|Home | About | Journals | Submit | Contact Us | Français|
Clostridium difficile is the most common cause of health care-associated diarrhea. Accurate and rapid diagnosis is essential to improve patient outcome and prevent disease spread. We compared our two-step diagnostic algorithm, an enzyme immunoassay for glutamate dehydrogenase (GDH) followed by the cytotoxin neutralization test (CYT) with a turnaround time of 24 to 48 h, versus the Cepheid Xpert C. difficile Epi assay, a PCR-based assay with a turnaround time of <1 h. In the first phase of the study, only GDH-positive stool samples were tested by both CYT and Xpert PCR. Discordant results were resolved by toxigenic culture. In the second phase, all stool samples were tested by GDH and Xpert PCR. Only GDH-positive stools were further tested by CYT. Genotypic characterization of 45 Xpert PCR-positive stools was performed by sequencing of the tcdC gene and PCR ribotyping. In phase 1, the agreement between the GDH-CYT and the GDH-Xpert PCR was 72%. The sensitivities and specificities of GDH-CYT and GDH-Xpert PCR were 57% and 97% and 100% and 97%, respectively. In phase 2, the agreement between GDH-CYT and Xpert PCR alone was 95%. As in phase 1, sensitivity of the Xpert PCR was higher than that of the GDH-CYT. The correlation between PCR-ribotyping, sequencing, and Xpert PCR for detection of NAP1 strains was excellent (>90%). The excellent sensitivity and specificity and the rapid turnaround time of the Xpert PCR assay as well as its strain-typing capability make it an attractive option for diagnosis of C. difficile infection.
Clostridium difficile is the most common cause of health care-associated diarrhea. The incidence and severity of C. difficile infection (CDI) has increased across many parts of the world (11, 18, 19, 22). Toxins A and B (especially toxin B), encoded by the tcdA and tcdB genes of C. difficile, are essential for the manifestation of the disease and responsible for the inflammation and colon damage that results in diarrhea and pseudomembranous colitis (17, 28).
In the last 10 years, several outbreaks of CDI associated with increased disease severity and increased mortality have been reported in Canada, the United States, and Europe (12, 16). The epidemic strain of C. difficile responsible for outbreaks has been referred to as either strain NAP1, BI, or 027, depending on whether typing was done using pulse-field gel electrophoresis, restriction endonuclease analysis, or PCR-ribotyping, respectively. The NAP1/BI/027 C. difficile strain produces increased levels of toxin A and toxin B and a third toxin called the binary toxin and also carries an 18-bp deletion and a 1-bp deletion (at nucleotide [nt] 117) in the tcdC gene, a putative negative regulator of tcdA and tcdB gene expression. The 18-bp and 1-bp deletions are proposed to result in a truncated TcdC protein, with loss of negative regulatory function that results in higher toxin production (22, 31). Although a recent report failed to detect a statistically significant correlation between disease severity and the NAP1 strain (30), others, including a recent report of a phase 3 clinical trial of fidaxomicin, support the increased virulence of the NAP1 strain (6, 9, 14).
Accurate and rapid diagnosis of CDI is essential for optimal patient care and to prevent hospital spread of the infection (3). The latest guidelines from the Society for Healthcare Epidemiology of America (SHEA) and the Infectious Diseases Society of America (IDSA) reemphasized the need to consider toxigenic culture, which includes anaerobic stool culture followed by toxin detection, as the gold standard for diagnosis of CDI (3). However, the turnaround time (TAT) for toxigenic culture is several days, severely limiting its value for early diagnosis and prevention. Alternative tests, including numerous enzyme immunoassays (EIA) for toxins A and B and the C. difficile common antigen glutamate dehydrogenase (GDH), cytotoxin neutralization assays (CYT), and real-time PCR assays, are available for diagnosis of CDI. The variations in sensitivities and specificities, TATs, and costs of these various techniques have resulted in laboratories adopting a variety of algorithms that rely on two or more tests to diagnose CDI (4, 24, 25).
Three FDA-approved real-time PCR tests, the GeneOhm Cdiff assay (BD Diagnostics, San Diego, CA), the ProGastro Cd test (Prodesse Inc., Waukesha, WI), and the Xpert C. difficile test (Cepheid, Sunnyvale, CA), are now available commercially. A few studies have reported an increased sensitivity of these assays for diagnosis of CDI (13, 21, 27). An alternative version of the FDA-approved Xpert C. difficile test (the Xpert C. difficile Epi) is available from Cepheid as a research-use-only (RUO) assay. The Xpert C. difficile Epi test can differentiate between toxigenic C. difficile strains and the epidemic NAP1/BI/027 C. difficile strains, providing additional epidemiologic information that may have an impact on patient care and infection control practices (8, 29).
At Memorial Sloan-Kettering Cancer Center (MSKCC), diagnosis of CDI is accomplished by using a two-step algorithm that includes an EIA for the detection of GDH followed by the CYT assay; this is an algorithm with a TAT of 12 to 48 h. For epidemiological studies, positive C. difficile isolates are further analyzed by PCR-ribotyping and/or direct sequencing of the tcdC gene to detect the 18-bp or nucleotide 117 deletions. The goals of the current study were (i) to compare the performance of the Cepheid Xpert C. difficile Epi assay to our current two-step CDI diagnostic algorithm and (ii) to compare the Cepheid Xpert C. difficile presumptive typing of NAP1 C. difficile strains to our typing methodologies, including PCR-ribotyping and direct sequencing.
A total of 560 stool specimens were tested in this study. Specimens tested were soft and liquid stools submitted to the clinical microbiology laboratory for routine CDI diagnosis between March 2010 and May 2010 and included duplicate specimens from the same patients. Specimens were tested daily or stored at 4°C and tested within 24 h.
The C. DIFF CHEK-60 assay (Techlab, Blacksburg, VA), an EIA for the detection of the GDH antigen, was performed according to the manufacturer's instructions. Briefly, 100 μl of diluted stool specimen (1:5 dilution) was added to each well containing 50 μl of the conjugate. The plate was incubated for 50 min at 37°C, the wells emptied, and wash buffer added. The procedure was repeated four times. After washing, 100 μl of substrate was added to each well and incubated at room temperature for 10 min followed by addition of a stop solution. Reactions (development of a yellow color) were read at 450 nm on a microplate reader against a blank at 620 nm.
For the cytotoxin neutralization assay, GDH-positive stool specimens were diluted (1:4) in phosphate-buffered saline (PBS) and centrifuged for 10 min at 2,000 × g. The resulting supernatant was filtered using a 0.45-μm filter (Acrodisc; Pall Corporation, Ann Arbor, MI), and 100 μl was inoculated into a tube of human lung fibroblast cells (1:40 final dilution; Diagnostics Hybrids, Athens, OH). To control for nonspecific toxicity, a second tube was inoculated with both the supernatant and 100 μl of Clostridium difficile goat antitoxin (Techlab Inc., Blacksburg, VA). The cells were incubated at 37°C and checked for cytopathic effect (CPE) at 24 h and 48 h. A positive result was defined as the presence of CPE in 50% of the cell monolayer and no CPE in the tube inoculated with the antitoxin.
All discordant results were tested by toxigenic culture. C. difficile selective agar (CDSA; BD BBL, Sparks, MD) plates were reduced overnight in an anaerobic chamber prior to use. A 100-mg stool sample was added to 500 μl of 100% ethanol, vortexed, and incubated at room temperature for 1 to 2 h. The solution was centrifuged at 1,200 × g for 5 min, ethanol was removed, and the stool sample was inoculated onto reduced CDSA medium using a cotton swab. The CDSA agar medium was incubated for 48 h under anaerobic conditions. Colonies resembling C. difficile (pale yellow to yellow) were subcultured on sheep blood agar (SBA) plates, and their identity was further confirmed by Gram staining and RapID ANA testing (Remel, Lenexa, KS). Toxin production was confirmed by performing the CYT test directly on a colony isolate diluted in 500 μl of PBS and used in cell culture at a final dilution of 1:50. Non-C. difficile strains identified by RapID ANA were confirmed by sequencing of the 16S rRNA gene at the Mayo Medical Laboratories (Rochester, MN). Enriched cultures in prereduced, anaerobically sterilized chopped meat carbohydrate broth (Remel, Lenexa, KS) were used when organisms failed to grow following direct culture. The enriched cultures were incubated in an anaerobic chamber for up to 48 h. The broth was subcultured to a CDSA plate and an SBA plate and incubated for an additional 48 h. Identification of toxigenic C. difficile was determined as described above.
The Xpert C. difficile Epi PCR assay is a multiplex real-time PCR that detects the toxin B gene (tcdB), the binary toxin gene (cdt), and the tcdC gene deletion at nt 117. The extraction, amplification, and detection steps take place in different chambers of a self-contained, single-use cartridge containing all the reagents necessary for the detection of C. difficile gene targets. The Xpert C. difficile Epi PCR (Xpert PCR) was performed according to the manufacturer's instructions. Briefly, a stool sample was collected on a swab (Cepheid collection device) from the container received in the laboratory and transferred into the sample reagent vial. The vial was vortexed for 10 s and the solution pipetted into the S chamber of the cartridge by using a Pasteur pipette. The cartridge was then placed on the GeneXpert instrument, and the test was performed using the GeneXpert C. difficile assay program. Potential results included the following: toxigenic C. difficile positive/presumptive 027-NAP1-BI negative, toxigenic C. difficile positive/presumptive 027-NAP1-BI positive, toxigenic C. difficile negative/presumptive 027-NAP1-BI negative, invalid, error, or no results. Testing of specimens with an invalid/error result or no result was repeated once.
Isolates were typed by PCR-ribotyping as previously described (2). Briefly, genomic DNA was isolated from C. difficile colonies growing on a CDSA plate, amplified, and fractionated on a 3% Amresco agarose plate (Fisher Scientific, Fairlawn, NJ) for 6 h at 85 V and 4°C in Tris-borate-EDTA buffer. The gel was stained with ethidium bromide, and an image of the gel was captured using a Gel Doc 2000 system (Bio-Rad Laboratories, Hercules, CA). The ATCC strain BAA-1805, a C. difficile strain confirmed to be NAP1 by the Centers for Diseases Control and Prevention (CDC), was used as the positive control.
DNA was purified from each C. difficile colony, and the entire tcdC gene was amplified using primers C1 and C2 as described previously (26). The PCR products were run on a 1.2% agar gel, and the ~800-bp band was gel purified and submitted for sequencing at the MSKCC DNA Sequencing Core Facility.
Statistical analysis was performed using the Fisher's exact test and a one-way analysis of variance (ANOVA; GraphPad Software, La Jolla, CA).
A total of 560 stool samples from 408 patients were analyzed during the 10-week study. The toxin B target (tcdB) was detected in all Xpert PCR-positive samples. The study was conducted in two phases. Phase 1 included only GDH-positive stools that were tested by CYT (GDH-CYT) and Xpert PCR (GDH-PCR). A total of 73 stool samples from 64 patients were tested, with 44 positive samples from 39 patients (Table (Table1).1). The agreement between the GDH-CYT algorithm and the GDH-PCR algorithm was 72.6%, with a kappa score (κ) of 0.48 (moderate agreement). Twenty results were discordant between the two methods: 19 stools were positive by PCR and negative by CYT, and 1 stool was positive by CYT and negative by PCR (Table (Table1).1). All 20 discordant stools were tested by toxigenic culture, and 18/20 stools were toxigenic culture positive and 2/20 stools were toxigenic culture negative (no growth) by both direct and enriched culture (data not shown). The sensitivities and specificities of the GDH-CYT and the GDH-PCR algorithms after resolution of discrepant results by toxigenic culture were 57% and 97% and 100% and 97%, respectively. The difference in sensitivity between the GDH-CYT and the GDH-PCR was statistically significant (P < 0.0001) as determined using Fisher's exact test.
For phase 2, all stool samples submitted to the microbiology laboratory for diagnosis of CDI were tested by GDH-CYT and by Xpert PCR (regardless of the GDH result). In this phase, 487 stool samples from 389 patients were tested, with a total of 60 positive samples from 47 patients (Table (Table2).2). The agreement between the GDH-CYT algorithm and the Xpert PCR was 94.3% with a κ value of 0.67 (good agreement). Twenty-eight results were discordant between the two methods: 27 stool samples were positive by Xpert PCR and negative by GDH-CYT, and 1 stool sample was positive by GDH-CYT and negative by Xpert PCR (Table (Table2).2). Discordant results were tested by toxigenic culture; 21/28 stool samples were toxigenic culture positive and 7/28 were toxigenic culture negative (data not shown). The seven toxigenic culture-negative samples included three stool samples (two from the same patient) with no organism growth by both direct and enriched culture and four stool samples that grew Clostridium innocuum and/or Clostridium beijerinckii/dolis as determined by RapID ANA and 16S RNA gene sequencing. Xpert PCR was performed directly on the four C. innocuum isolates to exclude the possibility of cross-reaction of the assay with other Clostridium species. All four isolates tested negative by Xpert PCR. Analysis of these seven results is summarized in Table Table3.3. The sensitivities and specificities of the GDH-CYT algorithm and the Xpert PCR after resolution of discrepant results by toxigenic culture in phase 2 were 61.1% and 99.7% and 100% and 98.6%, respectively. The difference in sensitivity was statistically significant (P < 0.0001) as determined using Fisher's exact test. There were no significant differences in sensitivity or specificity for the Xpert PCR compared with the use of the GDH assay as a screening test (100% and 97% in phase 1 versus 100% and 98.6% in phase 2). In phase 2, 10/27 Xpert PCR-positive and GDH-CYT-negative samples were GDH negative, and 7/10 were identified as true positive by toxigenic culture.
To evaluate the impact of duplicate specimens from the same patients on the performance characteristics of the Xpert PCR over the GDH-CYT algorithm, we reanalyzed the data, including only the first diarrheal specimen submitted for each patient. Table Table44 shows the results of the comparison between GDH-CYT and Xpert PCR for the 408 unique patient specimens. Resolution of discrepant results by toxigenic culture resulted in the following assay sensitivities and specificities: GDH-CYT, 58.1% and 99.7% sensitive and specific, respectively; Xpert PCR, 100% and 98.8% sensitive and specific, respectively. The difference in sensitivity was statistically significant (P < 0.0001) as determined using Fisher's exact test.
All C. difficile NAP1 strains identified by the Xpert PCR were positive for the three gene targets: toxin B, binary toxin, and the tcdC nt 117 gene deletion. All tcdC nt 117 gene deletion-positive samples were positive for both toxin B and the binary toxin, and all binary toxin-positive strains were also toxin B positive. Three strains were toxin B and binary toxin positive but negative for the tcdC nt 117 gene deletion, and this resulted in a classification as toxigenic C. difficile, not NAP1.
In order to evaluate the epidemiologic feature of the Xpert C. difficile assay, 45 Xpert PCR-positive stool samples were tested by both PCR-ribotyping and sequencing of the tcdC gene. The agreement between Xpert PCR and PCR-ribotyping was 93% (42/45) (Table (Table5).5). Of the three discordant stool samples, two were presumptive NAP1 by Xpert PCR (confirmed by sequencing), but the ribopattern was not consistent with a NAP1 strain. The third discordant isolate was NAP1 by PCR ribotyping (confirmed by sequencing) but not by Xpert PCR. The agreement between Xpert PCR and sequencing was 93% (42/45) (Table (Table4),4), with the three discordant isolates identified as NAP1 by sequencing but not by Xpert PCR. The ribopatterns of the three discordant isolates were not consistent with NAP1. The difference between the three methods was not significantly different (P = 0.749) as determined by one-way ANOVA.
Laboratory diagnosis of CDI continues to be an important issue, given the increased incidence and severity of CDI cases (11, 18, 19, 22). At MSKCC, a tertiary care cancer hospital, patients suspected of having C. difficile-associated diarrhea are placed in contact isolation until the result of their C. difficile test is obtained. If the test is negative, isolation is discontinued. If the test is positive, patients are kept in isolation until they have completed 7 days of treatment and have been asymptomatic for at least an additional 48 h following treatment. Since diarrhea is a common complication experienced by hematopoietic stem cell transplants and cancer patients, implementation of a rapid and sensitive test for CDI diagnosis is necessary to improve patient care and reduce the time patients remain in isolation for diarrhea caused by factors other than C. difficile infection.
In the present study, we compared the sensitivity and specificity of our current algorithm of GDH-CYT to those of the Xpert PCR C. difficile assay by using toxigenic culture to resolve discordant results. Using this methodology, GDH-CYT and Xpert PCR had excellent specificity but differed markedly in their sensitivity. Although often viewed as an alternative gold standard for CDI diagnosis, the cytotoxicity assay has been reported to have a sensitivity as low as 56% (5). We determined that the GDH-CYT algorithm in our laboratory has a sensitivity of up to 61%, consistent with what has been reported in the literature.
The sensitivity of the Xpert PCR was 100%, with the assay detecting all potential positive results as confirmed by toxigenic culture. Overall, 7 of the 103 Xpert PCR-positive results were not confirmed by toxigenic culture and could represent potential false-positive results. These seven Xpert PCR-positive results were from six patients with previous C. difficile test results. Two of these seven results were from patients who were previously positive (one 5 days prior and the other 10 days prior) by Xpert PCR, CYT, and/or toxigenic culture; one was from a patient who became positive the following day by CYT and Xpert PCR; and two were from the same patient, 5 days apart, and this patient was tested several times prior to the study and the CYT result was always negative. The last two PCR-positive results were from two patients with previous CYT-negative results as well. After review of these patients' histories, only four results from three patients (those with consistently CYT-negative results) could really be considered false-positive PCR results, as there was no history of C. difficile infection or other clinical indication of CDI aside from the diarrhea in these patients.
Detection of C. difficile organisms by Xpert PCR alone suggests that the assay might be detecting C. difficile carriers. However, determination of carrier state in symptomatic patients would be difficult, and in the MSKCC cancer patient population, where diarrhea is a common complaint, a positive C. difficile result will most likely be considered infection rather than colonization. We were not able to address the relevance or benefits of identifying and treating asymptomatic carriers with our current study, and a previous report showed limited or temporary value in identifying and treating asymptomatic carriers (10).
We excluded the possibility that the Xpert PCR-positive, toxigenic culture-negative results were due to a cross-reacting organism by directly testing the isolates that grew in culture (C. innocuum) by Xpert PCR. In all instances, the results of the Xpert PCR were negative, suggesting that the assay was specific and that the C. innocuum isolates, which are vancomycin resistant (1), probably represented accompanying stool flora. In the two cases where patients had been previously positive for C. difficile by both Xpert PCR and toxigenic culture, it is arguable that the positive PCR result was due to the presence of nonviable organisms. This observation supports the SHEA guidelines, which discourage repeat testing during the same episode of diarrhea, as it is of limited value (3).
The sensitivities and specificities of both the GDH-CYT and Xpert PCR were not affected by inclusion of duplicate specimens from the same patients. Of interest, in several cases, the first diarrheal episode was negative by GDH-CYT and positive by Xpert PCR and only on repeat testing did the GDH-CYT finally became positive. This suggests that the increased sensitivity of the Xpert PCR allowed earlier detection of C. difficile-associated diarrhea than did the GDH-cytotoxin algorithm assay. This is an important point for infection control, since these symptomatic patients were probably removed from isolation based on an initially negative GDH-CYT test result.
Several testing algorithms have been developed for the diagnosis of CDI, including two-step and three-step algorithms using various combinations of toxin A/B EIA, GDH EIA, CYT, and PCR (15, 23, 24). At MSKCC, screening with the GDH EIA was implemented to improve the TAT of C. difficile testing, which was previously done exclusively using the CYT assay. Since approximately 80% of our C. difficile tests were negative, introduction of the GDH EIA as a screening test resulted in a significant decrease in TAT. Although the Xpert PCR detected an additional seven cases when used alone, the differences in sensitivity and specificity between the GDH-PCR algorithm and PCR alone were not significantly different. The difference in the percent agreement between the GDH-CYT and Xpert PCR tests observed in phase 1 and phase 2 (72% versus 94%, respectively) was probably due to the higher number of specimens tested in phase 2. The agreement was calculated using both positive and negative results, and a greater number of negative results was concomitantly detected by both methods in phase 2. Thus, a testing algorithm comprising the GDH EIA followed by the Xpert PCR would represent a sensitive and specific option for C. difficile testing. Our findings are different from those reported by Novak-Weekly et al. (21), who showed a lower assay sensitivity when the Xpert PCR was used following an initial screen with the GDH EIA. Because those authors used toxigenic culture to test all stools, the observed differences between the two studies might be due to a difference in testing methodology.
The ability of the Xpert PCR to detect the NAP1 strain of C. difficile isolates was evaluated against the abilities of PCR-ribotyping and/or sequencing of the tcdC gene. Xpert PCR identifies the NAP1 strain by detecting the tcdB gene, the cdt gene, and the tcdC nt 117 gene deletion. All NAP1 strains identified were positive for all three markers. Agreement between Xpert PCR and the other two methods was greater than 90% (κ, 0.87). The added advantage of a rapid and direct result obtained when using Xpert PCR as opposed to PCR-ribotyping or sequencing makes it a valuable option for collecting surveillance data on NAP1 strain incidence. As drugs are being developed and evaluated for increased efficacy against epidemic strains, knowing that an isolate is a NAP1 strain might be useful in guiding the choice of therapy for treatment of CDI. Unlike typing and sequencing, the Xpert PCR will not provide any information about the prevalence of strains other than NAP1 that could be on the rise in hospital settings (7, 20). Currently, the Xpert C. difficile Epi assay is only available as an RUO assay, and its benefit at this time is mainly to obtain surveillance data on the NAP1 strain of C. difficile. Therefore, the FDA-approved version of the test, Xpert C. difficile, which only detects the toxin B gene, is sufficient and more economical for routine diagnosis of CDI ($50 for the RUO assay versus $36 for the FDA assay).
At MSKCC, when hands-on time, including accessioning of specimens and reporting of results, and material costs of the various algorithms are compared, the GDH-CYT algorithm has a cost of approximately $11.64 (GDH, $4.45; CYT, $7.19), with a hands-on time of 120 min for a GDH-positive sample and a cost of $4.45 with a hands-on time of 70 min for a GDH-negative sample. A GDH-Xpert PCR algorithm would cost $40. 45 (GDH, $4.45; Xpert PCR, $36) with a hands-on time of 95 min for a GDH-positive sample and $4.45 with a hands-on time of 70 min for a GDH-negative sample. Performing the Xpert PCR directly on each specimen would cost $36 with a hands-on time of 35 min. Taking into account the cost of labor and the TAT, we decided to perform the Xpert PCR test on all stool samples at MSKCC. Each laboratory considering the assay would have to determine the algorithm that best fits their environment.
Some of the limitations in our study include the use of the gold standard toxigenic culture only in cases where results were discrepant between GDH-CYT and Xpert PCR. This approach, although more efficient and less cumbersome, could affect the overall sensitivity and specificity of the two methods tested. However, the sensitivity and specificity reported in our study are in the ranges previous reported in the literature based on real-time PCR for detection of C. difficile (reviewed in reference 27). Second, only a subset of the 103 Xpert PCR-positive C. difficile isolates was typed by all three methods under investigation, which could also have affected the overall agreement between the three assays.
In addition to an increase in analytical sensitivity, the greatest impact of adopting the Xpert PCR assay will be its value in effectively reducing the time patients are kept in isolation. With our current algorithm, patients spend a minimum of 12 to 24 h (if the GDH screen is negative) and up to 72 h (if the GDH screen is positive and CYT is negative) in isolation. Outcome studies, measuring the impact of a random-access assay with a 1-hour TAT, will be important in assessing the impact of the Xpert Clostridium difficile PCR on health care-associated costs and patient satisfaction.
We thank the medical technologists and assistants in the Clinical Virology section of the Microbiology Laboratories at Memorial Sloan-Kettering for their assistance with the collection of clinical specimens and performance of the assays.
The study was funded by Memorial Sloan-Kettering Cancer Center.
Published ahead of print on 13 October 2010.