Clostridium difficile Testing in the Clinical Laboratory by Use of Multiple Testing Algorithms

Abstract

The incidence of Clostridium difficile infection (CDI) has risen almost 3-fold in the United States over the past decade, emphasizing the need for rapid and accurate tests for CDI. The Cepheid Xpert C. difficile assay is an integrated, closed, nucleic acid amplification system that automates sample preparation and real-time PCR detection of the toxin B gene (tcdB). A total of 432 stool specimens from symptomatic patients were tested by a glutamate dehydrogenase (GDH) assay, a toxin A and B enzyme immunoassay (EIA), the Xpert C. difficile assay, and a cell culture cytotoxicity neutralization assay (CCCN). The results of these methods, used individually and in combination, were compared to those of toxigenic culture. Results for the Xpert C. difficile assay alone showed a sensitivity, specificity, positive predictive value, and negative predictive value (NPV) of 94.4, 96.3, 84.0, and 98.8%, while the EIA alone gave corresponding values of 58.3, 94.7, 68.9, and 91.9%, respectively. An algorithm using the GDH assay and the EIA (plus the CCCN if the EIA was negative) showed corresponding values of 83.1, 96.7, 83.1, and 96.1%. The Xpert C. difficile assay was statistically superior to the EIA (P, <0.001 by Fisher's exact test) and to the GDH-EIA-CCCN algorithm (P, 0.0363). Combining the GDH and Xpert C. difficile assays lowered both the sensitivity and the NPV of the Xpert assay. The GDH-EIA-CCCN procedure required, on average, 2 days to complete testing on GDH-positive results, while testing by the Xpert C. difficile assay was completed, on average, in less than 1 h. Xpert C. difficile testing yielded the highest sensitivity and NPV, in the least amount of time, of the individual- and multiple-test algorithms evaluated in this study.

Clostridium difficile is the main cause of infectious health care-associated diarrhea in the United States and around the world. C. difficile infections (CDI) can vary from a mild diarrhea to the potentially fatal pseudomembranous colitis, toxic megacolon, and sepsis (7, 19). C. difficile colonization of the bowel often follows disruption of normal flora after the patient receives antimicrobial therapy. The incidence and severity of C. difficile has increased in both hospital and long-term care settings, due in part to the emergence of several novel strains, including the epidemic J strain described by Johnson et al. and the hypervirulent NAP1/027/BI strain (16, 19). Strain NAP1/027/BI produces high quantities of spores, which disseminate easily in the hospital environment, and is associated with high mortality rates (1, 5, 9, 11, 17, 32).

Historically, the cell culture cytotoxicity neutralization assay (CCCN), which detects cytotoxin production in monolayers of cells, such as human diploid fibroblasts, has been the gold standard for C. difficile detection in the laboratory. However, cell culture is labor-intensive, and many laboratories have adopted other testing methods, such as enzyme immunoassays (EIAs) for toxins A and B, which are easier and faster to perform than CCCN (22, 27, 33). However, recent reports have highlighted the lack of sensitivity of the toxin A/B EIAs, which show sensitivities as low as 48% (2, 28). Although toxigenic culture of the organism has now been reaccepted as the true gold standard (25), this method requires substantial laboratory resources, and results are not available in a short enough time frame to be clinically useful (18, 24, 28). Thus, other approaches to improving both the sensitivity and the cost-effectiveness of C. difficile testing have been introduced (2, 22). Testing algorithms using a glutamate dehydrogenase (GDH) assay (which has presumptively higher sensitivity but lacks specificity) to screen for C. difficile in stool samples, with reflex testing using a more specific assay, such as a toxin A/B EIA or the CCCN, have been proposed (26, 29, 31). GDH assays detect antigen present in both toxigenic and nontoxigenic strains of C. difficile directly in stool samples. The time necessary to perform the GDH assay with EIA or CCCN confirmation can be as long as 3 days (34). Gilligan noted that EIAs often lack sufficient sensitivity for confirmation of positive GDH assay results (14). In this algorithm, the need to confirm GDH-positive specimens increases the turnaround time (TAT) for positive results, delaying the notification of the physician ordering the test. PCR assays for various targets have been developed as a potential replacement for the less-sensitive (EIA) and less-specific (GDH) assays for C. difficile detection (3, 4, 6, 23, 30). Such assays include both “home brew” PCR assays and FDA-cleared commercial assays (15, 20, 28, 30). Cepheid (Sunnyvale, CA) has recently developed a GeneXpert cartridge-based assay for detection of the C. difficile toxin B gene (tcdB) directly from stool. In this study, we compared the sensitivity and specificity of the Xpert C. difficile PCR assay to those of the GDH assay and the EIA, individually and within specific testing algorithms, using toxigenic culture as the gold standard for a positive specimen.

(This work was previously presented as a poster at the 109th General Meeting of the American Society for Microbiology, 2009.)

MATERIALS AND METHODS

Study population and sample collection.

This was a prospective study conducted at the Southern California Permanente Medical Group Regional Reference Laboratories. Eligible patients included those with suspected CDI for whom toxin EIAs from unformed stool were ordered for C. difficile testing according to the institution's standard practices. All specimens enrolled and tested in this study represented excess, leftover stool, and therefore, informed consent was waived by the Institutional Review Board. Duplicate specimens from the same patient and patients under the age of 2 years were excluded.

Upon the establishment of eligibility and following standard testing at the institution, excess unformed stool specimen was placed in a sterile container. A portion of the specimen was shipped to a single central laboratory for reference culture. The remaining unformed stool was used for testing by the EIA, the GDH assay, and the Xpert C. difficile assay on site. Those specimens that were negative for C. difficile by the EIA but positive by the GDH assay were further tested for cytotoxin B by the CCCN.

Enzyme immunoassay.

Toxin testing was performed using the Premier Toxins A & B microwell EIA (Meridian Bioscience, Inc., Cincinnati, OH). The assay was performed on previously frozen stool specimens according to the manufacturer's instructions. Positive results for the Premier Toxins A & B EIA are indicated by optical densities at 450 and 630 nm (OD_450/630) of ≥0.100, and negative results are indicated by OD_450/630 of <0.100.

Glutamate dehydrogenase.

Specimens were tested for GDH using the TechLab C. DIFF CHEK-60 EIA (distributed by Inverness Medical Innovations, Inc., Princeton, NJ). Testing was performed on previously frozen stool specimens according to the manufacturer's recommendations. Negative results for GDH antigen are indicated by OD_450/630 of <0.80, and positive results are indicated by OD_450/630 of ≥0.80. Specimens that were positive for GDH were tested by EIA. GDH-positive, EIA-positive specimens were considered positive for toxin B-producing C. difficile. Those specimens that were GDH positive and EIA negative were tested by the CCCN. GDH-positive, EIA-negative, CCCN-positive specimens were considered positive for toxin B-producing C. difficile.

Xpert C. difficile PCR assay.

The Xpert C. difficile PCR assay (Cepheid, Sunnyvale, CA) was performed according to the manufacturer's instructions. Briefly, a swab was dipped into the unformed stool specimen container. The swab was placed in sample reagent and capped. The specimen was vortexed for 10 s, and all the liquid from the sample reagent was transferred to the “S” chamber of the cartridge using a large transfer pipette. Next, reagent 1 was added to chamber 1 of the test cartridge. Finally, reagent 2 was added to chamber 2 of the test cartridge, and the lid was closed. The cartridge barcode was scanned and placed in the GeneXpert instrument. Any Xpert C. difficile assay not yielding a result on the first attempt was repeated using new reagents and a new cartridge. If no result was obtained upon retesting, the specimen was reported as unresolved.

Cell culture cytotoxicity neutralization assay.

Approximately 1 ml of fecal material was transferred to M4 transport medium, centrifuged at 2,500 rpm for 15 min, and filtered with a 0.45-μm-pore-size Millex-HA filter unit. Next, 250 μl of filtered stool was inoculated into two tubes of Vero cells (Diagnostic Hybrids, Inc., Athens, OH). Simultaneously, a preincubated filtrate antitoxin mixture (TechLab, Blacksburg, VA) was inoculated into the second tube. Tubes were incubated at 35 ± 2°C for 3 days. Cells were observed daily for cytopathic effect in the first tube and neutralization (no toxic effect) in the second tube, using an inverted microscope at a total magnification of ×63. Specimens that were negative for C. difficile toxin (CDT) showed cells that remained normal in appearance, while specimens positive for CDT demonstrated cytotoxicity. The lack of specific cytopathic effect in the antitoxin-neutralized well confirms CDT.

Reference culture methods.

All bacterial cultures were performed at a single centralized reference laboratory that was blinded to the other test results. An aliquot of the unformed stool specimen was shipped on ice packs in anaerobic transport medium (Anaerobe Systems, Morgan Hill, CA). Toxigenic culture was initiated within 2 days of sample receipt; all manipulations took place under an anaerobic atmosphere. Stool was inoculated into prereduced cycloserine-cefoxitin-fructose direct agar (CCFA-D) and cycloserine-cefoxitin-mannitol broth with taurocholate lysozyme cysteine (CCMB-TAL) (Anaerobe Systems). The CCFA-D plate was incubated at 35°C to 37°C for 48 h; CCMB-TAL was incubated at 35°C to 37°C for 24 h. CCMB-TAL was subcultured at 24 h to a second CCFA plate (CCFA-E [enriched]). If the CCFA-D plate was positive for C. difficile (see the testing protocol below), the CCFA-E plate was discarded. If the CCFA-D plate was negative for C. difficile, the CCFA-E plate was considered as described below.

Aerotolerance testing was performed from the CCFA plate. The appearance of large Gram-positive rods that were obligate anaerobes from CCFA and were susceptible to 5 μg of vancomycin was considered presumptive evidence of C. difficile. These isolates were subcultured, using approximately 2 to 4 colonies, to chopped meat broth with carbohydrates (CMC) and were incubated for 2 days. The presence of iso acids as an end product of glucose fermentation by gas-liquid chromatography served as confirmation of identification. A cell-free supernatant was then tested by a cytotoxin assay similar to that described above (instead using tube cultures of human foreskin fibroblasts; Diagnostic Hybrids, Inc.) to detect C. difficile toxin B. A C. difficile antitoxin was used to demonstrate the specificity of the tissue culture reaction by neutralizing the cytotoxin present in the supernatant.

Analysis.

A sample was considered to contain toxigenic C. difficile if the CCCN for cytotoxin B from a culture supernatant was positive. Performance characteristics were calculated for the GDH assay, the EIA, and the Xpert C. difficile assay relative to the toxigenic culture result. Accuracy is defined as the percentage of overall agreement between the two tests or algorithms being compared. Performance was further investigated by considering the results of the GDH assay and EIA combined versus toxigenic culture; the GDH and Xpert C. difficile assays combined versus toxigenic culture; and GDH, EIA, and CCCN testing combined versus toxigenic culture.

Statistical methods.

This study was conducted in order to compare several different methods/algorithms for detecting C. difficile. The equivalency of these platforms/algorithms was categorically analyzed with 2 × 5 tables using the chi-square test and Fisher's exact test under the Freq procedure in the SAS system, version 9. Fisher's exact test calculates the actual (nonasymptotic) probability of observing the actual results (or something more extreme) given that the different methods are in fact equivalent. A P value of >0.05 by Fisher's exact test indicates that the categories are statistically equivalent. The chi-square test gives an asymptotic probability for the same data. The P value is the probability of obtaining the observed result, or something more extreme, given that all the tests are in fact equivalent. This gives the statistical significance of the test and is used to test the null hypothesis of no difference between the different methods.

RESULTS

Clinical performance.

The performances of the GDH assay, EIA, and Xpert C. difficile assay as individual indicators and as parts of testing algorithms for the presence of toxin B-producing C. difficile were assessed for 432 stool samples; the results of each assay were compared to the results of toxigenic culture. In this study, both direct toxigenic and enriched toxigenic cultures were performed. If the direct toxigenic culture was negative, then the results of the enriched toxigenic culture were evaluated. Four specimens failed to yield a valid PCR result by use of the Xpert C. difficile assay (n = 428). For the GDH-EIA-CCCN algorithm, one cytotoxin specimen gave no result, due to bacterial contamination (n = 431). Only four samples yielded nontoxigenic C. difficile isolates. Relative to toxigenic culture, the Xpert C. difficile assay demonstrated a sensitivity, a specificity, and positive and negative predictive values of 94.4, 96.3, 84.0, and 98.8%, respectively, while the EIA demonstrated a sensitivity, a specificity, and positive and negative predictive values of 58.3, 94.7, 68.9, and 91.9%, respectively (Table 1).

TABLE 1.

Summary of algorithm versus stand-alone testing options compared to direct/enriched toxigenic culture

Parametera	Test(s)
	EIA only	GDH + EIA	GDH + EIA + cytotoxinb	GDH + Xpertc	Xpert onlyd
No. of specimens	432	432	431	432	428
Sensitivity	58.3 (42/72)	55.6 (40/72)	83.1 (59/71)	86.1 (62/72)	94.4% (68/72)
Specificity	94.7 (341/360)	98.3 (354/360)	96.7 (348/360)	97.8 (352/360)	96.3 (343/356)
Accuracy	88.7 (383/432)	91.2 (394/432)	94.4 (407/431)	95.8 (414/432)	96.0 (411/428)
PPV	68.9 (42/61)	87.0 (40/46)	83.1 (59/71)	88.6 (62/70)	84.0 (68/81)
NPV	91.9 (341/371)	91.7 (354/386)	96.7 (348/360)	97.2 (352/362)	98.8 (343/347)

^aSensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV) are expressed as percentages followed by fractions in parentheses as follows: sensitivity, (number of true-positive results)/(sum of true-positive and false-negative results); specificity, (number of true-negative results)/(sum of true-negative and false-positive results); accuracy, (number of true results)/(total results); PPV, (number of true-positive results)/(sum of true-positive and false-positive results); NPV, (number of true-negative results)/(sum of true-negative and false-negative results).

^bCytotoxin assay results were not available for one specimen.

^cThe four specimens with no Xpert results were GDH negative, so according to the algorithm, no Xpert test was needed.

^dXpert results were not available for four specimens.

Compared to toxigenic culture, the algorithm of the GDH assay and the EIA yielded a sensitivity, a specificity, and positive and negative predictive values of 55.6, 98.3, 87.0, and 91.7%, respectively, and the algorithm of the GDH assay combined with the toxin EIA and the CCCN (if the confirmation EIA gave a negative result) demonstrated values of 83.1, 96.7, 83.1, and 96.7%, respectively (Table 1). The GDH assay combined with the Xpert C. difficile PCR assay demonstrated both lower sensitivity (86.1%) and a lower negative predictive value (97.8%) than the Xpert C. difficile assay alone (Table 1). The addition of the GDH screening test increased the turnaround time for most results. The GDH-EIA-CCCN results were available, on average, in 48 h, while the results of the Xpert C. difficile assay were available, on average, in less than 1 h.

Of the 13 patient specimens that were positive by the Xpert C. difficile PCR assay and negative by toxigenic culture, three specimens were positive by the GDH assay, the EIA, and the CCCN; two specimens were GDH positive, EIA positive, and CCCN negative; two specimens were GDH positive, EIA negative, and CCCN positive; one specimen was GDH positive, EIA negative, and CCCN negative; three specimens were GDH negative, EIA negative, and CCCN positive; and two specimens were negative by the GDH assay, the EIA, and the CCCN (Table 2). Of the four specimens that were negative by the Xpert C. difficile assay but culture positive, three were GDH positive, EIA negative, and CCCN negative (Table 2). Eleven of the 17 specimens for which discrepant results were obtained came from patients with histories of C. difficile infection.

TABLE 2.

Characterization of specimens with discrepant results by the Xpert C. difficile assay versus toxigenic culture

IDa	Resultb by:						Corrected resultc
	Xpert C. difficile PCR assay	Direct toxigenic culture	Enriched toxigenic culture	GDH assay	EIA	CCCN
203	Pos	Neg	Neg	Pos	Pos	Pos	TP
259	Pos	Neg	Neg	Pos	Pos	Neg	TP
285	Pos	Neg	Neg	Pos	Pos	Pos	TP
370	Pos	Neg	Neg	Pos	Pos	Neg	TP
408	Pos	Neg	Neg	Pos	Pos	Pos	TP
037	Pos	Neg	Neg	Pos	Neg	Pos	TP
164	Pos	Neg	Neg	Pos	Neg	Pos	TP
207	Pos	Neg	Neg	Pos	Neg	Neg	FP
156	Pos	Neg	Neg	Neg	Neg	Pos	TP
191	Pos	Neg	Neg	Neg	Neg	Pos	TP
229	Pos	Neg	Neg	Neg	Neg	Neg	FP
238	Pos	Neg	Neg	Neg	Neg	Pos	TP
411	Pos	Neg	Neg	Neg	Neg	Neg	FP
027	Neg	Pos	Pos	Pos	Neg	Neg	FN
110	Neg	Pos	Pos	Pos	Neg	Neg	FN
154	Neg	Pos	Pos	Pos	Neg	Neg	FN
036	Neg	Neg	Pos	Neg	Neg	Neg	FN

^aID, specimen identification number.

^bPos, positive; Neg, negative.

^cTP, true positive; FP, false positive; FN, false negative.

The results of a categorical analysis of the data across the different methods showed that the methods gave statistically different results for sensitivity (P, ≤0.001 [P, <0.0001 both by Fisher's exact test and by the chi-square test]). However, the P value for specificity was ≥0.05 (P, 0.0657 by Fisher's exact test and 0.0603 by the chi-square test), indicating that the different methods are statistically equivalent for specificity, although the lower specificity of the EIA approached significance (data not shown). There was no statistical difference between the sensitivity result for the toxin EIA (58.3%) and that for the combination of the GDH assay and the toxin EIA (55.6%) (P, 0.8664 by Fisher's exact test) or between the sensitivity result for the Xpert C. difficile assay (94.4%) and that for the GDH assay-Xpert C. difficile assay algorithm (86.1%) (P = 0.1578), which was surprising given the large gap between the sensitivities, although the lack of significant difference may have been due to the relatively small number of positive samples in this study. There were, however, statistically significant differences between the sensitivity result for the Xpert C. difficile assay alone and that for the toxin EIA alone (P < 0.001) and between the sensitivity result for the Xpert C. difficile assay alone and that for the GDH-EIA-CCCN three-test algorithm (P = 0.0363).

DISCUSSION

CDI is a major medical and infection control problem in many health care facilities, including hospitals, long-term care facilities, and nursing homes around the world (8, 20). Accurate and timely diagnosis is necessary both for appropriate clinical management of the patient and for the timely implementation of infection control and pharmacy measures (13, 21). Many hospitals are now required to report health care-associated transmission of pathogens, including C. difficile, to public health departments. Thus, it is imperative that the diagnosis of CDI be rapid and accurate.

Due to the poor sensitivity of the toxin A/B EIAs, patients with negative toxin EIA results who are still suspected of having CDI are often subjected to additional testing and procedures, adding substantial costs to hospitalization (14, 19). Perhaps more importantly, a significant number of active C. difficile carriers are undiagnosed, leading to unmitigated transmission in health care settings. Often one or more stool specimens are collected following the first EIA-negative result if the suspicion of CDI is high (10, 20). Algorithms for C. difficile testing in the hospital setting have evolved due to the need for more-accurate assays to better detect those patients with CDI (26, 31). Algorithms that appear to be more sensitive than testing by the toxin A/B EIA alone have been implemented in laboratories (26, 29, 31). These algorithms use the GDH assay as a screen and confirm GDH-positive results with the toxin A/B EIA, or with either culture or the CCCN. In previously published data, the GDH assay appeared to have good sensitivity in detecting C. difficile-positive specimens but lacked specificity (29). Two recent studies evaluated a multistep approach with the GDH assay as a screening assay and found that as many as 23 to 24% of the positive specimens were missed by using this algorithm (i.e., sensitivity was approximately 76%) (26, 28). In our study, using a multistep approach, we achieved a sensitivity of 83.1% compared to toxigenic culture by using the GDH assay as the initial screen. These results are slightly lower than those reported by Eastwood et al. (12), who noted a GDH assay sensitivity of 87.6% compared to toxigenic culture in their study.

Although these multitest algorithms improve the specificity of the GDH test, they delay the reporting of the results to the ordering physician (14). Multitest algorithms often require 2 or more days to validate a positive screening result, particularly if the CCCN is used for confirmation. Although many physicians start empirical therapy when a patient is suspected of having CDI, others wait for the results of the diagnostic tests to confirm the diagnosis of CDI before stopping other antibiotics and initiating specific CDI therapy. This delay, especially in combination with the low sensitivity of the EIA, can facilitate the spread of C. difficile to other patients and further contamination of the environment, in addition to increasing the potential for the patient to develop more-severe disease (21).

Our data demonstrate that the sensitivity and negative predictive value of the Xpert C. difficile assay exceed those of the other individual assays and testing algorithms, specifically the GDH-EIA-CCCN algorithm (P, 0.0363), which is superior to the GDH-EIA algorithm in terms of sensitivity (83.1% versus 55.6%, respectively). This study confirms other published data demonstrating that the sensitivity of the EIAs currently marketed is very limited; thus, they should not be considered reliable for the detection of CDI. A review of the other testing results for specimens that were positive by the Xpert C. difficile assay but negative by reference toxigenic culture shows that 7 of 13 specimens appear to be true-positive specimens, given the positive GDH, EIA, and/or CCCN results (Table 2). The corrected sensitivity and specificity of the PCR assay would be 95.1% and 99.4%, respectively. Negative results by reference toxigenic culture could be due to several factors, such as patients' receipt of antimicrobial agents at the time of specimen collection, the lag time between specimen collection and testing, or the presence of nonviable organisms in the sample. Lack of either heat or alcohol shock for spore enrichment in the culture protocol may also have contributed to these seemingly false-negative culture results. Conversely, three of the four specimens that were negative by the Xpert C. difficile assay but positive by reference toxigenic culture were positive by the GDH assay but both EIA negative and CCCN negative (Table 2), which may be explained by a sampling problem due to low concentrations of organisms in very heterogeneous samples. Taken together, however, these data suggest that the sensitivity and specificity of the Xpert C. difficile assay would allow the laboratory not only to enhance and speed up the detection of toxigenic C. difficile beyond that with EIAs and GDH assay-based algorithms but also to have even more confidence in an initially negative result based on one test only. Further studies will be needed to determine the significance of PCR-positive results for patients who test negative by other conventional methods.

Two other PCR assays for C. difficile have recently been cleared by the FDA for use in U.S. laboratories. The results of those FDA-cleared assays were compared to the results of the CCCN rather than to the results of toxigenic culture in the clinical trials used for FDA clearance (30). Compared to the CCCN, these assays performed well. However, Stamper et al. compared the GeneOhm assay from Becton Dickinson & Company (BD) to toxigenic culture and demonstrated a sensitivity of 83.6%, which was lower than stated in the package insert for the BD product (30). Eastwood et al. (12) reported a somewhat higher sensitivity (88.5%) for the BD assay compared to toxigenic culture. Gluck et al. (15) compared the Prodesse proGASTRO C. difficile PCR assay to toxigenic culture and reported a sensitivity of 73.3% and a specificity of 99.2%. Since toxigenic culture has better sensitivity than the CCCN alone for detecting C. difficile, it is not surprising that the sensitivities reported by Stamper et al. and Gluck et al. when comparing the assays to the gold standard are lower than those reported in the package inserts (30). The sensitivity of the Xpert C. difficile assay was determined in this study by comparing the results to those of enriched toxigenic culture. A study comparing the results of all three commercial PCR methods for the same stool specimen to those of enriched toxigenic culture would be of interest.

The cost-benefit ratio of the PCR-based assays, which will be more expensive than the GDH-EIA-CCCN algorithm, needs to be determined. Due to the improved sensitivity that PCR assays provide, theoretically, physicians would be able to order fewer C. difficile tests in order to obtain an accurate diagnosis (25). This potentially offsets some of the increased cost of the newer PCR assays. Hospitals need to consider what TAT is needed for C. difficile results, as well as the impact of TAT on the risk of transmission. Patient outcome should be considered in addition to other data in determining what approach to C. difficile testing is required in the laboratory setting. We conclude that when the various algorithms are compared, the Xpert C. difficile assay, both with regard to sensitivity and with regard to the negative predictive value, exceeds the performance levels of the individual tests and the GDH-EIA-CCCN algorithm and may also be able to reduce the practice of multiple test ordering, which has become commonplace in many hospitals.