Prospective Observational Study Comparing Three Different Treatment Regimes in Patients with Clostridium difficile Infection

Abstract

In a hospital-based, prospective cohort study, the effects of the three standard treatment regimens for mild Clostridium difficile infection (CDI), oral (p.o.) metronidazole at 500 mg three times/day, intravenous (i.v.) metronidazole at 500 mg three times/day, and oral (p.o.) vancomycin at 250 mg four times/day, were compared with respect to the risk of occurrence of complications, sequelae, and all-cause death within 30 days after the date of starting treatment. Differences in the incidence of these outcomes were tested by χ² or Fisher's exact tests. A Poisson regression model was performed to control for possible confounding effects of sex, age, and severity of comorbidity categorized according to the Charlson comorbidity index. The highest mortality was observed in the metronidazole i.v. group, with a mortality rate 38.1% (16/42) compared to mortality rates of 7.4% (9/121) in the metronidazole p.o. group and 9.5% (4/42) in the vancomycin p.o. group (P < 0.001). After adjustment for possible effects of sex, age (>65 years), and severity of comorbidity, the relative risk of a 30-day fatal outcome for patients receiving metronidazole i.v. was 4.3 (95% confidence interval [CI] = 1.92 to 10; P < 0.0001) compared to patients treated with metronidazole p.o. and 4.0 (95% CI = 1.31 to 5.0; P < 0.015) compared to patients treated with vancomycin p.o. There were no significant differences in the risk of complications between the three treatment groups. This study generates the hypothesis that treatment with i.v. metronidazole is inferior to the oral alternatives metronidazole and vancomycin.

INTRODUCTION

Discharge data have shown an increase in the percentage of patients with a diagnosis of Clostridium difficile infection (CDI) up to 26% per year (3, 7, 15, 18). Although guidelines for both diagnosis and therapy (2) are available, therapy of CDI remains complicated by recurrences, relapses, and failure in severe forms (4, 10, 11).

Metronidazole has been recommended as initial therapy since the late 1990s and continues to be the first choice for all but seriously ill patients and those with complicated or fulminant infections or multiple recurrences of CDI, for whom vancomycin is recommended. This is based on the fact that metronidazole is associated with a poorer clinical outcome than vancomycin in these patients (97% versus 76%) (19). Recently, fidaxomicin has also become available for CDI therapy (1). Both vancomycin and fidaxomicin have the limitation that only oral administration is feasible in patients with CDI. Few clinical data for intravenous metronidazole exist (5). Despite an increasing quantity of available data, controversy regarding treatment options still exists (2, 6).

In order to assess the three currently recommended treatment regimens for mild CDI (2), we conducted a prospective, clinical cohort study comparing oral (p.o.) metronidazole, intravenous (i.v.) metronidazole, and p.o. vancomycin with respect to the occurrence of complications, sequelae, and all-cause death within 30 days of starting treatment.

MATERIALS AND METHODS

Study design.

A hospital-based, prospective cohort study comparing three standard treatment regimens for mild CDI, i.e., a nonrandomized design, was applied.

Study population.

From 1 December 2008 to 15 March 2010, patients with mild CDI at a single center, the Kaiser Franz Josef Spital, an 800-bed hospital in Vienna, Austria, were consecutively included. According to the European Society of Clinical Microbiology and Infectious Diseases, a patient with mild CDI was defined as a patient (i) with a clinical picture compatible with CDI, (ii) with microbiological evidence of toxin-producing Clostridium difficile in stool without evidence of another cause of diarrhea, and (iii) without any signs of severe colitis (hemodynamic instability, signs of peritonitis or of ileus, marked leukocytosis [leukocyte count, >15 × 10⁹/liter], a marked left shift [band neutrophils, >20% of leukocytes], a rise in serum creatinine level [>50% above the baseline], elevated serum lactate, pseudomembranous colitis at onset [as diagnosed by endoscopy], distension of the large intestine, colonic wall thickening and pericolonic fat stranding [as diagnosed by imaging], ascites not explained by other causes [as diagnosed by imaging] [2]).

Exclusion criteria were detection of an additional diarrhea-causing pathogen in the stool sample, age under 18 years, and no treatment with any of the three treatment regimens under study.

The three treatment regimens under study were metronidazole 500 mg three times/day p.o. (treatment group I), metronidazole at 500 mg three times/day i.v. (treatment group II), and vancomycin at 250 mg four times/day p.o. (treatment group III), all for 10 days. Current hospital guidelines on CDI treatment recommend oral metronidazole at 500 mg three times/day as the primary choice of antimicrobial treatment. Oral vancomycin and i.v. metronidazole are recommended as alternative regimens. The choice of the antimicrobial treatment regimen was at the discretion of the individual treating physician.

Data collection.

Stool samples were taken whenever a nurse or a patient recognized loose or frequent stools. It is a nursing standard to ask the patients about their stool characteristics at least once daily. Sampling was repeated up to three times if results remained negative. In case of positivity, sampling was stopped. The data collected at the time of patient inclusion comprised demographics (date of birth, age, and sex), clinical signs and symptoms of CDI (fever, diarrhea), epidemiological CDI classification, and comorbidity. The comorbidity was categorized using the Charlson comorbidity index (5). A score of <3 points was defined as low, one of 3 to 5 points was defined as moderate, and one of >5 points was defined as severe comorbidity. For the analyses, the two categories low comorbidity (≤2 points) and moderate/severe comorbidity (>2 points) were used. The primary endpoints were all-cause death within 30 days following the date of starting one of the three treatment regimens; recovery was defined as no clinical signs or symptoms compatible with CDI at day 30. The secondary endpoints included severe CDI other than death, defined as admission to an intensive care unit (ICU) or surgical intervention due to CDI, complications including dehydration (as diagnosed by skin fold remaining standing), paralytic ileus, aerocoly, pseudomembranous colitis (as diagnosed by endoscopy), pancolitis, colon perforation, peritonitis, toxic megacolon (defined as radiological signs of distension of the colon and signs of a severe systemic inflammatory response), recurrent episode of diarrhea, and sequelae including colostomy and malabsorption.

Data were ascertained from the medical charts or, in case of postdischarge follow-up, by telephone or face-to-face interview with the study subjects.

Microbiological investigations.

Stool samples obtained from patients were tested for C. difficile toxins A and B by enzyme-linked immunosorbent assay (Vidas system; bioMérieux, Marcy l'Etoile, France). In parallel and irrespective of toxin results, stool cultures for C. difficile were performed. Stool samples were treated with an equal volume of absolute alcohol, homogenized using a vortex mixer, and rested at room temperature for at least 30 min. Thereafter, 2 drops of the sample were inoculated on CCFA agar (Oxoid, Cambridge, United Kingdom) and the sample was incubated anaerobically at 36°C ± 1°C for 48 to 72 h. If cultures were positive, a toxin assay was done from the cultured bacteria (the so-called toxigenic culture method). Only in case of both toxin tests being negative was a negative result for C. difficile confirmed. In case of unclear results, stool sampling was repeated. Presumptive colonies of C. difficile were tested with a rapid latex agglutination test for the presence of the specific somatic antigen: glutamate dehydrogenase (GLD) (Oxoid, Cambridge, United Kingdom).

Statistical analysis.

Differences in the incidence of recovery and all-cause death and in the incidence of severe CDI other than death, complications, and sequelae observed within the 30-day follow-up between the three treatment regimens were tested by use of a χ² test for a contingency table. χ² test or Fisher's exact test was used when comparing two treatment groups each (metronidazole i.v. versus metronidazole p.o., metronidazole i.v. versus vancomycin p.o., and vancomycin p.o. versus metronidazole p.o.) with respect to the primary and secondary endpoints, as given above. A Poisson regression model was performed in order to adjust for possible confounding effects of comorbidity, age, and sex. The equation used for the model was Y = a₀ + (a × X₁) + (b × X₂) + (c × sex) + (d × age) + (e × comorbidity), where a₀ is the intercept, a is the regression coefficient for x₁, b is the regression coefficient for x₂, c is the regression coefficient for sex, d is the regression coefficient for age, and e is the regression coefficient for comorbidity. X₁ and X₂ were established as dummy variables for determining the three treatment regimes in the model (group I, X₁ = 1 and X₂ = 0; group III, X₁ = 0 and X₂ = 1; group II as a reference, X₁ = 0 and X₂ = 0). Crude and adjusted incidence ratios were provided as effect measures.

RESULTS

Study population.

A total of 265 patients with mild C. difficile infection (stool frequency <4 times daily and no signs of severe colitis, as defined above) were identified in the study hospital between 1 December 2008 and 15 March 2010. Sixty-nine patients were admitted with CDI (community-acquired disease), and 196 patients acquired CDI during their hospital stay (nosocomial disease). The mean age was 77 years (range, 21 to 98 years), and 63.4% of the patients were females (n = 168). Fifty-five percent (n = 146) suffered from moderate/severe comorbidity, according to the Charlson comorbidity index of >2 points (Table 1). A predischarge all-cause fatal outcome was observed 3.33 times (95% confidence interval [CI] = 1.81 to 6.16; P > 0.0001) more frequently in CDI patients with moderate/severe comorbidity than in CDI patients with low comorbidity (Table 1).

Table 1.

Association of predischarge outcome with the characteristics sex, age, and severity of comorbidity among the 265 patients with mild CDI^a

Characteristic	No. (%) of CDI patients			RR (95% CI)	P
	Total (n = 265)	Predischarge all-cause death (n = 56)	Discharged alive (n = 209)
Age (>65 yrs)	213 (80.38)	46 (88.46)	167 (79.90)	1.73 (0.78–3.81)	0.15
Sex (male)	97 (36.6)	27 (48.21)	70 (33.49)	1.61 (1.02–2.56)	0.04
Comorbidity (moderate/severe)	146 (55.1)	45 (80.36)	101 (48.33)	3.33 (1.81–6.16)	<0.0001

^aThe characteristics were distinguished by predischarge outcome versus all-cause death or discharged alive, sex was male versus female, age >65 years versus ≤65 years, and severity of comorbidity versus moderate/severe versus low.

Out of the 265 patients with mild CDI, 60 patients did not receive any of the treatment regimens under study and were excluded from the study. A total of 121 study patients received metronidazole p.o., and 42 study patients each received metronidazole i.v. or vancomycin p.o. Oral medication was possible for all study patients. Most colleagues decided on the antimicrobial treatment regimen in compliance with the local hospital guidelines (i.e., the primary choice, metronidazole p.o.); some also preferred metronidazole i.v. or vancomycin p.o. for mild CDI due to personal experience, and others chose intravenous metronidazole in patients with i.v. lines in place. No significant differences in the median age or in the frequency distribution of sex and severity of comorbidity were observed between the three treatment groups and the CDI patients excluded from the study due to a treatment for CDI other than an antimicrobial treatment (n = 60) (Table 2). There were no significant differences between the three treatment groups in the frequency of comorbid conditions, such as cerebral dysfunction, cardiovascular disease, pulmonary disease, renal disease, liver disease, recent surgery (within 3 months), invasive medical device, current neoplasm, diabetes mellitus, or HIV infection (Table 2).

Table 2.

The three treatment groups and CDI patients receiving no antimicrobial treatment by demographics, severity of comorbidity, and underlying condition^a

Characteristic or underlying condition	Met p.o. (n = 121)	Met i.v. (n = 42)	Van p.o. (n = 42)	No antimicrobial treatment (n = 60)	P
Characteristic
Median (range) age (yr)	75.0 (22–95)	75.3 (43–92)	75.0 (27–98)	73.5 (21–94)	0.99
No. (%) of patients by:
Sex (male)	45 (37.2)	20 (47.6)	19 (45.2)	13 (21.7)	0.41
Comorbidity (moderate/severe)	62 (51.2)	24 (57.1)	24 (57.1)	36 (60)	0.71
Underlying condition
Cerebral dysfunction	52 (43.0)	17 (40.5)	19 (45.2)		0.91
Cardiovascular disease	92 (76.0)	31 (73.8)	37 (88.1)		0.2
Pulmonary disease	27 (22.3)	13 (31.0)	15 (35.7)		0.19
Renal disease	36 (29.8)	11 (26.2)	14 (33.3)		0.77
Liver disease	10 (8.3)	5 (11.9)	5 (11.9)		0.69
Surgery in previous 3 mo	21 (17.4)	6 (14.3)	5 (11.9)		0.68
Medical invasive device	11 (9.1)	3 (7.1)	4 (9.5)		0.91
Current neoplasia	11 (9.1)	5 (11.9)	3 (7.1)		0.75
Diabetes mellitus	39 (32.2)	14 (33.3)	11 (26.2)		0.73
HIV infection	0 (0)	0 (0)	0 (0)

^aThe three treatment groups were metronidazole (Met) p.o. (n = 121), metronidazole i.v. (n = 42), and vancomycin p.o. (n = 42).

Primary endpoints.

The highest all-cause 30-day mortality incidence was observed in the metronidazole i.v. group, with a 38.1% (16/42) mortality rate compared to mortality rates of 7.4% (9/121) in the metronidazole p.o. group and 9.5% (4/42) in the vancomycin p.o. group (P < 0.001; Table 3). When comparing two treatment groups each, the CDI patients in the metronidazole i.v. group were 4 times (95% CI = 1.45 to 11.1; P = 0.0001) more likely to die than the CDI patients in the vancomycin p.o. group and 5 times (95% CI = 2.43 to 11.1; P = 0.002) more likely to die than the CDI patients in the metronidazole p.o. group. After adjustment for possible effects of sex, age (>65 years), and severity of comorbidity, the relative risk of a 30-day fatal outcome was 4.3 (95% CI = 1.92 to 10; P < 0.0001) for patients receiving metronidazole i.v. compared to patients receiving metronidazole p.o. and 4.0 (95% CI = 1.31 to 5.0; P < 0.015) compared to patients treated with vancomycin.

Table 3.

Findings of the comparative analyses between the three treatment regimens with respect to the primary and secondary endpoints^a

Endpoint	No. (%) of patients			Pb
	Met p.o. (n = 121)	Met i.v. (n = 42)	Van p.o. (n = 42)
Primary
Recovery	100 (82.6)	22 (52.4)	34 (81)	<0.001
Continuing diarrhea	12 (9.9)	4 (9.5)	4 (9.5)	1
All-cause death	9 (7.4)	16 (38.1)	4 (9.5)	<0.001
Secondary
Paralytic ileus	0	1 (2.4)	0	0.14
Aerocoly	0	0	0
Pseudomembranous colitis	2 (1.7)	2 (4.8)	1 (2.4)	0.53
Pancolitis	0	0	0
Colon perforation	0	1 (2.4)	0	0.14
Peritonitis	0	0	0
Toxic megacolon	0	1 (2.4)	0	0.14
Recurrent episode of diarrhea ≥2 wkc	14 (11.6)	6 (14.3)	2 (4.8)	0.33
Recurrent episode of diarrhea <2 wkd	5 (4.1)	5 (11.9)	5 (11.9)	0.11
Dehydration	52 (43.0)	21 (50.0)	28 (66.7)	0.03
Severe diarrhea	14 (11.6)	10 (23.8)	8 (19.0)	0.13
BSId (clinical sepsis)	0	1 (2.4)	1 (2.4)	0.23
BSI (lab-confirmed sepsis)	3 (2.5)	2 (4.8)	4 (9.5)	0.16
At least one complication	64 (52.9)	29 (69.0)	32 (76.2)	0.01
Admission to ICU during hospital stay
Surgical intervention
Colostomy	0	0	0
Malabsorption	0	0	0

^aMet, metronidazole; Van, vancomycin; BSI, bloodstream infection.

^bχ² test for a contingency table.

^c≥2 weeks since the onset of the previous episode.

^d<2 weeks since the onset of the previous episode.

There was no significant difference in the all-cause 30-day mortality incidence observed between the two treatment groups metronidazole p.o. and vancomycin p.o. (P > 0.05).

CDI patients treated with metronidazole p.o. were 1.56 times (95% CI = 1.16 to 2.11; P = 0.0002) and CDI patients treated with vancomycin p.o. were 1.55 times (95% CI = 1.12 to 2.14; P = 0.005) more likely to recover than CDI patients treated with metronidazole i.v. (Table 4).

Table 4.

Findings of comparative analyses between two treatment regimens each^a

Endpoint	Met p.o. vs Met i.v. (ref)		Van vs Met p.o. (ref)		Van vs Met i.v. (ref)
	RR (95% CI)	P	RR (95% CI)	P	RR (95% CI)	P
Primary
Recovery	1.56 (1.16–2.11)	0.0002	0.99 (0.84–1.17)	0.9	1.55 (1.12–2.14)	0.005
Continuing diarrhea	1.04 (0.36–3.05)	1	0.96 (0.33–2.82)	1	1 (0.27–3.74)	1
All-cause death	0.2 (0.09–0.41)	0.0001	1.28 (0.42–3.94)	0.74	0.25 (0.09–0.69)	0.002
Secondary
Paralytic ileus
Aerocoly
Pseudomembranous colitis	0.35 (0.05–2.39)	0.27	1.44 (0.13–15.48)	1	0.5 (0.05–5.31)	1
Pancolitis
Colon perforation
Peritonitis
Toxic megacolon
Recurrent episode of diarrhea ≥2 wkb	0.35 (0.11–1.14)	0.13	2.88 (0.88–9.46)	0.13	1 (0.31–3.20)	1
Recurrent episode of diarrhea <2 wkc	0.81 (0.33–1.97)	0.64	0.41 (0.1–1.74)	0.25	0.33 (0.07–1.56)	0.26
Dehydration	0.86 (0.6–1.24)	0.43	1.55 (1.15–2.09)	0.008	1.33 (0.92–1.93)	0.12
Severe diarrhea	0.49 (0.23–1.01)	0.054	1.65 (0.74–3.64)	0.22	0.8 (0.35–1.83)	0.59
BSI (clinical sepsis)			∞	0.26	1 (0.06–15.47)	1
BSI (lab-confirmed sepsis)	0.52 (0.09–3.01)	0.6	3.84 (0.9–16.46)	0.07	2 (0.39–10.34)	0.68
At least one complication	0.77 (0.59–1)	0.07	1.44 (1.13–1.83)	0.008	1.1 (0.85–1.44)	0.46
Admission to ICU during hospital stay	1.04 (0.11–9.74)	1	2.88 (0.6–13.73)	0.19	3 (0.33–27.69)	0.62
Surgical intervention

^aThe comparisons were metronidazole (Met) p.o. versus Met i.v. (used as the reference [ref]), vancomycin (Van) versus metronidazole p.o. (used as the reference), and vancomycin versus metronidazole i.v. (used as the reference) with respect to the primary and secondary endpoints. RR, relative risk; BSI, bloodstream infection.

^b≥2 weeks since the onset of the previous episode.

^c<2 weeks since the onset of the previous episode.

Secondary endpoints.

The highest incidence of dehydration as a complication was observed in CDI patients treated with vancomycin p.o. (Table 3). CDI patients of the vancomycin group were 1.55 times (95% CI = 1.15 to 2.09; P = 0.0008) more likely to develop this complication than those of the metronidazole p.o. group. No difference in dehydration incidence was found between the treatment groups metronidazole i.v. versus metronidazole p.o. and metronidazole i.v. versus vancomycin p.o. There was no significant difference in the incidence of any other secondary endpoint between the three treatment groups (Tables 3 and 4).

DISCUSSION

The all-cause 30-day mortality from mild CDI was 29/205 (13%), which is in accordance with a recent European survey showing a mortality rate for all CDI cases of 101/455 (22%) after 3 months (3). Insufficient stool concentrations of metronidazole after infusion could be the reason for increased mortality in mild CDI compared to oral metronidazole or vancomycin. Unfortunately, only a single trial with a very limited number of CDI participants exists. In that trial, stool concentrations of metronidazole and the active hydroxy metabolite were directly compared after oral (n = 6) and intravenous (n = 3) therapy. No significant differences in the highly variable stool concentrations were found between the two treatment modalities (5). However, considering the high variability in measured metronidazole stool concentrations, the study might have been underpowered to detect any existing difference. Oral and intravenous metronidazole therapy resulted in low stool concentrations (range, 0.8 to 24.2 μg/g of stool; maximum, 1,212 μg/g) in these patients, and the concentrations became undetectable when diarrhea resolved (5). This finding is supported by another study in which metronidazole was not detectable in 9 of 10 patients treated for eradication of asymptomatic C. difficile fecal excretion as a means of controlling nosocomial outbreaks of C. difficile diarrhea (8). In comparison, the stool concentration after oral vancomycin was 1,406 ± 1,164 μg/g feces (8).

Metronidazole given orally is almost completely absorbed and is unaffected by infection. It is distributed widely and reaches 42 to 76% of plasma concentrations in colonic mucosa (9, 13). Metronidazole is extensively metabolized by the liver and excreted predominantly in feces, with only 12% excreted unchanged in urine (12). Thiercelin et al. (17) noted that metronidazole at 500 mg orally every 8 h resulted in steady-state area under the curve values which were 51% higher than those after the same dosage administered intravenously, even though the half-life is the same for both routes. This was also observed in a study by Mattila et al. (14). Two proposed hypotheses by the authors that may explain this phenomenon were that (i) metronidazole inhibits its own metabolism by intestinal microsomal mono-oxygenases or (ii) oral administration leads to excessive glucuronidation, with resulting enterohepatic recirculation (17). Therefore, in our study, patients receiving i.v. metronidazole could have had lower drug exposure than patients treated orally.

In this study, CDI patients receiving vancomycin p.o. were 1.55 times more likely to be dehydrated than CDI patients receiving metronidazole p.o., but there was no difference in the risk of dehydration between i.v. and p.o. metronidazole or between p.o. vancomycin and i.v. metronidazole. A possible explanation for this finding would be that vancomycin peroral is given in a fluid form. The vancomycin vial containing 1,000 mg dry substance is dissolved in 20 ml fluid. The daily dose therefore equals a fluid intake of 20 ml. Unlike administration in this insignificant amount of fluid, oral metronidazole is administered in tablet form. These tablets are swallowed twice daily with some additional sips of water on top of the normal fluid intake. Similarly, metronidazole intravenously means an additional intravenous fluid replacement of at least 300 ml per day. Of course, the considerations following the finding of higher dehydration with vancomycin may not be applicable to all patients. Many older patients have a high burden of comedications, which still have to be taken with water. Nevertheless, fluid balance and clinical signs of dehydration should be monitored closely in elderly CDI patients, especially when under therapy with oral vancomycin. As in this study, patients found to be dehydrated during the course of disease should receive fluid substitution.

Our study of patients with mild CDI showed similar incidences of cure and of mortality within 30 days when the standard therapy regimes of oral metronidazole and oral vancomycin were compared. In severe disease, metronidazole is associated with a worse clinical outcome than vancomycin (97% versus 76%) (19). These findings support the recommendations of current treatment guidelines to use oral metronidazole as the primary choice for all but seriously ill patients or those with complicated or fulminant infections or multiple recurrences of CDI, for whom vancomycin is recommended (2, 7).

A limitation of our study is its nonrandomized design. Although there are several reasons why randomized controlled trials (RCTs) are felt to provide the most robust evidence for treatment guidelines, they may suffer from insufficient duration of follow-up, inadequate study power to consider differences in important adverse events, and highly selected patient populations. Furthermore, as most RCTs are performed for licensing purposes, strategic treatment decisions often lack supportive evidence from RCTs. Although data from cohort studies may be used to complement information from RCTs, cohort studies themselves are susceptible to several biases (most notably, confounding), which may limit their findings (16). In respect to our study, one could argue that the treatment decision at the discretion of the physician on duty has introduced selection bias. For example, the decision for the intravenous treatment could have been favored in older, severely diseased patients or those with reduced vigilance. Although we cannot entirely rule out selection bias, the results are unlikely driven by these factors. First, only patients with mild CDI were included; the severely diseased were not eligible to participate in the study. Second, age was similar between the three treatment groups, and most importantly, there was no difference in comorbidity (Table 2). Third, all patients were able to take oral medication. Since no RCT comparing the three regimens described here is currently available for mild CDI, we see a cohort study to be an adequate approach to provide data, which should lead to further investigations, such as randomized controlled trials.

In conclusion, we observed a lower cure incidence and a higher all-cause 30-day mortality risk with the treatment regimen of intravenous metronidazole than with oral metronidazole and oral vancomycin in mild CDI. To confirm our hypothesis that intravenous metronidazole is inferior to the oral formulations in the treatment of mild CDI and should not be given whenever oral therapy is feasible, a prospective randomized controlled trial comparing these three treatment options is needed.