Loop-Mediated Isothermal Amplification Compared to Real-Time PCR and Enzyme Immunoassay for Toxigenic Clostridium difficile Detection

Abstract

Clostridium difficile infection is the primary cause of health care-associated diarrhea. While most laboratories have been using rapid antigen tests for detecting C. difficile toxins, they have poor sensitivity; newer molecular methods offer rapid results with high test sensitivity and specificity. This study was designed to compare the performances of two molecular assays (Meridian illumigene and BD GeneOhm) and two antigen assays (Wampole Quik Chek Complete and TechLab Tox A/B II) to detect toxigenic C. difficile. Fecal specimens from hospitalized patients (n = 139) suspected of having C. difficile infection were tested by the four assays. Nine specimens were positive and 109 were negative by all four methods. After discrepant analysis by toxigenic culture (n = 21), the total numbers of stool specimens classified as positive and negative for toxigenic C. difficile were 21 (15%) and 118 (85%), respectively. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were as follows: GeneOhm (95.2%, 100%, 100%, and 99.2%), illumigene (95.2%, 96.6%, 83.3%, and 99.2%), Tox A/B II (52.4%, 97.5%, 78.6%, and 92.4%), and Quik Chek Complete (47.6%, 100%, 100%, and 91.9%). The illumigene assay performed comparably to the GeneOhm assay with a slight decrease in test specificity; the sensitivities of both far exceeded those of the antigen assays. The clinical characteristics of the concordant and discrepant study patients were similar, including stool consistency and frequency. In the era of rapid molecular-based tests for toxigenic C. difficile, toxin enzyme immunoassays (EIAs) should no longer be considered the standard of care.

INTRODUCTION

Clostridium difficile infection is the primary cause of health care-associated diarrhea, with a prevalence rate of 13.1 per 1,000 inpatients in the United States and a disease spectrum ranging from mild diarrhea to fulminant colitis and death (7, 10). There has been an increase in the incidence and severity of C. difficile infection over the last 10 years, partly due to the emergence of the BI/NAP1/027 clone (2). Further, there are high rates of disease recurrence and treatment failure, and attributable health care costs are estimated to be in excess of $1 billion annually (2). Accurate and rapid diagnostic testing for C. difficile infection is essential for patient management and the timely implementation of infection control measures.

Toxigenic culture (TC) and cell culture cytotoxicity neutralization assay (CCNA) are considered reference standards but are not routinely used, as they are labor-intensive and have prolonged turnaround times (2). Currently, enzyme immunoassays (EIAs) dominate the diagnostic testing arena. They can detect glutamate dehydrogenase (GDH) (so-called common antigen) and/or major toxins A and B and are inexpensive, rapid, and easy to perform. A drawback of EIA toxin tests is a lack of sensitivity (8, 9, 13, 15, 16, 20, 22, 27, 30, 32). Conversely, EIA GDH tests have good sensitivity but lack specificity, as they cannot distinguish toxigenic from nontoxigenic C. difficile (8, 13, 15, 23–25, 30, 32). As such, GDH is a good screening test, but GDH-positive specimens must be subjected to another test that detects toxin A and/or B or the toxin gene(s) (13, 15, 23–25, 27, 30, 32). Molecular assays directed at one or more of five genes residing within the pathogenicity locus (PaLoc) have fostered much interest. Commercial kit-based real-time PCR tests targeting the tcdB gene have recently become available—the BD GeneOhm Cdiff assay (Becton Dickinson), proGastro Cd (Gen-Probe/Prodesse), and Xpert C. difficile (Cepheid)—with exceptional performance and relative ease of use documented (5, 8, 9, 13, 28–31). FDA clearance was recently granted for the Meridian illumigene C. difficile assay, which uses loop-mediated isothermal amplification (LAMP) to detect the toxin A gene (tcdA) within the PaLoc. The goal of this study was to compare the performance of the illumigene assay to those of another molecular test (BD GeneOhm Cdiff) and two EIAs (Wampole C. Diff Quik Chek Complete and TechLab C. difficile Tox A/B II) targeting toxins A and B with or without GDH antigen. It has been suggested that clinical presentation is important when interpreting C. difficile diagnostic assays (6). Therefore, patient clinical characteristics were analyzed to determine their influence on test performance.

MATERIALS AND METHODS

Ethics approval.

This study was approved by the Human Investigative Committee of Beaumont Hospitals.

Setting, specimen acquisition, and testing.

Fecal specimens for C. difficile testing at Beaumont Hospitals (Royal Oak, MI) had to be liquid or soft stool that conformed to the shape of the collection container. Specimens were held at 4 to 7°C and tested within 24 h of collection. During a 3-week period, stool specimens (n = 145) from hospitalized patients (n = 139; age range, 5 to 94 years; mean, 67 years; male, 45%; female, 55%) suspected of having C. difficile infection were prospectively subjected to four diagnostic tests: TechLab C. difficile Tox A/B II assay (a microwell EIA distributed by Alere [formerly Inverness Medical Innovations], Waltham, MA), GeneOhm Cdiff (Becton Dickinson Diagnostics, La Jolla, CA), illumigene C. difficile (Meridian Bioscience, Cincinnati, OH), and C. Diff Quik Chek Complete (a lateral-flow EIA card from Alere Medical Inc.). The last assay detects both GDH and toxins A and B; both components must be positive for the test to be positive for toxigenic C. difficile. Additionally, two 1-ml aliquots from each specimen were deidentified and frozen at −20°C for discrepant analysis. All testing was performed according to the manufacturer's instructions by two medical technologists dedicated to the study at the same physical location.

Discrepant analysis.

Discrepant testing was performed without knowledge of the prior test results. Each stool specimen was thawed, followed by ethanol shock and inoculation 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 visible growth was observed in the broth culture after 48 h or at 5 to 7 days (late growth), 0.1 ml was subjected to the Premier Tox A/B enzyme-linked immunosorbent assay (ELISA) according to the manufacturer's instructions (Meridian Bioscience, Cincinnati, OH). A positive result (optical density [OD] > 0.10) supported the detection of toxigenic C. difficile. Colonies on CCFA with typical characteristics of C. difficile (flat yellow colonies) were tested by PCR in an internally validated PCR assay targeting the putative toxin repressor gene tcdC (primers 5′-TCTAGCTAATTGGTCATAAG-3′ and 5′-AATAGCAAATTGTCTGAT-3′) and the GDH “common-antigen” gene (gdh) using published primers (33). Despite variability in other regions of the PaLoc, the tcdC primers bind conserved regions in all toxigenic strains of C. difficile in GenBank (evaluated in August 2005) and over 1,200 sequenced strains in a recent survey (4). All PCRs were performed with Qiagen HotStarTaq Master Mix PCR reagents (Qiagen, Valencia, CA) and MgCl₂ (2.5 mM 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. The PCR amplicons were resolved on 2% agarose gels stained with ethidium bromide. A visible band of approximately 200 bp from the tcdC PCR and 750 bp from the gdh PCR supported the detection of toxigenic C. difficile and C. difficile, respectively.

Specimen classification and statistical analysis.

In order to determine performance characteristics for each test, stool specimens were classified according to the following rules. If the results for all four tests were positive, then the stool specimen was considered positive for toxigenic C. difficile. If the results for all four tests were negative, then the stool specimen was considered negative for toxigenic C. difficile. If the stool specimen yielded a positive result for 1, 2, or 3 of the 4 tests, it underwent discrepant analysis as described above. The results of discrepant analysis were deemed definitive regardless of the results of any other tests.

Clinical evaluation.

Chart review was conducted on 139 patients to obtain the following clinical information: age; gender; number of stools per day, maximum white blood cell (WBC) count, and creatinine (each within 1 day of testing); prior C. difficile infection within 90 days; and prior use of antibiotics, proton pump inhibitor (PPI) medication, and cancer chemotherapy (each within 30 days of testing). Stool quality (liquid, mucoid, or semisolid that still conformed to the shape of the container) was recorded by the laboratory. Specimens were stratified into five groups (see Table 3). Differences in clinical parameters among the five groups were assessed by chi-square analysis or Fisher's exact test where appropriate for categorical variables and a 2-tailed t test for pairwise comparisons of continuous variables using Microsoft Excel.

Table 3.

Summary of patient clinical parameters

Parameter	Value
	Concordant positivesb	Concordant negativesc	Discordant false positivesd	Discordant false negativese	Nontoxigenicf
na	9	109	6	12	3
Mean age (yr)	75.8	66.5	68.7	68.3	45
Gender (% female)	55.5	54.1	50	66.7	33.3
Mean no. of stools/dayg (range)	4.1 (2–7)	4.1 (1–13)	3.9 (1–10)	3.8 (1–7)	2.3 (2–3)
No. (%) with <3 stools/dayg	3 (33)	26 (30)	2 (40)	2 (22)	2 (66)
Median WBC count (range, 4.0–10.1 billion/liter)	12.25	7.95	10.7	11.2	7.4
% WBC count > 10.1	88	27	50	58	0
Median creatinine (range, 0.6–1.40 mg/dl)	0.91	0.94	1.01	0.905	0.5
Prior C. difficile infection (7–90 days) [no. (%)]	5 (56)	4 (4)	0 (0)	4 (33)	0
Antibiotic use (last 30 days) [no. (%)]	8 (89)	89 (82)	5 (83)	9 (75)	2 (66)
PPI use (last 30 days) [no. (%)]	7 (78)	75 (69)	1 (17)	11 (92)	1 (33)
Chemotherapy use (last 30 days) [no. (%)]	2 (22)	16 (15)	2 (33)	3 (25)	1 (33)

^an (number of specimens) = 139.

^bConcordant positives, positive test results were obtained by all four initial testing modalities.

^cConcordant negatives, negative test results were obtained by all four initial testing modalities.

^dDiscordant false positives, false-positive test results (final classification) were obtained for any of the initial four testing modalities following discrepant analysis.

^eDiscordant false negative, false-negative test results (final classification) were obtained for any of the initial four testing modalities following discrepant analysis.

^fSpecimens (n = 3) originally considered false positive by the GDH component of the Wampole Quik Chek Complete Test. Discrepancy analysis yielded nontoxigenic C. difficile in all three specimens.

^gOnly 112 specimens had the number of stools per day recorded in the medical chart.

RESULTS

During the study period, 145 stool specimens were collected from 139 hospitalized patients clinically suspected of having C. difficile infection. Six patients had two stool specimens analyzed, and both were negative by all four tests. Thus, of the 139 unique patient specimens included in the final data set, 9 were positive and 109 were negative by all four tests and were classified as positive and negative, respectively, for toxigenic C. difficile. Twenty-one specimens required discrepant analysis, with the following results: toxigenic C. difficile detected (n = 12); nontoxigenic C. difficile detected (n = 3); no C. difficile detected (n = 6). Therefore, the total numbers of stool specimens classified as positive and negative for toxigenic C. difficile were 21 (15%) and 118 (85%), respectively.

Five stool specimens initially yielded an invalid test result: 1 with GeneOhm and 4 with illumigene. Upon repeat testing, a valid result was obtained for each specimen. Table 1 summarizes the results of each test and discrepant analysis. The positivity rates for GeneOhm, illumigene, Tox A/B II, and Quik Chek Complete were 14.4%, 14.4%, 7.9%, and 7.2%, respectively. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for each test (Table 2) were as follows: GeneOhm (95.2%, 100%, 100%, and 99.2%), illumigene (95.2%, 96.6%, 83.3%, and 99.1%), Tox A/B II (52.4%, 97.5%, 78.6%, and 92.0%), and Quik Chek Complete (47.6%, 100%, 100%, and 91.5%). The differences in sensitivity and NPV for the combined molecular tests versus the combined EIAs were both significant (P < 0.001; Fisher). There was no significant difference in the PPVs of the GeneOhm and illumigene assays (P = 0.11; Fisher).

Table 1.

Summary of individual testing modalities to detect toxigenic C. difficile^a

n	Test result summaryb					Discrepancy analysis		Final classificationc
	BD GeneOhm	Meridian illumigene	TechLab Tox A/B II	Wampole Quik Chek Complete				BD GeneOhm	Meridian illumigene	TechLab Tox A/B II	Wampole Quik Chek Complete
				GDH	Tox A/B	C. difficile Present	Toxigenic C. difficile Present				GDH	Tox A/B
109	−	−	−	−	−	Not performed	Not performed	TN	TN	TN	TN	TN
9	+	+	+	+	+	Not performed	Not performed	TP	TP	TP	TP	TP
7	+	+	−	+	−	Yes	Yes	TP	TP	FN	TP	FN
3	−	−	−	+	−	Yes	No	TN	TN	TN	TP	TN
3	−	+	−	−	−	No	No	TN	FP	TN	TN	TN
2	−	−	+	−	−	No	No	TN	TN	FP	TN	TN
2	+	+	+	+	−	Yes	Yes	TP	TP	TP	TP	FN
1	−	+	−	−	−	Yes	Yes	FN	TP	FN	FN	FN
1	+	+	−	+	+	Yes	Yes	TP	TP	FN	TP	TP
1	+	−	−	+	−	Yes	Yes	TP	FN	FN	TP	FN
1	−	+	+	−	−	No	No	TN	FP	FP	TN	TN

^an (number of specimens) = 139.

^b−, negative; +, positive.

^cTN, true negative; TP, true positive; FN, false negative; FP, false positive.

Table 2.

Performance characteristics of individual testing modalities to detect toxigenic C. difficile^a

Parameter	Value
	BD GeneOhm	Meridian illumigene	TechLab Tox A/B II	Wampole Quik Chek Completeb
				GDH	Tox A/B
True positive (no.)	20	20	11	23	10
True negative (no.)	118	114	115	115	118
False positive (no.)	0	4	3	0	0
False negative (no.)	1	1	10	1	11
Sensitivity (%)	95.2	95.2	52.4	95.8	47.6
Specificity (%)	100	96.6	97.5	100	100
PPV (%)	100	83.3	78.6	100	100
NPV (%)	99.2	99.1	92.0	99.1	91.5

^an = 139.

^bA positive test result requires that both the GDH and Tox A/B components be positive. The sensitivity, specificity, PPV, and NPV of the Wampole Quik Chek Complete test were identical to the data listed for the Tox A/B component. The GDH component was assessed for the ability to detect any (toxigenic or nontoxigenic) C. difficile.

During discrepant analysis, all TC-positive samples had growth in broth, and all but one specimen had extensive growth on CCFA within 72 h. The one specimen was negative by the GeneOhm PCR and GDH EIAs. There was no significant difference in in vitro toxin expression from the cultured isolates using the Meridian Premier Tox A/B EIA, again with one exception. This specimen was positive by three amplified methods (illumigene, GeneOhm, and tcdC PCR from culture); however, it was negative by initial stool EIA toxin testing (Tox A/B II and Quik Chek Complete EIA) and had a negative to low-positive reading from a cultured isolate using the Meridian Premier Tox A/B EIA, but when a pure culture was retested with a different EIA, Tox A/B II, a strong positive reading was obtained.

With patients categorized as concordant or discordant based on the results of their C. difficile laboratory testing (Table 3), there were no significant differences in the numbers of loose bowel movements (BMs) per day, prior use of antibiotic or chemotherapeutic agents, or renal insufficiency. However, those patients with stool specimens testing positive for toxigenic C. difficile were more likely to have had prior C. difficile infections (range, 7 to 90 days; P < 0.001; Fisher) and elevated WBC counts (P < 0.001; Fisher) and tended to be older than patients with negative test results (P = 0.05; t test). The three patients with nontoxigenic C. difficile were significantly younger than all other patients in the study (P = 0.03; t test) and tended to have fewer BMs. Overall, 31% of patients with known numbers of BMs (n = 112) had less than 3 BMs per day and potentially should not have been tested, as they did not fulfill 1 definition of diarrhea (21); this number is reduced to 21% if the definition of diarrhea includes liquid stool (data not shown). However, these patients without significant diarrhea were fairly evenly distributed among all diagnostic categories (Table 3). Interestingly, all patients in this study had high rates of PPI utilization, except those determined to have “false-positive” test results for C. difficile (P = 0.003; Fisher).

DISCUSSION

This study compares methods for detecting the genes and respective antigens for C. difficile toxins, incorporating clinical information and the results of toxigenic culture. During the latter analysis, it was shown that all strains with toxin genes actually expressed toxin protein in vitro, as was observed in testing of nearly 600 isolates (P. Riska, unpublished data) and in some published reports (24, 25), but not others (6, 8, 26). Notably, one isolate in our present study carried toxin genes, as demonstrated by multiple assays, yet failed to produce detectable toxin by the Meridian Premier toxin EIA while producing a very high toxin yield when retested by the Inverness Toxin A/B EIA. This suggests that toxin was expressed but the epitopes of the toxin were not readily detected by one of the toxin A/B assays.

Meridian illumigene C. difficile.

The illumigene assay is based upon LAMP (12, 17, 18), in which primers qualitatively amplify a 204-bp region of the conserved 5′ sequence of the tcdA gene within the PaLoc of toxigenic C. difficile via continuous isothermal amplification. Magnesium pyrophosphate is an amplification by-product that forms a white precipitate that is detected via turbidimetric measurement. The test is relatively simple to perform, requires minimal hands-on time, and can be completed in about 1 h, the shortest turnaround time of all the amplified assays. In our study, the illumigene assay performed exceptionally well, with a sensitivity and a specificity of 95.2% and 96.6%, respectively, comparable to the manufacturer's performance claims (with 95% confidence intervals) of 95.2% (89.2 to 97.9%) and 95.3% (92.3 to 96.7%), respectively. Our results are also congruent with those of Noren et al. (19), Dubberke et al. (6), and Lalande et al. (14), who demonstrated excellent sensitivity and specificity compared to TC and CCNA, respectively. Invalid results (2.8% overall; near the manufacturer's claim of 2.9%) occurred early in assay implementation, with semisolid stools too heavily inoculated on the collection brush. With experience, invalid results were minimized.

Four false-positive results and one false-negative result were obtained with the illumigene assay. Three of the false-positive stool specimens yielded negative results with the other three tests, as well as TC (discrepant analysis). The fourth false-positive result was also falsely positive with the Tox A/B II assay (OD = 0.176; cutoff > 0.08) and negative with all other assays, including TC. Specimen-to-specimen cross-contamination was excluded for the four false-positive results. Notably, these four patients (i) had no history of C. difficile infection within the previous 90 days, (ii) did not develop laboratory-confirmed C. difficile infection over the next 6 months, and (iii) had not received PPI therapy.

Of 41 study patients not using PPIs, 12% were false positive by illumigene (n = 4) or Tox A/B EIA (n = 1) versus 7% (n = 3) confirmed positive. Of 95 patients using PPIs, only 1% (n = 1) had a false-positive C. difficile test (by Tox A/B II), while 19% (n = 18) were confirmed positive. A mechanism for this apparent association between nonuse of PPI and false-positive tests for C. difficile (P < 0.01; Fisher) remains unclear. This finding may be due to a confounding effect of PPIs on the C. difficile infection risk: PPI nonusers could be considered a low-prevalence population and thus more likely to test false positive. However, the prevalence of C. difficile infection was not statistically different (P = 0.12; Fisher) between PPI users (19%) and nonusers (7%). It is possible that the illumigene assay was more sensitive than TC, as suggested by others (12, 19); however, this is unlikely, given the multiple and extended culture protocols used during discrepant analysis and the extended clinical follow-up on these patients.

The one false-negative result with the illumigene assay was positive by GeneOhm, the GDH component of the Quik Chek Complete, and TC (discrepant analysis) and negative by both toxin assays. In theory, this specimen may have contained a tcdA-negative tcdB⁺ strain of toxigenic C. difficile. While certain tcdA-negative strains in toxinotypes VIII and X (1, 3, 11) are still detected by illumigene (manufacturer's package insert; 3), there may be other primer-binding site mutations or deletions that are not recognized.

BD GeneOhm Cdiff.

The GeneOhm assay performed exceptionally well, with sensitivity of 95.2% and specificity of 100%, minimally exceeding the performance of the illumigene assay (P > 0.05 for the comparison of all parameters). Our findings are compatible with those of other evaluators of the GeneOhm assay who used CCNA and/or TC as a reference standard (13, 23, 30, 31). Our invalid-test rate (0.7%) is below the manufacturer's claims (mean, 4.6%; range 0.8% to 8.5%) and comparable to the 1.1% reported by Eastwood et al. (8) One false-negative result was obtained with the GeneOhm assay. For this specimen, toxigenic C. difficile was detected by the illumigene assay and TC (growth only in broth) but not with the Tox A/B II and Quik Chek Complete assays. This likely reflects a concentration of toxigenic C. difficile at or below the limit of detection of the GeneOhm assay.

TechLab Tox A/B II.

The Tox A/B II assay demonstrated poor sensitivity (52.4%), near the lower end of the ranges reported by others compared to TC (66 to 88.6%) (6, 8, 13, 20, 27) or CCNA (38 to 90.7%) (8, 13, 20, 32). The high specificity of the Tox A/B II assay of 97.5% is congruent with the work of others (95.7 to 100%) (6, 8, 13, 20, 27, 32).

Wampole Quik Chek Complete.

The Wampole Quik Chek Complete alternative EIA approach, targeting GDH antigen and toxin A/B together, performed similarly to the Techlab Tox A/B II. The toxin A/B component of the Quik Chek Complete assay had poor sensitivity (47.6%) and high specificity (100%) (Table 2). However, considering only the GDH component of the Quik Chek Complete to detect any C. difficile in this study, the calculated sensitivity, specificity, positive predictive value, and negative predictive value were 95.8%, 100%, 100%, and 99.1%, respectively (Table 2). One recent suggestion was that using fresh rather than frozen stool specimens increased the sensitivity of the assay (25); of note, we tested only fresh samples in our study. Interestingly, only three nontoxigenic C. difficile isolates were detected overall in our study versus 21 toxigenic isolates for a “false-positive” rate of 12.5%. However, the reported rate of false-positive GDH EIAs approaches 50% (8, 13, 15, 23–25, 30, 32), suggesting that the specificity and PPV of a GDH assay alone could be much lower in other settings. Nonetheless, the high sensitivity reported with this GDH EIA format supports the use of GDH assays as an initial screen, followed by a more specific 2nd test, such as a CCNA, a TC, or a DNA amplification assay, to confirm the presence of toxin A and/or B or their respective genes (23, 25, 30). The net benefit (i.e., the expense of personnel and supplies versus reimbursement) of performing multiple tests compared to DNA amplification alone remains to be determined, with stand-alone real-time PCR demonstrating equivalent to superior sensitivity and specificity (13, 15, 24, 32).

The primary limitation of this study was failure to perform TC on all specimens, potentially overestimating the sensitivity of each testing modality. However, recent analyses of the GDH EIA and its performance here support its high sensitivity as a screening tool, thus minimizing the likelihood of finding additional C. difficile-positive stools by culture-based methods (23, 25, 30). The strengths of this study are that all testing, except discrepant analysis, was performed by two dedicated medical technologists in the same physical location, avoiding test result bias due to variations in technical expertise or the testing environment (e.g., temperature or humidity). Additionally, chart reviews performed by an infectious-diseases fellow and a staff pathologist without knowledge of the test results confirmed that most patients had clinically significant diarrhea, as judged by stool quality and/or frequency. While the number of documented BMs per day is most likely underreported by nursing staff records, a majority of subjects still had documented diarrhea. Other clinical features were analyzed and were distributed evenly across the categories of patients. A larger sample size may have revealed other clinical predictors of false-negative and false-positive test results for C. difficile.

In summary, these results support the use of the illumigene C. difficile assay as a rapid, cost-effective, and technically simple test. Minor concerns are raised about the slightly higher false-positive and initial invalid-test rates of the illumigene assay; however, the latter was resolved with practice in obtaining the correct stool inoculum. The association of false-positive results with lack of PPI use needs further investigation. The theoretical risk of missing tcdA-negative tcdB⁺ strains of toxigenic C. difficile other than those belonging to toxinotype VIII or X should also be monitored. This study further confirms that toxin EIAs should no longer be the standard of care for detecting toxigenic C. difficile in the United States.