Evaluation of Four Different Diagnostic Tests To Detect Clostridium difficile in Piglets

Abstract

Clostridium difficile is emerging as pathogen in both humans and animals. In 2000 it was described as one of the causes of neonatal enteritis in piglets, and it is now the most common cause of neonatal diarrhea in the United States. In Europe, C. difficile infection (CDI) in both neonatal piglets and adult sows has also been reported. Diagnosis of this infection is based on detection of the bacterium C. difficile or its toxins A and B. Most detection methods, however, are only validated for diagnosing human infections. In this study three commercially available enzyme immunoassays (EIAs) and a commercial real-time-PCR (Becton, Dickinson, and Company) were evaluated by testing 172 pig fecal specimens (139 diarrheic and 33 nondiarrheic piglets). The results of each test were compared to those of cytotoxicity assays (CTAs) and toxigenic culture as the “gold standards.” Compared to CTAs, the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were, respectively, as follows: for real-time PCR, 91.6, 37.1, 57.6, and 82.5%; for Premier Toxins A&B (Meridian), 83.1, 31.5, 53.1, and 66.7%; for ImmunoCard Toxins A&B kit (ICTAB; Meridian), 86.6, 56.8, 66.9, and 80.7%; and for VIDAS (bioMérieux), 54.8, 92.6, 85.0, and 72.8%. Compared to toxigenic culture, the sensitivity, specificity, PPV, and NPV were, respectively, as follows: for real-time PCR, 93.0, 34.7, 50.0, and 87.5%; for Premier Toxins A&B, 80.3, 27.7, 43.8, and 66.7%; and for ICTAB, 80.0, 46.2, 52.8, and 75.4%; and for VIDAS, 56.4, 89.8, 77.5, and 76.7%. We conclude that all tests had an unacceptably low performance as a single test for the detection of C. difficile in pig herds and that a two-step algorithm is necessary, similar to that in cases of human CDI. Of all of the assays, the real-time PCR had the highest NPV compared to both reference methods and is therefore the most appropriate test to screen for the absence of C. difficile in pigs as a first step in the algorithm. The second step would be a confirmation of the positive results by toxigenic culture.

INTRODUCTION

Clostridium difficile is reported as the major cause of diarrhea in piglets from 0 to 7 days old (22). Nonetheless, some piglets with C. difficile infection (CDI) are nondiarrheic and even constipated or obstipated, although colitis is seen at necropsy (2, 29). CDI affects on average two-thirds of the litters, and within litters the morbidity can be as high as 97 to 100% (2, 21, 27). Mortality attributed to CDI in piglets is usually low, although outbreaks have been reported with mortality rates as high as 16% (2). Piglets recovered from CDI have growth retardation resulting in about half a kilogram lower average weaning weights (21).

Comparative analysis of piglet C. difficile isolates with isolates from humans suffering from CDI in the Netherlands showed overlapping antibiotic susceptibility profiles and a high genetic relatedness of the strains (8, 11). This has led to the assumption that transmission of C. difficile from piglets to humans and vice versa is likely to occur (8, 11). Since C. difficile is a potential zoonotic pathogen and a major cause of diarrhea in piglets, it is important to gain insight in the prevalence and transmission of C. difficile within and between pig populations. A prerequisite for these studies are reliable and validated detection methods.

No uniform consensus has been achieved on a “gold standard” to diagnose CDI in humans. Until recently, the cell cytotoxicity assay (CTA) has been used to evaluate the performance of new diagnostics tests, but now a known positive feces culture with a toxin-producing C. difficile strain (toxigenic culture) is more frequently used (6, 7).

No guidelines are available for diagnosing CDI in animals, and literature on this topic is scarce. Although commercially available detection methods for C. difficile are extensively evaluated for use to detect human infections, their performance in animal samples is largely unknown. Two commercial enzyme immunoassays (EIAs)—Clostridium difficile Tox A/B II (TechLab, Blacksburg, VA) and Gastro-Tect Clostridium difficile Toxin A+B (Medical Chemical Corp.)—for the detection of C. difficile toxins in humans have been evaluated for use with piglet fecal samples. A sensitivity of 91% was found for the TechLab Tox A/B II test compared to a cytotoxicity assay (19). The results of the Gastro-Tect assay were compared to the results of the TechLab A/B II in a study by Anderson and Songer (2), and a 39% sensitivity was found. Commercially available EIAs for the detection of C. difficile in human fecal samples were also described to have lower sensitivity when used in canine and horse fecal specimens compared to the use in human fecal samples (3, 5, 16).

Recently, commercially available molecular diagnostics, such as real-time PCR (RT-PCR) methods, for detection of the C. difficile toxin B gene (tcdB) have been introduced to diagnose human CDI (7). Except for an in-house-developed semiautomatic PCR method to detect the C. difficile toxin genes A (tcdA) and B (tcdB), and a C. difficile specific triose phosphate isomerase (tpi) housekeeping gene, real-time PCR methods have not been evaluated to diagnose animal CDI (1). The aim of the present study was to compare the test performances of three different immunological assays and one molecular test to the CTA and toxigenic culture as the gold standards. The following three EIAs for the detection of both toxins A and B were included: Premier Toxins A&B (Meridian), ImmunoCard Toxins A&B (ICTAB; Meridian), VIDAS (bioMérieux), and the GeneOhm Cdiff RT-PCR (Becton, Dickinson, and Company) for detection of the tcdB gene.

MATERIALS AND METHODS

Samples.

To obtain fecal samples from neonatal piglets, varying from 0 to 7 days in age, 18 pig breeding farms were visited between April 2009 and April 2010. The visited pig breeding farms were characterized by the presence of neonatal diarrhea for longer than 6 months. The cause of diarrhea was diagnosed, but not further specified as Clostridia diarrhea. In addition, both preventive vaccination and therapy with antibiotics were unsuccessful.

At each farm, fecal samples were taken from piglets from at least three different litters. In order to obtain fresh samples, piglets were gently squeezed in the abdomen to make them defecate in sterile 50-ml tubes (Falcon). At least 2.5 ml of feces was needed to perform all assays and culturing of a single sample. If the volume harvested from one piglet was insufficient to perform all of the assays and culturing, the samples from piglets were pooled until a sample volume of 2.5 ml was reached. From piglets with diarrhea 139 samples were investigated, including 89 pooled samples. From piglets without diarrhea 33 samples, which all consisted of pooled samples, were examined. Samples were stored immediately in a cool box. After transport, the samples were split. One part was used for the RT-PCR and stored at −20°C until use; the other part, used for culture and the EIA, was stored at 2 to 8°C and processed within 48 h. When processing was delayed, the samples were stored at −20°C. The frozen samples for culture and the EIAs were thawed only once, just before processing. Culturing and all of the EIAs were performed within the same day for each batch of samples.

Toxigenic culture, isolation, and identification of C. difficile.

All fecal samples were cultured for presence of C. difficile using C. difficile agar (CLO; bioMérieux) and Colombia CNA agar (CNA; bioMérieux) with ethanol shock pretreatment as described previously (13, 26). Inoculated plates were incubated for 48 h at 37°C in an anaerobic environment. Characteristic C. difficile colonies were picked from CLO plates, recultured upon on Schaedler agar (SCS; bioMérieux), and anaerobically incubated for another 48 h at 37°C. In cases where no growth on CLO plates was observed, colonies were picked from CNA plates and inoculated on SCS plates. Suspected C. difficile colonies on the SCS plates were selected by their typical horse manure smell, colony morphology, and by Gram staining.

PCR analyses.

Colonies were cultured in an anaerobic environment on CLOS plates (Becton, Dickinson). DNA was isolated from single C. difficile colonies by using a QIAamp DNA mini-blood kit (Qiagen) according to the manufacturer's protocol. C. difficile isolates were confirmed to be C. difficile by using an in-house developed PCR (18) for the detection of the C. difficile glutamate dehydrogenase gene (GluD). Confirmed C. difficile isolates were PCR ribotyped and characterized for the presence of genes encoding toxin A and toxin B as described previously (4, 18).

CTA.

The CTA for the detection of C. difficile toxins A and B was performed upon Vero cell monolayers grown in 24-well plates (Greiner) as described earlier (25). Fecal samples were diluted 1:4 in Eagle minimum essential medium (EMEM) supplemented with 10% fetal bovine serum (FBS) and centrifuged at 12,000 × g. The resulting supernatant was filtered through a 0.45-μm-pore-size filter. Subsequently, the filtrate was diluted 1:10 and 1:100 in EMEM with 10% FBS. Undiluted filtrate and both serial dilutions were added onto the Vero cell monolayers. Parallel samples were incubated with specific C. difficile antitoxin, which was diluted 1:25 (TechLab). The cells were examined both after 24 and 48 h of incubation at 37°C in a 5% CO₂ incubator. The result of the cytotoxicity assay was considered positive if cell rounding and detachment of cells was seen only in monolayers without antitoxin (25).

EIAs.

In the present study three different types of EIAs that detect both toxins A and B were evaluated: ICTAB (Meridian), which is a membrane-type enzyme-linked immunosorbent assay (ELISA); the Premier Toxins A&B (Meridian) kit, which is a well-type ELISA; and the VIDAS-AB (bioMérieux), an enzyme-linked fluorescence immunoassay.

All EIAs were carried out according to the manufacturers' recommendations. Plates of the Premier Toxins A&B were read at 450/630 nm (EL800 Universal Microplate Reader [Bio-Tek Instruments, Inc.). The optical densities (ODs) were recorded, and results were calculated according to the manufacturers' instructions. When the OD of a sample was above the cutoff value, the sample was recorded positive. If ODs were equal to the cutoff value, the results were recorded as equivocal. The ICTAB was read by eye according to manufacturer's protocol. When the entire reaction port was a very light blue, this was recorded as an equivocal result.

The VIDAS-AB assay was performed on the miniVIDAS, provided by bioMérieux. The VIDAS assay used algorithms set by the manufacturer to calculate the results. The results were recorded as positive, negative, or equivocal.

Real-time PCR.

The real-time PCR assay (GeneOhm; Becton, Dickinson) for detection of the tcdB gene was performed according to the manufacturers' instructions on a SmartCycler (Cepheid, United Kingdom; supplied by Becton, Dickinson at the time of the study). The software of the SmartCycler recorded the results of the PCR assay as positive, negative, or unresolved.

Statistical analysis.

The sensitivity and specificity were calculated for each kit against both of the gold standard assays (CTA and toxigenic culture). The sensitivity and specificity data were used to calculate the positive predictive value (PPV) and the negative predictive value (NPV). The data were analyzed using SPSS.16 software to determine whether storage conditions influenced the results of the detection methods. Therefore, data were stratified based on whether the samples were fresh or frozen, and Pearson chi-square tests were performed for every detection method compared to the reference methods. The Breslow-Day test was used to examine the homogeneity of the odds ratios of the strata. Furthermore, the data of the frozen samples were stratified based on the storage time of the samples. Pearson chi-square and Breslow-Day tests were used to examine whether the storage time at −20°C influenced the results of the detection methods.

Ethical approval.

Ethical approval for this study was granted from the Animal Experiments Committee of the University of Utrecht.

RESULTS

Refrigerated and frozen samples.

In all, 172 samples were analyzed using the methods described above. A total of 35 samples were processed within 48 h, and 137 samples were frozen upon arrival and processed after thawing once. Statistical analysis showed that the Breslow-Day test was not significant for any of the tests, indicating that the storage conditions did not influence the results of the detection methods. Therefore, it was not necessary to calculate the specificity and sensitivity for the fresh and frozen samples separately. The storage times of the frozen samples varied from 1 to 6 months. When the test results of samples processed within 2 months (n = 38) were compared to test results of samples processed after 2 months (n = 99), no significant differences between the test results were found for any of the tests. Also, when the test results of the samples obtained within 2 months (n = 38) were compared to the test results of the samples obtained between 4.5 to 6 months (n = 39), no significant differences were found.

Equivocal values.

Using the VIDAS assay, 31 samples (18.0%) were recorded as equivocal. The ICTAB yielded 9 (5.2%) equivocal values. These equivocal values were excluded from the analysis of the performance characteristics of both tests.

The manufacturer's recommendation, in the case of equivocal results, was not followed in the present study since insufficient substrate was left to repeat the test with the original specimen, and repeated sampling was not possible due to practical considerations.

Isolation and characterization of C. difficile.

C. difficile was isolated from 71 (41.3%) of the 172 fecal samples: 59 (34%) isolated strains were derived from diarrheal piglets (n = 139), and 12 (36%) isolates were cultured from nondiarrheal piglets (n = 33). A total of 70 of the 71 isolates were characterized as PCR ribotype 078, and one isolate belonged to PCR ribotype 045. An overview of these results is given in Table 1. The difference in prevalence of C. difficile between the diarrheal and nondiarrheal piglets was not significant (P = 0.52).

Table 1.

Results of CTA and toxigenic culture

Animal group	Toxigenic culture		PCR ribotype	CTA
	Result	No. of samples (%)		Result	No. of samples (%)
Diarrheal piglets	Positive	59 (42.4)	078/045	Positive	68 (48.9)
	Negative	80 (57.6)		Negative	71 (41.3)
Nondiarrheal piglets	Positive	12 (36.4)	078	Positive	15 (45.5)
	Negative	21 (63.6)		Negative	18 (54.5)
Total piglets	Positive	71 (41.3)	078/045	Positive	83 (48.3)
	Negative	101 (58.7)		Negative	89 (51.7)

CTA.

The results of the CTA are given in Table 1.

In all, 83 (48.3%) of the 172 fecal samples were determined to be positive by CTA. All positive results were observed within 24 h of incubation. Of the fecal samples of 139 diarrheal piglets, 68 (48.9%) were positive with CTA, whereas 15 (45.5%) of the 33 nondiarrheal piglets fecal samples were positive. No difference in prevalence between the diarrheal and nondiarrheal piglets was found (P = 0.72).

Concordant results.

Concordant results with all detection methods, including the reference methods, were observed in 29 (17%) samples. Concordant positive results for all methods were recorded for 24 samples (14%), and concordant negative results were obtained for 5 samples (3%).

A positive result for all of the toxin detection methods, including CTA, was observed in 30 samples (17%).

Sensitivity, specificity, PPV, and NPV of the EIAs and the PCR assay compared to toxigenic culture and CTA.

The sensitivity and specificity data of the EIAs and the RT-PCR assay calculated versus CTA and toxigenic culture are shown in Tables 2 and 3.

Table 2.

Comparison of various detection methods to CTA (n = 172)^a

Assay	Assay result (no. of samples)			CTA result (no. of samples)		% (range)
	Pos.	Neg.	Eq.	Pos.	Neg.	Sensitivity	Specificity	PPV	NPV
GeneOhm PCR	132	40	0	83	89
Premier Toxins A&B	130	42	0	83	89	83.1 (75.1–91.2)	31.5 (21.8–41.1)	53.1 (44.5–61.7)	66.7 (52.4–80.9)
ImmunoCard Toxins A&B	106	57	9*	82	81	86.6 (79.2–94.0)	56.8 (46.0–67.6)	66.9 (58.0–75.9)	80.7 (70.5–90.9)
VIDAS	40	103	29*	62	81	54.8 (42.5–67.2)	92.6 (86.9–90.3)	85.0 (73.9–96.1)	72.8 (64.2–81.4)

^aPos., positive; Neg., negative; Eq., equivocal. Equivocal values were excluded from the analysis of the performance characteristics of the tests.

Table 3.

Comparison of various detection methods to toxigenic culture (n = 172)^a

Assay	Assay result (no. of samples)			CTA result (no. of samples)		% (range)
	Pos.	Neg.	Eq.	Pos.	Neg.	Sensitivity	Specificity	PPV	NPV
GeneOhm PCR	132	40	0	71	101	93.0 (87.0–98.9)	34.7 (25.4–43.9)	50.0 (41.5–58.5)	87.5 (77.3–97.7)
Premier Toxins A&B	130	42	0	71	101	80.3 (71.0–89.5)	27.7 (19.0–36.5)	43.8 (35.3–52.4)	66.7 (52.4–80.9)
ImmunoCard Toxins A&B	106	57	9	70	93	80.0 (70.6–89.4)	46.2 (36.1–56.4)	52.8 (43.3–62.3)	75.4 (64.3–86.6)
VIDAS	40	103	29	55	88	56.4 (43.3–69.5)	89.8 (83.4–96.1)	77.5 (64.6–90.4)	76.7 (68.5–84.9)

^aPos., positive; Neg., negative; Eq., equivocal. Equivocal values were excluded from the analysis of the performance characteristics of the tests.

The GeneOhm PCR assay was more sensitive than the EIAs compared to both reference methods. The GeneOhm assay was less specific than the ICTAB or VIDAS assays. The specificity of the tests ranged from 27.7 to 92.6%. The VIDAS assay performed best, but none of the tests demonstrated a high specificity.

The PPVs and NPVs for the EIAs and the GeneOhm compared to both reference methods are shown in Tables 2 and 3. The highest NPV was obtained with the GeneOhm assay. The PPVs of the GeneOhm, ICTAB, and Premier Toxins A&B assays were comparable. The highest PPV was obtained with the VIDAS assay.

DISCUSSION

We show here that the current commercially available tests for detection of C. difficile or its toxins in humans with CDI have a much lower sensitivity and specificity when used in porcine fecal samples than in human stool samples. For the evaluation of these tests, 172 fecal samples of piglets were analyzed using three EIAs, one RT-PCR, and two reference methods considered to be gold standards. Even though there was a high prevalence of CDI in the study population, the PPVs and NPVs of the tests were unacceptably low. The concordance between the tests and the reference methods was only 16.9%. The lower sensitivities and specificities of the tests than in human samples is in accord with the findings of other studies, wherein the performance of EIAs in fecal samples of pigs, dogs, and horses was also described to be less than in human fecal samples (2, 3, 5, 16). The lower specificity of the detection methods when used to detect CDI in animals may be caused by the presence of an inhibitor in animal feces that reduces the specificity of binding of the toxins (2, 5). However, published data on defined inhibitors in animal fecal specimens are not available. This seems to be an important focus for further research.

Discordant results with the reference methods were obtained with all of the tests examined here. One test (VIDAS) also had a high number (18.0%) of equivocal results. The samples that yielded discordant results with the reference methods varied widely between the different tests, indicating that it was not a core group of samples repeatedly giving false-positive or false-negative results. This implies that incorrect results were due neither to the inaccuracy of the interpreter of the assays nor due to the presence of disturbing substances in a group of samples. Negative toxin tests of feces that harbor C. difficile could indicate colonization with a nontoxigenic C. difficile strain. Since all of the recovered isolates examined here were toxigenic, this possibility can be excluded for our study. Another reason for the discrepancy between negative toxin tests and positive culture results could be toxin degradation due to inappropriate storage, but all manufacturers' instructions state that freezing of fecal samples at −20°C will not affect test results. A study by Weese et al. (28) showed that toxins remain detectable by ELISA (the Clostridium difficile Tox A/B Test; TechLab) when feces samples from horses inoculated with C. difficile isolates were stored at −20°C for 60 days. Nonetheless, contrasting results were reported by Freeman and Wilcox (10), who diluted human feces 1/20 in prereduced phosphate-buffered saline (pH 7.4) and inoculated each of these fecal emulsions with one of three different C. difficile strains. Toxin titers of the inoculated fecal emulsions that were frozen at −20°C and measured by CTA were significantly lower than the toxin titers of refrigerated fecal emulsions (P < 0.01) by day 56 of the experiment in fecal emulsions inoculated for two of the three strains used. These two studies suggest that storage at −20°C might have resulted in a loss of cytotoxicity, but not of immunological recognition of the toxins (10, 28). Nonetheless, in our study no significant differences were found between samples that were frozen for <2 months and samples that remained frozen for longer times (Table 4). Furthermore, Chouicha and Marks (5) suggest that the presence of protease activity in animal fecal specimens may also lead to increased toxin degradation, and toxins may not even reach detection levels. Finally, it remains possible that the presence of toxin-producing strains in feces samples and the absence of free toxins A and B reflect an asymptomatic carriership, as is known for humans. Another reason for a discrepancy between a negative CTA and a positive culture of toxin-producing strain is a lack of a standardized CTA. No standard protocol for the CTA exists. This results in differences in cell line used for the assay, with different sensitivities, and in different methods, such as the dilution of stool specimens (15).

Table 4.

P values of the Breslow-Day test for all test results stratified on storage conditions

Storage condition comparison and assay	Pa
	CTA	Toxigenic culture
Refrigerated vs frozen samples
Premier Toxins A&B	0.332	0.525
ImmunoCard Toxins A&B	0.773	0.793
VIDAS	0.053	0.840
Processed within 2 months vs processed after 2 mo
RT-PCR	0.906	0.569
Premier Toxins A&B	0.237	0.345
ImmunoCard Toxins A&B	0.249	0.742
VIDAS	0.632	0.969
Processed within 2 months vs processed after 4.5 to 6 mo
RT-PCR	0.835	0.416
Premier Toxins A&B	0.594	0.065
ImmunoCard Toxins A&B	0.110	0.672
VIDAS	0.244	0.668

^aThat is, the Breslow-Day test statistical P value.

In our study the concordance between the two reference methods was 79.1%. The sensitivity of the CTA compared to toxigenic culture was 83.1%. A low sensitivity of the CTA has previously been reported by several authors, ranging from 56.7% (9) to 61.7% (24) to 74.0% (14). The sensitivity of toxigenic culture compared to CTA was 71.1%. It is difficult to compare reported data on the sensitivity of culture due to the differences in culture methods, e.g., with or without enrichment, heat, or alcohol shock and the use of different C. difficile-specific media. In our study, the samples were pretreated with an alcohol shock, followed by inoculation onto standard commercial media to improve the comparability of our results. Reller et al. (20) followed a similar culture method for human samples, although the pretreatment was with heat instead of alcohol, and they reported a sensitivity of 87%.

The low sensitivity of the reference methods used in the present study hampers the PPV values of the four studied test methods considerably. The high prevalence of positive samples observed here increases the PPV of the tests. Because the PPV and NPV of a test are codetermined by the prevalence of the pathogen in the sampled population, it is important to evaluate a test in a population with a prevalence that reflects the normal situation. The high prevalence of C. difficile (41.3% with toxigenic culture and 48.3% with CTA) that we observed here is in accord with the prevalences of C. difficile reported in piglets in other studies in different countries (17, 22, 30).

The main difference between diagnosing CDI in humans and in pigs is the fact that in humans the individual patient is diagnosed, whereas in pigs CDI is diagnosed in groups of pigs or herds. The sensitivity, specificity, and predictive values of the tests are too low for the diagnosis of CDI in an individual piglet; however, the tests can be used to screen for the presence of C. difficile in pig herds. A test with a high sensitivity ensures that farms where C. difficile is present can be identified. The fact that C. difficile is a possible zoonotic pathogen may lead to future surveillance programs for C. difficile in pigs on farms. This underscores the importance of a high NPV of a test to reliably declare a pig herd free of C. difficile. The GeneOhm PCR assay showed the highest sensitivity and NPV and is therefore the most suitable test for screening a large number of samples for C. difficile on pig farms. However, the PPV of the GeneOhm PCR assay is rather low. Consequently, a two-step algorithm is necessary, with as a second step confirmation of samples that were determined to be positive by the GeneOhm PCR. The confirmatory test could be a reference test, such as toxigenic culture. An additional advantage of culture would be that more information about the isolate, such as the ribotype and antibiotic susceptibility can be obtained. A two-step algorithm is also recommended in human medicine for the diagnosis of CDI (6, 7).

A limitation of our study is that the strains were almost exclusively ribotype 078, which could have influenced the sensitivity of the tests. Ribotyping of the isolates that were obtained here showed that 70 isolates (99% of all isolates) belonged to PCR ribotype 078, and one isolate was identified as type 045. The predominance of type 078 could have attributed to the low sensitivity of the assays, because recent research by Tenover et al. (23) shows that the PCR ribotype of strains can have an impact on the sensitivity of molecular diagnostics and EIAs of human feces samples. A lower sensitivity (81.8% with the RT-PCR and 63.6% with EIA) was found for the C. difficile PCR ribotype 078 than for many other ribotypes (for example, ribotype 027, 100% with the RT-PCR and 78.4% with EIA) using molecular diagnostics and EIAs (23). The high prevalence of ribotype 078 reflects the current situation in the Netherlands, where 078 is the predominant ribotype in piglets (12). Furthermore, it would have been interesting to include another RT-PCR in the study. The strengths of our study are the study design with the use of both gold standards as a reference, and it is the first study to evaluate a commercially available RT-PCR and three EIAs for the use of detection of C. difficile in pigs.

We conclude that all tests in our study had an unacceptably low performance and that a two-step algorithm is necessary, similar to the situation in human medicine.