The utility of repeat enzyme immunoassay testing for the diagnosis of Clostridium difficile infection: A systematic review of the literaturePS Garimella, R Agarwal, A Katz
Department of Internal Medicine, John H. Stroger, Jr. Hospital of Cook County, 1900 W. Polk St, Chicago, IL 60607, USA
Over the last 20 years, the prevalence of healthcare-associated Clostridium difficile (C. diff) disease has increased. While multiple tests are available for the diagnosis of C. diff infection, enzyme immunoassay (EIA) testing for toxin is the most used. Repeat EIA testing, although of limited utility, is common in medical practice. To assess the utility of repeat EIA testing to diagnose C. diff infections. Systematic literature review. Eligible studies performed >1 EIA test for C. diff toxin and were published in English. Electronic searches of MEDLINE and EMBASE were performed and bibliographies of review articles and conference abstracts were hand searched. Of 805 citations identified, 32 were reviewed in detail and nine were included in the final review. All studies except one were retrospective chart reviews. Seven studies had data on number of participants (32,526), and the overall reporting of test setting and patient characteristics was poor. The prevalence of C. diff infection ranged from 9.1% to 18.5%. The yield of the first EIA test ranged from 8.4% to 16.6%, dropping to 1.5-4.7% with a second test. The utility of repeat testing was evident in outbreak settings, where the yield of repeat testing was 5%. Repeat C. diff testing for hospitalized patients has low clinical utility and may be considered in outbreak settings or when the pre-test probability of disease is high. Future studies should aim to identify patients with a likelihood of disease and determine the utility of repeat testing compared with empiric treatment.
Keywords: Clostridium difficile , enzyme-linked immunosorbent assay, healthcare-associated diarrhea, meta-analysis
Over the last 20 years, the prevalence and virulence of healthcare-associated Clostridium difficile (C. diff) disease has increased,  and it is the most common cause of infectious diarrhea in the healthcare setting.  While there are multiple tests commercially available for the diagnosis of C. diff infection, enzyme immunoassay (EIA) testing for toxins A and B is most commonly used. Early studies recommended performing three EIA tests for increased sensitivity  but, with improvement in diagnostic methods, there is evidence that a two-step strategy to diagnose C. diff infections may be more useful. ,, Some reports also recommend using the most sensitive rapid test first and not repeating it for 7-14 days if initially negative. 
There have been numerous studies evaluating the accuracy and validity of different testing methods, including one large systematic review that compared different testing strategies.  Studies evaluating the utility of repeat EIA testing in diagnosis have been small, with relatively modest effect. Prior studies that evaluated the utility of repeat EIA testing for C. diff have not identified whether there is a specific population that is at increased risk of disease or a special setting in which repeat testing is beneficial. Although the Infectious Disease Society of America (IDSA) does not recommend repeat EIA testing during the same diarrheal episode,  repeat testing is common in medical practice. We performed this systematic review to synthesize the existing evidence on the utility of repeat EIA testing for diagnosing C. diff infections and to describe patients and settings in which there may be a benefit in repeat EIA testing.
We searched the MEDLINE and EMBASE databases (1950 to February 2011) using various combinations for the following terms: "clostridium difficile" OR "Enterocolitis" OR "Pseudomembranous" and "Enzyme-Linked Immunosorbent Assay" OR "Cytotoxicity Tests." The detailed search strategy is shown in [Table 1]. In addition, bibliographies of retrieved trials, review articles and conference abstracts were examined.
Studies were eligible for inclusion if they (i) performed >1 EIA test for C. diff toxin with or without comparison with any other diagnostic modality, (ii) reported results for either number of samples or patients tested and (iii) were published in English. For each study, two reviewers (PSG and RA) independently assessed study eligibility. Differences in eligibility assessments were minimal and were resolved by discussion.
For each study, information on study characteristics, patient demographics, inclusion and exclusion criteria, patient risk factors, number of testing episodes and results of testing were extracted by two independent reviewers (PSG and RA) in a standardized fashion. The primary outcome of this review was to evaluate if repeat EIA testing after an initial negative result significantly increased the yield of a positive result. Secondary outcomes included (i) identifying and describing patients or circumstances in which repeat EIA testing was likely to be beneficial and (ii) costs associated with repeat EIA testing.
For assessing the utility of repeat EIA testing, we determined the proportion of subjects/samples that tested positive for toxin on repeat testing overall and within specific time frames where possible. Because of the limitations and heterogeneity of the reported studies, we were unable to perform a metaanalysis or any further subgroup analysis of the data. In this report, we described the overall yield and predictive value of repeat testing for C. diff toxin.
Eight hundred and five citations were identified through an electronic literature search up until February 2010 [Table 1]. After review of all titles and abstracts from both searches, 30 articles were retrieved for a detailed review. In addition, one study was identified by reviewing conference abstracts. After full text review of these 31 trials, nine were included in the final review [Figure 1].
Characteristics of trials
Nine studies ,,,,,,,, that evaluated the utility of repeat EIA testing in the diagnosis of C. diff infections were included in the review [Table 2]. Seven studies ,,,,,, reported number of participants (n=32,526; range 268-17,921). Eight were published in peer-reviewed English language journals ,,,,,,, and one was a conference abstract.  All studies except one were retrospective, the data for which were obtained from reviewing hospital charts or laboratory records of patients who had at least one EIA for C. diff performed. One study was a prospective study of patients admitted to a hospital who had an episode of diarrhea.  Three studies reported the utility of repeat testing within a specific testing period: 7 days,  10 days  and 5 days.  All studies were performed in the US except for one that was performed in The Netherlands.  Only one study was performed during an outbreak of C. diff.  Three studies reported using the Premier Toxins A & B EIA kit (Meridian Bioscience, Cincinnati, OH, USA), ,, one used the ICTAB assay (Meridian Bioscience, Halderheiweg, The Netherlands),  one used the Wampole Tox A/B II EIA kit (Techlab, Blacksburg, VA, USA),  one used the Cytoclone A+B assay (Cambridge Biotech, Worscester, MA, USA Worscester, MA, USA),  while the remaining studies did not report the type of assay used. ,, Five studies ,,,, reported data on the validity (sensitivity, specificity) of the assay used. The sensitivity ranged from 80% to 95% and the specificity from 80% to 100% among these trials. Only one study reported data on costs associated with repeat EIA testing.  C. diff infection was diagnosed by EIA testing in all studies, and confirmatory testing was used in one study during the outbreak setting. 
Six studies reported data on inpatients while two did not report the setting (outpatient versus inpatient). , Three studies reported that all tests were performed in adults. ,, In three trials, the mean age ranged from 53 to 67 years. ,,, Most studies did not provide additional data about the patients who were tested. One study reported that most patients tested were from the medicine service as compared with the surgical services, and also provided broad diagnostic categories for all patients tested.  One study reported results of a subgroup analysis performed on patients from its bone marrow transplant unit, who were immunosuppressed,  while another provided data on percentage of patients who previously had a documented C. diff infection. 
Yield of repeat testing
The results for EIA test outcomes among the included trials are shown in [Table 3]. The overall proportion of the population that tested positive for C. diff by EIA ranged from 9.1% to 18.5%. In a retrospective review of 29,373 episodes of diarrhea among 17,971 patients, 91% of all positive results were obtained with one test itself,  and the probability of a second and third test being positive after an initial negative test was <2.5%. In the study by Nemat and colleagues, nearly 90% (618/689) of all cases were diagnosed with one EIA test, with the yield decreasing to 2% (15/507) with the third test.  In one study of patients who tested positive for C. diff by EIA, 83.6% were from one EIA and 11.9% were detected by a second subsequent test.  In the study by Aichinger et al., 7.7% of the patients tested positive by EIA, of which 7.4% were in the first test alone, 2.2% with two tests and 0.2% with three tests.  Cardona et al. analyzed the number of samples retested within 10 days of an initial negative result ("testing episodes"), and found that 2.6% (96/3749) of the overall samples tested positive after an initial negative result.  They also noted that the percent conversion to positive was approximately 3-4% in the first 3 days after the initial negative result, increasing up to 10.6% when repeated 7-10 days after the initial negative result. These results were similar when they performed the analysis in patients from their bone marrow unit.  Among 237 patients who tested positive during an outbreak of C. diff, 86% (204/237) were diagnosed on the first stool sample and 5% (12/237) on the second or third stool sample within 1 week of an initial negative result.  In this same study, 10% (20/202) of the samples that tested positive by EIA were negative on culture for C. diff. Drees and colleagues reported a 12.2% (189/1547) infection rate among patients tested with EIA.  Of the 2938 tested sampled, 7.3% (217/2938) were positive and 6.2% (184/2938) were from the first EIA itself. The remaining 1% tested positive with two more tests, of which half were later than 7 days from the initial negative result.  In a small study of 474 EIA tests for C. diff, Mohan et al. reported that only one of 78 repeat samples turned positive after an initial negative result.  The available data did not permit comparison of the yield between the types of assay used.
Cost of repeat EIA testing
Testing costs were not reported in most studies, and varied in the ones reported. While Gade and colleagues reported $85 per EIA test,  in the study by Mohan, the investigators reported the cost of a single EIA as $128, and reported that nearly $10,000 were spent on repeat testing for the one additional positive result over a 10-month period. 
To the best of our knowledge, this is the first systematic review examining the utility of repeat EIA testing to diagnose C. diff infections. Our analyses suggest that such testing has limited utility in hospitalized patients with diarrhea in the absence of additional risk factors or an outbreak. Using exactly one test is adequate to rule out disease in symptomatic hospitalized patients, where the pre-test probability ranges from 9% to 12%. Repeat testing should only be considered in outbreak settings or when the pre-test prevalence of disease is very high. The consistently low diagnostic yield of repeat testing across the studies identified adds to our confidence that the findings are real.
In the study by Manabe and colleagues in 1995, the sensitivity of a singly EIA was reported at 72%, increasing to 84% with a second test.  Interestingly, in 13% (3/39) patients who tested positive after two initial negative EIA tests, the third sample was collected 4 days after the preceding negative test. It may be possible that this represented a new episode of diarrhea in these patients, in which case the sensitivity of a single test may in fact be higher than that reported. Peterson and colleagues in a review addressing diagnostic testing for C. diff infection highlight the fact that repeat testing after a negative test decreases the positive predictive value (PPV), resulting in a PPV of 0.48 after three EIA tests.  They also hypothesize that continued negative tests may in fact be false negatives, resulting from an undetected substance in the stool, although there have been no studies to support this. Based on this hypothesis, they suggest using a different testing modality if repeat testing is necessary within 7-14 days after an initial negative EIA.  Newer testing modalities such as glutamate dehydrogenase (GDH) antigen and the quantitative real-time polymerase chain reaction (qPCR) tests have been studied for diagnosis of C. diff infection. Initial studies of the GDH reported much higher sensitivities (>95%) compared with the EIA test , ; however, further studies have shown conflicting data with a lower sensitivity of 75%.  A recent meta-analysis of studies evaluating GDH in the diagnoses of C. diff-associated diarrhea reported high sensitivity and specificity (>90%) when compared with culture, which was the gold standard.  However, because the GDH test detects both toxigenic and non-toxigenic strains of C. diff, the specificity for toxigenic strain is lower, with a false-positive rate of approximately 20%. In recent times, the qPCR too has shown a higher sensitivity, ranging from 84% to 95% in comparison with the stool EIA, cytotoxin culture, but the specificities of these to qPCR were comparable. ,, This is of clinical significance as it may direct the need for therapy by differentiating asymptomatic colonization from C. diff infection. However, this does not take away the importance of rational clinical decision - the clinical scenario and need for the test should guide the choice of test and decisions on therapy.
The current review is limited by the available evidence. The studies that we included, although the best available, were of relatively poor quality, precluding a formal quality assessment. Statistical pooling was not possible given the heterogeneity of study designs and data reporting. None of the studies compared repeat testing to the better diagnostic tests such as culture for toxigenic C. diff or cell culture cytotoxicity assays. Patient characteristics and predictors of increased risk could not be accurately ascertained based on data provided. Only one study reported additional analysis on immunosuppressed patients  and another limited data on patient diagnosis.  A formal cost-effectiveness analysis was beyond the scope of this review as none of the studies reported data. The cost of performing a single EIA test could range from $5 to $17, depending on the manufacturer. , Using these estimates and an example from our study, to detect an additional 2.3% cases on repeat testing, the excess cost could range from $40,330 to $137,122 annually.  However, additional studies will be needed to evaluate whether the expense of empiric antibiotic therapy and isolation precautions will be offset by the savings on repeat EIA testing.
Our study however represents the first attempt to systematically evaluate the evidence supporting the practice of repeat EIA testing to diagnose hospitalized patients. Our findings are in concordance with the recent IDSA statement recommending against the repeat testing in a single episode of diarrhea. Further studies need to be conducted to evaluate the utility of repeat EIA in specific populations at high risk of C. diff infection, e.g. patients with prior antibiotic use, patients in the Intensive Care Unit, immunosuppressed patients or those with previously documented C. diff infections. The choice of testing and interpretation of results must be individualized, and depend on the clinical scenario at hand.
Dr. David Schwartz, Chairman, Division of Infectious Diseases, is greatly acknowledged for his review and comments on earlier drafts of this article.
The data in this manuscript has not been presented or published before.
[Table 1], [Table 2], [Table 3]