Comparison of BD GeneOhm Cdiff Real-Time PCR Assay with a Two-Step Algorithm and a Toxin A/B Enzyme-Linked Immunosorbent Assay for Diagnosis of Toxigenic Clostridium difficile Infection

Abstract

The BD GeneOhm Cdiff assay, a real-time PCR assay for the detection of the Clostridium difficile toxin B (tcdB) gene, was compared with the toxin A/B (Tox A/B) II enzyme-linked immunosorbent assay (ELISA) and a two-step algorithm which includes a C. Diff Chek-60 glutamate dehydrogenase (GDH) antigen assay followed by cytotoxin neutralization. Four hundred liquid or semisolid stool samples submitted for diagnostic C. difficile testing, 200 GDH antigen positive and 200 GDH antigen negative, were selected for analysis. All samples were tested by the C. Diff Chek-60 GDH antigen and cytotoxin neutralization assays, the Tox A/B II ELISA, and the BD GeneOhm Cdiff assay. Specimens with discrepant results were tested by toxigenic culture as an independent “gold standard.” Of 200 GDH-positive samples, 71 were positive by the Tox A/B II ELISA, 88 were positive by the two-step method, 93 were positive by PCR, and 96 were positive by the GDH antigen assay only. Of 200 GDH-negative samples, 3 were positive by PCR only. Toxigenic culture was performed for 41 samples with discrepant results, and 39 were culture positive. Culture resolution of discrepant results showed the Tox A/B II assay to have detected 70 (66.7%), the two-step method to have detected 87 (82.9%), and PCR to have detected 96 (91.4%) of 105 true positives. The BD GeneOhm Cdiff assay was more sensitive in detecting toxigenic C. difficile than the Tox A/B II assay (P < 0.0001); however, the difference between PCR and the two-step method was not significant (P = 0.1237). Enhanced sensitivity and rapid turnaround time make the BD GeneOhm Cdiff assay an important advance in the diagnosis of toxigenic C. difficile infection.

Clostridium difficile infection (CDI) is emerging as the most common infectious cause of nosocomial diarrhea, yet sensitive and specific commercially available diagnostic tests with rapid turnaround times are lacking (10). Toxigenic culture is considered to be the ultimate reference standard but is tedious, takes up to a week to complete, and is considered too time-consuming for clinical use. While the cytotoxin neutralization assay is the current clinical “gold standard,” it is utilized only by a minority of clinical laboratories because it requires cell culture expertise and up to 48 h to report some positive and all negative results (4). Enzyme-linked immunosorbent assays (ELISA) for detection of toxins A and B (Tox A/B) are the most commonly employed tests, since they use readily available technology, are inexpensive, and have rapid turnaround times, but they lack sensitivity (3, 19). Recently, a two-step protocol has been recommended: testing for an abundant C. difficile antigen, glutamate dehydrogenase (GDH), by a rapid and sensitive ELISA, followed by cytotoxin testing of GDH-positive samples to confirm toxin production in vivo (8, 20, 25, 27). This method achieves relatively high sensitivity and specificity and can rapidly report results for most samples that are negative for C. difficile but can still take up to 48 h to report low-level cytotoxin positivity.

In December 2008, the Food and Drug Administration (FDA) approved the first commercially available real-time PCR assay (the BD GeneOhm Cdiff assay; BD Diagnostics, San Diego, CA) to directly detect the toxin B (tcdB) gene in stool to aid in the diagnosis of CDI. Reports of two prospective studies comparing the BD GeneOhm Cdiff assay, a cytotoxicity assay, and toxigenic culture have been published (2, 24). Both found the BD GeneOhm Cdiff assay to have higher sensitivity than the cytotoxicity assay by using toxigenic culture as the gold standard. Neither study compared the BD GeneOhm Cdiff assay to a toxin ELISA, which is the most widely used diagnostic method for CDI, or to a two-step testing algorithm. Other studies of PCR assays reported in the literature utilized in-house tests with small numbers of positive results, making it difficult to propose general recommendations (1, 5, 9, 15, 17, 23, 26).

The objective of this study was to compare the performance of the BD GeneOhm Cdiff PCR assay for detection of the C. difficile toxin B gene with that of a two-step method (the C. Diff Chek-60 GDH antigen assay followed by cytotoxin neutralization) and that of the Tox A/B II ELISA. Toxigenic culture was used to resolve findings for samples with discrepant results.

(This research was presented at the 109th General Meeting of the American Society for Microbiology, Philadelphia, PA, 17 to 21 May 2009.)

MATERIALS AND METHODS

Clinical samples.

Liquid or semisolid stool samples obtained from patients hospitalized at Yale-New Haven Hospital and submitted for C. difficile testing from August to December 2008 were entered into the study. All samples were tested within 24 h of receipt with the C. Diff Chek-60 GDH antigen ELISA as part of the hospital's standard two-step diagnostic routine. On each study day, all samples testing positive for the C. difficile GDH antigen with a sufficient amount of available stool, as well as an equivalent number of stool samples testing negative for the GDH antigen, were selected for further analysis. All study samples were subsequently tested by the cytotoxin neutralization method, the Tox A/B II ELISA, and the BD GeneOhm Cdiff PCR assay (Fig. 1). ELISA and PCR analyses were performed by study personnel blinded to the results of the two-step method. When ELISA or PCR analysis could not be performed on the same day as the cytotoxin neutralization assay, samples were frozen and thawed only once according to the assay manufacturers' instructions. An aliquot of each original stool sample was saved at −70°C for further testing. Samples that did not have positive results from all four tests or negative results from all four tests, excluding samples positive by the GDH antigen assay only, were sent for toxigenic culture. Samples with discrepant results from patients who were receiving treatment for CDI at the time of sample collection were excluded from analysis. Only two samples per patient in a 7-day period were included. Repeat samples sent on the same day were excluded.

FIG. 1.

Algorithm for testing of stool samples.

Two-step method: C. Diff Chek-60 and cytotoxicity assays.

The C. Diff Chek-60 assay was performed according to the instructions of the manufacturer (TechLab, Blackburg, VA). Briefly, 0.1 ml of a specimen was added to 0.4 ml of diluent, and the sample was subjected to a vortex and then centrifuged at 5,000 × g for 10 min. To each test well, 0.05 ml of a conjugate solution was added, followed by 0.1 ml of centrifuged specimen. After incubation for 50 min at 37°C, wells were washed for seven cycles, a substrate was added, and wells were incubated at room temperature for 10 min. Stop solution was added, and the optical density at the spectrophotometric dual wavelengths of 450 and 620 nm was measured. Stool samples received before 1:00 p.m. were tested on the same day. Samples received after 1:00 p.m. were stored at 4°C and tested within 24 h. A positive result corresponded to an optical density of ≥0.080, and a negative result corresponded to an optical density of <0.080.

Stool samples positive for GDH antigen were tested by the cytotoxicity assay. Stool samples (0.5 ml) were added to 0.5 ml of phosphate-buffered saline with antibiotics (vancomycin, gentamicin, and amphotericin B) and then subjected to a vortex, and the toxin was allowed to elute for 5 min. After centrifugation for 10 min in a microcentrifuge, the supernatant was removed and passed through a 0.45-μm-pore-size filter. Then 20-μl aliquots of filtrate in serial 10-fold dilutions (1:10 to 1:10,000) were inoculated in duplicate onto foreskin fibroblast monolayers (MRHF cells; BioWhittaker, Walkersville, MD) in 96-well plates. C. difficile antitoxin (20 μl; TechLab, Inc., Blacksburg, VA) was added to one of the duplicate wells inoculated with the 1:10 and 1:100 dilutions. Thus, after the addition of antitoxin, the final dilution in the first culture well was 1:20. Monolayers were read at 4, 24, and 48 h after inoculation by using an inverted microscope. A known positive control, run with each assay, was required to show cytotoxicity in the expected range. A positive result consisted of cytotoxicity that was neutralized by C. difficile antitoxin. Results were recorded as the highest dilution showing specific cytotoxicity. All study samples underwent the cytotoxicity assay on the same day of receipt or within 24 h of receipt if the sample was received after 1:00 p.m.

Tox A/B II ELISA.

The Tox A/B II ELISA was performed according to the instructions of the manufacturer (TechLab, Blacksburg, VA). Briefly, 0.05 ml of a specimen was added to 0.2 ml of specimen diluent, and the sample was then centrifuged at 5,000 × g for 10 min. To each test well, 0.05 ml of a conjugate solution was added, followed by 0.1 ml of centrifuged specimen. After incubation for 50 min at 37°C, wells were washed for seven cycles, a substrate was added, and wells were incubated at room temperature for 10 min. Stop solution was added, and the optical density at the spectrophotometric dual wavelengths of 450 and 620 nm was measured. A positive result corresponded to an optical density of ≥0.080, and a negative result corresponded to an optical density of <0.080. Samples not tested within 24 h of receipt were stored at −20°C and tested within 72 h.

BD GeneOhm Cdiff PCR assay.

The BD GeneOhm Cdiff assay was performed directly on stool specimens according to the instructions of the manufacturer (BD Diagnostics, San Diego, CA). It utilizes real-time PCR to amplify the toxin B (tcdB) gene from C. difficile and fluorogenic target-specific hybridization probes for the identification of amplified target DNA. An internal control which is based on amplification of a 277-bp DNA sequence not found in C. difficile is employed. Amplification and detection and the interpretation of the assay results are done using the SmartCycler instrument (Cepheid, Sunnyvale, CA). Results were reported as positive, negative, or unresolved. Unresolved results can be due to technical errors or inhibitors in the sample lysate. According to the manufacturer's instructions, lysates of samples with unresolved results were frozen at −70°C and retested after thawing. The entire procedure required about 2 h, depending on the number of samples being run. All samples not tested within 24 h were stored at −20°C and tested within 5 days.

Toxigenic culture.

One 2-ml vial of stool from each of the 400 samples was saved and stored at −70°C until the completion of the study, after which coded samples with discrepant results were tested by toxigenic culture at Montefiore Medical Center, Bronx, NY, by personnel blinded to prior test results. Stool samples were treated with ethanol to kill nonspore flora and inoculated in parallel onto selective cycloserine-cefoxitin-fructose agar (CCFA) supplemented with 0.1% taurocholate (Sigma, St. Louis, MO) and into chopped meat broth (BD BBL, Sparks, MD) supplemented with 0.1% taurocholate, 250 μg/ml cycloserine, and 16 μg/ml cefoxitin. If there was visible growth in the broth after 48 h or at 5 to 7 days (late growth), 0.1 ml was subjected to the Premier toxin A and B ELISA according to the protocol of the manufacturer (Meridian Bioscience, Cincinnati, OH). A positive result (optical density, >0.10) supported the detection of toxigenic C. difficile. Colonies of growth on the agar that had the appearance of C. difficile (flat yellow colonies) were scored from 1 to 10 individual colonies or the number of quadrants with confluent growth (with a score of 1+ to 4+). Colonies were then tested by PCR in an internally validated PCR analysis for the putative toxin repressor gene tcdC (using primers 5′-TCTAGCTAATTGGTCATAAG-3′ and 5′-AATAGCAAATTGTCTGAT-3′), as well as an analysis for the GDH gene (gdh) using published primers (27). The tcdC gene is tightly linked to tcdB on the pathogenicity locus uniquely found in toxigenic strains of C. difficile and thus is a reliable indicator of the presence of such strains. All PCRs were performed with FastStart Hi-Fidelity Taq PCR reagents (Roche, Indianapolis, IN) and MgCl₂ (2.5 mM final concentration of Mg) on a Perkin-Elmer 2400 thermocycler using a multiplex PCR protocol consisting of 5 min at 95°C followed by 45 cycles of 94°C for 1 min, 52°C for 1 min, and 72°C for 2 min. PCR amplicons were resolved on 2% agarose gels stained with ethidium bromide. A positive ∼200-bp band from the tcdC PCR supported the detection of toxigenic C. difficile, and a ∼750-bp gdh band confirmed the presence of C. difficile. Specimens that did not have consistent results from the Tox A/B ELISA for broth culture samples and the toxin gene PCR had all tests with discordant results repeated; in addition, the broth culture was subcultured onto agar to identify individual colonies which could be analyzed by PCR. If there was a negative toxin ELISA result with the broth culture but PCR-positive colonies were detected on agar, the colonies were directly inoculated into broth and tested by toxin ELISA after 48 h.

Discrepancy analysis.

Results were considered to be discrepant if one, two, or three tests were positive, except for samples positive by the GDH antigen assay only. Samples positive by the GDH antigen assay only were considered to represent colonization with nontoxigenic strains of C. difficile. To resolve discrepant results, three steps were taken. First, all samples for which PCR results were discordant with two-step assay results were retested by PCR. Second, all samples with discrepant results were submitted for toxigenic culture. Lastly, chart review was conducted for all patients whose samples yielded discrepant results.

Statistical analysis.

Statistical analysis was performed using McNemar's test.

RESULTS

A total of 434 samples were initially tested. Of these, 18 were excluded because the patients were found to be receiving treatment for CDI and 16 were excluded because there were more than 2 samples per patient sent in a 7-day period. Four hundred stool samples from 341 patients were included in the final analysis.

Overall, 66 samples were positive by all four tests. Of the 200 GDH-positive samples, 71 were positive by the Tox A/B II assay, 88 were positive by the two-step method, 93 were positive by the BD GeneOhm Cdiff assay, and 96 were positive by the GDH antigen assay only. Of 200 GDH-negative samples, 3 were positive by PCR only and 197 were negative by all four tests. Results are shown in Table 1. Results for 2 of 400 samples (0.5%) were initially unresolved by PCR but became negative upon repeat testing. Of the 88 cytotoxin-positive samples in this study, 29 (33.0%) were positive at 4 h, 48 (54.5%) were positive at 24 h, and 11 (12.5%) were positive at 48 h. Six of the 11 samples positive at 48 h were positive only at the 1:20 starting dilution.

TABLE 1.

Results showing discrepancies

No. of tests with initially positive result	GDH antigen result	Cytotoxin result	Tox A/B ELISA result	PCR result	No. of samples with indicated resultsa	No. positive by toxigenic culture	No. of true positives
4	+	+	+	+	66	NDb	66
1	+	−	−	−	96	ND	0
0	−	−	−	−	197	ND	0
3	+	+	+	−	4	4	4
3	+	+	−	+	12	12	12
2	+	+	−	−	6	5	5
2	+	−	−	+	15	15	15
2	+	−	+	−	1	0	0
1	−	−	−	+	3	3	3
Total					400	39	105

^aIn total, 41 samples with discrepant results were tested by toxigenic culture and 39 were positive.

^bND, not determined.

Forty-one of the samples were considered to have discrepant results (Table 1) and were cultured. Thirty-nine of these were toxigenic culture positive, and two were culture negative. Fifteen cytotoxin-negative samples were GDH antigen and PCR positive, and three were PCR positive only. All of these 18 samples were toxigenic culture positive. Ten samples with discrepant results were cytotoxin positive and PCR negative, and nine of these were found to be toxigenic culture positive (true positives). Five of the nine PCR-negative samples were cytotoxin positive at 4 or 24 h, and four were positive at 48 h. Two of the nine eluates stored at −70°C were found to be PCR positive upon retesting. Twelve of 41 samples with discrepant results were negative by the Tox A/B II assay and positive by all other tests, including toxigenic culture. Of the remaining two toxigenic culture-negative samples, one was positive by the GDH antigen and Tox A/B II assays and the other was positive by the GDH antigen and cytotoxin assays at 1:20 only. These were designated false positives.

At the time of the study, the BD GeneOhm Cdiff assay was not an FDA-approved test, so PCR results were not clinically reported. Eight patients with PCR-positive, cytotoxin-negative samples were treated for CDI for a median of 1.5 days until the negative cytotoxin result was reported, and then treatment was stopped. Ten patients with these results received no treatment. Only 5 of the 18 patients with PCR-positive, cytotoxin-negative samples had additional testing over 6 to 8 months of follow-up, and only 1 of these had a subsequent cytotoxin-positive sample (22 days after collection of the sample included in the study).

The results of all tests after resolution by toxigenic culture are given in Table 2. After the resolution of results by culture, the Tox A/B II assay was found to have detected 70 (66.7%), the two-step method was found to have detected 87 (82.9%), and PCR was found to have detected 96 (91.4%) of 105 true positives. There was 93.0% concordance between results from PCR and those from the two-step method and 91.3% concordance between results from PCR and those from the Tox A/B II assay. The average turnaround times and materials cost per test in our laboratory, based on a quantity of 5,000 samples a year, are given in Table 3.

TABLE 2.

Comparison of GeneOhm Cdiff real-time PCR assay, a two-step algorithm, and a Tox A/B II ELISA after resolution of discrepant results by toxigenic culture

Assayc	Result	No. with result after resolution by toxigenic culturea		Assay performance characteristicb
		Positive	Negative	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
GeneOhm real-time PCR				91.4	100	100	97.0
	Positive	96	0
	Negative	9	295
	Total	105	295
Two-step GDH/cytotoxin algorithm				82.9	99.7	98.9	94.2
	Positive	87	1
	Negative	18	294
	Total	105	295
Tox A/B II ELISA				66.0	99.7	98.6	89.4
	Positive	70	1
	Negative	35	294
	Total	105	295

^aResults for samples with discrepant results were resolved by toxigenic culture.

^bCalculations make the assumptions that samples with all test results positive would be toxigenic culture positive and that samples that were either positive by GDH antigen analysis only or negative by all tests would be toxigenic culture negative. Only a subset of GDH-negative samples were tested. PPV, positive predictive value; NPV, negative predictive value.

^cThe difference between results from PCR and those from the two-step algorithm was not significant (P = 0.1237; McNemar's test). The difference between results from PCR and those from the Tox A/B II ELISA was significant (P < 0.0001; McNemar's test).

TABLE 3.

Comparison of the turnaround times for results and the material costs for different methods

Test	Turnaround time (h)	Materials cost per testa
BD GeneOhm Cdiff assay	2-24	$25.83
Tox A/B II ELISA	2-24	$4.12
Two-step (GDH/cytotoxin) algorithm	6-48b	$5.02
Cytotoxin test only	4-48c	$3.97

^aActual costs in our laboratory. Costs for labor, controls, and repeats are not included.

^bWith the two-step method, GDH antigen-negative results are reported within 2 to 24 h. GDH-positive samples are tested for cytotoxin, and cytotoxin results are reported after 4, 24, and 48 h of incubation.

^cCytotoxin-positive results are reported at 4, 24, and 48 h, and all negative results are reported at 48 h.

DISCUSSION

This is the first study comparing the performance characteristics of the commercial BD GeneOhm Cdiff assay to those of the Tox A/B II assay and the two-step GDH antigen ELISA/cytotoxin protocol for diagnosis of CDI. The BD GeneOhm Cdiff assay was more sensitive than the Tox A/B II assay (91.4% versus 66.0%; P < 0.0001). There was an absolute increase in the number of cases detected by the BD GeneOhm Cdiff assay compared to the number detected by the two-step method (sensitivity, 91.4% versus 82.9%), but the difference was not statistically significant in our study (P = 0.1237). The level of concordance of the BD GeneOhm Cdiff assay with the two-step method (93.0%) is similar to those of the BD GeneOhm Cdiff assay with the cytotoxicity assay as reported by Stamper et al. (94.8%) and Barbut et al. (92%) (2, 24).

All 18 two-step assay-negative, PCR-positive samples with discrepant results in our study were toxigenic culture positive. In contrast, Stamper et al. (24) reported that 6 (35.3%) of 17 PCR-positive, cytotoxin-negative samples failed to yield toxigenic C. difficile upon culture. While this may reflect differences in the culture methods used, further study is merited.

Since PCR results were not clinically reported during our study, the 18 PCR-positive samples were reported to be C. difficile negative and patients received brief or no treatment for CDI, with no apparent adverse consequences. Only 1 (5.5%) of 18 patients had subsequent C. difficile disease 22 days later. This outcome highlights a concern for highly sensitive molecular amplification tests that target toxin genes, rather than in vivo toxin production. Patients may be colonized with toxigenic C. difficile but have diarrhea due to other causes. There is evidence that carriers of toxigenic C. difficile, who may be inadvertently identified by PCR due to inappropriate sample submission, may have immune mechanisms that inhibit the toxin effects of their strain (11, 12, 13). The effects of eradicating asymptomatic carriage by antibiotics on the immune response of the carrier are not known. Thus, clinical correlation is even more essential for accurate diagnosis of CDI in patients diagnosed by PCR to avoid treating patients unnecessarily (18).

It is also possible that these PCR-identified patients have low titers of toxin that are below the detection limits of toxin ELISA or cytotoxin neutralization assays. These cases may represent a small pathogen burden. In the 15 to 23% of CDI patients with mild disease, the symptoms may be cleared by simply removing the inciting antibiotics (22). However, these patients may still be at risk of spreading C. difficile spores in the hospital setting, because they test negative by conventional methods and are neither isolated nor treated. In theory, if these patients are detected by PCR, isolated, and/or treated, nosocomial transmission to other, more susceptible patients could be reduced (7). It should also be noted that identification of these additional patients by PCR may increase the number of patients requiring isolation and prolong their hospitalization. In summary, the costs, the benefits, and the consequences of treating patients positive only by PCR merit further study.

Although PCR detected more positives than other methods in this study, 10 samples were two-step method positive and PCR negative. Nine of these 10 samples yielded toxigenic C. difficile in culture, and 1 was culture negative. The one sample that was two-step method positive but toxigenic culture negative had a low toxin titer of 1:20 at 48 h. This was considered a false-positive cytotoxin result. Reviews of charts for the nine patients with cytotoxin-positive, PCR-negative samples revealed multiple prior or subsequent C. difficile cytotoxin-positive stools from four (44.4%) of these nine patients. Eight of the nine toxigenic culture-positive samples had very low yields of bacteria upon agar culture (with a score of 1+ or lower by a semiquantitative scoring system as described in Materials and Methods) and thus may have had bacterial levels below the detection limit of the PCR assay. Upon retesting of the frozen PCR lysates, 2 of the 10 two-step assay-positive, PCR-negative samples were PCR positive, implying some degree of inhibition or sampling variability in the initial testing. Stamper et al. (24) reported only one cytotoxin-positive, toxigenic culture-positive sample missed by PCR, and Barbut et al. (2) reported none. These differences may be explained by differences in cytotoxin testing methods. Cytotoxicity testing in our institution is performed by virology laboratory personnel, using freshly prepared cell culture plates and starting at a lower final dilution (1:20) than most laboratories employ, with serial 10-fold dilutions of the sample, and results are read at 4, 24, and 48 h, with almost 90% of positive results reported within 4 to 24 h. Our cytotoxin results may be superior to those obtained using higher starting dilutions of 1:50 (24) or 1:100 (2) and commercially prepared cell cultures. Thus, the BD GeneOhm Cdiff assay may perform even better compared to commercial cytotoxin neutralization assays.

Nevertheless, the failure of PCR to detect nine true positives detected by the two-step method was a concern. For several reasons, it is unlikely that the BD GeneOhm Cdiff PCR assay failed to detect these nine samples due to inhibitors alone. First, each reaction mix has an internal amplification control to monitor reagent integrity and PCR inhibition. Inhibitory samples have an internal control result of “fail” and a BD GeneOhm Cdiff assay result of “unresolved”. The nine samples passed the test for internal control amplification. Second, eight of the nine samples had very low yields of C. difficile upon direct anaerobic culture, with <10 CFU recovered from five of those samples. In addition, two of the stool samples were positive upon repeat testing of the sample lysates, suggesting sample variability in the runs. These findings suggest that the inability of the BD GeneOhm Cdiff assay to detect C. difficile in these samples was primarily due to bacterial loads below the level of detection of the assay. Thus, dilution of the samples would not likely improve PCR detection.

In addition to low numbers of bacteria and the presence of inhibitors, another possible explanation for the failure of the BD GeneOhm Cdiff assay to detect samples that were positive by the two-step method is genetic variance at the tcdB locus, leading to mismatch of PCR primers. While the majority of isolates of toxigenic C. difficile come from toxinotype variants with an intact tcdB gene, mutations and deletions in tcdB have been documented previously (6, 21). Researchers in a recent study discovered a pathogenic C. difficile strain that contains the toxin A and binary toxin genes but is negative for the toxin B gene (16). Reports of such toxigenic C. difficile variants are very rare, as toxin B appears to be an essential virulence factor (14). The tcdB-negative strain is genetically distinct from the current North American epidemic BI/NAP1/027 strain. Preliminary data from one of us (P.F.R.) shows that, in testing over 70 isolates of either epidemic (ribotype 27) or nonepidemic strains, the BD GeneOhm PCR is equally efficient at detecting both types. However, monitoring for the evolution of new variants of tcdB is warranted.

The BD GeneOhm Cdiff assay was simple to perform and produced results in approximately 2 h, compared to 2 h for the Tox A/B II ELISA and 6 to 48 h for the two-step method. The Tox A/B II ELISA failed to detect 35 toxigenic culture-positive samples and yielded one false-positive result, consistent with the previously documented low sensitivities of this class of tests (3, 19). Cytotoxin neutralization testing as performed in our virology laboratory detected nine true positives missed by PCR. The PCR method is more expensive than other methods but may reduce nosocomial transmission of toxigenic C. difficile and thus lead to long-term savings for hospitals and patients. Of note, the costs for each test will vary among institutions depending on test volume, contracts with suppliers, and other factors. In our virology laboratory, cytotoxin neutralization is an inexpensive test. Thus, to test 5,000 samples a year using the BD GeneOhm Cdiff assay as the sole assay would cost an additional $100,000 in reagents.

There were several limitations of this study. Only the samples with discrepant results were cultured, and not all samples were included in the study. The performance characteristics of the various assays were compared to toxigenic culture as the gold standard, with the assumptions that samples for which all four tests were positive would be toxigenic culture positive and that those for which all four tests were negative would be toxigenic culture negative. All GDH antigen-positive samples that were submitted on study days and met study criteria, but only a subset of GDH antigen-negative samples, were included. In addition, samples were sent to a separate institution for toxigenic culture, and different assays were used to determine whether strains of C. difficile were toxigenic. Specifically, a Meridian Tox A/B ELISA and an in-house PCR assay to detect tcdC (the toxin repressor gene) were utilized, while the TechLab Tox A/B ELISA and the tcdB-based PCR assay were used in the initial analysis. Thus, some of the divergent results may be due to different performance attributes of these particular assays. However, it is likely that a greater difference was attributable to the fact that the toxigenic culture assays were applied to an amplified culture rather than a crude stool sample.

Review of charts for patients with samples yielding discrepant results revealed hospital-wide problems with stool sample submission for C. difficile analysis, including submission of samples from patients with minimal diarrhea and from patients who were already receiving treatment and submission of multiple samples from the same patient on the same day or within a 7-day period. Though samples with discordant results obtained from patients found to be on therapy were excluded from analysis in the study, these problems were illustrative of inappropriate clinical use of testing. To both avoid unnecessary treatment and reduce costs of PCR testing, clinicians will need to be educated about limiting C. difficile testing to patients with a reasonable probability of having disease, such as those patients having three or more loose stools per day for 1 to 2 days (18).

In conclusion, the BD GeneOhm Cdiff assay was more sensitive than a commonly used toxin ELISA and as sensitive and specific as a two-step method for detection of toxigenic C. difficile. It is more expensive than these methods but has a faster turnaround time than the two-step assay, which could lead to earlier diagnosis of CDI and reduce nosocomial transmission.