Factors associated with prolonged symptoms and severe disease due to Clostridium difficile

Transcription

1 Age and Ageing 1999; 28: Factors associated with prolonged symptoms and severe disease due to Clostridium difficile LORRAINE KYNE, CONCEPTA MERRY 2,BRIAN O CONNELL 2,ALAN KELLY 3,CONOR KEANE 2, DESMOND O NEILL , British Geriatrics Society Beth Israel Deaconess Medical Center, Gerontology Division, RABB-440, 330 Brookline Avenue, Boston MA 02115, USA 1 Age-Related Health Care, Meath Hospital, Dublin, Ireland 2 Department of Clinical Microbiology, Trinity College, Central Pathology Laboratory, St James s Hospital, Dublin, Ireland 3 Department of Community Health and General Practice, Trinity College, Dublin, Ireland Address correspondence to: L. Kyne. Fax: (+1) lkyne@bidmc.harvard.edu Abstract Objective: toxigenic Clostridium difficile is responsible for a spectrum of disease severity ranging from mild diarrhoea to fulminant colitis. This study attempts to determine the proportion of patients in each category of severity and evaluate the risk factors for a more prolonged and complicated course. Design: prospective cohort study. Setting: university teaching hospital. Subjects: all patients with symptomatic C. difficile infection during 4 months of an outbreak (January April 1995); n = 73; median age 74 years (range 17 91). Measurements: incidence of C. difficile-associated disease (CDAD); severity of disease; percentage of patients in each category of severity; risk factors for severe disease/prolonged symptoms (univariate and multivariable analyses). Results: the incidence of CDAD was 0.93%. Of the cases identified, 18 (24.7%) had mild, self-limiting disease; 26 (35.6%) had moderately severe disease; 23 (31.5%) had prolonged symptoms and six (8.2%) had a complicated course. Although CDAD was more common in older patients (P < 0.001), increasing age was not a risk factor for severity. Significant risk factors for severe CDAD included low Barthel and abbreviated mental test scores (P < 0.01, P < respectively) and recent endoscopy (P = 0.03). Logistic regression analysis revealed the following independent predictors of severe CDAD: endoscopy [odds ratios (OR) 4.0, P = 0.03] and cognitiveimpairment (OR 11.0, P < 0.01). A trend towards significance was noted for nasogastric tube insertion (OR 3.1, P = 0.08). Complications of infection included dehydration, malnutrition and faecal incontinence (which was statistically significantly associated with more severe disease; P < 0.01). Conclusion: risk factors for severity of CDAD include functional disability, cognitive impairment, and recent endoscopy. Anticipation of severe CDAD may limit morbidity and mortality. Keywords: Clostridium difficile, diarrhoea, elderly, outbreak, pseudomembranous colitis Introduction Fulminant, life-threatening colitis is the extreme end of a spectrum of disease caused by Clostridium difficile [1, 2]. Fortunately such severe disease occurs in only 3 5% of those affected [3 5]. Most patients experience diarrhoea which usually resolves within 10 days with appropriate medical therapy [6]. The morbidity and mortality associated with severe disease is important but, although risk factors for the acquisition of C. difficile have been extensively investigated [7 10], the host and bacterial factors responsible for prolonged symptoms and the progression to severe colitis are poorly understood. During a recent epidemic we had the opportunity to prospectively study 73 patients with symptomatic C. difficile infection. The aims of this study were to document the spectrum of disease caused by this 107

2 L. Kyne et al. organism and to evaluate the factors associated with a protracted course and/or the progression to severe disease. Methods The study was conducted over 4 months at St James s Hospital, Dublin, Ireland. As part of a larger epidemiological investigation of an outbreak, infection control policy in the hospital at this time advised that a stool sample should be submitted to the laboratory for C. difficile toxin assay if a patient had one loose bowel motion. All patients between January and April 1995 with a positive stool cytotoxin test (which detects C. difficile toxin B) were reviewed. If patients did not have diarrhoea (defined as at least three loose bowel motions per day for at least 2 consecutive days) [7] or other symptoms suggestive of C. difficile infection (abdominal pain, tenderness or distension) [2], they were considered to be asymptomatic carriers but were monitored during their hospital stay for signs of progression to symptomatic disease. All patients with symptomatic disease were reviewed and followed throughout their hospital course. The following information was recorded: demographics; duration and nature of symptoms; number of underlying diagnoses; presence of diabetes mellitus, immunosuppression, malignancy, chronic gastrointestinal conditions, respiratory disease or alcohol dependency; and medications, including the total number of days and total doses of antimicrobial therapy, treatment with immunosuppressive agents, laxatives/suppositories, antacids, histamine-2 receptor (H2) antagonists and proton pump inhibitors. In addition, we documented whether affected patients were fed via nasogastric or percutaneous endoscopic gastrostomy tubes or were receiving total parenteral nutrition. Gastrointestinal instrumentation (oesphago-gastroduodenoscopy, colonoscopy or barium enema) or surgery before disease onset was also recorded. At the onset of symptoms a measure of ability to perform activities of daily living (Barthel score) [11] and an abbreviated mental test (AMT) [12] score was recorded for each patient. Patients were monitored until discharge or death; associated morbidity including dehydration (assessed clinically), malnutrition (loss of greater than 10% of body weight), faecal incontinence, progression to toxic megacolon or death attributable to C. difficile infection were documented. Additional information recorded included the total number and age profile of admissions to hospital during the study period. As part of the epidemiological investigation of this outbreak, selected faecal isolates were typed using pyrolysis mass spectrometry. Definitions C. difficile-associated diarrhoea was defined as diarrhoea (three or more loose bowel motions per day for at least 2 days) not attributable to any other cause and occurring in association with a positive stool cytotoxin test for C. difficile. Disease severity was defined as follows: mild, diarrhoea lasting 4 days; moderate, diarrhoea lasting 5 10 days and severe, diarrhoea or other symptoms (abdominal pain, tenderness or distension) lasting longer than 10 days or associated with a complicated course [i.e. development of toxic megacolon or fulminant pseudomembranous colitis (diagnosed by endoscopy)]. Statistical methods The clinical characteristics of patients with mild to moderate disease severity were compared with those with severe disease using x 2 test for categorical data. Comparison of continuous variables between disease severity was by means of the Wilcoxon s rank sum test as these data were highly skewed. Severity was additionally modelled as a function of patient characteristics and interventions employing multiple logistic regression. Factors found to be significant by univariate analysis, age, sex, surgery, use of laxatives, H2 antagonists and the median values for the total days and doses of antimicrobial therapy were included in this model. Results Incidence and disease spectrum There were 7845 admissions to hospital during the 4- month study period. Eighty-seven patients with positive stool cytotoxin tests for C. difficile were identified via the microbiology department. Fourteen of these patients had 1 2 loose bowel motions only and did not develop symptoms of C. difficile-associated disease (CDAD) for the remainder of their hospital stay. These patients were considered to be asymptomatic cytotoxin-b-positive carriers. Seventy-three patients had symptomatic C. difficile infection, making the overall incidence of disease 0.93 cases per 100 admissions. All other causes of diarrhoea were excluded in these patients. Cases were identified throughout the hospital but there was heavy clustering of cases (35 in total) in four medicine for the elderly wards which were located in a building separate from the main hospital. The median age of all cases was 74 years (range 17 91). Eighteen cases (24.7%) had mild, self-limiting disease; 26 (35.6%) had moderately severe disease and 29 (39.7%) had prolonged symptoms. Of the group with prolonged symptoms, 18 had prolonged diarrhoea with a median duration of 18 days (range days), five experienced abdominal pain and tenderness which persisted for longer than 10 days and six had a 108

3 Severity of Clostridium difficile-associated disease complicated course. The six patients with a complicated course (8.2% of the total 73) progressed to fulminant colitis associated with severe inflammation with pseudomembrane formation and toxic megacolon, necessitating emergency subtotal colectomy in four patients. Surgery was associated with 100% mortality: three patients died in the immediate postoperative period and the other patient had a complicated postoperative course with progressive deterioration and death within 3 weeks. Two other patients with fulminant pseudomembranous colitis were managed medically with antibiotics and colonic decompression via a flatus tube. Sixty-two patients (85%) were treated with either metronidazole or vancomycin; 11 were managed conservatively. The median number of days from onset of symptoms to start of treatment was 3.5 (range 1 10). Factors associated with disease severity In Table 1 the characteristics of patients with mild to moderate disease severity (n = 44) are compared with those of patients with severe disease (n = 29). Debility (lower Barthel scores) and cognitive impairment (low AMT scores) at the onset of symptoms were highly significantly associated with prolonged symptoms and more severe disease (P < 0.01 and P < Table 1. Univariate analysis of factors associated with severity of Clostridium difficile-associated disease (CDAD) C. difficile-associated disease severity... All Mild/moderate Severe P value... No. of subjects Male:female ratio 36:37 23:21 13: Median age (range) 74 (17 91) 72.5 (17 91) 74 (25 89) 0.38 Median Barthel score (range) 13 (0 20) 14 (0 20) 8 (0 20) < 0.01 Median AMT score (range) 8 (0 10) 8.5 (0 10) 4 (0 10) < Underlying diagnosis No. conditions (median) Respiratory condition 34 (46.6%) 19 (43.2%) 15 (51.7%) 0.47 Malignancy 8 (11%) 2 (4.5%) 6 (20.7%) 0.37 Diabetes 2 (2.7%) 2 (4.5%) Chronic GI condition 8 (11%) 4 (9%) 4 (13.8%) 0.53 Stroke 11 (15%) 6 (13.6%) 5 (17.2%) 0.67 Alcohol dependency 4 (5.5%) 1 (2.3%) 3 (10.3%) 0.14 Co-existent MRSA 15 (20.5%) 7 ((15.9%) 8 (27.6%) 0.23 Medications Antibiotics: total days (median) 9 (0 33) 9.5 (0 33) 7 (0 27) 0.47 Antibiotics: total doses (median) 23 (0 188) 24 (0 188) 22 (0 144) 0.38 Immunosuppressive therapy 18 (24.7%) 11 (25%) 7 (24.1%) 0.93 Laxatives/suppositories 15 (20.5%) 9 (20.5%) 6 (20.7%) 0.98 Antidiarrhoeal agents 16 (22%) 12 (27.3%) 4 (13.8%) 0.19 H2 antagonists 22 (30.1%) 11 (25%) 11 (38%) 0.24 Antacids 1 (1.4%) 1 (2.3%) 1 (3.4%) 0.41 Proton pump inhibitors 6 (8.2%) 4 (9%) 2 (6.9%) 0.74 Interventions Surgery (any) 23 (31.5%) 15 (34%) 8 (27.6%) 0.56 Surgery (gastrointestinal) 11 (15%) 5 (11.4%) 6 (20.7%) 0.32 Endoscopy (OGD or lower) 18 (24.7%) 7 (15.9%) 11 (38%) 0.03 Barium enema 4 (5.5%) 1 (2.3%) 3 (10.3%) 0.30 Feeding Total parenteral nutrition 3 (4.1%) 1 (2.3%) 2 (6.9%) 0.56 NG or PEG 22 (30.1%) 10 (22.7%) 12 (41.4%) 0.09 Complications Dehydration 16 (22%) 7 (22%) 9 (15.9%) 0.13 Malnutrition 9 (12.3%) 4 (12.3%) 5 (9%) 0.30 Faecal incontinence 34 (46.6%) 15 (46.6%) 19 (34%) < 0.01 Death 9 (12.3%) 5 (12.3%) 4 (11.4%) 0.76 AMT, abbreviated mental test; GI, gastrointestinal, MRSA, methicillin-resistant Staphylococcus aureus; H2, histamine-2 receptor; OGD, oesphago-gastroduodenoscopy; NG, nasogastric; PEG, percutaneous endoscopic gastrostomy. 109

4 L. Kyne et al. Figure 1. Total number of admissions in the 4-month period (A, n = 7845) and number of patients developing Clostridium difficile-associated disease ( ), by age group. respectively). Although a highly significant trend for CDAD incidence with increasing age existed (P < 0.001; Figure 1), there was no association between disease severity and age. Patients with more severe disease had a median of four underlying medical conditions compared with three in the mild/moderate group; there were no significant differences between the nature of these conditions although both groups had a high prevalence of respiratory disease (mild/ moderate group, 43%; severe group, 52%). The duration of antimicrobial therapy, route of therapy (oral versus intravenous), or the total number of doses administered did not differ significantly between the two groups. Twenty-two percent of cases received anti-diarrhoeal medications. Although the use of the latter appeared to be associated with mild/moderate disease, one in four cases receiving these agents developed severe CDAD evidenced by abdominal pain, distension and radiological features consistent with toxic megacolon formation. Surgery (of any kind) before the onset of disease was not a risk factor for disease severity. Gastrointestinal instrumentation however was associated with more severe disease (P = 0.03); endoscopy was performed in 18 cases before disease onset, oesphagogastroduodenoscopy (nine), left-sided colonoscopy (six), total colonoscopy (three). A trend towards significance was also noted for nasogastric or percutaneous endoscopic gastrostomy feeding (P = 0.09). The results [as odds ratios (ORs) 95% confidence intervals] of an exploratory logistic regression model in which patient characteristics and typical interventions are present are shown in Table 2. Although significant in the univariate analysis, Barthel score has been excluded from the general model as it was highly correlated with AMT score. In this model the only factors that were statistically significant were endoscopy and AMT score, while nasogastric tube insertion/feeding was almost significant (P = 0.06). The ORs for the reduced model are shown in Table 3. Patients who underwent endoscopy were four times more likely to develop severe disease compared with patients who did not undergo this procedure (OR 4.0). Patients with AMT scores < 7 were also at greater risk of severe disease (OR 11.0). Although not significant at the traditional 0.05 level, patients in whom a nasogastric tube had been inserted seemed to have a slightly elevated risk of developing severe CDAD (OR 3.1). Complications On analysis of the complications, there were no statistically significant differences in dehydration or malnutrition between patients with mild/moderate disease and those with severe disease. Faecal incontinence was a common complication and highly significantly associated with severe disease (P < 0.01). The overall relapse rate was 20%, and was not significantly different in either group. Table 2. Multiple logistic-regression model for factors independently associated with disease severity Risk factor P value Odds ratio 95% confidence interval... AMT score (increasing) Age (increasing) Gender (female vs male) Nasogastric tube insertion PEG tube insertion Endoscopy Antibiotics Total days (0 9 vs 10þ) Total doses (0 25 vs 26þ) Laxatives/enemas H2 antagonists Surgery AMT, abbreviated mental test; H2, histamine-2 receptor; PEG, percutaneous endoscopic gastrostomy. 110

5 Severity of Clostridium difficile-associated disease Table 3. Reduced logistic regression model Risk factor P value Odds ratio 95% confidence interval... Endoscopy Abbreviated mental test score (< 7) Nasogastric tube insertion Mortality The overall mortality in the group of patients studied was 12.3% (nine patients). Five patients classified as having mild to moderate disease based on symptom duration died (11.4% of 44); however, it was felt that the primary cause of death in these patients was related to severe underlying disease and not CDAD, although this may have been a contributory factor. There were four deaths directly attributable to C. difficile infection in the severe group (14% of 29): these four patients had fulminant pseudomembranous colitis. Pyrolysis mass spectrometry typing Pyrolysis mass spectrometry identified two clusters of isolates representing two strains of toxigenic C. difficile. One strain type was identified predominantly in isolates from patients hospitalized in the four medicine for the elderly wards and the other strain type was located in the main hospital. There was no trend in severity as a function of strain type (P = 0.56, Mantel Haenszel x 2 trend test). Discussion Infection with toxigenic C. difficile results in a spectrum of disease ranging from mild self-limiting diarrhoea to fulminant colitis [13 15]; asymptomatic toxin-positive carriers have also been identified [16 18]. Few studies have evaluated the proportion of patients in each category of severity or have examined the risk factors for prolonged symptoms or severe disease [4, 19, 20]. By prospectively studying all new patients with positive stool cytotoxin tests for C. difficile during an outbreak we were able to construct a model of disease severity and identify factors which may place affected individuals at risk for a protracted or complicated course. The prevalence of CDAD varies depending on the institution and patient population studied [21]. We found that almost 1% of hospitalized patients developed symptomatic C. difficile infection during an epidemic and, of these, approximately 60% had diarrhoea which resolved within 10 days and 40% had prolonged symptoms or more severe disease. One in five patients in the severe group had fulminant colitis which was associated with a 67% mortality rate. Our definition of severe disease was less stringent than that used by other investigators [4]. Rubin et al. retrospectively examined a cohort of patients with CDAD and found that 3% of these patients developed severe colitis, defined as clinical decompensation that was clearly precipitated by C. difficile and resulted in intensive care unit admission or death [4]. The mortality in these patients was identical to that reported in our subset of patients with fulminant pseudomembranous colitis but almost five times higher than the overall mortality of patients whom we categorized as having severe disease. In comparison, Talbot et al. defined colitis as severe if stool frequency was more than 10 times a day or associated with abdominal pain, swelling or fever [19]. Using this definition, 15% had severe colitis; mortality rates were not given. The reported mean duration of diarrhoea due to C. difficile ranges from 2 to 10 days [7, 22 24]. We considered diarrhoea or other symptoms, such as abdominal pain, tenderness or distension, lasting longer than this to be important and therefore we included these cases in the severe group. Older age is considered a risk factor for the acquisition of C. difficile [7, 25]. This may be due to age-related changes in the faecal flora, immune system dysfunction or more severe underlying disease [26, 27]. Although the incidence of CDAD was significantly higher in older patients in our study, age was not a risk factor for disease severity. Debility and cognitive impairment at the onset of symptoms, regardless of age, were highly significant predictors for a more prolonged or complicated course. Our findings are similar to those reported by Annan et al. [28] who, in a retrospective review of patients with CDAD, found that age by itself did not influence mortality but that debilitated patients who acquired CDAD had a significantly higher mortality [28]. Other investigators have reported that CDAD in elderly patients is more serious and carries a higher mortality than in younger patients [29, 30]. In these studies functional impairment as an independent risk factor was not considered. Debility and cognitive impairment are common in older patients and the former often reflects severity of underlying disease. Functional dependency is a risk factor for nosocomial infections in general [31], therefore this may be a useful marker for identifying patients at risk for acquisition of C. difficile and determining who may experience prolonged symptoms or severe disease. Almost one in four patients had a gastrointestinal 111

6 L. Kyne et al. endoscopy prior to the onset of disease. Without case controls we can only speculate on the significance of this. To date, endoscopies have not been implicated in the transmission of this organism, but the potential for spread via this mechanism exists unless there is adherence to cleaning protocols [21]. Immersion of endoscopes in 2% alkaline gluteraldehyde solution for 4 min, as recommended by the British Society of Gastroenterology [32], is probably sufficient to inactivate vegetative bacteria [32] but it may take several hours for this disinfectant to be sporicidal [33]. We did not identify any link between patients who underwent gastrointestinal endoscopic examination in our hospital, except that they all attended the endoscopy unit, which is located in an area of the hospital which was heavily contaminated with C. difficile. In this unit gastrointestinal flexible endoscopes are normally disinfected between patients in accordance with the British Society of Gastroenterology recommendations [32] but for high-risk patients or if infection with C. difficile is suspected, the endoscopes are disinfected for 1 h after the procedure. Endoscopy and nasogastric tube insertion/feeding were also associated with more severe disease in this cohort. These are also risk factors for the acquisition of C. difficile [8, 18, 34]. Perhaps their association with disease may reflect greater risk of exposure to the organism via hands of personnel, contaminated environmental sites or inanimate objects [8, 18, 35]. We speculate that these procedures were associated with more severe disease due to greater alteration of the faecal flora by gastrointestinal manipulation, and hence loss of colonization resistance. Alternatively, they may be markers for co-morbid diseases which may be risk factors for severe CDAD. We did not record the indication for, or results of, endoscopic examinations. It is possible that the patients who underwent these procedures had gastrointestinal disease which predisposed them to severe diarrhoea or other abdominal symptoms. In the case of nasogastric tube feeding, prolonged symptoms may have been be caused by other mechanisms of diarrhoea in addition to C. difficile infection, such as osmotic diarrhoea secondary to feeding formulations. The lack of association of disease severity with the number of doses or days of antibiotic treatment is perhaps surprising. C. difficile acquisition can occur after a single antibiotic dose. However, longer and more continuous antimicrobial exposure increases the risk for C. difficile diarrhoea [36, 37]. Our results suggest that antimicrobial therapy, of any duration or intensity, merely serves to alter the faecal flora allowing overgrowth of C. difficile and that excess antibioticuse is not associated with more severe disease. We only considered antimicrobial therapy prior to disease onset. For this reason we have no evidence to suggest that continued antimicrobial use does or does not influence the outcome in affected individuals. Virulence factors of C. difficile (other than strain type and toxin production) may contribute to the spectrum of disease seen in colonized individuals [38]. In addition, the mucosal and systemic immune response of the host may also affect the outcome of acquisition [39]. Unfortunately, the assessment of these factors was beyond the scope of this study. In conclusion, C. difficile is an important nosocomial pathogen which is responsible for an increasing incidence of disease in many hospitals and long-term care facilities. It will probably continue to be a problem, given an increasingly elderly inpatient population and the widespread use of broad-spectrum antimicrobial therapy [25]. During an outbreak we have shown that this organism is responsible for a spectrum of disease in older patients, in whom symptomatic disease was more common, although age was not a predictor of disease severity. Factors associated with more severe disease included debility (lower Barthel/AMT scores) and recent gastrointestinal endoscopy. Nasogastric tube insertion and enteral feeding was also associated with severe CDAD, although not significantly. Awareness of these factors may help in the anticipation of severe CDAD and prompt treatment which in turn may limit morbidity and mortality. Key points Clostridium difficile-associated disease is common in elderly patients. Infection results in a broad spectrum of disease. In a cohort of cases identified during an outbreak, 60% of patients had symptoms which resolved within 10 days, 32% had prolonged symptoms (mean diarrhoea duration 18 days) and 8% had severe colitis. Severe disease was associated with a 67% mortality rate. Age was not a risk factor for disease severity, but debility, cognitive impairment and recent endoscopy were. A trend for more severe disease was associated with nasogastric feeding. Early identification of patients at risk for severe disease may limit morbidity and mortality. References 1. Tabaqchali S, Jumaa P. Diagnosis and management of Clostridium difficile infection. Br Med J 1995; 310: Fekety R, Shah AB. Diagnosis and treatment of Clostridium difficile colitis. JAMA 1993; 269: Jobe BA, Grasley A, Deveney K, Sheppard B. Clostridium difficile colitis: an increasing hospital-acquired illness. Am J Surg 1995; 169: Rubin M, Bodenstein L, Kent K. Severe Clostridium difficile colitis. Dis Colon Rectum 1995; 38:

7 Severity of Clostridium difficile-associated disease 5. Pothoulakis C, LaMont JT. Clostridium difficile colitis and diarrhoea. Gastroenterol Clin North Am 1993; 22: Kelly CP, Pothoulakis C, LaMont JT. Clostridium difficile colitis. N Engl J Med 1994; 330: McFarland LV, Surawicz CM, Stamm WE. Risk factors for Clostridium difficile carriage and Clostridium difficile-associated diarrhoea in a cohort of hospitalised patients. J Infect Dis 1990; 162: Brown E, Talbot G, Axecrod P, Provencher M, Hoegg C. Risk factors for Clostridium difficile toxin-associated diarrhoea. Infect Control Hosp Epidemiol 1990; 11: Thibault A, Miller M, Gaese C. Risk factors for the development of Clostridium difficile-associated diarrhoea during a hospital outbreak. Infect Control Hosp Epidemiol 1991; 12: Zimmerman RK. Risk factors for Clostridium difficile cytotoxinpositive diarrhoea after control for horizontal transmission. Infect Control Hosp Epidemiol 1991; 12: Mahony FI, Barthel DW. Functional evaluation: the Barthel index. Md State Med J 1965; 14: Hodkinson HM. Evaluation of a mental test score for assessment of mental impairment in the elderly. Age Ageing 1972; 1: Fekety R. Antibiotic-associated diarrhoea and colitis. Current Opin Infect Dis 1995; 8: Tedesco FJ, Barton RW, Alpers DH. Clindamycin-associated colitis. A prospective study. Ann Intern Med 1974; 81: Gerding DN. Disease associated with Clostridium difficile infection. Ann Intern Med 1989; 110: Gerding DN, Olsen MM, Peterson LR et al. Clostridium difficileassociated diarrhoea and colitis in adults: a prospective case-controlled epidemiological study. Arch Intern Med 1986; 146: Thomas DR, Bennett RG, Laughon BE, Greenough WB III, Bartlett JG. Postantibiotic colonisation with Clostridium difficile in nursing home patients. J Am Geriatr Soc 1990; 38: Simor AE, Yake SL, Tsimidis K. Infection due to Clostridium difficile among elderly residents of a long-term care facility. Clin Infect Dis 1993; 17: Talbot RW, Walker RC, Beart RW. Changing epidemiology, diagnosis, and treatment of Clostridium difficile toxin-associated colitis. Br J Surg 1986; 73: Prendergast TM, Marini CP, D Angelo AJ, Sher ME, Cohen JR. Surgical patients with pseudomembranous colitis: factors affecting prognosis. Surgery 1994; 116: Gerding DN, Johnson S, Peterson L, Mulligan M, Silva J. Clostridium difficile-associated diarrhoea and colitis. Infect Control Hosp Epidemiol 1995; 16: Bennett RG. Diarrhoea among residents of long-term care facilities. Infect Control Hosp Epidemiol 1993; 14: Nolan NPM, Kelly CP, Humphreys JFH et al. An epidemic of pseudomembranous colitis: importance of person to person spread. Gut 1987; 28: Zaidi M, Ponce de Leon, Oritz RM et al. Hospital-acquired diarrhoea in adults: a prospective case controlled study in Mexico. Infect Control Hosp Epidemiol 1991; 12: Joint Working Group. Report: the prevention and management of Clostridium difficile infection. London: Department of Health and Public Health Laboratory Service, Borriello SP, Barcley FE. An in vitro model of colonisation resistance to Clostridium difficile infection. J Med Microbiol 1986; 21: Nakamura S, Mikawa M, Nakashio S et al. Isolation of Clostridium difficile from the faeces and the antibody in sera of young and elderly adults. Microbiol Immunol 1981; 25: Anand A, Bashey B, Mir T, Glatt A. Epidemiology, clinical manifestations and outcome of Clostridium difficile-associated diarrhoea. Am J Gastroenterol 1994; 89: Mogg GAG, Keighley MRB, Burdon DW et al. Antibioticassociated colitis-a review of 66 cases. Br J Surg. 1979; 66: Lesna M, Parham DM. Mortality due to C. difficile colitis in elderly people has been underestimated (Letter). Br Med J 1996; 312: Hussain M, Oppenheim BA, O Neill P, Trembath C, Morris J, Horan MA. Prospective survey of the incidence, risk factors and outcome of hospital-acquired infections in the elderly. J Hosp Infect 1996; 32: British Society of Gastroenterology Endoscopy Committee. Aldehyde disinfectants and health in endoscopy units. Gut 1993; 34: British Society of Gastroenterology. Cleaning and disinfection of equipment for gastrointestinal flexible endoscopy: interim recommendations of a working group. Gut 1988; 29: Pierce PF Jr, Wilson R, Silva J Jr et al. Antibiotic-associated pseudomembranous colitis: an epidemiological investigation of a cluster of cases. J Hosp Infect 1982; 145: Samore MH, DeGirolami PC, Tlucko A, Lichenberg DA, Melvin ZA, Karchmer A. Clostridium difficile colonisation and diarrhoea at a tertiary care hospital. Clin Infect Dis 1994; 18: Ambrose NS, Johnson M, Burdon DW, Keighley MR. The influence of single dose intravenous antibiotics on faecal flora and emergence of Clostridium difficile. J Antimicrob Chemother 1985; 15: Nelson DE, Auerbach SB, Baltch AL et al. Epidemic Clostridium difficile-associated diarrhoea: role of second and third generation cephalosporins. Infect Control Hosp Epidemiol 1994; 15: Borriello SP, Davies HA, Kamiya S et al. Virulence factors of Clostridium difficile. Rev Infect Dis 1990; 12 (suppl. 2): S Kelly CP. Immune response to Clostridium difficile infection. Eur J Gastroenterol Hepatol 1996; 8: Received 10 February 1998; accepted 10 March

8 Sally and Richard Greenhill.