Comparison of commercial assays and two-step approach to detect Clostridioides difficile in South Africa

Background Clostridioides difficile is the number one cause of hospital-acquired diarrhoea. Accurate diagnosis of C. difficile is of utmost importance as it guides patient management and infection control practices. Studies evaluating the performance of commercially available nucleic acid amplification tests (NAATs) versus algorithms are lacking in resource-limited settings. Objective This study assessed the performance of three commercially available tests and a two-step approach for the diagnosis of C. difficile infection using toxigenic culture (TC) as the gold standard. Methods Two hundred and twenty-three non-duplicate loose stool samples were submitted to the National Health Laboratory Service Microbiology Laboratory at Tygerberg Hospital, Cape Town, South Africa, from October 2017 to October 2018. The samples were tested in parallel using the C. DIFF QUIK CHEK COMPLETE enzyme immunoassay (EIA) and two NAATs (Xpert C. difficile and BD MAX Cdiff), and the results were compared to TC. The performance of a two-step approach consisting of the C. DIFF QUIK CHEK COMPLETE followed by the Xpert C. difficile was also determined. Results Of 223 faecal specimens tested, 37 (16.6%) were TC-positive. The sensitivity and specificity of the C. DIFF QUIK CHEK COMPLETE were 54.1% and 98.9%; Xpert C. difficile, 86.4% and 96.8%; BD MAX Cdiff, 89.2% and 96.8%; and two-step approach, 89.2% and 96.2%. Conclusion The C. DIFF QUIK CHEK COMPLETE, in a two-step approach with the Xpert C. difficile, performed similarly to the NAATs on their own and offer advantages in terms of cost and workflow in low-resource settings.


Introduction
C. difficile are toxigenic culture (TC) and the cell culture neutralisation assay; however, these are labour-intensive and have extended turnaround times of 2-3 days. 10 Current guidelines recommend a two-step approach: an enzyme immunoassay (EIA), followed by a molecular test to increase the diagnostic yield. 11 This approach, compared to toxin EIA only, is much more sensitive. Algorithm-based testing is recommended to optimise the positive predicative value of laboratory results. 11 Commercial EIAs utilise monoclonal antibodies to detect glutamate dehydrogenase (GDH), an antigen common to all C. difficile strains irrespective of toxin production, and Toxin A and/or Toxin B (Tox A/B). The C. DIFF QUIK CHEK COMPLETE EIA detects GDH as well as Tox A/B. Commercial nucleic acid amplification tests (NAATs) are usually real-time polymerase chain reaction (PCR) assays which target tcdB.
In this study, the performance of C. DIFF QUIK CHEK COMPLETE (QUIK CHEK) (Alere Techlab, Blacksburg, Virginia, United States) and two commercial NAATs, the Xpert C. difficile (Xpert) (Cepheid, Sunnyvale, California, United States) and BD MAX Cdiff (BDM) (Becton Dickinson, San Jose, California, United States) were compared to TC. The performance of a two-step approach consisting of the QUIK CHEK and Xpert, that is, the current standard of care (SOC) at our institution, was also evaluated.

Ethical considerations
Ethics approval was obtained from the Health Research Ethics Committee of Stellenbosch University, Cape Town, South Africa (reference number S17/03/064). A waiver of informed consent was obtained as no patient identifiers were published and no invasive procedures were performed as a result of this study. Only the results of the current SOC tests were reported to physicians. After the results were recorded, a study number was assigned to the specimen and all patient identifiers were removed.

Study design
This was a prospective diagnostic test accuracy study comparing three different assays, namely the QUIK CHEK, Xpert and BDM, to TC (reference method) for the detection of toxigenic C. difficile in faecal samples. The performance of the current SOC (a two-step approach consisting of the QUIK CHEK and Xpert), at our institution was also compared to TC. A composite reference standard (CRS) analysis was performed to account for limitations in TC as a reference method (Supplementary Material). 12 The CRS composite positive was defined as TC-positive or positive by two commercial assays in a TC-negative sample.

Sample size
Tygerberg Hospital near Cape Town, South Africa, is a 1380bed tertiary hospital which delivers specialist services to approximately half the population of the Western Cape province (total population 6.2 million). The Microbiology Laboratory of the National Health Laboratory Service at Tygerberg Hospital performs C. difficile testing on patient samples from Tygerberg Hospital as well as peripheral hospitals and clinics within the Tygerberg drainage area. Assuming a C. difficile prevalence of 16% based on previous studies, and using a 95% confidence interval with a 5% error rate on both sides, a sample size of 207 was calculated using the Open Epi sample size calculator (G Dean & KM Sullivan, Atlanta, Georgia, United States). 13

Sample processing
Non-duplicate loose stool samples (defined as taking the shape of the container) from adult and paediatric patients older than two years of age submitted to the National Health Laboratory Service Microbiology Laboratory at Tygerberg Hospital from October 2017 to October 2018 for routine C. difficile testing were tested in parallel with the four assays. None of the samples was frozen and thawed prior to testing. In the rare event that samples could not be tested on the day of collection, they were kept at 2 °C -8 °C and processed within 48 h of collection.
The QUIK CHEK EIA was performed as per instructions provided by the manufacturer and read by multiple laboratory technologists as part of their routine daily work. The results were interpreted as follows: GDH-positive and Tox A/B-positive samples were regarded as positive, GDH-negative and Tox A/B-negative samples were regarded as negative, and GDH-positive and Tox A/B-negative samples were regarded as negative.
The Xpert NAAT was performed as per the instructions of the manufacturer. The test was interpreted as positive for toxigenic C. difficile if the cytotoxin gene (tcdB) was detected within the valid cycle threshold range and above the minimum endpoint setting, and as toxigenic C. difficile negative if the tcdB gene was not detected, provided the sample processing control and probe check controls met the manufacturer's requirements. Testing was repeated on any samples with invalid or error results due to failure of the sample processing control or probe check controls.
The BDM NAAT was performed by following the instructions of the manufacturer. Test results were automatically interpreted by the BDM instrument. Positive, negative and unresolved results were based on the target's and sample processing control's amplification status. The test was interpreted as toxigenic C. difficile positive if the cytotoxin gene (tcdB) was detected within the valid cycle threshold range and above the minimum endpoint setting, and as toxigenic C. difficile negative if the tcdB gene was not detected, provided the sample processing and probe check controls met the manufacturer's requirements. Indeterminate and incomplete results are obtained when the BDM system fails. Any unresolved, indeterminate and incomplete samples were repeated.
The two-step algorithm was performed as follows: The QUIK CHEK was done and samples that were GDH-positive and Tox A/B-positive were regarded as positive; GDH-negative and Tox A/B-negative samples were regarded as negative. Samples that were GDH-positive and Tox A/B-negative were interpreted in conjunction with the Xpert results to establish if toxin genes were present or absent.
Toxigenic culture was used as the reference method and performed by culturing C. difficile from stool samples on chromID C.diff agar (bioMérieux, Marcy l'Etoile, France), a differential and selective medium, followed by testing for the organism's ability to produce toxin by PCR. 14,15,16 A swab was dipped into the stool sample and inoculated onto the agar medium. The inoculum was streaked to enhance the recovery of single colonies. The plates were placed in a jar with an anaerobic sachet, AnaeroPack-Anaero (Mitsubishi Gas Chemical Company, Inc., Tokyo, Japan) and incubated at 35 ºC. After 48 h of incubation, the plates were examined for grey and black colonies; if none were present, the culture was considered negative for C. difficile. A multiplex PCR was performed on a streak of grey and black colonies to detect the C. difficile-specific tpi gene, as well as the tcdA and tcdB genes. 17 Cultures that were PCR-positive for the tpi and tcdB genes were considered TC-positive, while those that were PCR-negative for the tpi or tcdB genes were considered TC-negative. In humans, the tcdA gene has been reported in tcdB-positive C. difficile strains only. 18

Statistical analysis
Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were calculated for each assay against both TC and the CRS using Epicalc 2000 version 1.00 software (Brixton Books, London, England).

Results
A total of 223 samples were included in the study. Of these, 37 (16.6%) were TC-positive and 19/37 (51.4%) were also positive from all three commercial assays. The SOC positivity rate was 17.9% (40/223).  Table 1).
The QUIK CHEK performed poorly while the Xpert and BDM and the two-step approach had similar sensitivities, specificities, PPV and NPV when compared to TC. Results were also compared to a CRS (Supplementary Table 1).
The Xpert, BDM and two-step approach showed higher sensitivities, specificities and PPV in comparison with the CRS, but the NPV of all the assays were similar.

Discussion
There is a lack of consensus regarding the optimal diagnostic C. difficile laboratory assays. High-quality evidence for the best diagnostic testing strategy is scarce and researchers rarely use either of the two accepted reference standards, that is, cell culture neutralisation assay or TC, for assessment of diagnostic accuracy, but rather their own laboratory-defined CRS criteria. In this study, we determined the diagnostic accuracy of three different commercial assays for the detection of toxigenic C. difficile in stools compared to both TC and a CRS to account for any limitations of the TC. Clostridioides difficile toxin EIAs lack sensitivity. All C. difficile contain the GDH antigen, whether toxin genes are present or absent. Glutamate dehydrogenase EIAs have high sensitivities but poor specificities; therefore, an additional test (most commonly a toxin assay) must be performed. The GDH EIA is the first test performed in a two-step or threestep approach where a GDH-positive result is followed by a toxin assay or a NAAT to detect toxin genes. 11 Our findings for the QUIK CHEK showed a sensitivity, specificity, PPV and NPV of 54.1%, 98.9%, 90.9% and 91.5%, respectively. These findings were similar to a study conducted in 2012 in Kuwait comparing the Xpert, QUIK CHEK and TC, which found the sensitivity, specificity, PPV and NPV of the QUIK CHEK to be 53.85%, 100%, 100% and 98.51%, respectively, when using a CRS defined as two tests being in agreement. 19 A 2009-2018 study conducted by Chung and Lee in Korea compared the diagnostic performance of the QUIK CHEK to Xpert as a reference test, with sensitivity, specificity, PPV and NPV of 55.4%, 100.0%, 100.0% and 80.0%, respectively. 20 However, they had a higher prevalence of C. difficile infection (35.9%) than our study population (16.6%), which could explain the difference in PPV (100.0% vs 90.9%). 20 In contrast, a study conducted in 2013-2014 by Seo et al. in Korea, found a lower sensitivity of 45.7% when using either TC or the combination of QUIK CHEK and Xpert as a reference standard. 21 Nucleic acid amplification tests that target chromosomal toxin genes show high sensitivity and specificity, provide rapid results, and are amenable to both batch and on-demand testing. 22 Reported estimates of sensitivity for NAATs range from 77% to 100% when compared to TC. Specificity ranges from 83% to 100% in comparison to TC. 23 nucleic acid amplification tests may detect asymptomatic carriage due to possible non-expression of the toxin encoding gene and therefore the clinical specificity may be lower than reported. 24 In their meta-analysis conducted in 2019, Kraft 10 In contrast, the Seo et al. study found a higher sensitivity, specificity, PPV and NPV of 94.0%, 100.0%, 100.0% and 100.0% for the two-step approach, which could be attributed to the CRS in this study being either TC-positive or a combination of Xpert-and QUIK CHEK-positive. 21 Our algorithm performed similarly to the Xpert or the BDM alone in terms of sensitivity, specificity, PPV and NPV.
The Xpert, BDM and two-step algorithm performed similarly, with overlapping confidence intervals when compared to TC and CRS. From these findings it is evident that TC is a robust reference test for the statistical measures of sensitivity, specificity, PPV and NPV for the Xpert, BDM and two-step approach.

Limitations
Limitations of our study include the use of stool sampless conforming to the shape of the container as a surrogate for diarrhoea, as well as not excluding other causes of diarrhoea. In addition, we did not collect pre-analytical data such as prior or current antibiotic use or determine any other risk factors for C. difficile. Post-analytical patient outcome data was also not collected. The sensitivity of lateral flow assays is limited by the dissociation constant of the antibodyantigen conjugate and by user interpretation of the colorimetric read-out. The TC method used in this study did not include heat-shock treatment prior to the inoculation of samples onto media, which may have improved the detection of C. difficile.