Previously identified risk factors for acquiring C. difficile include the administration of antibiotic agents (especially cephalosporins, ampicillin, and clindamycin); duration of hospitalization (a surrogate marker for nosocomial acquisition and severity of underlying illness); and abdominal surgery [2, 8-12]. The clinical signs of C. difficile disease may include fever, leukocytosis, and systemic illness, in addition to diarrhea [3, 7]. Despite these retrospective clinical observations, no prospective study has quantified which clinical and laboratory features of an illness best predict the likelihood of C. difficile disease. Nor has any study determined the optimal number of stool samples required to exclude this diagnosis. Standard practice at some hospitals has been to send three stool samples per patient for evaluation, but no published data support this approach.

We did a prospective study of hospitalized adult patients to 1) define readily available clinical and laboratory parameters that can help predict the presence of C. difficile disease and 2) assess the effectiveness of laboratory tests for C. difficile toxin in excluding disease related to C. difficile.

Methods

Top Methods Results Discussion Author & Article Info References

Patients and Specimen Collection

Our study was approved by the Johns Hopkins Medical Institutions Human Subjects Research Committee. From 1 January to 1 March 1993 and from 1 May to 1 July 1993, 268 consecutive hospitalized adult inpatients who had a stool specimen sent to the Johns Hopkins Hospital Microbiology Laboratory for C. difficile toxin assay were enrolled in the study. Patients were enrolled only if they had diarrheal stool (we accepted only unformed stool samples; rectal swabs were excluded) and if they had been on medical and surgical inpatient services for more than 24 hours. We defined a stool specimen as diarrhea if it took the shape of its container; no effort was made to quantify stool output. Patients were excluded if they were on a pediatric service, if they were in the Johns Hopkins Oncology Center, if they were discharged within 24 hours of stool specimen acquisition, or if they had had C. difficile toxin in a stool sample within the previous 3 months. Oncology patients on the general medical service were included because chemotherapy is only administered in the Johns Hopkins Oncology Center (chemotherapy can induce, worsen, or mimic C. difficile disease and thus would confound our results when they were applied to a general hospital population). Of the original 347 patients, 268 consecutive patients met the study criteria.

Three stool specimens were to be obtained from each patient at intervals of at least 12 hours. Floor nurses had in-service education about the purpose of the study, proper specimen collection, and the importance of obtaining three stools, and investigators returned to encourage patients and nurses to send all specimens. Despite these efforts, the targeted number of stools were not sent for some patients [see Results] for the following reasons: 1) the patient no longer produced unformed stool, 2) the patient was noncompliant with collection; or 3) a positive specimen on the first or second stool resulted in treatment.

Definition of Clostridium difficile Disease

Patients were considered to have disease related to C. difficile if they had both diarrhea and the presence of C. difficile toxin by either EIA or the in vitro cytotoxin tissue-culture assay. In our experience, neither EIA nor the tissue-culture assay yields false-positive results.

Clinical Evaluation of Patients

All patients enrolled in the study were interviewed and examined, and their medical records were reviewed according to a standardized protocol. All clinical evaluations were done by one of two of the authors, who had no knowledge of the cytotoxin assay results. Demographic data, including age, race, sex, hospital floor and service, primary and secondary diagnoses, vital signs, antibiotic history, and pertinent laboratory test values, were recorded. Patients were specifically asked about onset, character, and frequency of diarrhea and about the presence or absence of pain. If patients were unable to provide information, primary nurses or family members, or both, were interviewed. Physical examination of the abdomen was done.

Laboratory Evaluation of Stools

A total of 692 stool specimens were studied. Laboratory technicians recorded the color of the stool (brown, clear, yellow, or green) and determined whether the stool specimen was formed (excluded from study), semiformed (solid material in liquid), or watery (no solid material). The stool specimens (< 24 hours from the time of collection) were refrigerated at 4 °C until they were tested; an aliquot of each sample was also frozen at –70 °C for future testing. Testing was done within 24 hours of the time the specimens were received in the laboratory. Smears for fecal leukocytes were made at the time the specimens were processed for toxin assay.

Enzyme-linked immunoassay for cytotoxins A and B was done using the Cambridge Biotech Cytoclone A + B EIA kit (Cambridge Biotech, Worcester, Massachusetts) according to the manufacturer's instructions.

The cytotoxin tissue-culture assay was done on human foreskin fibroblast monolayers (in-house propagation of ATCC Hs68 [American Type Culture Collection, Rockville, Maryland]). The frozen stool specimen was thawed and then diluted 1:5 in phosphate buffer (pH, 6.9), vortexed, and centrifuged (4100 rpm in a Jouan CT422 centrifuge) for 30 minutes at 4 °C. The supernatant was filter-sterilized (0.45-µm pore Millipore filtration unit [Millipore Products, Bedford, Massachusetts]), and the filtrate was diluted 1:2 and 1:10. The wells of a 96-well microtiter plate containing the cell line were inoculated with 20 µL of each of the two dilutions, making final dilutions of stool in the wells of 1/100 and 1/500, respectively. Plates were incubated in 5% CO₂ at 37 °C and were examined at 24 and 48 hours. A positive result was defined as at least 10% fibroblast rounding that could be neutralized by the addition of C. difficile antitoxin (Ana-Tox, Inc., Blacksburg, Virginia) during retesting at the dilutions described above. When the neutralization was done, the original specimens were tested again.

The presence of fecal leukocytes was determined using both light microscopy and an in vitro test for lactoferrin, an enzyme that is a marker for degraded polymorphonuclear leukocytes. Thin smears of stool were prepared, heat-fixed, and stained with Gram stain. Twenty fields (times 100) were examined to determine the presence of polymorphonuclear leukocytes.

The TechLab Fast'N'Flammatory Test (Leuko-Test, Techlab, Blacksburg, Virginia) was used to determine the presence of lactoferrin. The kit consists of latex beads coated with polyclonal antibody to lactoferrin. The test was done according to the manufacturer's instructions. Latex agglutination was recorded as present or absent.

Statistical Evaluation

The chi-square test was used to assess the univariate statistical association of demographic and clinical variables with C. difficile disease. Logistic regression was used to assess the multivariate statistical associations of these variables with C. difficile disease. A backward stepwise method was used for the logistic regression, starting with all variables that showed a univariate association with C. difficile of P < 0.50. We used SAS software (SAS, Inc., Cary, North Carolina) for the statistical analysis.

No gold standard exists for diagnosing C. difficile disease, but, for the purposes of our sensitivity analyses, finding toxin at any time on either EIA or tissue-culture assay was considered to indicate the presence of C. difficile disease.

Results

Top Methods Results Discussion Author & Article Info References

Study Sample

During two 2-month periods in 1993, we enrolled 268 patients in our study (Table 1). Fifty-one percent of the patients were women, and 59% were white. The mean age of the entire group was 52.8 years; the mean age of those persons who had C. difficile toxin in their stool was 43 years. One hundred sixty-seven patients were on internal medicine and neurology services, 95 were on surgical services, and 6 were on the gynecology (but not the obstetrics) service. Patient diagnoses were divided into broad groups Table 1 and were similar to those of the population of a tertiary care teaching hospital. No single diagnostic group dominated. The diagnoses of those patients whose stools were positive for C. difficile toxin were not significantly different from those of the overall study sample (data not shown). However, patients with pulmonary diseases and patients seropositive for the human immunodeficiency virus were relatively overrepresented in the group that was positive for C. difficile.

Of the 268 patients enrolled, 64 (24%) each had two stool specimens submitted; 98 (37%) each had three specimens submitted; and 106 (40%) each had only 1 specimen submitted. The specific reasons for failure to obtain three stool specimens from each patient were not recorded; generally, failure occurred because patients were unable to produce successive stool specimens. Forty-three patients were positive for toxin by either EIA or tissue-culture assay. Thirty patients were positive by both assays, 9 were positive by EIA alone, and 4 were positive by tissue-culture assay alone. The overall prevalence of toxin positivity was 16.4%.

Clinical and Laboratory Predictors of Positivity for Clostridium difficile Toxin

The most important univariate clinical and laboratory predictors of positivity for toxin Table 2 were cephalo-sporin use, prolonged hospital stay (> 15 days), onset of diarrheal symptoms 6 or more days after the administration of antibiotic agents, the presence of fecal leukocytes, and the presence of semiformed (as opposed to watery) stool. The presence of fecal leukocytes was highly predictive of the presence of C. difficile toxin when assessed by both the lactoferrin assay and the traditional fecal leukocyte stain (P = 0.001). The presence of fecal leukocytes by microscopy, although less sensitive for predicting C. difficile toxin, suggested more severe clinical disease. Excluding the diagnostic category of abdominal surgery (because persons in this group could have had a high fecal leukocyte count without having C. difficile disease) still resulted in a strong association with the presence of C. difficile toxin (P = 0.004).

Of the broad range of antimicrobial agents administered to study patients, only cephalosporins were found to be a significant predictor of C. difficile disease (P = 0.02). Penicillin and its derivatives, clindamycin, aminoglycosides, ciprofloxacin, trimethoprim-sulfamethoxazole, and oral and intravenous metronidazole were not associated with increased detection of C. difficile toxin. Stool color, abdominal cramps or tenderness, leukocytosis, fever, patient recognition of gross blood, and the presence of occult blood were not associated with the presence of toxin. Onset of diarrhea 6 or more days after the initiation of antimicrobial therapy predicted the presence of C. difficile toxin (P = 0.03). Length of stay longer than 15 days was also a significant independent predictor (P = 0.03). We found no clustering of cases by time, hospital floor, service, or diagnosis.

Multivariate logistic regression analysis showed that the presence of fecal leukocytes as shown by either lactoferrin assay or microscopy (Gram stain), the administration of cephalosporin, and the presence of semiformed stool were independently associated with C. difficile toxin positivity (Table 3). No other variables were associated after we adjusted for these four variables.

Sensitivity of Successive Stool Specimens

Of the 43 patients who had C. difficile toxin, 34 were diagnosed from the first stool specimen, 5 were diagnosed from the second specimen, and 4 were diagnosed from the third specimen. All four patients who had toxin detected only in the third stool specimen had at least 4 days elapse between the collection of the second and third stool specimens. Overall, in the 268 patients, the first stool was only moderately useful for predicting the presence of C. difficile toxin: If EIA assay alone was used, only 72% of persons were diagnosed on the first stool specimen; if either EIA or tissue-culture assay was used, 81% of persons were diagnosed on the first stool specimen (Table 4). Of the 225 patients who were negative for C. difficile toxin, none became symptomatic from diarrhea or had sequelae from C. difficile disease during short-term follow-up, suggesting that no clinically manifest cases were missed.

Discussion

Top Methods Results Discussion Author & Article Info References

In this prospective study of hospital inpatients, we combined readily available clinical and laboratory data to predict the likelihood of C. difficile disease. We also evaluated the number of stool samples necessary to exclude the diagnosis of disease related to C. difficile. Clostridium difficile is a treatable pathogen and an important cause of nosocomial and iatrogenic disease. Consequently, it is important to be able to establish its presence quickly. Efforts to contain cost by minimizing the number of tests done must be weighed against the morbidity and mortality caused by undiagnosed but treatable C. difficile disease. We show that the combination of focused clinical evaluation and simple, inexpensive, readily available laboratory data can enhance the clinician's ability to both exclude and diagnose disease caused by C. difficile.

All patients in this study in whom C. difficile toxin was detected had had recent antibiotic exposure; this is consistent with previous observations. Clinical and laboratory predictors of positivity for toxin were the presence of fecal leukocytes (assessed by microscopy or lactoferrin assay), semiformed (as opposed to watery) stools, use of cephalosporin, and the onset of diarrhea more than 6 days after the initiation of antibiotic therapy. The finding that semiformed stool was a predictor of C. difficile colitis was unexpected and may simply suggest that the diarrhea is due to the lack of colonic water absorption caused by the cytolytic action of the toxin on the colonic epithelium rather than to the presence of a secretory diarrhea. The first stool specimen was positive for toxin in almost all patients with C. difficile colitis and two or more predictors; this probably reflects a somewhat higher disease prevalence rather than greater true test sensitivity in this subgroup. Interestingly, the four patients who had C. difficile toxin detected only in the third stool specimen had at least 4 days elapse between the collection of the second and third specimens. Thus, the third specimens that were positive may actually have been the first specimens of a new complication.

Several risk factors for C. difficile disease have been found [2-12], but most of the epidemiologic studies in which they were established were retrospective [8, 9, 13] or laboratory-based. These studies did not prospectively follow patients clinically despite the prospective laboratory assessment [10, 11], and many previous studies have focused on nosocomial outbreaks rather than on sporadic disease [8, 9, 14]. Furthermore, although the cytotoxin tissue-culture assay is often considered the gold standard for identifying C. difficile toxin in stool specimens, no gold standard exists for determining disease caused by C. difficile on the basis of clinical criteria, either with or without endoscopy. Short-term follow-up of all patients in our study did not produce any additional cases of positivity for C. difficile toxin, and no cases of pseudomembranous colitis occurred that required surgical intervention. This finding suggests that our study design and the in vitro assays we used to detect C. difficile did not miss any clinically manifest cases of C. difficile disease. Our study is unique because it combined clinical evaluation with two optimal toxin detection methods in multiple specimens.

We used the submission of stool specimens to a clinical laboratory as the primary method of selecting patients for the study. We intended the study to be inclusive and to reflect current clinical practice patterns. In general, specimens were submitted from patients who had diarrhea accompanied by current or recent exposure to antibiotics. The incidence of positivity was 16.4%, which is similar to that seen in a study done at our institution in 1993 [6]. Analysis of the hospital locations of patients who had C. difficile toxin in their stool showed no point outbreaks of disease (data not shown). Thus, our data appear to reflect sporadic, rather than epidemic, disease.

Assessment of the presence of fecal leukocytes is an inexpensive, simple, and predictive adjunct to the cyto-toxin assays. This applies especially to the lactoferrin assay, which appears to improve the sensitivity of detection of leukocytes in the stool and also may be done on stored specimens or swabs [15, 16]. In our study, we separately analyzed the microscopic and the lactoferrin examinations for fecal leukocytes because no study has directly compared the two tests in combination with the prospective clinical diagnosis of C. difficile colitis. We found, in agreement with others [17], that the lactoferrin test is moderately predictive of, but not very specific for, the presence of C. difficile toxin. Nevertheless, the presence of fecal leukocytes is obviously not a sensitive indicator for the presence of C. difficile disease; in most studies, substantially fewer than half of patients with documented C. difficile-associated enteric disease are found to have fecal leukocytes by microscopy. In addition, the lactoferrin assay has been shown to be positive more often in symptomatic than in asymptomatic carriers of C. difficile (4% compared with 40%; P < 0.01) [5, 18] and is probably at least as sensitive as microscopy for detecting polymorphonuclear leukocytes in the stool [17].

Previous reports have associated C. difficile disease with ampicillin and clindamycin, as well as with cephalosporins, but the use of these agents did not predict the presence of C. difficile in our study. We presume that the reason for this difference is our lack of statistical power to detect the effect of these antibiotics; this lack was caused by the infrequent use of these agents. Most studies show that cephalosporins are the most frequent offending agents in antibiotic-associated colitis caused by C. difficile. This reflects high rates of use; the incidence of this colitis with clindamycin is much higher, and rates are as high as 10% [14].

Certain aspects of our study design must be considered when interpreting our results. We studied consecutive patients in whom clinicians suspected disease instead of entering all hospital inpatients into the study. Notably, the incidence of positivity for C. difficile toxin was similar to the overall toxin positivity rate reported by the laboratory for 1993. We believe that our data apply to other hospitalized populations who have signs and symptoms similar to those seen in our study sample.

We used both EIA for cytotoxins A and B and the cytotoxin tissue-culture assay. However, although EIA was done within 24 hours of the collection of each stool specimen, the tissue-culture assay was done later on batched, frozen specimens. Nonetheless, there was a high concordance between positivity on EIA and positivity on tissue-culture assay. The tissue-culture assay was still more sensitive, even when testing frozen stored specimens, than EIA; it detected an additional 10% of patients who were negative on EIA. Freezing stool specimens has been reported to have little effect on the yield of cytotoxin assays [15, 16]. Our results are consistent with the known relative sensitivities of the two tests [6].

The efficient diagnosis of C. difficile disease relies on the combination of clinical suspicion and the laboratory detection of toxin. We provide the clinician at the bedside with specific criteria for suspecting C. difficile colitis in patients and for directing an efficient work-up of those patients.

Dr. Moore: Johns Hopkins School of Medicine, Division of Internal Medicine, 1830 East Monument Street Room 8059, Baltimore, MD 21205.

Ms. Merz: Johns Hopkins Hospital, Clinical Microbiology Laboratory, Meyer B1-191, 601 North Wolfe Street, Baltimore, MD 21287.

Dr. Charache: Johns Hopkins School of Medicine, Department of Pathology, Carnegie 422, 601 North Wolfe Street, Baltimore, MD 21287.