Detection of Toxigenic Clostridium difficile in Stool Samples by Real-Time Polymerase Chain Reaction for the Diagnosis of C. difficile-Associated Diarrhea

Abstract

Background. Clostridium difficile-associated diarrhea (CDAD) is the major cause of health care-associated infectious diarrhea. Current laboratory testing lacks a single assay that is sensitive, specific, and rapid. The purpose of this work was to design and validate a sensitive and specific real-time polymerase chain reaction (PCR) diagnostic test for CDAD.

Methods. This observational validation study of a new real-time PCR assay occurred from July 2004 through April 2006 and involved the testing of 1368 stool samples. As the final validation portion of the investigation, 350 inpatients were prospectively interviewed for clinical findings for 365 episodes of diarrheal illness. Test results and clinical criteria were used to assess the performance of 4 assays.

Results. Using clinical criteria requiring at least 3 loose stools in 1 day as part of the reference standard for a positive test result supporting CDAD, the sensitivity, specificity, and positive and negative predictive values were 73.3%, 97.6%, 73.3%, and 97.6%, respectively, for enzyme immunoassay; 93.3%, 97.4%, 75.7%, and 99.4%, respectively, for real-time PCR; 76.7%, 97.1%, 69.7%, and 97.9%, respectively, for cell culture cytotoxin assay; and 100.0%, 95.9%, 68.2%, and 100.0%, respectively, for anaerobic culture (for toxigenic C. difficile strains). The real-time PCR and anaerobic culture assays were significantly more sensitive than the enzyme immunoassay (P < .01 to P < .05).

Conclusions. With an assay turnaround time of <4 h, real-time PCR is a more sensitive and equally rapid test, compared with enzyme immunoassay, and is a feasible laboratory option to replace enzyme immunoassay for toxigenic C. difficile detection in clinical practice, as well as for use during the development of new therapeutic agents.

Clostridium difficile is responsible for the majority of cases of infectious antibiotic-associated diarrhea and pseudomembranous colitis and is rapidly increasing in prevalence [1–6]. Recently, there have been 2 major clinical developments: (1) emergence of a hypervirulent form of the disease [7–10] and (2) the suggestion that metronidazole therapy is currently considered to be less effective than it was thought to be in the past [11, 12]. However, in most reports, the description of diagnostic laboratory testing for detection of toxin-producing C. difficile is minimal and overlooks the fact that current diagnosis relies on assays with poor sensitivity and specificity. This raises the concern that the appearance of the increasing number of cases of disease, disease severity, and frequency of therapeutic failure may be based on uncertain diagnosis of C. difficile—associated diarrhea (CDAD), in which symptoms can be similar to those of diarrhea (or at least “loose stool”) caused by a change in diet (in the hospital setting) or as an effect of antibiotic administration or an infection with another pathogen (e.g., Salmonella species or norovirus).

The pathogenic effects of C. difficile are mucosal damage to the colon that is caused by the production of toxin A and/or toxin B [13]. The diagnostic methods that target 1 or both of these toxins include EIA and cell culture cytotoxicity assay performed on stool samples. Although the various EIA methods have proven to be less than optimal diagnostic tests, they are the assays that are most commonly used. EIA methods offer a rapid turnaround time, compared with cell culture cytotoxicity or culture for toxigenic C. difficile organisms—tests for which the time to final result can be 48–72 h [1, 14]. However, EIA is associated with widely varying sensitivity (50%–99%) and specificity (70%–100%)—with performance largely dependent on which “gold” reference standard is used for comparison—making its reliability questionable for an accurate diagnosis of CDAD [15–18]. The purpose of this investigation was to develop a real-time PCR for the rapid, sensitive, and specific detection of toxigenic C. difficile in stool samples for the diagnosis of CDAD.

Methods

Setting and specimens. Evanston Northwestern Healthcare is a tertiary care, university-affiliated organization composed of 3 acute care hospitals in suburban Chicago and cares for ∼40,000 inpatients annually. Sequential stool specimens were obtained from the Evanston Northwestern Healthcare Clinical Microbiology Laboratory (Evanston, IL) for 2 investigations. The first evaluation, in which clinical data were retrospectively assessed, occurred from July 2004 through January 2005. The second investigation, in which clinical data were prospectively collected, was performed from December 2005 through April 2006. An unformed stool sample was defined as a room temperature sample that took the form of the collection container.

Real-time PCR testing. Real-time PCR testing was the same for both investigations. The procedure is shown in table 1. Primers were designed in-house to target a region of the toxin B gene (tcdB) that is highly conserved among multiple strains of toxigenic C. difficile. The primers effectively detect toxin A-negative, toxin B-positive, and toxin A and B-positive C. difficile strains, which comprise the vast majority of toxigenic C. difficile disease-causing isolates [19–21]. Our method was developed on the basis of preliminary experiments to provide a complete extraction that can be performed on bloody samples. During the test development stage, detection of (human) p53 gene in samples was used to assess inhibition; p53 genome was detected in all samples using the described assay.

Table 1

Procedure for real-time PCR testing for detection of Clostridium difficile.

To determine the detection sensitivity of the PCR assay, 10-fold serial dilutions were used with C. difficile ATCC strain 9689. Target specificity was confirmed by testing strains CF2-5340 (a toxin A-negative, toxin B-positive C. difficile strain) and P1 (nontoxigenic C. difficile).

Anaerobic culture for toxigenicC. difficileisolates. Anaerobic culture was performed on all stool specimens by plating specimens onto prereduced cycloserine-cefoxitin-fructose agar media (CCFA-VA formulation; Remel) [22, 23]. Plates were incubated anaerobically using the Anaero Pouch System (Mitsubishi Gas Chemical Company) at 35°C for up to 5 days before a final interpretation of a negative result was determined. Yellow, spreading colonies that were characteristic of C. difficile were further analyzed. Gram staining was performed to confirm the organisms' morphology (large gram-positive or gram-variable bacillus). A Pro Disc test (Key Scientific Products) was performed on isolated colonies. Organisms that did not grow aerobically and yielded positive results using Pro Disc were considered to be C. difficile. All C. difficile isolates were grown in anaerobic chopped-meat broth (Remel), and supernatant passed through a 0.45-µm filter (Spin-X centrifuge tube filter; Costar) was used to determine toxigenicity. Only toxigenic strains were considered to be consistent with the diagnosis of CDAD.

EIA. The C. difficile Tox A/B II (Wampole) was the EIA used in this study. We adhered to the kit protocol for double wavelengths, which identify all specimens with fluorescence; an optical density <0.08 was considered to be a negative result, and an optical density ⩾0.08 was considered to be a positive result.

Cell culture cytotoxicity assay. To determine the presence of toxin B in all of the direct stool specimens or recovered C. difficile isolates, 2 tubes of MRC-5 cells (ViroMed Laboratories) were used for each sample. For stool samples, 0.5 mL was suspended in 1.5 mL of EMEM buffer (Eagles-MEM 2% Media/Vancomycin; ViroMed). Stool suspensions were vortexed and then centrifuged at > 10,000 × g for 10 min. The supernatant was filtered through a 0.45-µm filter (Spin-X). One MRC-5 tube contained 100 µL of filtrate with 100 µL EMEM buffer, and the other 100 µL of filtrate was mixed with 100 µL of C. difficile antitoxin (160-fold final dilution). The specimen was considered to be toxigenic if a cytopathic effect was observed in the first tube and no effect was observed in the second, antitoxin tube [23].

Clinical evaluation. Two sequential investigations were performed to compare laboratory tests. An evaluation checklist was developed (figure 1) on the basis of published recommendations for validated criteria required for consideration of a diagnosis of CDAD [24, 25]. The checklist was used retrospectively for the first set of test data and prospectively for the second investigation. Clinical findings and laboratory results were examined by separate investigators and were not compared until each investigation was complete.

Figure 1

Clinical scoring criteria used in prospective patient evaluation

Anaerobic (toxigenic) culture was used as the reference standard for the first investigation (in which the clinical information regarding the patients was retrospectively collected), and the other tests were compared with that standard. The assay results for the prospective evaluation (the second investigation) were analyzed in 2 ways, using distinct gold reference standards. First, we defined any sample having at least 3 positive test results as the gold standard for assay performance and evaluated the various tests against this standard. The other measure of assay performance considered clinical criteria requiring ⩾3 loose stools for at least 1 day and ⩾2 positive test results as the gold reference standard for a CDAD diagnosis. This latter definition as a reference standard is a minor modification of the accepted definition for CDAD, which requires 3 loose stools per day for 36–48 h, in addition to a laboratory test result indicating the presence of toxigenic C. difficile [1]. This latter definition comprised our final recommended reference standard for test evaluation.

We also assessed the association of CDAD with the use of proton pump inhibitors (indicated in the medical records) by matching patients with any positive test result to another patient hospitalized at the same time who reported the same clinical scoring results indicating potential CDAD but for whom all 4 tests yielded negative results.

We determined the cost of each test (based on the manufacturer's suggested retail price of reagents) and the amount of technologist time required, assuming that a batch of 10 specimens were tested together. Statistical evaluation was performed using Web-based calculators for the Fisher's exact probability test, as well as for sensitivity, specificity, positive predictive value, and negative predictive value for clinical laboratory testing [26, 27]. Student's t test was performed using Excel 2000 (Microsoft). The analysis of laboratory test performance was approved by the Institutional Review Board of Evanston Northwestern Healthcare.

Results

Sensitivity of the real-time PCR assay was 5 colony-forming units per reaction. A positive PCR result was determined by an amplification cycle threshold value <36 and a melting temperature (T_M) of 78.5°C ± 1°C.

Investigation 1. Nine hundred ninety-eight sequential stool specimens were obtained. Of these, 380 were eliminated because they were formed stool, leaving 618 samples for comparing EIA and real-time PCR with anaerobic culture. Retrospective analysis of the patient records revealed that documentation was too poor to permit recording the clinical criteria required for CDAD; therefore, no clinical correlation was determined. Compared with culture, EIA had a sensitivity of 66.7% and a specificity of 91.8%, and real-time PCR had a sensitivity of 94.4% and a specificity of 96.8%.

Investigation 2. In the second investigation, only unformed stool samples from inpatients were accepted for processing, which resulted in 370 acceptable specimens from 350 patients experiencing 365 unique episodes of potential CDAD. Samples were examined using 4 diagnostic methods: EIA, real-time PCR, direct stool cytotoxicity, and anaerobic culture for toxigenic C. difficile. Prospective subject interviews revealed that, in 142 (39%) of 365 episodes, the patient denied having sufficient diarrhea to warrant consideration of CDAD as a clinical diagnosis.

Use of the clinical scoring evaluation (figure 1) demonstrated that the best statistical separation of the testing occurred when the results of ⩾3 tests (rather than ⩽2 tests) were positive (P < .001; data not shown). Thus, in 1 comparison, the reference standard was defined on the basis of ⩾3 positive test results (among 228 patients with diarrhea); 30 positive samples were used for analysis of the 4 assays. The results are shown in table 2, and the analysis is shown in table 3. In the clinically-based evaluation, the reference standard for the diagnosis of CDAD was defined on the basis of combined history findings (requiring the presence of diarrhea) and ⩾2 positive tests results; 370 samples were used for this analysis. Results of this evaluation are shown in table 2, and the analysis is shown in table 4. The clinical findings for 12 patients who had a single positive test result in the prospective, clinically-based investigation are shown in table 5.

Table 2

Test results for the comparison analysis using the 2 reference standards for the diagnosis of Clostridium difficile-associated diarrhea (CDAD).

Table 3

Results of tests performed on stool samples from 228 patients with clinically significant diarrhea to detect the presence of toxigenic Clostridium difficile or its toxins.

Table 4

Results of tests of 370 nonformed stool samples to detect toxigenic Clostridium difficile or its toxins.

Table 5

Characteristics of 12 patients who had 1 test of stool samples that yielded positive results in the prospective clinical assessment for investigation 2.

As expected, test performance was dependent on whether clinical criteria were used, which particularly affected the sensitivity of the EIA and direct cytotoxin test. Importantly, in the analysis of test performance using only the results of other assays (e.g., when ⩾3 tests yielded positive results), 20% of the patients denied having significant diarrhea.

The frequency of proton pump inhibitor use was high in our population; 47 (92%) of 51 patients with positive test results received therapy with these agents, and 46 (90%) of 51 control subjects were similarly treated (the P value is not statistically significant). The reagent cost of each assay and the amount of technical time required was $5.12 and 3 min, respectively, for EIA; $6.60 and 2 min, respectively, for real-time PCR; $6.85 and 5 min, respectively, for direct cytotoxicity; and $2.69 and 1 min, respectively, for anaerobic culture.

Discussion

The difficulty of choosing an optimal test for the diagnosis of CDAD has long been known. Anaerobic culture has sensitivity approaching 100%, but the false-positive rate exceeds 10% [28] because of the high rate of asymptomatic carriage of the organism that is associated with prolonged hospitalization [1, 29]. Although cell culture cytotoxicity is considered to have good specificity, the sensitivity, compared with anaerobic culture for CDAD, can be as low as 71% [28]. Confounding the current approach of diagnosis of CDAD is the test performance of commercial EIAs, with sensitivities of 50%–99% and specificities of 70%–100% [1, 15–18]. This test performance poses challenges for clinicians caring for patients with possible CDAD, because EIA accounted for nearly 95% of the toxin assay methods used in the United States by 2003 [3]. Snell et al. [30] recently reported higher specificities (⩾98%) and positive predictive values (>80%) for EIA tests, particularly when the tests were used in combination, but the authors' definition of a true-positive case of CDAD was solely based on laboratory tests, thereby enhancing apparent assay performance. In addition, the samples used by Snell and colleagues had a very high prevalence of positive results (18.7%), which improves the positive predictive value. Perhaps most importantly for the accurate diagnosis of this infectious disease is the fact that the rapid, real-time PCR assay had a sensitivity similar to that of culture for detecting toxigenic C. difficile, while retaining the specificity of the direct cytotoxicity test when assessed using an accepted clinical definition of CDAD.

Several reports of use of real-time PCR for detection of toxigenic C. difficile have appeared over the past few years, but our report is the only assay comparison to use a prospective interview assessment of testing for patients with suspected CDAD in >2 decades [29]. Belanger et al. [31] were perhaps the first to have an article published about a diagnostic approach using real-time PCR. More recently, van den Berg et al. [20] reported their evaluation of real-time PCR and a new immunoassay for C. difficile. The sensitivities of the new assays were 87% and 91%, respectively, with specificities of 96% and 97%, respectively. van den Berg et al. [32] also reported another assessment of their real-time PCR test, compared with a cell culture cytotoxin assay, and found the sensitivity and specificity of PCR to be 100% and 94%, respectively, compared with other laboratory tests. In the most recent multicenter trial by van den Berg et al. [33], the authors substantiated the use of PCR by demonstrating that PCR is most concordant with toxigenic culture and suggested that PCR is the preferred method for diagnosing CDAD. Guilbault et al. [21] also targeted the toxin B gene in their 2002 report, noting that the selection of this target was based on the presence of toxin A-negative, toxin B-positive C. difficile, which is capable of causing disease, and that target selection was facilitated by the characterization of these strains [34]. Importantly, this genotypic test is not influenced by the amount of toxin produced; only the presence or absence of the gene target and the ability to detect it influence the assay result.

Adding to the need for accurate laboratory diagnosis of CDAD on a daily basis for the practicing physician are the emerging clinical imperatives that are described in table 6. A critical question regarding the diagnosis of CDAD is the definition of diarrhea, which was addressed by the Society for Healthcare Epidemiology of America guideline in 1995 [1]. The recent Centers for Disease Control and Prevention recommendation for epidemiologic tracking of CDAD gives no guidance as to what constitutes diarrhea [43]. However, the Centers for Disease Control and Prevention did this to develop an epidemiological definition that could be used in retrospective analyses in which the recorded history is insufficient to determine the number of diarrheal stools—a problem we documented in our investigation. The contemporary article by Berrington et al. [44] supports our contention that standard clinical criteria are required to enhance accurate CDAD diagnosis. Interestingly, in our study, the patients who did not report having sufficient diarrhea to clinically suggest CDAD fulfilled this standard by diarrhea duration but not by the number of stools per day, having ⩽2 loose stools per day. Because of the short length of hospital stay in today's health care practice, assuring that someone has ⩾3 stools on the first day of diarrheal symptoms appears to be a reasonable criterion for the consideration of CDAD testing in most instances. Thus, for our final gold reference standard we considered only those patients who had ⩾3 stools for 1 day and positive results of ⩾2 assays. Although the emergence of a new epidemic strain of C. difficile has caused concern, there is no distinction between BI/NAP1 and other C. difficile strains in relation to tcdB.

Table 6

Clinical needs for accurate testing for Clostridium difficile-associated diarrhea (CDAD).

The major limitation of our study (prospective investigation 2) was the relatively small number of positive CDAD cases. However, we believe this deficiency was offset by the careful, prospective clinical data collection via patient interviews. A second limitation is the fact that we did not perform extensive testing of other clostridial species for possible cross-reactivity of our PCR primers targeting tcdB. tcdB has been sequenced; thus, we were able to select our primers to specifically detect C. difficile toxin B [45]. In addition, other researchers have performed extensive analysis of clostridial species and enteropathogenic bacteria, with no findings of cross-reactivity of genomic targets [32]. A third potential limitation is the fact that our “clinical scoring criteria” tool has not been validated. However, it was designed in accordance with the Society for Healthcare Epidemiology of America guideline for CDAD, as well as that by Katz et al. [1, 24, 25], and was solely intended as an instrument to separate patients who had clinically likely CDAD from those who did not have clinically likely CDAD. Unanswered questions remain, including the use of rectal swab samples versus stool samples (for infection control surveillance) and multiplexing with other primers indicative of CDAD or other stool pathogens. However, multiplexing tends to lower the sensitivity of PCR assays, and currently, there are no targets beyond tcdB that are thought to be clinically important for CDAD diagnosis.

We realize that there are always exceptions to test sample screening approaches and that these approaches should be used as guides rather than absolute “rules.” In the context of CDAD, this is exemplified by the clinical scenario of toxic megacolon, in the context of which stool production is often nonexistent. However, in our prospective evaluation, we found no cases of toxic megacolon, and the fact that such cases typically do not produce fecal material adds credence to the use of applying clinical criteria to samples prior to testing in clinical laboratories. The criteria we used for a clinical scenario consistent with CDAD are in concert with the current case definition of the epidemic strain [46].

In conclusion, we have developed and clinically validated (through prospective interviews) a real-time PCR test for the sensitive and specific detection of toxin B-producing C. difficile for the laboratory confirmation of CDAD that outperforms the most widely used diagnostic test for toxigenic C. difficile detection and is cost competitive. Once validated in a laboratory, such testing can be applied to confirm the diagnosis of CDAD in patients with International Classification of Diseases, Ninth Revision, Clinical Modification code 008.45, which is specific for intestinal infection due to C. difficile. Commercial random-access, real-time PCR assays are being developed that have the potential to be very useful tests—perhaps, even with point-of-care in application. Finally, with ∼50% of submitted stool samples either being formed or obtained from patients without significant diarrhea, we observed the need for ongoing medical education with regard to the clinical criteria suggesting CDAD in patients reporting loose stools to their physicians. We suggest that ⩾3 loose stools in a single day should be the minimum clinical criteria for considering CDAD. Without appropriate interviewing for CDAD risk factors and use of accurate diagnostic tests, there is a meaningful potential for a high false-positive detection rate that can lead to mistaken diagnosis, delayed directed therapy, and confusing epidemiology.

Acknowledgments

We thank the Clinical Microbiology Laboratory of Evanston Hospital and the Infection Control Professionals of Evanston Northwestern Healthcare (Evanston, IL), for their expertise and commitment to the improved understanding of Clostridium difficile-associated diarrhea, and D. N. Gerding, for providing test strains.

Financial support. Department of Pathology and Laboratory Medicine at Evanston Northwestern Healthcare (Evanston Hospital).

Potential conflicts of interest. All authors: no conflicts

Presented in part: 10th Annual Meeting of the Association for Molecular Pathology, Los Angeles, November 2004 (abstract I43), and the 12th Annual Meeting of the Association for Molecular Pathology, Orlando, November 2006 (abstract ID48).