Laboratory diagnosis of Clostridioides difficile infection in symptomatic patients: what can we do better?

Abstract

The laboratory diagnosis of Clostridioides difficile infection (CDI) is challenging since this bacteria may be detected in healthy people and toxin production detection is not sensitive enough to be used alone. Thus, there is no single test with adequate sensitivity and specificity to be used in laboratory diagnosis. We evaluated the performance of tests used in the diagnosis of CDI in symptomatic patients with risk factors in hospitals in southern Brazil. Enzyme immunoassays (EIA) for glutamate dehydrogenase antigen (GDH) and toxins A/B, real-time polymerase chain reaction (qPCR), GeneXpert system, and a two-step algorithm comprising GDH/TOXIN EIA performed simultaneously followed by GeneXpert for outliers were evaluated. Toxigenic strain in stool culture was considered CDI positive (gold standard). Among 400 samples tested, 54 (13.5%) were positive for CDI and 346 (86.5%) were negative. The diagnosis of the two-step algorithm and qPCR had an excellent performance with an accuracy of 94.5% and 94.2%, respectively. The Youden index showed that GeneXpert as a single test (83.5%) and the two-step algorithm (82.8%) were the most effective assays. Diagnosing CDI and non-CDI diarrhea could be successfully attained by the combination of clinical data with accuracy of laboratory tests.

Introduction

Clostridioides difficile is a sporogenic anaerobic enteric pathogen found in animals and human intestinal microbiota. Recognized as the leading cause of healthcare-associated infectious diarrhea, they multiply and express their virulence factors when normal microbiota is disrupted, as seen during dysbiosis caused by antimicrobial use (the main risk factor). Diarrhea may occasionally evolve into pseudomembranous colitis [1].

Asymptomatic colonization in healthy people varies across different population groups [2]. The pathogenicity is related to the production of toxin A (TcdA) and toxin B (TcdB), and, in some strains, there is a production of a third toxin, known as a binary toxin (CDT) [3]. Overall, the mortality rate ranges from 6% to 13.5% among older adults [4].

Diagnosis of Clostridioides difficile infection (CDI) is confirmed by the presence of toxigenic C. difficile in samples of diarrheal stools [5, 6]. Some diagnostic methods that are currently available include tests that detect (i) the presence of the microorganism through toxigenic culture (TC), (ii) glutamate dehydrogenase enzyme (GDH), (iii) specific genes through nucleic acid amplification tests (NAAT), and (iv) the presence of the toxin in stool through cell cytotoxicity neutralization assay (CCNA) or immune enzymatic assays (EIA) [7].

The toxin detection using CCNA is performed to observe the cytopathic effect caused by toxins present in feces in cell culture and the neutralization of these toxins by antibodies. In TC, microorganisms are isolated in culture and tested for toxin production using CCNA or EIA. These techniques are the gold standard in studies of methodological comparisons. Although they are laborious and time-consuming for routine application in CDI diagnosis [8, 9], toxin detection by automated immunological methods or point-of-care tests is fast, easy to perform, and highly specific. Nevertheless, the limited sensitivity of these techniques, which may be attributed to the test’s limit of detection and the toxin’s rapid degradation at room temperature, restricts their application [5].

Molecular tests or NAATs are highly sensitive but may overestimate cases because they also detect asymptomatic patients colonized by toxigenic strains. Similarly, GDH detection has high sensitivity and low specificity because it does not differentiate between toxigenic and non-toxigenic strains or between infection and colonization [5, 10].

Therefore, a single test does not have sufficient sensitivity and specificity for an accurate diagnosis; hence, the recommendation to perform an algorithm that starts with a highly sensitive test and with a high negative predictive value [5, 6]. The combination of GDH and toxin detection has been used in several institutions around the world either as a standalone test or as a two-step algorithm, with TC or NAAT being used to resolve any discrepant cases. Although this method addresses the test limitations, it also raises the costs.

The two main issues when selecting and implementing diagnostic tests are underdiagnosis (failure to detect CDI) and overdiagnosis (detection of asymptomatic colonization). The lack of standardized laboratory practices makes it challenging to determine the best test to utilize and to interpret the findings [11]. The adequacy and performance of a test are significantly influenced by the selection criteria used to accept or reject samples [5–7, 10].

This study aimed to assess the effectiveness of performance tests used to diagnose CDI in symptomatic patients with risk factors. Procedures such as patient screening, control of the sample collection, sample preservation, and rapid test execution were performed to the greatest extent to reduce confounding factors in interpreting these tests. This procedure made it possible to assess the tests’ performance individually as well as in the combinations established by international guidelines, allowing the proposal of a viable strategy for CDI laboratory diagnosis in symptomatic patients.

Methods

Population, specimens, and ethical aspects of the study

A multicenter prospective cohort study was conducted in nine media to large-scale hospitals in Curitiba, Southern Brazil, from September 2017 to January 2020. The Institutional Research and Ethics Committee from each participating institution approved the study (Approval no. 59027716.5.1001.0096). All subjects signed the Informed Consent Form. Stool samples were collected from all aged groups who had diarrhea (≥3 episodes of unformed stool within 24 h) [7] after 48 h of admission and who previously used antimicrobial or chemotherapy up to 90 days before sample collection. Subjects with chronic diarrhea or laxative use, non-diarrheal, or repeated samples within the same diarrheal episode were excluded.

Laboratory diagnosis

Stool samples were transported under refrigeration (4–8°C) to the Laboratory of Bacteriology and the Laboratory of Molecular Biology of Infectious Diseases-Hospital de Clínicas Complex from the Federal University of Paraná for analyses. All samples that met the eligibility criteria were tested within 24 hours of collection. All samples were analyzed using EIA for GDH and toxin A/B detection (EIA–GDH/TOXIN), TC, and in-house qPCR tests (n = 400). In addition, 92 samples with GDH-positive results and 108 samples with GDH-negative results (n = 200) were also evaluated by the GeneXpert method (Fig. 1).

Fig. 1.

Flowchart of CDI diagnosis

GDH and toxin A and B detection (EIA–GDH/TOXIN)

EIA C. DIFF QUIK CHEK COMPLETE® (C. diff Quick Chek Complete, Alere Inc., Waltham, MA, USA) was performed according to the manufacturer’s instructions.

Toxigenic culture (TC)

Part of the sample was seeded on cefoxitin cycloserine fructose CCFA agar (Oxoid, Thermo Scientific, Basingstoke, UK) supplemented with 5% horse blood [10, 12]. The remaining part was shock-treated with alcohol and inoculated on Brucella agar supplemented with 5% horse blood and enriched with taurocholate (Sigma), before being incubated under anaerobic conditions at 37°C for 5 days [13]. The identification of the isolates was confirmed by matrix-assisted laser desorption/ionization mass spectroscopy (MALDI-TOF/MS, Bruker Daltonics, Germany) according to the manufacturer’s instructions and by tpi (triose phosphate isomerase) gene detection using PCR [14]. Toxin production from the isolates was confirmed using a thioglycolate broth (Oxoid, Thermo Scientific, Basingstoke, United Kingdom) with bacterial suspension and incubated overnight in the C. DIFF QUIK CHEK COMPLETE® (C. diff Quick Chek Complete, Alere Inc., Waltham, MA, USA) immune enzymatic test.

In-house real-time polymerase chain reaction (qPCR)

The primers used for toxin B (tcdB gene) detection were previously described [13] and validated by our group [15]. The sequences were as follows: F-5′AGCAGTTGAATATAGTGGTTTAGTTAGAGTTG3′, R-5′CATGCTTTTTTAGTTTCTGATTGAA3′. Reaction conditions were 20 μL of PowerUP™ SYBR™ Green Master Mix (Applied Biosystems, Foster City, California, USA), Primer mix (300nM), and 5 μL of target DNA (5–50ng DNA/ml). Denaturation cycles occurred for 120 s at 95 °C, followed by an amplification of 40 cycles performed in 3 steps: 15 s at 95 °C, 15 s at 55 °C, and 1 min at 72 °C. A melting curve was constructed for 15 s at 95 °C, 1 min at 60 °C, and 15 s at 95 °C. Specific products with a single peak temperature of melting (Tm) of 73 °C were considered positive.

GeneXpert® CEPHEID System (Xpert C. difficile test, Cepheid Inc., Sunnyvale, CA, USA)

Fully automated multiplex real-time PCR assay detects the toxin B (tcdB), binary toxin (cdt) genes, and possible deletion in the accessory Paloc gene (tcdC), used as a marker to Ribotype 027 hypervirulent strains. This test was performed according to the manufacturer’s instructions.

Two-step algorithm

The algorithm selected for the assessment included an initial screening with EIA for GDH and toxin A/B detection, followed by GeneXpert C. difficile test for discrepant samples (Fig. 2).

Fig. 2.

Algorithm for CDI diagnosis using GDH and toxin A/B EIA as a screening test

Severity of diarrhea evaluation

Clinical data were collected on a specific form. It was completed by the assisting physician and was sent with the patient sample to the Bacteriology Laboratory. Severe diarrhea was defined as bloody diarrhea or diarrhea with hypotension or hypoalbuminemia (albumin level <20 g/L), fever (temperature >38°C), and leukocytosis (white blood cells (WBC) > 15×10⁹ cells/L) or pseudomembranous colitis diagnosis [5, 6].

Statistical analysis

All samples with isolated toxigenic strains in their culture were classified as CDI cases (gold standard). Methods were evaluated individually and in algorithms. Fisher’s exact test was used to assess the differences in sensitivity, specificity, positive (PPV) and negative predictive values (NPV), 95% confidence intervals between each assay and the algorithm, and the presence of severe diarrhea and positivity in the tests used. Kappa index is a measure of interrater reliability, and its value is always less than or equal to 1. A value from 0.80 to 1 represents a very good agreement, and less than 0.20 represents a poor agreement [16]. The Kappa agreement index compared the two qPCR methods. The Youden index was used to assess the diagnostic tests’ accuracy [17]. The tests’ operational characteristics were performed using https://www.medcalc.org/calc/diagnostic_test.php, and the statistical analysis was performed using IBM SPSS Statistics 22.0 (IBM Corporation, Armonk, NY, USA), Minitab 16, and Excel Office 2010.

Results

A total of 426 samples were collected, 26 of which were removed from further analysis (14 non-diarrheal, nine repeated, one chronic diarrhea, and two in laxative usage). Among 400 included samples, 78 have C. difficile isolated from culture, but only 54 strains showed in vitro toxin production. As a result, 54 (13.5%) CDI cases and 346 (86.5%) negative samples met the established criteria. Table 1 summarizes the operational characteristics of the diagnostic tests evaluated, and Fig. 3 shows the tests’ individual performance CDI diagnosis according to the reference standard.

Table 1.

Operational characteristics results of CDI tests

		C. difficile infection*			Test performance
Test results		CDI	Non-CDI	Total	Accuracy	Sensitivity	Specificity	PPV	NPV
GDH	Positive	50	42	92	88.5%	92.6%	87.9%	54.3%	98.7%
	Negative	4	304	308
Toxin	Positive	23	4	27	91.2%	42.6%	98.8%	85.2%	91.7%
	Negative	31	342	373
qPCR	Positive	47	16	63	94.2%	87.1%	95.4%	74.6%	97.9%
	Negative	7	330	337
GeneXperta	Positive	51	16	67	91.0%	94.4%	89.1%	76.1%	97.7%
	Negative	3	130	133
Algorithmb	Positive	47	15	62	94.5%	87.1%	95.7%	75.8%	97.9%
	Negative	7	331	338

PPV positive predictive value, NPV negative predictive value

*According pre-established criteria—toxigenic culture positive (gold standard)

^aOnly 200 samples tested by GeneXpert were used in the calculation

^bEIA for GDH/TOXIN detection followed by GeneXpert

Fig. 3.

Performance characteristics of various assays for detection of toxigenic strains

All samples were initially tested by EIA for GDH/TOXIN detection, and 308 (77.0%) were negative for both test components (GDH−/TOXIN−). The GDH component was detected in 92 samples, 27 (29.4%) of which were classified as toxigenic (GDH+/TOXIN+). The remaining 65 samples (70.6%) exhibited “discrepant” findings (GDH+/TOXIN−) and were considered inconclusive based only on these two tests. Among the discrepant samples, 35 (53.8%) were toxin positive, and 30 (46.2%) were toxin negative when tested by the GeneXpert system. When test findings were compared to the gold standard, GDH was not detected in 4 samples that were positive for TC, and the toxigenic strains were not isolated in four (4) GDH+/TOXIN+ samples.

In-house qPCR detected C. difficile in 47 out of the 54 (87.1%) confirmed cases, and seven subjects with negative qPCR were considered CDI positive by the established criteria.

The GeneXpert system evaluated the two-step algorithm as a commercial molecular method. It was also tested in similar amounts on GDH-negative samples for better assessment. Thus, 200 samples comprising 92 GDH (+) and 108 GDH (−) were tested by this method. The test returned 47 positives out of the 92 GDH (+) samples and four positives out of the 108 GDH (−) samples. When compared with in-house PCR, the kappa correlation index between both tests had a very good agreement (0.87). The two-step algorithm test returned 24 positive samples with previous EIA discrepant results (GDH+/TOXIN−).

It has been recommended that patients with GDH (+) and TOXIN (−) tests be reassessed for the need for treatment. In this study, 11 out of the 65 patients with discrepant results (GDH+/TOXIN−) had severe diarrhea and were successfully treated. There was no statistical difference between subjects with severe diarrhea and those whose stool toxin was detected by EIA or TC. The recurrence rate was 16.7% (9/54), and it was also not related to toxin detection by EIA or TC.

The two-step algorithm and qPCR performed best in CDI laboratory diagnosis, with an accuracy of 94.5% and 94.2%, respectively. According to Youden’s index, which measures the balance between sensitivity (detection of disease) and specificity (absence of disease) of a diagnostic test, the most robust assay was the GeneXpert as a single test (83.5%), followed by the two-step’s algorithm (82.8%). Youden’s indices for sensitivity and specificity between qPCR and EIA were 82.5% and 41.4%, respectively.

Discussion

We examined the performance of several available molecular and immunological methods using samples collected from hospitalized patients with diarrhea. The operational characteristics of the tests were very similar, which is an essential aspect since it allows service providers to select methods that best suit the available laboratory equipment and personnel expertise.

There is no consensus about standardized test for the effective detection of C. difficile infection, and combination tests are susceptible to variations in performance. As a result, diagnosis may become challenging and sometimes controversial due to the variety of targets used in diagnostic methods [11, 18]. Unlike in developed countries, where standardized protocols for detecting CDI have been proposed on a regular basis [5, 6, 8, 19], there are no specific guidelines in developing countries [19].

The EIA for GDH, a rapid test that is simple to perform, showed satisfactory results for sensitivity and negative predictive values (92.6% and 98.7%, respectively), as previously reported [7, 10]. Only four CDI patients tested negative for GDH, probably due to low microbial load, below the detection threshold of the assay. Although only symptomatic individuals were tested, the non-toxigenic strains accounted for 41% of the positive samples, and these strains had an effect on the test’s specificity. This is consistent with previous findings, limiting the usefulness of the GDH as a single test in the diagnosis of CDI [10, 20]. Therefore, the GDH test is recommended as the first step in screening samples, either alone or in combination with a toxin detection test [7, 21].

The EIA for toxins showed adequate specificity, allowing its use as a reference test, despite a sensitivity of 42.6%. This assay is the most used in diagnosing CDI in developing countries due to its ease of use and low cost. However, its operational characteristics show a substantial variability in sensitivity (29–98.7%) and specificity (75–100%) values [5, 10], suggesting that it should not be used as an independent test in the diagnosis of CDI [5]. The divergent performance of EIA tests in detecting toxigenic strains occurs due to differences in the expression of toxins of geographically circulating strains (antigenic characteristic), the gold standard assay used, the binding of host antibodies to the toxins in the gastrointestinal tract, and the degradation of toxins during transport and storage [10, 22]. In this study, degradation of the toxins was minimized by transporting samples under refrigeration and rapid execution of the tests (within 24 h after collection). Despite this care with the samples, the detection of fecal toxin by the EIA test was not related to disease severity, as observed previously [23].

Despite specific protocols and previously documented advances for C. difficile recovery, four (4) GDH+/TOXIN+ samples were TC-negative, which is uncommon [10, 24–26]. Some possible explanations for growth culture inhibition include the following: (i) at the time of collection, the patients were under the effect of different antibiotic regimens, including vancomycin (1 patient), beta-lactams (2 patients), and combination of vancomycin and beta-lactams (1 patient) that may have inhibited in vitro growth, (ii) the time of exposure to air may have had an effect on C. difficile ribotype vegetative cell survival [26, 27], and (iii) the inability of a given sample strain to survive the culture during the enrichment process [28].

Although the specificity of the in-house qPCR (95.4%) was adequate, the sensitivity was lower than expected (87.1%), probably due to difficulties in DNA isolation from mucoid samples or non-detection of the target due to variation in the toxin B genes. Mehlig and colleagues suggested that the tcdB gene is not as conserved as expected [29], and similar results have been previously reported [27]. The four cases detected by qPCR that were not considered cases of CDI by the pre-established criteria could potentially be reclassified since all the evaluated patients were symptomatic and had risk factors for CDI. In addition, the reference methods had limitations in their sensitivity. Jensen et al. compared various NAATs (commercial and in-house assays) and found that, in two assays, the GeneXpert and the two-step methodology incorporating in-house qPCR had higher sensitivity for toxigenic culture [30].

Commercial molecular tests have improved diagnostic speed and accuracy, [31] but they may overestimate CDI incidence as they cannot differentiate colonization from infection [32]. The American Society for Microbiology recommends NAATs as a standalone test for adults only in patients with clinical signs of CDI [33]. As a more cost-effective option than commercial ones, in-house molecular assays are equally sensitive and accurately quantify human pathogens in fecal samples [34]. However, they are more expensive than EIA and require specific equipment, team skills, and infrastructure [7]. The automatized molecular assays, such as the GeneXpert system, are easy to perform and do not require infrastructure and trained personnel, but they are expensive.

The GeneXpert system as an independent test (used alone) had a sensitivity of 94.4%, and when used as an algorithm in combination with GDH/TOXIN, it detected 28 cases of CDI that were negative for the EIA toxin, thus significantly improving the diagnosis. Carroll and Mizusawa reported a variation of 10 to 100 times higher in the analytical sensitivity of NAATs for cytotoxin assays [7], and its use is indicated in cases of discrepant GDH/TOXIN tests. According to Grein et al., the NAATs, compared to the two-step test (GDH followed by PCR) or EIA toxin detection alone, significantly increased the number of patients testing positive for CDI [35]. As for in-house qPCR, the cases with positive results that are not considered CDI cases should be reevaluated due to the higher sensitivity of molecular methods [30]. The patients with qPCR positive in this study were reevaluated and two of them had several diarrheal symptoms. All patients received treatment for CDI and had symptom resolution. One patient had an infection recurrence after two months.

The correlation between NAATs (in-house qPCR and GeneXpert) was 87%, and the disagreements between the tests may be explained by factors inherent to the sample that affect the quality of DNA extraction [36] and the different primer sensibility in detecting tcdB variants [26, 29]. The European Society of Clinical Microbiology and Infectious Diseases and Infectious Diseases Society of America/Society for Healthcare Epidemiology of America (IDSA/SHEA) guidelines state that algorithmic assays should be preferred to those used independently, especially in institutions where there are no pre-agreed criteria for sending patients’ stool samples [5, 6, 18, 19]. Reller et al. showed the advantages of using two-step algorithms for the diagnosis of CDI, such as negative results reported on the same day and positive results within 24 h to 48 h [20]. A European study found an increase from 19% to 46% between 2011 and 2014 in CDI detection when adopting an optimal diagnosis [37].

Institutions face several obstacles in implementing CDI diagnosis, including high costs and the need to send samples to distant reference laboratories, delaying treatment time. The toxin A/B EIA test is the most commonly available [38]. The algorithm proposed in our cohort meets the requirements demanded by international entities, including starting with a highly sensitive test, such as GDH [5, 6]. We used the combination between simultaneous detection of GDH and toxins both by EIA (considered a rapid test) because the high negative predictive value (VPN) of GDH associated with the high positive predictive value (VPP) for toxins makes it possible to report the CDI (GDH+/TOXIN+) and non-CDI (GDH−/TOXIN−) cases, without additional tests. Recent studies report that GDH/EIA allows rapid identification and treatment of CDI patients, hence enabling disease control [39, 40].

Resolution of discrepant cases (GDH+/TOXIN−) proved necessary, even with adequate patient screening. Our findings agree with Cançado et al., who compared eleven different diagnostic algorithms and concluded that the GDH+/TOXIN− immunoassay arbitrated by NAAT was the best cost-effective strategy [37]. Other reports showed that this algorithm approach is fast and reliable [7, 41–44].

Screening samples using accurate and rapid tests has a direct influence on the prevention of pathogen transmission in healthcare services and therefore reducing clinical complications. In addition, the immediate exclusion of CDI diagnosis in patients with diarrhea significantly reduces the days of empirical treatment [35].

The performance of toxigenic C. difficile isolation (gold standard) test was one of the main limitations of this study. Samples that tested positive for toxins and failed to be isolate in culture were not considered true positives. Although all the included samples met the pre-analytical criteria, only patients with diarrhea and risk factors for CDI were included.

Conclusion

All evaluated methods showed adequate accuracy using specific inclusion criteria, suitable collection, and transport protocols. However, combining two or more methods appears to be a better cost-effective strategy. Laboratories should adopt criteria for accepting or rejecting samples to increase the probability of detecting C. difficile regardless of the testing method or algorithm used. Combining clinical data with precise laboratory tests allows differentiation between CDI and non-CDI diarrhea. A fast and accurate CDI laboratory diagnosis may lead to prompt and specific treatment. The algorithm starting with GDH screening is an effective method for routine usage and is cheaper than screening all samples first with NAAT.

Author contributions

All authors have accepted responsibility for the entire content of this manuscript and approved its submission.

Funding

This research was supported by a Brazilian Research Program for the Unified Health System: Management Shared Health—PPSUS 2015 edition—Fundação Araucária-PR/SESA-PR/CNPq/MS-Decit.

Data availability

All data generated or analyzed during this study are included in this published article.

Declarations

Ethics approval and consent to participate

The Institutional Research and Ethics Committee approved the study, and the need for informed consent was waived for all participating patients (CEAE = 59027716.5.1001.0096).

Competing interests

The authors declare no competing interests.