|Home | About | Journals | Submit | Contact Us | Français|
The continuing rise in the incidence of Clostridium difficile infection is a cause for concern, with implications for patients and health care systems. Laboratory diagnosis largely relies on rapid toxin detection kits, although assays detecting alternative targets, including glutamate dehydrogenase (GDH) and toxin genes, are now available. Six hundred routine diagnostic diarrheal samples were tested prospectively using nine commercial toxin detection assays, cytotoxin assay (CYT), and cytotoxigenic culture (CYTGC) and retrospectively using a GDH detection assay and PCR for the toxin B gene. The mean sensitivity and specificity for toxin detection assays were 82.8% (range, 66.7 to 91.7%) and 95.4% (range, 90.9 to 98.8%), respectively, in comparison with CYT and 75.0% (range, 60.0 to 86.4%) and 96.1% (91.4 to 99.4%), respectively, in comparison with CYTGC. The sensitivity and specificity of the GDH assay were 90.1% and 92.9%, respectively, compared to CYT and 87.6% and 94.3%, respectively, compared to CYTGC. The PCR assay had the highest sensitivity of all the tests in comparison with CYT (92.2%) and CYTGC (88.5%), and the specificities of the PCR assay were 94.0% and 95.4% compared to CYT and CYTGC, respectively. All kits had low positive predictive values (range, 48.6 to 86.8%) compared with CYT, assuming a positive sample prevalence of 10% (representing the hospital setting), which compromises the clinical utility of single tests for the laboratory diagnosis of C. difficile infection. The optimum rapid single test was PCR for toxin B gene, as this had the highest negative predictive value. Diagnostic algorithms that optimize test combinations for the laboratory diagnosis of C. difficile infection need to be defined.
Clostridium difficile is a major nosocomial pathogen causing a range of symptoms from mild to severe diarrhea and is the etiological agent of pseudomembranous colitis. The incidence of C. difficile infection has increased markedly in many countries, notably associated with the epidemic spread of PCR ribotype 027 (NAP1) since its recognition in the United States and Canada (6, 7, 13). It is essential to have accurate laboratory diagnosis of C. difficile infection to ensure patients receive appropriate treatment and that correct infection control measures are put in place. Also, inaccurate testing will potentially lead to poor quality surveillance data that may lead to inappropriate infection prevention measures.
The cytotoxin assay (CYT), first described by Chang et al., detects the toxins produced by C. difficile in the supernatants of patient feces, using both antitoxin-protected and nonprotected cell monolayers (2). This assay is commonly used as the gold standard method for comparison in toxin kit evaluations, although its use in routine microbiology laboratories has largely been superseded. Cytotoxogenic culture (CYTGC) has been used as an alternative gold standard method to CYT testing, i.e., where CYT testing is performed using culture supernatants instead of directly from the fecal sample (1). These are lengthy assays, however, with results delayed for 24 to 48 h for the CYT and for more than 72 h for the CYTGC assay.
Rapid, commercially available, toxin detection kits removed the need for laboratories to maintain the cell lines necessary for CYT testing. Although originally designed to detect either toxin A or toxin B, the kits currently available detect both toxins to enable detection of toxin A-negative, toxin B-positive strains. Alternative detection methods have now been developed, including an assay that detects a surface-associated enzyme of C. difficile, glutamate dehydrogenase (GDH). Zheng et al. reported that the Techlab C. diff Chek-60 GDH assay had good sensitivity compared to CYT testing of 92%, but it had a low specificity of 89.1% and poor positive predictive value (PPV) of 57.7% (21). Commercial molecular diagnostic tests, such as the BD GeneOhm C. difficile PCR assay, which detects the tcdB toxin gene of C. difficile, are now available. A recent study compared this assay to CYT testing and found a sensitivity and specificity of 90.9% and 95.2%, respectively (15). The PPVs of the BD GeneOhm C. difficile PCR assay were only 70.2% compared with CYT testing and 89.5% compared with CYTGC (15), with a prevalence of toxin-positive fecal samples of 15.2%.
Despite numerous evaluations of C. difficile testing methods, no evaluation has compared all methods on the same sample set. This study compared six commercially available enzyme immunoassays (EIAs) and three lateral-flow assays for detection of C. difficile toxins A and B, a PCR assay for detection of the tcdB gene of C. difficile, and an assay for detection of C. difficile-specific GDH, with CYT testing and CYTGC.
Ethical approval for this study was granted from the North Sheffield Research Ethics committee. Part of this study was completed on behalf of the Centre of Evaluations and Purchasing, part of the Purchasing and Supply Agency of the United Kingdom National Health Service (NHS) (18).
Ten fecal samples were selected daily for inclusion in the study, between April and September 2008, from those received by the routine enteric laboratory at the Microbiology Department of Leeds Teaching Hospitals NHS Trust. All samples included were diarrheal (adopting the shape of the container), <48 h old, submitted for CYT testing, had been stored at 2 to 5°C, and had sufficient fecal material to allow testing with all the assays (4). The 10 daily samples were randomly chosen from each daily cohort of samples (~20 to 25) that had sufficient volume of fecal matter to permit multiple testing by the biomedical scientist working in the enteric laboratory. Each sample was dispensed into a new container labeled with only a study number, with no link to patient details, before the repackaged samples were distributed to the investigators. To increase the statistical robustness of the study, the number of positive samples was increased by including samples from patients with a previous C. difficile-positive fecal sample (n = 2) and samples identified as toxin positive at another hospital (n = 13). Specimens were stored at 4°C for 1 week (in case retesting was required) and then frozen at −20°C.
After identifying all commercially available C. difficile toxin detection kits in the United Kingdom, manufacturers were approached and asked to supply the kits and supplementary equipment needed for the study. No manufacturer approached declined to be included in the study (Table (Table1.).1.). The Vidas C. difficile toxin A and B assay was performed on the mini Vidas, provided by Biomérieux, and the Premier toxin A + B assay was performed on the DS2 instrument, provided by Launch Diagnostics. An automated washer (Wellwash; Labsystems) was used to complete the wash steps in the four manual toxin detection EIAs, Techlab Toxin A/B II, Remel ProSpecT, Ridascreen toxin A/B, and GA Clostridium difficile antigen, and the manual GDH EIA Techlab C. diff Chek-60 kit. The final optical density (OD) values for these five assays were read on a microplate reader (230s; Organon Teknika) at the wavelength specified for the assay. The PCR assay GeneOhm was performed on the Smartcycler (Cephid, United Kingdom; supplied by BD at the time of the study). All assays were performed according to the manufacturers’ operational instructions, although use of freeze-thawed samples for the PCR and GDH assays is not recommended.
Every sample was tested using each toxin detection assay and by C. difficile culture. All C. difficile toxin detection assays were performed on the same day for each batch of 10 samples. A prestudy evaluation elucidated the optimum order in which to perform the assays. Selected samples were tested using the PCR (n = 554) and GDH (n = 558) assays, due to lack of sample volume in a minority of cases. The PCR and GDH assays were performed together at a later date on batched, previously frozen fecal samples (stored for <8 months at −20°C) that had never been defrosted before this occasion.
Fecal samples were diluted 1:5 in phosphate-buffered saline and centrifuged before 20 μl of supernatant was added to duplicate Vero cell monolayers, one of which had been protected by the addition of 20 μl Clostridium sordelli antitoxin (Prolab Diagnostics, United Kingdom). Vero cells were grown in 96-well flat-bottomed microtiter trays in 160 μl of Dulbecco medium. Samples were filtered before testing if the supernatant was cloudy. A positive result was recorded if cell rounding was seen, only in the unprotected cells, after 24 or 48 h of incubation in a 37°C CO2 incubator.
All samples were cultured, following alcohol shock in 50:50 [vol/vol] absolute ethanol and water, on CCEYL (Braziers CCEY agar base; Bioconnections, Wetherby, United Kingdom), supplemented with 8 mg/liter cefoxitin and 250 mg/liter cycloserine, 2% lysed horse blood (E & O Laboratories, United Kingdom), and 5 mg/liter lysozyme (Sigma, United Kingdom). Egg yolk was not added as a supplement. Plates were incubated in a MK3 anaerobic workstation (Don Whitley, Shipley, United Kingdom) and inspected for growth after 48 h. Gray-brown colonies growing on CCEYL with an irregular edge and a characteristic horse manure odor were identified as C. difficile. The Microgen C. difficile latex agglutination kit (Microgen Bioproducts Ltd., Camberley, United Kingdom) was used for confirmation of C. difficile identity where there was doubt about an isolate.
Spores were harvested into 10% glycerol broth from isolates subcultured onto fresh blood agar (FBA [E & O Laboratories, United Kingdom]), incubated in the anaerobic cabinet at 35°C for 7 days, and stored at −70°C. All isolates were typed by PCR ribotyping following the protocol from the Clostridium difficile Ribotyping Network for England (CDRNE) laboratory which is situated in the same department.
C. difficile isolates that grew on CCEYL from CYT-negative samples were tested for the production of toxin. Isolates were inoculated into brain heart infusion broth (BHI) (Oxoid, Basingstoke, United Kingdom) and incubated for 48 h in an anaerobic cabinet. Culture supernatants were centrifuged before adding to duplicate Vero cell monolayers, one protected by C. sordelli antitoxin. A positive result was recorded if cell rounding was seen only in the unprotected cells after 24 or 48 h of incubation in a 37°C CO2 incubator.
Optical density readings were recorded for the manual EIAs and compared with the cutoff value for the assay to determine the result. Three assays used a fixed cutoff value; any sample with an OD above this cutoff was positive. Results were recorded as equivocal if the OD was equal to the cutoff value. For the remaining two EIAs, Ridascreen toxin A/B and GA Clostridium difficile antigen, a correction factor set by the manufacturer had to be added to the OD value for the negative control each time the assay was performed to calculate the cutoff. Any value higher than 1.1 × cutoff was recorded as positive; below 0.9 × cutoff was recorded as negative. Any value between 0.9 and 1.1 × cutoff was recorded as equivocal. The lateral-flow assays were read visually by three evaluators to ensure an objective result. Results were recorded as positive (line or color change), negative (no line or color change), or equivocal (if unclear). When the operators could not agree, the majority result was recorded. The Vidas and DS2 instruments used algorithms set by the manufacturer to calculate the result and were recorded as positive, negative, or equivocal. The software on the Smartcycler (Cepheid, United Kingdom) recorded the results of the PCR assay as positive, negative, or unresolved.
Samples with equivocal results from the toxin detection kits were rediluted and retested; if a result was still equivocal, it was recorded as such. Specimens that gave discordant results when tested by the assays under evaluation were retested in duplicate on the same specimen, where sufficient specimen was available, to exclude the possibility of technical error. The BD GeneOhm C. difficile PCR assay was repeated when the result was unresolved, as recommended by the manufacturer. None of the samples were repeated on the Techlab C. diff Chek-60 GDH kit (due to insufficient sample).
Sensitivity and specificity were calculated for each kit against both gold standard assays (CYT and CYTGC). The difference in both sensitivity and specificity between each pair of toxin detection assays was determined using McNemar's test for paired proportions, with exact binomial P values, due to the potentially small numbers of discordant samples. The sensitivity and specificity data were used to calculate the PPV and negative predictive value (NPV) for different prevalence rates of C. difficile toxin-positive fecal samples to reflect the prevalences seen in the community and hospital settings. No statistical comparisons were performed between the toxin detection assays and the PCR and GDH assays, as these measure different targets.
Six hundred fecal samples were included in the evaluation, of which 108 were positive by CYT and 125 were positive by CYTGC. All 600 samples were tested by the toxin detection kits, but due to insufficient sample volume, only 558 and 564 samples were tested using the BD GeneOhm PCR and Techlab C. diff Chek-60 GDH assays, respectively. Four samples were removed from the CYT gold standard analysis, as these samples were considered to be false-positive CYT results and may have skewed the data. Sensitivity and specificity data against the CYT gold standard are shown in Table Table2,2, and sensitivity and specificity data against CYTGC are shown in Table Table3.3. The CYT was also evaluated in comparison with CYTGC (Table (Table33).
Statistical analysis showed that the Premier toxin A+B, Vidas C. difficile Tox A/B, Techlab toxin A/B II, and Remel ProSpecT assays were more sensitive (P = >0.05) than the other five toxin detection assays (Tables (Tables22 and and3;3; statistical data not shown). The Ridascreen toxin A/B assay was the least sensitive assay compared against either gold standard method, while the GA Clostridium difficile antigen assay was the least specific (Tables (Tables22 and and3).3). The two lateral-flow assays demonstrated better specificity than any of the EIAs (Tables (Tables22 and and3),3), although the differences were not significant (statistical data not shown). The cytotoxin assay had higher sensitivity and specificity values than any of the commercial toxin detection assays, except the Remel Xpect, when compared with CYTGC (Tables (Tables22 and and33).
The BD GeneOhm PCR assay was more sensitive than any of the toxin detection assays in comparison with both gold standard methods (Tables (Tables22 and and3).3). The assay did not perform as well for specificity, however, being in the middle of the range of specificity values seen with the toxin detection kits (Tables (Tables22 and and33).
The sensitivity of the Techlab GDH assay was similar to that of the Techlab toxin detection assay (which had the highest sensitivity of the EIAs), when using the CYT gold standard, but it is less specific (Table (Table2).2). When assessed against CYTGC, the GDH assay is more sensitive than the Techlab toxin detection assay, but it remains less specific (Table (Table33).
Table Table44 shows the changing PPV values for each assay tested in both the community and the hospital setting (2% and 10% prevalence of C. difficile-positive fecal samples, respectively). When CYT is used as the gold standard, the two lateral-flow assays give the highest PPVs across the range of prevalence values. It should be noted, however, that at a prevalence of 2%, the highest PPV is still only 56.3%, rising to 87.5% at 10% prevalence. When CYTGC is used as the gold standard, the CYT and the Remel Xpect lateral-flow assay give the highest PPVs, 92.0% and 92.4% for 10% prevalence, respectively.
The NPVs for all the assays were much higher than the PPVs, ranging from 99.3 to 99.8% at 2% prevalence to 96.3 to 99.1% at 10% prevalence, against the CYT gold standard (Table (Table44).
All nine of the commercial toxin detection assays tested gave discordant results against both gold standards (n = 23 to 69 against CYT and n = 29 to 65 against CYTGC). In the majority of cases, these results were unaffected by repeat testing (mean of 68.6% and range of 48.9 to 93.3% against CYT; mean of 66.5% and range of 28.1 to 95.0% against CYTGC). There were 30 equivocal results generated by the commercial toxin detection assays, of which 19 (63%) remained equivocal on repeat testing.
Initial and repeat unresolved rates were 1.1% (6/554) and 0%, respectively. Five of the six samples that gave an “unresolved” result on first testing gave a negative result on repeat testing, with one sample giving a positive result. These results matched the CYTGC results for these samples. Of the 16 samples that gave a false-positive result against CYTGC, two could not be repeated due to lack of sample, five were repeated and gave a negative result, and 10 were repeated and gave a positive result. Two of the 10 repeat false-positive results were positive by CYT, but these results had been regarded as false CYT positives in the original analysis, as no other test, including culture, was positive for these samples.
The OD values for each sample, in each of the commercial toxin detection assays, were recorded and initially plotted in scatter grams to show the distributions of false-positive and false-negative results around the respective assay cutoff values (data not shown). The results demonstrated that the samples that yielded false-positive and false-negative results varied markedly between assays. Thus, it was not a core group of samples that repeatedly yielded false-positive results. The data also show that the false-positive results were not solely associated with samples that yielded low OD results in some assays.
There were 128 culture-positive samples, 125 of which were CYT positive. The most common ribotypes identified during this evaluation were 106 (26.6%), 027 (18.8%), and 002 (6.3%). The total numbers of each ribotype were 001 (n = 4), 002 (n = 8), 003 (n = 1), 005 (n = 7), 014 (n = 1), 014/20 (n = 5), 015 (n = 7), 018 (n = 3), 023 (n = 3), 027 (n = 24), 044 (n = 1), 050 (n = 2), 054 (n = 2), 070 (n = 1), 078 (n = 4), 084 (n = 1), 094 (n = 1), 106 (n = 34), 118 (n = 4), 140 (n = 1), and sporadic types (n = 14). There did not appear to be any differences between the assays at identifying positive results in different ribotypes, but the numbers were too small to calculate statistical significance (data not shown).
There have been many reports on the performance of C. difficile (toxin, GDH, or toxin gene) detection kits, but these invariably involve a very small number of comparator methods (11). Planche et al. recently carried out a systematic review of such studies and concluded that single assays had unacceptably low PPVs (11). Our evaluation, which we understand is the largest study of C. difficile detection methods performed up to this point, supports the findings of Planche et al. Thus, combining the results from multiple small studies provides a very similar outcome to our evaluation. NPVs are generally much higher in both our evaluation and the recent systematic review, as would be expected when using routine diagnostic samples, the great majority of which will yield negative results for C. difficile, despite the presence of diarrhea. Importantly, the PPV of any test is affected by the prevalence of the disease in the population (i.e., by the test positivity rate). In most of the published comparison studies, the prevalence of toxigenic C. difficile-positive samples is higher than routinely seen in the clinical setting (9, 14, 15), leading to falsely inflated PPVs (11). Although our study had a prevalence of toxigenic C. difficile-positive samples of 20%, we used our data to calculate the PPV and NPV with changing levels of prevalence. In the hospital setting, with an expected prevalence of 10%, the mean PPV of a commercial toxin detection kit was 68.7% (range, 48.6 to 86.8%), which is comparable to the findings of Planche et al. (11). In the community setting, the prevalence is nearer 2% (19); as such, the mean PPV falls to 32.3% (range, 14.8 to 56.3%), making testing for C. difficile using current methods, especially single tests, extremely unreliable. The scatter grams of the OD values for the EIA toxin detection assays indicate that the false-positive and false-negative results were generally not seen in the same sample tested by different assays (data not shown). This implies that in general the incorrect results were due to the inaccuracy of the assay and not due to the sample or the evaluator. Although only one evaluator processed the samples on each day, two different evaluators worked on the project, and the number of discordant results was similar between them (data not shown).
As a result of increased awareness of C. difficile infection, including the epidemic spread of virulent C. difficile clones, additional testing has likely occurred. A recent United Kingdom survey showed that testing for C. difficile has increased by 39% in the microbiology laboratories of 18 hospitals; the median number of tests increased from 3,613 in 2005 and 2006 to 5,020 in 2007 and 2008 (8). Consequently, if the prevalence of true-positive samples tested decreases, this potentially increases the chance of obtaining a false-positive result using methods that have suboptimal accuracy. Furthermore, many laboratories have abandoned CYT testing in favor of the newer rapid toxin detection assays. As the reference laboratory for the CDRNE, we noted a high level of toxin detection assay-positive samples being sent for ribotyping that were culture negative for C. difficile. Conversely, local data show that we successfully isolate C. difficile from >95% of CYT-positive samples. We therefore became concerned about the apparent level of false-positive results for toxin detection kits, which stimulated our interest in conducting this study.
It should be noted that the clinical criteria used in definitions of C. difficile infection vary, notably, the frequency of diarrhea that is needed to satisfy a positive case diagnosis. Interpretation of clinical symptoms will clearly be made more difficult by inaccurate laboratory results, thus potentially affecting patient management. A false-positive result may lead to unnecessary treatment and isolation. The true cause of the patient's diarrhea may also not be further investigated if a diagnosis of C. difficile infection is made. In hospitals where C. difficile infection patients are isolated together within one ward, due to insufficient availability of single-room isolation facilities, a false-positive result could lead to a patient being at increased risk of cross-infection from patients who have true C. difficile infections. Conversely, false-negative results may lead to cross-infection to other patients and overtreatment with empirical antibiotics. This has implications for patients not receiving appropriate treatment and for the hospital.
It is important to note that there is not a universally accepted method for the CYT. Sample selection and preprocessing, choice of cell line, and different interpretive end points may all affect test performance. Such issues, coupled with the long waiting time for a test result and the need to maintain a cell line, have contributed to the decreased availability of the CYT in diagnostic laboratories. In this evaluation, the CYT was the best performing toxin detection assay, compared with CYTGC, with a sensitivity and specificity of 86.4 and 99.2%, respectively. The reported performance of the CYT has, however, been variable. Peterson et al. reported poor sensitivity of the CYT (76.7%) using a gold standard of at least two positive tests plus the presence of diarrhea; such discordant analysis may affect the calculated sensitivity of the CYT (10). Stamper et al. also reported a low sensitivity of 67.2% for CYT (Wampole tox-B assay; Techlab) in comparison with CYTGC (15). Higher sensitivity and specificity for the CYT (98% and 99%, respectively) have also been reported (9). Interpretation of such published results is clearly difficult, however, given the difficulty of assessing the accuracy of a gold standard method. For example, comparison of results of a CYT with CYTGC may be misleading, as although these tests both measure toxin, the former samples the toxin present in vivo, while the latter involves in vitro production.
The BD GeneOhm PCR assay was more specific than the toxin detection kits, as would be expected for a molecular detection method. The sensitivity, specificity, PPV, and NPV found in this evaluation are comparable to those found by Stamper et al. when comparing the GeneOhm PCR assay to both CYT testing and CYTGC (15). The PCR assay lacks specificity, however, leading to low PPVs, comparable with the toxin detection assays. The PCR assay has potential as a negative screening assay, as it has the highest NPV (99.1%) of any assay in this evaluation (at 10% prevalence, versus CYT). Other PCR assays have been developed and are becoming commercially available. Sloan et al. used a PCR assay to detect the tcdC gene of C. difficile and found sensitivity and specificity values of 86% and 97%, respectively, which are comparable with those in our study (14). The reported PPV of 90% is much higher than we found for the BD assay, but their study had a prevalence of toxin-positive stool of 22%, approximately double that found in hospital clinical samples (14). Any PCR assay can detect only the presence of the toxin gene, however, not the presence of toxin, and the relevance of a positive result has yet to be elucidated. Peterson et al. evaluated a real-time PCR assay for tcdB detection in diarrheal samples and found good correlation with the clinical status of the patients with a PPV of 75.7% and NPV of 99.4% when the prevalence of toxin-positive samples was 13.1% (10). The clinical details of the patients in our study were unavailable, as the samples were placed into new containers with no link to patient details before testing was performed. It was therefore not within the scope of this evaluation to investigate the diagnosis of those patients in whom only the PCR assay was positive. Isolates in this evaluation were typed using PCR ribotyping. The ribotypes seen and their proportions are representative of the local epidemiology. There did not appear to be any difference in the abilities of the assays to detect different ribotypes of C. difficile.
The GDH detection assay (Techlab C. diff Chek-60) also has potential as a negative screening tool with an NPV of 98.8% (at 10% prevalence, versus CYT). An assay PPV of 63.1% (at 10% prevalence, versus CYTGC) was comparable with those of the toxin detection assays and echoes the results of Zheng et al. (21). The GDH assay did not perform as well compared to the CYT. Clearly, the assays are detecting different targets; one detects an enzyme of C. difficile (GDH), and the other detects toxins produced by the organism. Although GDH may be useful in establishing the presence of C. difficile in a fecal sample, it does not indicate whether the organism has the potential to cause disease, as it cannot detect the presence of toxin. This assay can therefore be used only as part of a two-step testing algorithm. Fenner et al. showed that a rapid two-step algorithm using GDH followed by toxin testing could produce a result for 92% of samples in 4 h (3). However, GDH assays are reported to have variable sensitivities. In comparison with CYTGC, the Techlab GDH test had a sensitivity of 88% in our hands (using freeze-thawed samples). Using a different GDH assay (Triage [no longer available]), Sloan et al. reported a sensitivity of only 76% for C. difficile culture-positive (fresh or freeze-thawed) fecal samples (14).
Two- or three-step approaches to the diagnosis of C. difficile infection, e.g., rapid detection of toxin, bacterium, or toxin gene, with subsequent confirmation of the presence of toxin, will increase laboratory costs, but these might be offset by reduced total health care costs for C. difficile infection. A two-step approach will also increase the time for a final result, but this may be acceptable, particularly if interim results are made available. The algorithms tested so far include GDH testing as the first, negative screening step, due to its high NPV. In the algorithm used by Wren et al., the GDH assay performed well as the initial screen, with only four false-negative results (20). The authors hypothesize that these are due to colonization with C. difficile, rather than C. difficile infection, as the lactoferrin test (a marker of inflammatory diarrhea) gave negative results in these samples (20). There has been no consensus on the best assay for the second step, with studies using EIA toxin assays, cell culture CYTs, and toxin detection plus lactoferrin detection (3, 5, 12, 20). Algorithms including a toxin detection EIA as the second, confirmatory test, may not be optimal, as it relies on an assay which we have shown to have poor PPV. Published studies have compared one algorithm to the current methods of single assay diagnosis. There has been no evaluation of different combinations of assays, and further work is required to assess which assays would provide the optimal two-step laboratory diagnosis of C. difficile infection.
Planche et al. indicated that many of the published comparative studies made modifications to the manufacturer's instructions and used various methods for the cell CYT which makes comparison between studies difficult (11). We ensured that all assays in this evaluation were performed according to manufacturers’ instructions and compared all assays to one method of CYT. We have conducted the first study to compare all the currently available (in the United Kingdom) C. difficile detection kits on the same sample set. Previous comparison studies have used smaller sample sets and high prevalence rates of C. difficile-positive samples, not reflective of the clinical setting (9, 14, 15). All kits had low PPVs compared with CYT, assuming a positive sample prevalence of 10% (representing the hospital setting), which compromises the clinical utility of single tests for the laboratory diagnosis of C. difficile infection. As the prevalence of C. difficile infection decreases, this will exacerbate the issue of false-positive results in suboptimal assays. The optimum rapid single test was PCR for toxin B gene, as this had the highest negative predictive value. Diagnostic algorithms that optimize test combinations for the laboratory diagnosis of C. difficile infection need to be defined.
We thank all of the suppliers involved for supplying kits and equipment for the evaluation. Part of this evaluation was funded by the Centre of Evaluations and Purchasing, Purchasing and Supply Agency, National Health Service (NHS).
We also thank Keith Perry at the Health Protection Agency's Microbiological Diagnostics Assessment Service, London, United Kingdom, for invaluable advice.
Published ahead of print on 26 August 2009.