Clostridium difficile Testing Algorithms Using Glutamate Dehydrogenase Antigen and C. difficile Toxin Enzyme Immunoassays with C. difficile Nucleic Acid Amplification Testing Increase Diagnostic Yield in a Tertiary Pediatric Population

Abstract

We evaluated the performance of the rapid C. diff Quik Chek Complete's glutamate dehydrogenase antigen (GDH) and toxin A/B (CDT) tests in two algorithmic approaches for a tertiary pediatric population: algorithm 1 entailed initial testing with GDH/CDT followed by loop-mediated isothermal amplification (LAMP), and algorithm 2 entailed GDH/CDT followed by cytotoxicity neutralization assay (CCNA) for adjudication of discrepant GDH-positive/CDT-negative results. A true positive (TP) was defined as positivity by CCNA or positivity by LAMP plus another test (GDH, CDT, or the Premier C. difficile toxin A and B enzyme immunoassay [P-EIA]). A total of 141 specimens from 141 patients yielded 27 TPs and 19% prevalence. Sensitivity, specificity, positive predictive value, and negative predictive value were 56%, 100%, 100%, and 90% for P-EIA and 81%, 100%, 100%, and 96% for both algorithm 1 and algorithm 2. In summary, GDH-based algorithms detected C. difficile infections with superior sensitivity compared to P-EIA. The algorithms allowed immediate reporting of half of all TPs, but LAMP or CCNA was required to confirm the presence or absence of toxigenic C. difficile in GDH-positive/CDT-negative specimens.

INTRODUCTION

Clostridium difficile infection (CDI) is a leading cause of antibiotic-associated colitis in children and is associated with significant morbidity in vulnerable pediatric populations (4, 11). The optimal testing method for laboratory detection of CDI remains controversial in both adults and in children (18). The suboptimal sensitivity and specificity of many commercial enzyme immunoassays have limited their utility to the extent that the 2010 Infectious Diseases Society of America-Society for Health Care Epidemiology of America (IDSA-SHEA) clinical practice guidelines for Clostridium difficile infection in adults discouraged their use (2, 3). The cell culture cytotoxicity neutralization assay (CCNA), the traditional gold standard, performed with superior sensitivity but was labor-intensive and required 48 h to finalize. Nucleic acid amplification tests (NAATs), including the BD GeneOhm Cdiff (BD Diagnostics, San Diego, CA), Prodesse ProGastro Cd (Gen-Probe Inc., San Diego, CA), Xpert C. difficile (Cepheid, Sunnyvale, CA), and illumigene C. difficile (Meridian Biosciences, Cincinnati, OH) tests, appeared to perform with sensitivities equal to or exceeding that of CCNA.

In the literature regarding infections in adults, multistep diagnostic algorithms have been evaluated in an effort to use NAATs more judiciously (1). The most recent iteration of this approach involves initial testing of stool specimens submitted for C. difficile toxin testing with the rapid C. diff Quik Chek Complete assay (10, 12, 15). This product has two enzyme immunoassay components that detect glutamate dehydrogenase antigen (GDH) and toxin A/toxin B (CDT) in a lateral flow device. Typically, specimens that test positive or negative by both GDH and CDT are reported immediately as positive or negative for C. difficile toxin, respectively. Because this assay detects the presence of both toxigenic and nontoxigenic C. difficile strains, GDH-positive/CDT-negative specimens require additional testing to adjudicate such specimens. Data in the literature suggest that algorithmic approaches have superior sensitivity and specificity compared to commercial enzyme immunoassays, with sensitivities and specificities ranging from 98 to 100% and 88 to 97%, respectively (10, 12, 15). The negative predictive value (NPV) of GDH and the positive predictive value (PPV) of GDH-positive/CDT-positive results have been reported to be close to 100% (10, 12, 15), supporting the practice of immediately reporting these results without further testing. The high NPV of GDH also obviates the need to pursue the controversial practice of repeat testing of negative patients (1, 2). Finally, while some centers that process large volumes of C. difficile toxin tests may find performance of molecular testing routinely for all submitted specimens to be cost-effective (compared to use of two tests), an algorithmic approach may be more cost-effective for other centers (1, 18).

To our knowledge, algorithmic approaches to CDI detection have not been evaluated in children. This study evaluated the performance of two algorithms for detection of CDI in a tertiary pediatric population: (i) initial testing with GDH/CDT followed by loop-mediated isothermal amplification (LAMP) to resolve discrepant GDH-positive/CDT-negative specimens (algorithm 1); (ii) initial testing with GDH/CDT followed by CCNA to resolve discrepant specimens (algorithm 2).

MATERIALS AND METHODS

Patients and specimens.

This was a prospective observational study. Consecutive, fresh, liquid or nonformed stool specimens collected from patients 1 to 18 years of age and submitted to the microbiology laboratory for C. difficile toxin testing at The Children's Hospital of Philadelphia between 1 March and 30 March 2011 were included. Specimens were excluded from analysis if the patient had been previously tested by the study protocol.

All specimens were tested prospectively using the C. diff Quik Chek Complete test (GDH/CDT; Techlab, Blacksburg, VA), the Premier C. difficile toxin A and B enzyme immunoassay (P-EIA; Meridian Biosciences, Cincinnati, OH), the CCNA (Tox-B test; Techlab, Blacksburg, VA), or LAMP, using the illumigene C. difficile assay (Meridian Biosciences, Cincinnati, OH). All tests were performed according to the instructions and within the time frames recommended by the assay manufacturers. Specimens were stored at 2 to 8°C while awaiting testing. External positive and negative controls were included for each assay for each new lot and each time the assay was used.

GDH/CDT results were interpreted as follows. The test device was examined for the appearance of blue dots in the middle of the reaction window, representing the internal positive control. The test result was interpreted as invalid if the internal control dots were not present. The device was then examined for the appearance of blue lines on the “Ag” and “Tox” sides of the reaction window. The appearance of a blue line on the Ag side was read as GDH positive. The appearance of blue lines on both the Ag and Tox sides was interpreted as GDH positive and CDT positive.

CCNA was performed using the C. difficile Tox-B test kit. Briefly, following dilution, stool was centrifuged and the supernatant filtered using a 0.45-μm syringe filter. The stool filtrate was then inoculated onto a human foreskin fibroblast cell line (Diagnostic Hybrids, Athens, OH) with and without C. difficile antitoxin. The cell line was incubated at 37°C in 5 to 10% CO₂ for 48 h. Light microscopy was used to read the cell line plates at 24 and 48 h. Cytotoxic activity was considered positive if ≥50% of the cells in a well were rounded. Stool filtrates that produced cell cytopathic effect that was inhibited by C. difficile antitoxin were considered positive for C. difficile toxin. Specimens were tested by CCNA within 24 h of collection.

LAMP is a nucleic acid amplification test that targets a 204-bp region of the tcdA gene, which resides within the pathogenicity locus of toxigenic C. difficile. It was performed according to the manufacturer's instructions for the illumigene C. difficile assay. At the conclusion of the run, results were reported as positive, negative, or invalid.

Patient charts were reviewed retrospectively to understand the clinical relevance of specimens that tested LAMP positive/CCNA negative and CCNA positive/LAMP negative. This component of the study received approval by the institutional review board at The Children's Hospital of Philadelphia.

Statistical analysis.

A composite gold standard was initially used for analysis. A specimen was considered “true positive” for C. difficile toxin if it was positive by CCNA. If the CCNA was negative, the specimen could also be considered true positive if the specimen tested positive by LAMP and by another test (GDH, CDT, or P-EIA). For the purpose of analysis, specimens were categorized as true negative if they did not meet the criteria for a true positive. Prevalence, sensitivity, specificity, PPV, and NPV were calculated for the GDH, CDT, P-EIA, CCNA, and LAMP tests and by using two algorithms, GDH/CDT followed by LAMP to resolve discrepant GDH-positive/CDT-negative specimens (algorithm 1) and GDH/CDT followed by CCNA to resolve discrepant specimens (algorithm 2). The McNemar test was used to compare the sensitivity of each algorithm with P-EIA, our current testing method. The analysis was then repeated using CCNA positivity as the sole gold standard criterion.

RESULTS

A total of 141 stool specimens meeting inclusion criteria were collected and tested from 141 consecutive patients during the study period. Testing was completed for all 141specimens. The mean and median time intervals from specimen collection to testing were 17.1 and 18. 2 h for GDH/CDT, 23.0 and 19.6 h for P-EIA, 16.7 and 17.7 h for CCNA, and 49.3 and 49.5 h for LAMP. The sensitivity, specificity, PPV, and NPV of the GDH, CDT, P-EIA, CCNA, and LAMP tests and algorithms 1 and 2 using the composite definition of “true positive” and the CCNA definition of “true positive” are summarized in Table 1.

Table 1.

Performance of four C. difficile toxin assays in a tertiary pediatric population

Assaya	Gold standardb	No. ofc:				Performance statisticd (%)
		TP	TN	FP	FN	SN	SP	PPV	NPV
GDH	Composite	22	94	20	5	81	82	52	95
	CCNA	21	94	21	5	81	82	50	95
CDT	Composite	13	114	0	14	48	100	100	89
	CCNA	13	115	0	13	50	100	100	90
EIA	Composite	15	114	0	12	56	100	100	90
	CCNA	15	115	0	11	58	100	100	91
CCNA	Composite	26	114	0	1	96	100	100	99
	CCNA
LAMP	Composite	24	111	2	3	89	98	92	97
	CCNA	23	111	3	3	88	97	89	97
Algorithm 1	Composite	22	114	0	5	81	100	100	96
	CCNA	21	115	0	5	81	100	100	96
Algorithm 2	Composite	22	114	0	5	81	100	100	96
	CCNA	21	115	0	5	81	100	100	96

^aAssays evaluated: GDH, C. diff Quik Chek Complete glutamate dehydrogenase antigen enzyme immunoassay; CDT, C. diff Quik Chek Complete C. difficile toxin A and toxin B immunoassay; EIA, Premier C. difficile toxin A and B enzyme immunoassay; CCNA, cell cytotoxicity neutralization assay; LAMP, loop-mediated isothermal amplification; algorithm 1, initial testing with C. diff Quik Chek Complete, followed by LAMP for GDH-positive/CDT-negative specimens; algorithm 2, initial testing with C. diff Quik Chek Complete, followed by CCNA for GDH-positive/CDT-negative specimens.

^bFor the composite gold standard, a specimen was considered true positive if it was positive by CCNA. If the specimen was CCNA negative, the specimen would also be considered true positive if it was LAMP positive and positive by another test (GDH, CDT, or EIA). For the CCNA gold standard, a specimen was considered true positive if it was positive by CCNA.

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

^dSN, sensitivity; SP, specificity; PPV, positive predictive value; NPV, negative predictive value.

Using the composite definition of “true positive,” 27 specimens were true positive for C. difficile toxin, resulting in a prevalence of 19%. Of these, 26 met the definition through a positive CCNA result. The remaining specimen was positive by LAMP and GDH. GDH was positive in 42 specimens, performing with a sensitivity of 81% and specificity of 82%. PPV was 52% and NPV was 95%. The CDT component detected 13 true positives with a sensitivity of 48% and specificity of 100%. All specimens that were positive by CDT were also positive by GDH, CCNA, and LAMP. P-EIA was positive in 15 specimens, with a sensitivity of 56% and specificity of 100%. PPV and NPV were 100% and 90%, respectively. CCNA detected 26 positives with a sensitivity of 96% and specificity of 100%. LAMP detected 24 positives with a sensitivity of 89% and specificity of 98%. The sensitivities of both GDH-based algorithms were significantly higher than that of P-EIA by McNemar testing (P < 0.001). Using CCNA as the sole gold standard criterion, sensitivity, specificity, PPV, and NPV were all within 3% of those calculated using the composite gold standard definition.

With respect to discordance, three specimens were CCNA positive/LAMP negative. All three were GDH and P-EIA negative. There were also three LAMP-positive/CCNA-negative specimens. One was GDH positive, but the remaining two specimens were negative by all other tests. CCNA and LAMP were repeated on all six discordant specimens and yielded identical test results. Chart review of the discordant cases revealed that all six cases had diarrhea that met the definition provided by the 2010 SHEA-IDSA guidelines (3 or more diarrheal stools within 24 h or less) (2). None of the six cases had any other documented positive C. difficile toxin test results. Review of the LAMP-positive/CCNA-negative cases suggested false-positive LAMP results. None of the LAMP-positive/CCNA-negative cases had antibiotic or antineoplastic exposure. All three experienced symptom resolution without CDI therapy. In addition, one case had a positive rotavirus antigen EIA, and a second case had Blastocystis hominis in his stool. All three CCNA-positive/LAMP-negative cases had risk factors for CDI (exposure to antineoplastic agents in one and β-lactam antibiotics in the other two), within 8 weeks of the positive CCNA result. One case responded clinically to metronidazole therapy and therefore appeared to be a true CDI case. The other two tested positive for norovirus by PCR, providing an alternative diagnosis, making CDI less likely but still possible. All three patients had negative stool cultures.

DISCUSSION

In our pediatric cohort, CCNA and LAMP alone performed with the highest test sensitivities as stand-alone tests. The GDH-based algorithms performed with lower sensitivities (81%) than results reported for studies of adults in the literature (98 to 100%), suggesting that this approach may fail to diagnose 19% of pediatric patients with CDI (10, 12, 15). The reason for the lower sensitivity is unclear. It has been suggested that GDH antigen-based assays may demonstrate superior performance in populations with a high incidence of CDI mediated by hypervirulent C. difficile strains (16). Indeed, ribotyping of a small series of C. difficile isolates from our institution in 2007 indicated that the prevalence of the NAP1/027/III strain among P-EIA-positive stool specimens was only 14% (17). We note, however, that the sensitivity of the GDH-based algorithms was still statistically superior to P-EIA.

In this study, we observed a PPV of 100% for GDH-positive/CDT-positive specimens and an NPV of 95% for GDH-negative/CDT-negative specimens, strongly supporting the practice of reporting these specimens as “positive” or “negative” for C. difficile toxin immediately, without further testing. Eighty percent of all specimens tested in our cohort were immediately reportable following GDH/CDT testing (8% were GDH positive/CDT positive and 72% were GDH negative/CDT negative). The implications of the rapid GDH/CDT reporting are significant in our population. While syndromic infection control precaution assignment (13) is the current standard in our institution when the etiology of an inpatient with diarrhea is yet to be determined (i.e., a child with diarrhea will be isolated with contact precautions while symptomatic, regardless of C. difficile toxin status), our data indicate that about half of all CDI cases could be rapidly reported following GDH/CDT testing. With a 19% CDI prevalence, we anticipate that timelier initiation of CDI therapy and necessary environmental cleaning and disinfection procedures in even half of all CDI cases would be beneficial for the control of CDI in our patient population. Moreover, while rapid reporting of negative results may not affect resources allocated to the application of contact precautions for symptomatic patients, it may facilitate separation of inpatients into cohorts and encourage clinicians to consider other etiologic causes of antibiotic-associated diarrhea.

In our study cohort, the PPV of 52% for GDH alone confirmed the need to pursue additional testing of GDH-positive/CDT-negative specimens to resolve these specimens as positive or negative for C. difficile toxin. A PPV of 100% and NPV of 96% were observed for the entire algorithm when CCNA and LAMP were used in this capacity, indicating that both algorithmic approaches performed well in our population. The meaning of positive molecular CDI assays has been questioned in some quarters on the grounds that they do not detect C. difficile toxin or even the presence of organisms with the capacity to produce toxin in vitro but merely demonstrate the presence of the genes that code for toxins A and B (tcdA and tcdB) (18). In our study, the virtually identical performances of CCNA- and LAMP-confirmed algorithms provided reassurance that the user-friendly and relatively rapid LAMP procedure is as appropriate as CCNA in a confirmatory role. LAMP performed with sensitivity and specificity comparable to other studies evaluating this assay (7, 9).

With regard to discordance, LAMP did not detect three specimens that were CCNA positive. Chart review suggested that at least one of these cases was consistent with CDI, suggesting that the LAMP result may have been falsely negative. From the assay standpoint, the illumigene C. difficile assay uses an internal control to detect LAMP reaction failures related to inhibition reagent problems. Failure of the internal control reaction is indicated by an “invalid” test result. All three CCNA-positive, LAMP-negative results yielded valid negative results, indicating that the LAMP reactions were successful in each case. With respect to target design, while the illumigene C. difficile assay's tcdA (as opposed to tcdB) target has raised questions about its ability to detect toxin A⁻ B⁺ strains resulting from tcdA mutations, the target sequence appears to be conserved (5). Limited data demonstrating this assay's ability to detect these strains have been presented, supporting the manufacturers' assertion that this assay has the capacity to detect toxin A⁻ strains (5). Mutation at the actual LAMP target site, however, needs to be entertained as a possible explanation for these cases (5). Finally, false-negative LAMP results in specimens that were positive by toxigenic culture have been documented in the literature (7). While there are reports in the literature of an inferior sensitivity of CCNA compared to LAMP and other NAATs (7, 9, 14), the sensitivity of CCNA is highly dependent on a number of factors, including the quality of the cell line used and the time interval between specimen collection and inoculation of the cell culture plates (6). We feel that when carefully performed, CCNA may detect CDI with a sensitivity that compares with molecular techniques. While to our knowledge CCNA-positive/LAMP-negative specimens have not been reported in the literature, this phenomenon has been described with other commercial molecular assays (6) and may be explained by the presence of low bacterial concentrations of hypervirulent strains of toxigenic C. difficile producing disproportionately high levels of cytotoxin. Unfortunately, we did not have the means to pursue molecular typing of C. difficile isolates to investigate this hypothesis further.

There were three LAMP-positive/CCNA-negative specimens. Chart review suggested that these cases did not have CDI. None of the cases had clinical risk factors for CDI, and all three cases experienced symptom resolution without CDI therapy. In addition, two cases appeared to have alternative diagnoses. Possible false-positive LAMP results (using toxigenic culture as the gold standard) have been reported in the literature. Norén et al. described four such cases, with only one confirmed with an in-house PCR assay (9). End product detection in the loop-mediated isothermal amplification technology relies on pyrophosphate ion, released from deoxynucleoside triphosphate during DNA polymerization, reacting with magnesium ion to produce a precipitate. This is followed by detection of the resulting increase in turbidity of the reaction mixture (8). We speculate that false-positive LAMP results may be related to nonspecific DNA amplification generating magnesium pyrophosphate or substances in stool specimens that cause increases in turbidity of the reaction mixture.

This study's potential issue of controversy is the omission of toxigenic culture in our gold standard definition. The optimal gold standard method remains an issue of contention (18). While the SHEA-IDSA guidelines support the use of toxigenic culture for comparative studies (2), discussion in the literature continues to debate the merits of toxin detection versus the detection of C. difficile isolates with in vitro capacity to produce toxin (18). In this study, we opted to use CCNA as opposed to toxigenic culture, on the basis that detection of toxin (the sine qua non for CDI) is preferable (18). Also, while there are data suggesting that correlation between toxigenic culture and molecular testing is superior to CCNA with the latter (15), evidence also suggests that specimens that are negative by GDH/CDT and CCNA but positive by PCR and toxigenic culture may represent colonization with toxigenic C. difficile as opposed to actual CDI (6).

In summary, an algorithmic approach to CDI detection with the C. diff Quik Chek Complete product for initial CDI testing of pediatric stool specimens allowed swift reporting of 80% of C. difficile toxin results while increasing diagnostic yield over P-EIA in a tertiary pediatric population. The illumigene C. difficile assay provided an operator-friendly and accurate method of resolving GDH-positive/CDT-negative results. The sensitivity of the C. diff Quik Chek Complete test, however, may not be sufficient to allow its use as a stand-alone rule-out test in a pediatric population.