Clostridium difficile is a spore-forming bacterium that is the leading cause of nosocomial infection in the developed world 1,2 and as such is a healthcare priority 3. Imbalances in the gut microbiota allow C. difficile to proliferate and subsequently produce toxins, which in turn can lead to colitis. The infection extends hospital stays by 3–7 days and increases the mortality rate by three-fold 4.
Metronidazole and vancomycin are the main therapeutic agents, and although both antibiotics are equally efficacious in mild disease 5, vancomycin is superior in severe cases 6. However, metronidazole is cheaper, and may be less likely to prolong carriage 7. For these reasons, the UK Department of Health, North American, and local hospital guidelines recommend stratifying patients by severity and treating them accordingly 8,9. Patients with mild or moderate disease whose symptoms do not improve with metronidazole are usually switched to vancomycin. Such patients have prolonged hospital admission 10. It follows that accurate prediction of metronidazole failure at the outset could shorten hospital stay and possibly reduce morbidity and mortality in those with mild-to-moderate disease.
A number of metrics have been proposed as indicators of disease that will be resistant to metronidazole. Published guidelines typically use white cell count (WCC), creatinine, frequency of bowel movements, temperature, blood pressure, ileus, abdominal tenderness and radiological parameters 8,9. In addition, retrospective studies have found ileus, raised WCC 11,12 and low albumin levels (≤25 g/l) 6,10 to be associated with metronidazole treatment failure. A separate study found that serum creatinine levels more than 200 μmol/l were associated with C. difficile disease complicated by megacolon, perforation, colectomy, shock requiring vasopressors, or death 12. Until this study there has been no analysis of whether C-reactive protein (CRP) may be a useful indicator of metronidazole resistant disease. However, a multivariate analysis of patients with a first documented episode of C. difficile diarrhoea found that a high CRP was associated with death 13. Another study showed that those with severe C. difficile at diagnosis, as defined by systemic signs of hypovolaemia, had a higher CRP than those with nonsevere disease 14. Similarly, in ulcerative colitis there is evidence that a CRP of more than 45 mg/l on day 3 of treatment is associated with an 85% chance of colectomy 15.
In this study we carried out a logistic regression analysis to determine whether CRP or any of the other previously proposed measures might be useful in predicting metronidazole failure.
Materials and methods
St Mary’s Hospital is a 500-bed acute care teaching hospital serving a local population of 300 000 people, part of the Imperial NHS Trust, and affiliated with the medical school at Imperial College, London. We reviewed the case notes of 102 consecutive inpatients with a positive C. difficile toxin assay between 1 January 2009 and 20 April 2010. A list of these cases was kindly provided by the infection control team; this list did not include patients who tested C. difficile toxin positive, but were treated wholly in the community. The Meridian Premier C. difficile toxin A and B enzyme immunoassay was used to determine positivity. Treating clinicians are informed of a positive result on the same day so that therapy can be started promptly.
Patients were excluded if: they had severe disease and were therefore treated with vancomycin from the outset (n=10); the notes were incomplete (n=13); the episode was a recurrence of prior infection (n=7); there was another recognized aetiology for the diarrhoea, for example ulcerative colitis (n=2); they were discharged before the result of the assay being known and were treated on an outpatient basis (n=2); they were receiving palliative care and discharged for treatment at home (n=2); they died before the C. difficile toxin result was known and were not yet on treatment (n=1). This left a total of 65 cases that were reviewed.
We defined resolution of diarrhoea as passage of formed stool for 48 h. Where no bowel movement was recorded, this was only considered to be resolution of the diarrhoea if the subsequent stool was formed (type 4 or less on the Bristol Stool Chart).
Patients whose infection resolved with metronidazole therapy alone were considered successes. This included patients who were discharged before complete resolution of symptoms (as per the definition above), with cessation of diarrhoea confirmed as an outpatient.
Patients whose symptoms had not improved after 6 days of metronidazole treatment (as per previous studies 10,11) and in whom vancomycin was then started, or whose infection did not resolve and who died (n=2; although C. difficile was not implicated as the cause of death), were considered metronidazole treatment failures. Patients who were treated with vancomycin within the first 6 days because they progressed to meet the criteria for severe disease were also included as metronidazole treatment failures. The baseline characteristics of the two groups are shown in Table 1.
The data were drawn from multiple sources. The clinical parameters came from the entries of doctors and nurses in patient notes, observation charts and stool charts. The biochemical, haematological and radiological parameters were sourced from the electronic systems at St Mary’s Hospital. The values recorded are from the time of the C. difficile toxin result, as these reflect the information available to clinicians when assessing severity and making treatment decisions. For temperature and systolic blood pressure, we took the highest and the lowest readings, respectively, over the previous 24 h.
The distribution of patient characteristics between treatment successes and failures were compared. Unpaired t-tests and Kruskal–Wallis tests were used to analyse normally and non-normally distributed data, respectively. For categorical data, a Fisher’s exact test was used as expected values were small. The tests used for each comparison and P-values are reported in Table 1.
Subsequently, any variables that were significantly different between outcome groups were further analysed by univariate logistic regression to determine the strength of the association. For the regression analysis, the variable was categorized (such as WCC<15×109/l and WCC≥15×109/l) and Table 2 reports the odds ratios, 95% confidence intervals and P-values of the regression.
Table 1 summarizes the baseline characteristics of the treatment success and failure groups. Of the 65 cases included, 48 (74%) resolved with metronidazole alone. The other 17 cases reviewed were given vancomycin: 10 of these after failing a 6-day treatment with metronidazole, the other seven treated with vancomycin as they met the criteria for severe disease (within 6 days of metronidazole treatment).
Regression analysis of the individual variables showed that CRP, WCC and creatinine levels were significantly different in patients who failed metronidazole treatment (P=0.002, 0.012 and 0.028, respectively). This is shown in Table 1.
Our study examined the outcomes of 65 patients with C. difficile infection treated with metronidazole. Regression analysis suggests that most of the metrics currently used to assess severity are unrelated to metronidazole treatment failure. However CRP, not previously assessed, WCC and creatinine may be more closely related to treatment failure.
Four other studies have previously examined patients with C. difficile infection started on metronidazole to identify predictors of treatment failure 6,10,11,16. There are discrepancies with respect to their findings. All, like ours, are retrospective, observational and have a relatively small sample size (the largest sample, n=102), which may in part account for the inconsistencies. There are also subtle differences in methodology. For example, Fernandez et al. 10 defined treatment success as the absence of fever (<100.4°F), abdominal pain and two or less bowel movements a day at day 5. This differs from our definition, which allows a sixth day and does not consider fever or abdominal pain. Belmares et al. 11 use the same definition as our study; whereas, Zar et al. 6 additionally required a negative C. difficile toxin A assay at days 6 and 10.
Our findings with regard to creatinine are consistent with an earlier study that found levels more than 200 μmol/l were associated with complicated disease 12. The authors hypothesized that, in the context of C. difficile disease, creatinine serves as a marker of diarrhoea severity. In keeping with this, creatinine levels are significantly higher in those with C. difficile disease associated with systemic signs of hypovolaemia at diagnosis 14. However, two other studies found no difference in creatinine levels between metronidazole treatment successes and failures 11,16. The patients enrolled in these studies had a higher median creatinine at diagnosis (>118 vs. 84 μmol/l in our series), suggesting that they were more volume deplete. It is also possible that there were differences in the proportion of patients with chronic renal impairment. Either may have obscured any differences in creatinine. Other investigators who related creatinine levels to baseline found that a rise of more than 50% above baseline predicted colectomy or death 17.
The other main divergence from previous published results is the lack of an association with albumin levels. Ours is not the only study that failed to find this 11,16. Both studies that found a relationship with low albumin levels did so using a cut-off of 25 g/l or less 6,10. In total, 45/99 (45%) and 15/38 (39%) of their patients had albumin levels 25 g/l or more, compared with just 17/65 (26%) of our series. It is unclear what the cause of this may be, but it hints at differences in the underlying populations that may be the principal driver of any difference in albumin levels across metronidazole responders and nonresponders.
The finding related to CRP is particularly interesting as this variable has not previously been analysed and does not form a part of existing scoring systems 8,9. CRP is a plasma protein that is found in increased volumes during inflammation and tissue damage, therefore it is not surprising that it is elevated during severe C. difficile infection. The same applies to a raised WCC, a finding replicated by Belmares et al. 11. A very high CRP (>150 mg/l) has been found to predict mortality in another recent study 13. Our data would suggest that a lower level should be used to trigger early vancomycin therapy. Specifically, if we applied a threshold of 100 mg/l we would have commenced vancomycin unnecessarily in seven (out of 48 metronidazole successes, 15%) and had seven metronidazole failures (41% of the current 17 treatment failures).
Current approaches require the setting of a boundary; for example, WCC more than 15×109/l signifying severe disease. Perhaps a calculator that allowed risk to be stated on a continuum with the absolute values of metrics being used rather than a binary approach would be more accurate. Certainly our data would suggest that CRP should be included in such a model.
The authors thank Virginia Poulton (infection prevention and control support manager) who gave us the list of toxin positive patients.
Conflicts of interest
There are no conflicts of interest.
1. McFarland LV. Antibiotic-associated diarrhoea: epidemiology, trends and treatment. Future Microbiol. 2008;3:563–578
2. Bartlett JG. Narrative review: the new epidemic of Clostridium difficile
-associated enteric disease. Ann Intern Med. 2006;145:758–764
3. Mayor S. Battle against hospital acquired infections has been too limited, MPs’ report says. BMJ. 2009;339:b4680
4. Kyne L, Hamel MB, Polavaram R, Kelly CP. Health care costs and mortality associated with nosocomial diarrhea due to Clostridium difficile.
Clin Infect Dis. 2002;34:346–353
5. Teasley DG, Gerding DN, Olson MM, Peterson LR, Gebhard RL, Schwartz MJ, et al. Prospective randomised trial of metronidazole versus vancomycin for Clostridium difficile
associated diarrhea and colitis. Lancet. 1983;2:1043–1046
6. Zar FA, Bakkanagari SR, Moorthi KM, Davis MB. A comparison of vancomycin and metronidazole for the treatment of Clostridium difficile
-associated diarrhea, stratified by disease severity. Clin Infect Dis. 2007;45:302–307
7. Johnson S, Homann SR, Bettin KM, Quick JN, Clabots CR, Peterson LR, Gerding DN. Treatment of asymptomatic Clostridium difficile
carriers (fecal excretors) with vancomycin or metronidazole. A randomized, placebo-controlled trial. Ann Intern Med. 1992;117:297–302
9. Cohen SH, Gerding DN, Johnson S, Kelly CP, Loo VG, McDonald C, et al. Clinical practice guidelines for Clostridium difficile
infection in adults: 2010 update by the Society for Healthcare Epidemiology of America (SHEA) and the Infectious Diseases Society of America (IDSA). Infect Control Hosp Epidemiol. 2010;31:431–455
10. Fernandez A, Anand G, Friedenberg F. Factors associated with failure of metronidazole in Clostridium difficile
-associated disease. J Clin Gastroenterol. 2004;38:414–418
11. Belmares J, Gerding DN, Parada JP, Miskevics S, Weaver F, Johnson S. Outcome of metronidazole therapy for Clostridium difficile
disease and correlation with a scoring system. J Infect. 2007;55:495–501
12. Pépin J, Valiquette L, Alary ME, Villemure P, Pelletier A, Forget K, et al. Clostridium difficile
-associated diarrhea in a region of Quebec from 1991 to 2003: a changing pattern of disease severity. CMAJ. 2004;171:466–472
13. Bhangu S, Bhangu A, Nightingale P, Michael A. Mortality and risk stratification in patients with Clostridium difficile
-associated diarrhoea. Colorectal Dis. 2010;12:241–246
14. Hardt C, Berns T, Treder W, Dumoulin FL. Univariate and multivariate analysis of risk factors for severe Clostridium difficile
-associated diarrhoea: importance of co-morbidity and serum C-reactive protein. World J Gastroenterol. 2008;14:4338–4341
15. Travis SP, Farrant JM, Ricketts C, Nolan DJ, Mortensen NM, Kettlewell MG, Jewell DP. Predicting outcome in severe ulcerative colitis. Gut. 1996;38:905–910
16. Al-Nassir WN, Sethi AK, Nerandzic MM, Bobulsky GS, Jump RL, Donskey CJ. Comparison of clinical and microbiological response to treatment of Clostridium difficile
-associated disease with metronidazole and vancomycin. Clin Infect Dis. 2008;47:56–62
17. Gujja D, Friedenberg FK. Predictors of serious complications due to Clostridium difficile
infection. Aliment Pharmacol Ther. 2009;29:635–642