Clinical Application of Polymerase Chain Reaction to Diagnose Clostridium difficile in Hospitalized Patients With Diarrhea

CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2004;2:669 674 Clinical Application of Polymerase Chain Reaction to Diagnose Clostridium difficile in Hospitalized Patients With Diarrhea MICHAEL S. MORELLI, SUSAN D. ROUSTER, RALPH A. GIANNELLA, and KENNETH E. SHERMAN Division of Digestive Diseases, University of Cincinnati, Cincinnati, Ohio Background & Aims: Clostridium difficile is a common cause of diarrhea in hospitalized patients and is associated with significant morbidity and cost. The current diagnostic standard, enzyme immunoassay (EIA), has low sensitivity, leading to duplicate testing and empiric treatment. We sought to show the usefulness and potential cost effectiveness of polymerase chain reaction (PCR) amplification of toxin B gene for diagnosis of C. difficile induced diarrhea. Methods: A total of 148 stool samples from academic and community-based hospitals were sent for EIA testing and were evaluated prospectively for the presence of toxin B gene by PCR. Results were compared with EIA regarding sensitivity, specificity, and predictive values. Medical charts were reviewed to determine the following: (1) number of EIAs sent per admission, (2) number sent within a 24-hour time period, and (3) how caregivers practiced based on EIA results. Results: The mean age of 130 patients was 55 years. EIA and PCR were positive in 6.8% and 13.6% of patients, respectively. EIA sensitivity was 40%, specificity was 98%, and positive and negative predictive values were 80% and 91%, respectively. The cost of the PCR was $22/sample. Empiric treatment for C. difficile was given unnecessarily in 42% of EIA-negative results. Thirty percent of patients had 3 or more EIAs sent during their hospital admission. Of patients with multiple samples sent, 57% had more than 1 sample sent in a 24-hour period. Conclusions: Many physicians do not conform to practice guidelines regarding recommended diagnosis and empiric treatment of C. difficile. Toxin B gene PCR represents a more sensitive and potentially cost-effective method to diagnose C. difficile induced diarrhea than EIA and should be considered for use as an alternative diagnostic standard. Clostridium difficile is a common cause of diarrhea in hospitalized patients, accounting for roughly 10% 24% of cases of antibiotic-associated diarrhea. 1 Its prevalence among hospitalized patients is estimated to be between 0.4% and 1%, the majority of cases are acquired in the hospital. 2 The financial burden and morbidity associated with the development of C. difficile infection are enormous. Both case-control and cohort studies showed increased lengths of stay (4 20 days) and increased cost per case of C. difficile. The total cost of the disease in the United States has been estimated to be in excess of $1.1 billion per year. 3,4 Multivariate analyses of case control and cohort studies show that risk factors associated with the development of C. difficile diarrhea include age 65, residence in an intensive care unit, antibiotic use for 10 days, the use of enteral feeding, and a CD4 count of 50/mm. 5 The clinical diagnosis of C. difficile is difficult because the disease presentation can range from asymptomatic colonization to fulminant colitis and toxic megacolon. 6 Because of this, several methods of stool analysis including stool culture, cytotoxic assays, and enzyme immunoassay (EIA) toxin testing have been used to help establish the diagnosis. Stool culture suffers from a lack of specificity for toxigenic C. difficile strains, whereas the cytotoxic assay, the current diagnostic standard, suffers from high cost, the need for specialized equipment, and prolonged turn-around time. Most facilities currently use EIA for identification of toxins A and/or B testing for diagnosis of C. difficile, which is less expensive and easier to perform. However, EIA testing has decreased sensitivity compared with the cytotoxic assay and culture (range, 65% 75%). 7 It has therefore become common clinical practice to send more than one stool sample for EIA to increase the yield of the test. 8 This leads to increased cost and time necessary to ascertain a C. difficile diagnosis. In addition, empiric treatment for C. difficile has become increasingly prevalent. In recent years, a polymerase chain reaction (PCR)- based test for C. difficile has been proposed. PCR amplification of the gene segment encoding for toxins A and/or B avoids the detection of patients who simply are colonized with organisms that do not contain the gene sequence responsible for toxin production (i.e., organisms that are not toxigenic). This test has been found to Abbreviations used in this paper: EIA, enzyme immunoassay; PCR, polymerase chain reaction. 2004 by the American Gastroenterological Association 1542-3565/04/$30.00 PII: 10.1053/S1542-3565(04)00290-3

670 MORELLI ET AL. CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 2, No. 8 be highly sensitive and specific when compared with cytotoxic assay and stool culture for C. difficile. 9 11 The role of this test in clinical practice has not been well described. The primary goal of this study was to show the potential clinical usefulness and potential cost effectiveness of PCR for the diagnosis of C. difficile in hospitalized patients with diarrhea compared with the standard EIA technique. In addition, the ordering and interpretation of EIA and practice habits based on EIA results in our health system were compared with practice guideline recommendations for testing and treatment of C. difficile. Materials and Methods The proposed study was approved by the institutional review board of the University of Cincinnati. Patients and Practice Sites Stool specimens were obtained from hospitalized patients who had diarrhea sent to the laboratory for EIA testing for the presence of toxins A and B. Samples from adult patients ( 18 yr) prospectively collected over a 4-month period (November 2002 March 2003) were used. Repeat stool specimens from the same patient were eligible for testing but no more than 2 specimens from the same patient were tested for C. difficile by PCR. Practice sites included 5 hospitals representing both academic and community-based settings: University of Cincinnati, Jewish Hospital (Cincinnati), Christ Hospital (Cincinnati), St. Luke s East (Northern Kentucky), and St. Luke s West (Northern Kentucky). DNA Extraction and Testing The central laboratory uses the Premier EIA testing for toxins A and B (Meridian Bioscience, Cincinnati, OH) and follows manufacturer guidelines for test methodology. Stool specimens were stored in a freezer at a temperature of 70 C. Patient name, medical record number, date and time of submission, and stool color were recorded and coded prospectively. Collected samples were evaluated in the research laboratory as follows. Total DNA was extracted from each stool specimen using the QIAamp DNA Stool Mini Kit per manufacturer recommendations (Qiagen Group, Venlo, The Netherlands). Extracted DNA was stored at 20 C before analysis via PCR. The PCR methodology was based on that described by Guilbault et al. 12 wherein 4 L of each eluted sample was used directly for amplification. Primers CDTB1 (5 -GTGGC- CCTGAAGCATATG-3 ) and CDTB2 (5 -TCCTCTCTCTG- AACTTCTTGC-3 ) derived from the nonrepeating portion of the C. difficile toxin B gene 9 were used to amplify a 322-bp fragment. Amplification was performed in a 50- L reaction volume including 0.1 g/ L bovine serum albumin, 0.5 mol/l of each primer, 1 PCR reaction buffer, 200 mol/l of each deoxynucleoside triphosphate, 1.5 mmol/l MgCl 2, and 2.5 U platinum DNA taq polymerase (Invitrogen, Carlsbad, CA) in a PTC-100 thermal cycler (MJ Research, Inc, Watertown, MA). The cycling conditions consisted of an initial denaturation at 94 C for 5 minutes, followed by 40 cycles of 94 C for 45 seconds, 56 C for 45 seconds, and 72 C for 75 seconds, with a final extension at 72 C for 10 minutes. DNA purified from C. difficile strain ATCC 9689 served as a positive control for the amplification. Amplicons generated were visualized under ultraviolet illumination in a 1.4% agarose TBE stained with ethidium bromide. A 100-bp DNA ladder was run alongside the specimens to verify the correct 322-bp size of the DNA amplified. Representative sample DNA was sequenced directly on an ABI Model 377 Prism automated DNA sequencer (Applied Biosystems, Foster City, CA) and the resulting nucleotide sequence was analyzed for homology to the published sequence of the C. difficile toxin B gene. PCR of stool specimens for the gene segment encoding for toxin B was performed without knowledge of corresponding EIA results. Chart Review The medical records of each patient were evaluated by a single chart reviewer (M.M.). The following data were collected: (1) result of EIA of the specific stool later analyzed via PCR, (2) number and result of EIAs sent per patient during their hospital admission, (3) number of EIAs sent per patient within a 24-hour period, and (4) the action of caregivers based on the results of the EIA, with specific attention focused on the use of metronidazole in an appropriate (i.e., after a positive EIA or in a patient about whom there was an index of suspicion for the presence of C. difficile who was medically unstable) or inappropriate (i.e., empiric treatment in patients not ill enough to warrant empiric treatment and/or treatment despite negative EIA result) manner based on practice guideline recommendations. The empiric use of metronidazole or oral vancomycin was determined to have occurred based on the presence of one of these medications on the patient medication list proximate to the time of EIA testing. Chart documentation was used to determine why the medication was selected for use. If other potential uses of these medications were seen, this was noted in the Results section. Patients were excluded if they were less than 18 years of age. Statistical Methods Data were entered into the Statistix 7.0 software program (Analytical Software, Tallahassee, FL). Exact test methodology was used for analysis of nonparametric associations. Correlation was performed using the method of Pearson. All inferential testing was analyzed using a 2-tailed hypothesis with an of 0.05. Results A total of 148 stool samples was collected from 131 unique patients. No more than 2 specimens were analyzed from each patient who had more than one sample studied. One sample was excluded because the

August 2004 PCR FOR DIAGNOSIS OF C. DIFFICILE 671 Table 1. Demographics Age Mean (yr) 55.4 Range (yr) 21 88 Gender Male 67 (51.5%) Female 63 (48.5%) Race White 87 (66.9%) Black 34 (26.1%) Unknown 9 (6.9%) patient was found to be less than 18 years of age. Therefore, a total of 147 specimens from 130 patients and their associated demographics were included in the final analysis. Demographics The mean age of the 130 patients was 55.4 years (range, 21 88 yr). Sixty-seven (51.5%) patients were men and 63 (48.5%) patients were women. Caucasians represented 67% of the population and African Americans comprised 26%, the remaining 7% were unable to be determined after chart review (Table 1). Stool Characteristics Stool color was brown in 62.3% of cases, green in 12.3% of cases, and yellow in 12.3% of cases. In the remaining 10.7% of cases, the stool color was black, red, or not recorded. No association between stool color and EIA or PCR result was noted. EIA and PCR Results EIA results were available in all cases. The EIA toxin assay was found to be positive in 10 stool samples (6.8%) and negative in 137 stool samples (93.2%). PCR amplification for the toxin B gene was found to be positive in 20 samples (13.6%) and negative in 127 samples (86.4%) (Table 2). The presence of the toxin B gene was confirmed via sequence analysis of 1 specimen. One patient who had 2 initial negative EIAs later had a positive test on repeat testing performed within 36 hours of the initial 2 tests. The PCR performed on the positive EIA also was positive. Only one patient had multiple negative EIAs with a positive PCR. By using PCR as the gold standard, the sensitivity of the EIA in this study was 40%. The specificity was Table 2. Toxin Result Versus PCR Result PCR (n 20) PCR (n 127) EIA (n 10) 8 (5.4%) 2 (1.4%) EIA (n 137) 12 (8.2%) 125 (85%) Figure 1. Number of patients vs. effect of EIA result on patient management., Toxin positive;, toxin negative. 98.4%, the positive predictive value was 80%, and the negative predictive value was 91%. The positive likelihood ratio was 25.4, and the negative likelihood ratio was 0.61. No association was found between age, race, or sex, and a positive result on PCR or EIA. Ordering Style and Practice Habits A total of 120 of 130 medical records were available for review. In 13 cases (10 unfound records and 3 unclear or incomplete records), it was unable to be determined if empiric treatment was given. Thus, determination of appropriate or inappropriate treatment was based on data from the 117 patients whose treatment status could be determined. Of these, toxin was not detected by EIA in 110 patients. However, 46 patients (42%) were treated inappropriately (empiric treatment in patients not ill enough to warrant empiric treatment and/or treatment despite negative EIA result). No statistical association between treatment and toxin result was identified (P 0.237). Empiric treatment was given appropriately in one case and inappropriately was not given in 2 cases in which EIA was positive. In 9 patients, metronidazole was prescribed but for indications that were not clearly related to empiric treatment of C. difficile (e.g., aspiration pneumonia, inflammatory bowel disease, pouchitis). Recalculating the data including these 9 patients in the final analysis yielded a nonsignificant P value of 0.107 (Figure 1). Of note, vancomycin was not used for treatment of C. difficile in any of the reviewed cases. Sixty-seven of 130 patients (51.5%) had one stool sample sent for EIA, 23 (17.7%) had 2 sent, 21(16.2%) had 3 sent, and 19 (14.6%) patients had more than 3 samples sent for EIA during their hospital admission. The average number of EIAs sent per patient was 2.2. Excluding the 67 patients who only had 1 toxin sent,

672 MORELLI ET AL. CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 2, No. 8 Figure 2. Number of patients whose stool was sent for EIA in a 24- hour period vs. number of EIAs sent during admission. because these patients obviously only had one sent in 24 hours, yielded 63 remaining patients who had more than 1 EIA sent during their hospital admission. Of these 63 patients, 27 patients (43%) had only 1 sent in a 24-hour time period, and the remaining 36 (57%) had more than 1 sent in a 24-hour period. This includes 5 patients who had 3 or more sent within a 24-hour period (Figure 2). Discussion Antibiotic-associated diarrhea, although very common in hospitalized patients, is attributable to C. difficile only 15% 20% of the time. 13 The source usually is unidentified but is presumed to be secondary to the effect of antibiotics or other medications on the gastrointestinal tract. C. difficile, however, is the most commonly identified cause of infectious diarrhea in hospitalized patients. 14 It is associated with a wide range of reported mortality from 0.6% 83% in observational and case-control studies. 15 17 The financial burden of C. difficile is large and is caused by increased frequency of laboratory testing, extended stay in the hospital, the need for institution of isolation precautions, and costs associated with complications arising from the infection such as dehydration and toxic megacolon. 3 Thus, the diagnosis of this entity is extremely important. The most frequently used modality is the EIA for toxins A and/or B. This has a reported sensitivity ranging from 65% 75%. 7 Published practice guidelines 13 for the diagnosis of C. difficile recommend sending a single stool specimen for EIA testing in anyone who has received antibiotics within the previous 2 months and/or whose diarrhea began 72 hours or more after hospitalization. Follow-up testing is recommended if the initial testing is negative but diarrhea persists and clinical suspicion remains high by sending 1 to 2 more samples of stool for the same or different test. Stool samples should not be sent on the same day because sensitivity for the diagnosis is not increased and cost is added. Empiric treatment should be reserved for those patients who are seriously ill and cannot tolerate diarrhea owing to hemodynamic instability. 13 PCR as a diagnostic tool for C. difficile was developed more than 10 years ago. Gumerlock et al. published early work on this topic in 1993, describing PCR as 100-fold more sensitive than conventional anaerobic culture. The investigators showed that PCR was positive in 4 patients with antibiotic-associated colitis in which the cytotoxic assay was negative. No amplified product was found in asymptomatic patients and the assay did not cross-react with other Clostridial organisms. 18 Despite these results, the use of PCR for diagnosis of C. difficile has not become routine in clinical practice and appears to be reserved for use in basic science laboratories and in some cases for epidemiologic purposes during outbreaks. The reason the test is not used routinely is not clear, but may relate to the unavailability of PCR equipment, laborious nature of extracting DNA from stool in the past, and concerns regarding cost and specificity of the test. A recent review article on the epidemiology of C. difficile associated infections showed variable carrier rates (6% 11%) and acquisition rates (4% 21%) of C. difficile in hospitalized patients depending on the length of hospital stay and the facility involved in these studies. 19 It is more difficult to discern the differences in toxigenic and nontoxigenic C. difficile carrier rates because most studies do not specifically address this issue. However, one study of 192 symptom-free carriers revealed that 95 patients were colonized with a toxigenic strain of C. difficile, 76 were colonized with a nontoxigenic strain, 12 with both types of strain, and 9 with an indeterminate strain. Only 2 of these 192 patients went on to develop C. difficile associated diarrhea. 20 Since that time, several well-designed studies have been published on the test characteristics of PCR for the diagnosis of C. difficile. These studies showed that the test has a very high sensitivity and specificity compared with the gold standard of cytotoxic assay. Boondeekuhn et al. 11 reported a sensitivity of 94% and a specificity of 95% comparing PCR with both cytotoxic assay and culture. Alonso et al. 9 reported a sensitivity of 96% and a specificity of 100% for PCR when compared with cytotoxic assay and culture, with PCR being negative in all 5 samples in which culture was positive and cytotoxic assay was negative. Most recently, Guilbault et al. 12 showed a sensitivity of 91.5% and a specificity of 100% comparing PCR with cytotoxic assay results on 118 stool samples submitted for C. difficile detection. These studies consistently showed the high sensitivity and specificity of this test when compared with a well-accepted standard.

August 2004 PCR FOR DIAGNOSIS OF C. DIFFICILE 673 By using PCR for the diagnosis of C. difficile associated diarrhea, this study showed that approximately 14% of specimens were positive for the infection. This is somewhat less than prior reports of the prevalence of the infection as a cause of diarrhea in hospitalized patients. 14 The sensitivity of the EIA in this study compared with PCR was only 40%, which is well below the reported performance ability of this test. The reason for this is not clear; however, it is possible that clinical practice differs significantly from controlled sample collection generally used in clinical trials. In any event, C. difficile cases are not being diagnosed appropriately owing to the lack of sensitivity of the current method of testing. It also is readily apparent that the guidelines for the proper evaluation of patients suspected of having C. difficile are not being followed. Fully 42% of patients were treated empirically with metronidazole inappropriately as defined by established practice guidelines. 13 Statistical analysis failed to reveal any difference in treatment decision with regard to test results. This suggests that the test result is not influencing practice habits among a diverse population of academic and community-based physicians. Empiric treatment may expose patients to unnecessary side effects, increase the potential for future bacterial resistance to antibiotics, and increase health care costs. Furthermore, inappropriate treatment may lead to additional unnecessary antibiotic intervention if the patient is labeled as a C. difficile relapser. It also is clear that physicians are not adhering to the practice guidelines for the diagnosis of C. difficile because roughly 30% of patients had more than 3 EIAs sent per admission and 57% of patients with more than 1 EIA sent during an admission had more than 1 EIA sent within a 24-hour period. When pursuing the use of a new diagnostic test it is important to consider not only the seriousness of the disease being detected but also the sensitivity, specificity, cost, and practicability of the new test being evaluated. It is clear that C. difficile associated diarrhea is a potentially life-threatening and costly infection. A new diagnostic test for this should provide maximum sensitivity to avoid false-negative results. The use of PCR for the toxin B gene avoids the detection of nontoxigenic strains, which do not harbor the toxin-producing gene segment. In addition, our study used toxin B as the target of PCR rather than toxin A because toxin A positive/toxin B negative strains are exceedingly rare, if not unknown. 21 A lack of confidence in the test characteristics of EIA may explain why physicians in this study often empirically treated for C. difficile even in the face of a negative test. Other studies also suggest that inappropriate empiric treatment with metronidazole is common. 22 The development of a more simple method for DNA extraction of stool makes PCR, in this situation, more practical than in the past. In the current study, the DNA from 10 stool samples was retrieved in approximately 2.5 hours. Experienced clinical technicians may have even higher rates of productivity. Performing extraction from 20 stools adds 1 hour to the total time. PCR amplification was performed and gel electrophoresis was reviewed in less than 5 hours. The total turnaround time for 10 stool samples is roughly 7.5 hours, a time period that would allow for the test to be performed in one shift or batched and run overnight. This is the current practice at our institution when using EIA. The total cost to run an EIA at Health Alliance is $12.50 per sample, with a charge of $42 per test. Reimbursement rates for this amount obviously will vary based on insurance coverage. The cost for the supplies needed to extract DNA to run PCR was $2.70 per sample. The cost of PCR and gel electrophoresis with respect to supplies such as primers, taq polymerase, and gel reagents is, on average, $12.20 per sample. The remainder of the cost will vary based on the hourly wage of laboratory technicians running the test providing that a PCR machine already is available (which it most likely is at large hospitals in this country). At our institution the average wage for a technician is $18 per hour for hands-on work. This cost would not count toward the 3.5-hour cycling time for PCR during which the technician is free to do other jobs. Thus, running 10 stools would require, on average, 4 hours of manpower, or $72, which would average to $7.20 per stool sample. Our hospital currently runs EIA in batches of 10. If 5 samples were run instead of 10, the average cost would increase by an additional $7.20 per sample. Our estimates at overall cost are approximately $22 per sample when run in batches of 10, and $29 per sample when run in batches of 5. Although slightly more expensive than EIA, this is well within acceptable bounds for a test showing improved sensitivity. This does not take into consideration the money saved and improvement in medical care that would result from more reliable test results, what the true charge of EIA is ($42) at our institution, and the fact that, on average for this study, patients had 2 EIAs sent (cost of $25 per patient vs. $22 for one PCR). Our clinical laboratory tests 12,000 samples of stool for C. difficile by EIA per year. Extrapolation of these findings suggests that the health system may miss approximately 900 cases of C. difficile annually. The use of PCR may have significant implications on hospital length of stay,

674 MORELLI ET AL. CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 2, No. 8 avoidance of unnecessary exposure to antibiotics, confidence that we are not missing a potentially life-threatening disease, ability to move patients out of costly isolation rooms, less frequent and costly laboratory testing, and better overall medical care for our patients including potential improvements in mortality. Weaknesses of this study include potential information bias owing to the retrospective nature of the chart review and the lack of specific mean and median lengthof-stay data. However, bias was minimized by review of objective databases (medication chart) and by noting instances in which it was not clear that metronidazole was used for empiric treatment of C. difficile. The results also may be subject to documentation error or bias by primary physicians in charge of patient care. The overwhelming majority of patients had a length of stay of less than 2 weeks, with an estimated range of 1 90 days. These data were not recorded specifically, and could not be included in our analysis. It is possible that some patients may have been hospitalized long enough for the diagnosis of C. difficile to have been sought more than once in an appropriate manner. The relatively low rate of C. difficile toxin detection may reflect an inappropriate practice by physicians as well. Patients with diarrhea may be tested without clear risk factors for C. difficile. We feel that PCR is best used in those patients who are felt to have active C. difficile infection with known risk factors for the disease. In conclusion, PCR of stool for the presence of C. difficile associated diarrhea represents the best combination of sensitivity, specificity, and potential cost effectiveness needed for the diagnosis of this serious condition and should be strongly considered for general clinical use. Physicians in local medical institutions are not following established clinical guidelines for the proper use of the current standard diagnostic testing and for empiric treatment for C. difficile associated diarrhea and internal education programs are recommended. References 1. Bartlett JG. Clostridium difficile: history of its role as an enteric pathogen and current state of knowledge about the organism. Clin Infect Dis 1994;18:S265 S272. 2. Olson MM, Shanholtzer CJ, Lee JT, Gerding DN. Ten years of prospective Clostridium difficile disease surveillance and treatment at the Minneapolis VA medical center, 1982-91. Infect Control Hosp Epidemiol 1994;15:371 381. 3. Kyne L, Hammel MB, Polavaram R, Kelly C. Health care costs and mortality associated with nosocomial diarrhea due to Clostridium difficile. Clin Infect Dis 2002;34:346 353. 4. Wilcox MH, Cunniffe JG, Trundle C, Redpath C. Financial burden of hospital-acquired Clostridium difficile infection. J Hospital Infect 1996;34:23 30. 5. Bignardi GE. Risk factors for Clostridium difficile infection. J Hospital Infect 1998;40:1 15. 6. McFarland LV, Stamm WE. Review of Clostridium difficile-associated diseases. Am J Infect Control 1986;14:99 109. 7. Barbut F, Kajzer C, Planas N, Petit JC. Comparison of three enzyme immunoassays, a cytotoxic assay, and toxigenic culture for diagnosis of Clostridium associated diarrhea. J Clin Microbiol 1993;31:963 967. 8. Aronsson B, Mollby R, Nord CE. Diagnosis and epidemiology of Clostridium difficile enterocolitis in Sweden. J Antimicrob Chemother 1984;14(suppl D):85 95. 9. Alonso R, Munoz C, Gros S, Garcia de Viedman D, Pelaez T, Bouza E. Rapid detection of toxigenic Clostridium difficile from stool samples by a nested PCR of toxin B gene. J Hospital Infect 1999;41:145 149. 10. Kato N, Chin-Yih O, Kato H, Bartley S, Luo C, Kilgore G, Ueno K. Detection of toxigenic Clostridium difficile in stool specimens by the polymerase chain reaction. J Infect Dis 1993;167:458 461. 11. Boondeekhun HS, Gurtler V, Maryann LO, Wilson VA, Mayall BC. Detection of Clostridium difficile enterotoxin gene in clinical specimens by the polymerase chain reaction. J Med Microbiol 1993; 38:384 387. 12. Guilbault AC, Labbe LP, Busque L, Beliveau C, Laverdiere M. Development and evaluation of a PCR method of detection of the Clostridium difficile toxin B gene in stool specimens. J Clin Microbiol 2002;40:2288 2290. 13. Fekety R. Guidelines for the diagnosis and management of Clostridium difficile-associated diarrhea and colitis. Am J Gastroenterol 1997;92:739 750. 14. Siegel DL, Edelstein PH, Nachamkin I. Inappropriate testing for diarrheal diseases in the hospital. JAMA 1990;263:979 982. 15. Lesna M, Parham DM. Risk of diarrhoea due to Clostridium difficile during cefotaxime treatment; mortality due to C. difficile colitis in elderly people has been underestimated. BMJ 1996; 312:778. 16. Thomas DR, Bennet RG, Laughon BE, Grennough WB 3rd, Bartlett JG. Postantibiotic colonization with Clostridium difficile in nursing home patients. J Am Geriatr Soc 1990;38:415 420. 17. Eriksson S, Aronsson B. Medical implications of nosocomial infection with Clostridium difficile. Scand J Infect Dis 1989;21: 733 734. 18. Gumerlock PH, Yajarayma JT, Meyers FJ, Silva J Jr. Use of polymerase chain reaction for the specific and direct detection of Clostridium difficile in human feces. Rev Infect Dis 1991;13: 1053 1060. 19. Barbut F, Petit JC. Epidemiology of Clostridium difficile-associated infections. Clin Microbiol Infect 2001;7:405 410. 20. Shim JK, Johnson S, Samore MH, Bliss DZ, Gerding DN. Primary symptomless colonization by Clostridium difficile and decreased risk of subsequent diarrhea. Lancet 1998;351:633 636. 21. Rupnik M. How to detect Clostridium difficile variant strains in a routine laboratory. Clin Microbiol Infect 2001;7:417 420. 22. Vasa CV, Glatt AE. Effectiveness and appropriateness of empiric metronidazole for Clostridium difficile-associated diarrhea. Am J Gastroenterol 2003;98:354 358. Address requests for reprints to: Michael S. Morelli, M.D., Indianapolis Gastroenterology and Hepatology, 8051 South Emerson Drive, Suite 200, Indianapolis, Indiana 46237. e-mail: ammorelli@msn.com. The authors thank Norah Shire, MPH, for her critical review of this manuscript, and Sarah Dameron and Carmen Meier, MD, for their laboratory work. Dr. Morelli is currently in private practice.