protocol was approved by the Institutional Review Board of the Cleveland Clinic Foundation Hospitals. A cohort of 12,026 patients was generated that was defined as high-risk: age 55 or older, receiving treatment with broad-spectrum antibiotics (piperacillin-tazobactam or ciprofloxacin), and concurrent treatment with a gastric acid suppressant (either a proton pump inhibitor or a histamine-2 receptor blocker). This population then was divided into 2 subgroups: high-risk patients who received metronidazole before broad-spectrum antibiotics during their hospitalization and high-risk patients who did not receive metronidazole during their hospitalization (Figure 1). The length of treatment or the dose of antibiotic was not determined because the focus of the study
Clinical Gastroenterology and Hepatology Vol.
-,
No.
-
Figure 1. Flow diagram of patient population evaluated for CDI.
was on whether patients received these antibiotics categorically. In the metronidazole group, the antibiotic was initiated 1 to 3 days before starting piperacillin-tazobactam or ciprofloxacin. This time interval was chosen to ensure that patients were not being placed on metronidazole empirically for suspected CDI and also to determine a role for CDI prophylaxis. The primary outcome was the rate of development of CDI for all patients at any time from day 1 to day 7 after the initiation of broad-spectrum antibiotics. This time interval was chosen because most cases of CDI develop within 14 days of antibiotic initiation.32 Longer time intervals would require outpatient follow-up evaluation, and our study was not designed to perform this type of follow-up evaluation. CDI cases diagnosed before or concomitantly to the initiation of broadspectrum antibiotics were excluded. Explorys generated the cohorts that met our specific inclusion specifications. Piperacillin-tazobactam and ciprofloxacin were chosen on the basis of hospital formulary availability; these were the commonly prescribed antibiotics in the hospitals to treat respiratory, genitourinary, and gastrointestinal infections. As for metronidazole, the most common diagnoses related to its use were intra-abdominal infections (ie, abscesses, empiric treatment for diarrhea). Patients with a prior history of CDI were included as well as patients with repeated admissions if they met the study criteria. The development of CDI was determined from hospital diagnosis codes during admission. In the Cleveland Clinic Foundation Hospitals, a diagnosis of CDI is made by stool collection. More specifically, before March 2012, diagnosis was made by detection of toxin A and/or toxin B via enzyme-linked fluorescent immunoassay on 3 separate stool samples. Protocol for testing then changed to include detection of toxin A and/or toxin B via enzyme-linked fluorescent immunoassay performed one time in addition to glutamate dehydrogenase antigen detection. This was followed by polymerase chain reaction, if testing was indeterminate. Patient demographics including age, sex, and race; comorbidities; and clinical data including a medication list and clinical outcomes were retrieved from the Explorys database. A t test was used to compare groups with relation to continuous variables (such as age).
-
2014
CDI Antibiotic Prophylaxis
Chi-square analysis was used to compare groups with regards to categoric variables (including incidence of CDI), and to produce univariate predictors of CDI. Linear regression was used to estimate CDI prevalence among age groups. Logistic regression was used to control for patient demographics and comorbidities. Odds ratios were reported along with the corresponding 95% confidence interval. A Kaplan–Meier curve was produced to compare the age-dependent incidence of CDI as a function of a cohort, particularly in those patients who received metronidazole and those who did not. A P value of less .05 was considered statistically significant. All statistical calculations were performed using Medcalc 2011 software (version 13.1.2; Ostend, Belgium).
3
Table 2. Comparison of Demographics and Clinical Characteristics Based on Diagnosis of CDI
Mean age, y Male sex Caucasian DM HTN CHF CKD COPD Pneumonia UTI Ciprofloxacin Piperacillin-tazobactam
No CDI n ¼ 11287
CDI n ¼ 739
P value
72.02 8.45 51% 80% 35% 60% 13% 13% 10% 8% 17% 82% 19%
73.177 8.88 47% 78% 49% 74% 19% 20% 16% 29% 30% 88% 12%
.0003 .0114 .1965
