Toxin-positive Clostridium difficile cases per 10 000 hospital admissions per year (1984-2000). Linear regression; R2 = 0.27. 1993 was an anomalous year, with 45 cases total. When this point was excluded from analysis, the slope remained positive but decreased by 12%. R2, a measure of linear regression fit indicating accuracy of the plotted line, increased more than 2-fold.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Morris AM, Jobe BA, Stoney M, Sheppard BC, Deveney CW, Deveney KE. Clostridium difficile Colitis: An Increasingly Aggressive Iatrogenic Disease? Arch Surg. 2002;137(10):1096–1100. doi:10.1001/archsurg.137.10.1096
The diagnosis of Clostridium difficile colitis is increasing in frequency, with worsening patient outcomes.
Retrospective cohort study.
One hundred fifty-seven patients diagnosed with C difficile colitis between 1994-2000.
Main Outcome Measures
Resolution of disease, operative intervention, and death.
Compared with our previous 10-year experience, overall cases of C difficile colitis have risen by more than 30%, and immunocompromised patients comprise a larger proportion of those affected. One third of patients were receiving posttransplantation medication, chemotherapy, or had human immunodeficiency virus. Of these, 2 (4%) of 51 required surgical intervention and 10 (20%) of 51 died. An additional 18.5% of patients had diabetes, renal failure, or both. Of these, 2 (7%) of 30 required surgery and 4 (13%) of 30 died. Only 9.5% of patients had prophylactic perioperative antibiotics as a sole risk factor; 2 (13%) of 15 required surgery and 3 (20%) of 15 died. The overall mortality rate was 15.3%, increased from 3.5% in our previous series. Neither need for surgery nor mortality differed among these patient groups.
The frequency of C difficile colitis remains high and seems to be associated with increasing mortality. Among patients with positive C difficile toxin assay results, immunocompromise and delayed diagnosis no longer seem to be associated with higher risk for death. All patients taking antibiotics are at risk and require early recognition and aggressive medical intervention.
THE ASSOCIATION between antibiotic use and pseudomembranous colitis was suggested as early as 1952 by Reiner and colleagues,1 and widely accepted by 1974 when Tedesco and others2 prospectively studied the colonic sequelae of clindamycin use. Bartlett and associates3 subsequently identified Clostridium difficile as the offending pathogen. These investigators were further able to demonstrate that a bacterial toxin rather than a viral or direct bacterial process mediated antibiotic-associated colitis. Since then, accumulated evidence has established antibiotic use as the primary risk factor for pseudomembranous colitis and opportunistic C difficile infection as the cause of antibiotic-associated colitis, regardless of pseudomembrane formation.4-6
Although the paucity of prospective or population-based studies has obscured the true incidence of this disease, recent reports indicate increasing overall case numbers.7-10 Furthermore, recent case series and anecdotal reports of more debilitating disease course, increasing frequency of need for operative treatment, and higher mortality rates suggest that the clinical course of the disease may be changing.11-13
To determine whether cases in our institution have been exhibiting similar trends in frequency and outcome of disease, we examined patients with positive toxin assay results and compared their characteristics and outcomes with those from an earlier study at the same hospital.7
We conducted a retrospective review of all patients diagnosed with C difficile colitis by hospital discharge codes during 6 years, excluding those who had a negative toxin assay result (toxin A– and B–specific enzyme immunoassays [EIA]; Biowhittaker-Cambridge Biotech Corp, Worcester, Mass). The remaining cases were compared with patients in a previous 10-year study in our hospital. Because the number of cases per year can be affected by the size of the population at risk, we determined the total number of patients admitted during the years of interest.
Patient and treatment characteristics were recorded, including age, sex, number of hospitalization days preceding the onset of symptoms, indication for antibiotics, and specific antibiotics taken. Potential outcome predictors, treatment modalities, and main outcome measures were noted. Main outcome measures included resolution of disease, need for operation, and death without resolution of disease and within 90 days of diagnosis.
Outcome predictors were chosen based on historical association with increased risk for C difficile colitis and increased C difficile– related mortality, and included: (1) immunocompromised state (receiving posttransplantation medications, undergoing chemotherapy, or having human immunodeficiency virus [HIV]–positive status)8,14 (2) comorbid illness (renal disease or diabetes mellitus [DM]),8 and (3) time from symptoms to treatment.7,15 We also recorded service of origin, cross-referenced with indications for antibiotics, to study surgical patients receiving prophylactic antibiotics, who previously comprised a large proportion of patients infected with C difficile in our institution.
A 2-sample t test and ϑ2 analysis for noncontinuous variables were used to identify statistically significant associations between predictor and outcome variables. The Fisher exact test was used to confirm results for smaller sample sizes. Statistical analysis was performed using STATA statistical software package (StataCorp, College Station, Tex).
From July 1994 to June 2000, 157 patients hospitalized at Oregon Health Sciences University Hospital (OHSU) (Portland) were diagnosed with C difficile infection by discharge diagnostic codes and positive C difficile toxin assay results. This represents a 30.2% increase in average yearly cases for these 6 years compared with the previous 10 years at OHSU. The previously reported increase in cases has persisted, based on linear regression of cases/10 000 hospital admissions per year (Figure 1).
Patient and disease characteristics are presented in Table 1. The mean age at diagnosis was 46.8 years (range, 0.5-91 years). Mean number of hospital days at onset of symptoms was –2.3. Eighty-nine percent of patients' symptoms began after antibiotic use. Among the 11% of patients who received no antibiotics prior to symptoms, 12 (67%) of 18 were immunocompromised by medication or HIV infection. Overall, 33% of patients were immunocompromised, representing a substantial increase from our earlier study, which cited a 20% prevalence of immunocompromise among patients with positive toxin assay results. Among the patients who were not otherwise immunocompromised, 18.5% had renal disease, DM, or both.
Surgical services continued to account for the largest proportion of patients with a positive toxin assay result, but the percentage of surgical patients decreased from 55% previously to 40% in the current patient cohort. Surgical patients whose sole risk factor for C difficile colitis was prophylactic perioperative antibiotic treatment comprised only 9.6% of the total cohort. Patients in this group were of particular interest because their infections were iatrogenic, resulting from a decision to use prophylactic antibiotics perioperatively and not from treatment of a known or suspected infection.
Table 2 presents the relationship between antibiotic indications and outcomes. Eighty-nine percent of patients whose C difficile infection was not preceded by antibiotic use recovered with medication only. Among patients whose colitis was preceded by antibiotic use, indications for antibiotic use (prophylaxis, empiric treatment, and documented infection) seemed to have no association with outcome; rates of resolution without operation were 74%, 78%, and 80%, respectively, and mortality rates were 17%, 23%, and 16%, respectively.
Overall, 15.3% of patients in the current cohort died compared with 3.5% in the previous cohort. We were unable to identify a correlation between traditionally accepted predictors of operative intervention or mortality and either of these outcomes (Table 3). Contrary to our previous findings, length of time between symptom onset and treatment had no significant association with mortality. Patients who survived had a mean duration of 8.3 days between symptom onset and treatment, while those who died had a mean duration of 9.5 days (P =
.61). All patients who underwent operative intervention had an exploratory celiotomy, subtotal colectomy, and ileostomy formation.
Among the one third of patients who were immunocompromised, 2 (4%) of 51 required surgery and 10 (20%) of 51 died. Neither need for surgery nor mortality differed statistically from that in nonimmunocompromised patients. Among the 18.5% of patients who had renal disease and DM, 2 (7%) of 30 required operative intervention and 4 (13%) of 30 died. Again, neither outcome was significantly different from that in patients without chronic comorbid disease (renal disease/DM). Patients on surgical services are presumed to be at higher risk for C difficile infection because of the higher rate of antibiotic use for surgical incision prophylaxis but at lower risk for poor outcomes due to less comorbidity. Among the 10% of patients with prophylactic perioperative antibiotics as a sole risk factor, 2 (13%) of 15 required operation and recovered, while 3 (20%) of 15 died. Once again, neither outcome was statistically different from that of patients on nonsurgical services.
For nearly 25 years, C difficile has been recognized as the major cause of antibiotic-associated colitis, manifested by disease ranging from self-limited diarrhea to pseudomembrane formation to cytokine-mediated cardiovascular collapse and death.2-4,6 Although incompletely understood, the cause of this illness remains strongly linked to antibiotic use and nosocomial transmission. Moreover, recent reports suggest that the incidence of C difficile colitis is increasing throughout the United States and Great Britain.7,9-11,16 Given these findings and this infection's potentially iatrogenic origin, we were compelled to investigate recent changes in the disease course within our institution.
In our recent series, 157 patients had positive C difficile toxin assay results. It must be noted that the previous cohort's more rapid rise in yearly cases may have been partially due to improved or increasingly frequent testing during that 10-year period. The number of cases per year has been increasing at a much slower rate or remaining stable more recently, perhaps reflecting saturation of provider awareness regarding C difficile or stability of test kit usage (a single toxin A– and B–specific EIA throughout the 6-year period). Alternatively, the current slowed rate of increase may be due to yearly variation and a smaller sample size (Veterans Affairs hospital patients not included).
In previous reports, length of hospital stay has been associated with increased risk for C difficile colitis, presumably due to prolonged exposure to contaminated environmental surfaces by more susceptible patients.15,16 Among our current cohort, the mean hospital day when the first symptoms appeared was actually 2 days prior to admission, reflecting the development of symptoms in the community setting. Because more than 89% of patients had been under the recent care of a physician, we postulate that this timing simply underscores a current trend toward fewer overall hospital days per patient. Additionally, we found that 58% of patients were treated with multiple antibiotics. The most commonly used class of single-agent antibiotics continued to be cephalosporins (all generations).
We examined the relationship between indications for antibiotic use and outcomes, anticipating that those patients who developed C difficile infection after receiving antibiotics for prophylaxis would fare better than those undergoing actual treatment. Unexpectedly, the same percentage of patients in each category required operative intervention or died. Considering the length of hospitalization and antibiotic indication data together, it would seem that antibiotic use, whether prophylactic or for treatment, is a more important risk factor for disease and potentially poor outcome than horizontal transmission via exposure to contaminated surfaces.
The overall mortality rate among patients diagnosed with C difficile colitis in our institution has increased from 3.5% to a disquieting 15.3%. Although this increase may reflect hospitalization of only the sickest patients, we sought to identify patient-related factors predicting mortality. Multiple studies have described an association between immunocompromise and susceptibility to this opportunistic organism as well as poorer outcomes of disease.6,8,9,14 Although the proportion of patients in our cohort who were immunocompromised increased from 20% to 33%, their frequency of operation and mortality (3.9% and 19.6%, respectively) did not statistically differ from patients who were not immunocompromised.
Similarly, DM has been linked to increased risk for C difficile colitis and renal disease has been associated with increased mortality in those infected.6,9 After analysis of the immunocompromised subgroup, an additional 18.5% of our patients had DM, renal disease, or both. Four (13%) of 30 of these patients died and 2 (7%) of 30 required an operation; again, these patients did not differ significantly from their counterparts without renal disease or DM.
We were particularly interested in patients cared for by surgical services, which comprised the largest service subgroup (40% of patients). Typically, patients undergoing surgical procedures receive antibiotics, even when no infection is present, for wound-infection prophylaxis; however, this also renders them susceptible to C difficile colitis.5 Accordingly, the Surgical Infection Society has established guidelines to mitigate inappropriate antibiotic use while optimizing prevention of surgery-associated infection.17 Our data do not allow an accurate determination of the appropriateness of antibiotic use but we were able to identify those surgical patients whose sole risk factor was the receipt of perioperative antibiotics (ie, those without documented infection, immunocompromise, DM, or renal disease). These patients made up only 10% of our overall cohort but represented the patient group in whom C difficile colitis may have been the most avoidable. Among the patients who received prophylactic perioperative antibiotics, 2 (13%) of 15 required operation and another 3 (20%) of 15 died. Once again, the patients in this subgroup did not seem to differ statistically in adverse outcomes from their counterparts but this comparison is limited by low power due to small sample size.
The high mortality rate among our cohort may imply that additional factors are influencing mortality. Our study is limited by its retrospective design, leading to a lack of comparable serologic, immunologic, and histologic data. Recent evidence indicates that, although there is no host immunity against colonization, patients who develop higher levels of antitoxin A immunoglobulin G antibody after colonization experience fewer symptoms.18 Future investigations might include stimulation of antibody production and assessment of its effect on C difficile– associated morbidity and mortality.
Conversely, changing phage-specific characteristics may be responsible for worsened outcomes. Moncrief and colleagues19 found multiple strains of toxin A–negative, toxin B–positive C difficile that eluded detection by toxin A laboratory tests. Furthermore, Cohen and colleagues20 mapped gene loci for toxin A and B as well as 3 newly identified toxins, C, D, and E, of uncertain pathogenicity. While our laboratory performs tests for both toxins A and B, thereby reducing the risk of false-negative results (sensitivity = 89%, specificity = 99%), it does not culture C difficile to examine phage type. Further investigation of the exact toxins elaborated and their pathogenicity would be a fruitful area for future research.
In summary, the mortality rate for C difficile colitis has increased compared with the previous 10 years. Moreover, patients with minimal risk factors fared as poorly as those with profound immunocompromise or chronic diseases, implying that host-specific risk factors were less important than in previous studies. Antibiotic use continues to precede nearly all cases of C difficile colitis, but now patients are experiencing initial symptoms in the outpatient setting. Given the progressive increase in mortality after C difficile infection, even patients not traditionally expected to be at high risk for poor outcomes are best served by judicious antibiotic use as well as a high index of suspicion for disease when symptoms occur.
This paper was presented at the 2001 Annual Meeting of the American Society of Colon and Rectal Surgeons, San Diego, Calif, June 4, 2001.
Corresponding author and reprints: Arden M. Morris, MD, Robert Wood Johnson Clinical Scholars Program, Box 357183, H220-HSB, University of Washington, Seattle, WA 98195-7183 (e-mail: email@example.com).