- Email: [email protected]
Comparison of culture, cytotoxicity assays, and enzyme-linked immunosorbent assay for toxin A and toxin B in the diagnosis of Clostridium difficile-related enteric disease
DIAGNMICROBIOLINFECTDIS 1986;5:61-69
61
Comparison of Culture, Cytotoxicity Assays, and Enzyme-Linked Immunosorbent Assay for Toxin A and Toxin B in the Diagnosis of Clostridium difficile-Related Enteric Disease Randall C. Walker, Peter J. Ruane, Jon E. Rosenblatt, David M. Lyerly, Curt A. Gleaves, Thomas F. Smith, Phillip F. Pierce Jr., and Tracy D. Wilkins
Clostridium difficile culture, test tube, and micretiter cytotoxicity assays, and enzyme-linked immunosorbent assays (ELISAs) for toxin A and toxin B, were simultaneously performed on 113 fresh diarrheal steal specimens randomly selected from those submitted to our clinical laboratory for routine C. difficile testing. The performance of these tests in diagnosing C. difficile-related enteric disease (CDBED) was based on a clinical assessment of the likelihood of CDRED as determined by a systematic review of case histories blinded from the test results. Among 61 antibiotic recipients, both the microtiter cytotoxicity assay and the toxin A ELISA were highly specific for CDRED (95% and 100%, respectively). Specificities for the ether procedures were much lower (tube cytotoxicity assay, 79%; culture, 74%; and toxin B ELISA, 56%). The high sensitivities of the culture (89%) and toxin B ELISA (83%) were somewhat negated by their low specificities. The only test that was both specific and had acceptable sensitivity (78%) was the microtiter cytotoxicity assay. This study indicates that ELISAs for detection of C. difffieile toxins are net as reliable as the cytotoxicity assay in the laboratory diagnosis of CDRED, and that clinical correlation is essential in the evaluation of any new test for CDRED. INTRODUCTION The e x a m i n a t i o n of fecal specimens from large n u m b e r s of patients suspected of having Clostridium difficile-related enteric disease (CDRED) has become a major task for m a n y clinical microbiology laboratories. There is a need for a rapid, simple, and efficient test for C. difficile that might replace culture of stool specimens and tissue culture cytotoxicity assay. Clostridium difficile produces two toxins: toxin A, which is primarily an enterotoxin (Taylor et al., 1981} and toxin B, w h i c h is 1000 times more cytotoxic than toxin A and is responsible for the cytotoxicity observed in tissue culture assays
From the Divisions of Clinical Microbiology and Infectious Diseases, Mayo Clinic and Mayo Foundation Rochester, MN and Department of Anaerobic Microbiology, Virginia Polytechnic Institute and State University, Blacksburg, VA. Address reprint requests to: Jan E. Rosenblatt, M.D., Divisions of Clinical Microbiology and Infectious Diseases, Mayo Clinic and Mayo Foundation, Rochester, MN 55905. Received June 18, 1985; revised and accepted February 20, 1986. © 1986 Elsevier Science PublishingCo., Inc. 52 VanderbiltAvenue,New York, NY 10017
0732-8893/86/$03.50
62
R.C. Walker et al.
(Sullivan et al., 1982). The exact role of each toxin in producing CDRED has not been determined. The isolation and purification of these toxins has led to the development of enzyme-linked immunosorbent assays (ELISAs) using antisera directed at the purified toxins (Lyerly et al., 1983; Laughon et al., 1984). A problem in evaluating the clinical significance of newer tests for CDRED is the relationship between laboratory results and the clinical diagnosis (George et al., 1982). Culture and cytotoxicity assays can be falsely positive in patients who do not have enteric disease (George et al., 1982; Varki et al., 1982). While test results in patients with antibiotic-associated pseudomembranous colitis (PMC) can be used to determine its sensitivity, PMC is uncommon and C. difficile has been shown to cause enteric disease other than proctosigmoidoscopic-positive PMC (George et al., 1982; Tedesco, 1979). Therefore, clinical correlation is important in the evaluation of any new test. We performed C. difficile culture, two tissue culture cytotoxicity assays (tube and microtiter methods), and ELISAs directed against purified toxin A and toxin B on flesh diarrheal specimens randomly selected from clinical specimens submitted for routine testing. The patients' histories were systematically reviewed and classified in order to compare test results on the basis of the clinical likelihood of CDRED.
MATERIALS AND METHODS Stool Specimens During a 5-month time period, diarrheal stool samples from adult patients submitted for C. difficile testing were randomly selected for C. difficile toxin determination by ELISA and a microtiter cytotoxicity assay (MCTXA) method, in addition to routine bacteriologic culture and tube cytotoxicity assay. Besides the requirement that stools be diarrheal, there was no prior screening or selection of samples on the basis of any clinical data.
Stool Culture Upon arrival in the microbiology laboratory, culture plates containing cycloserine, cefoxitin, fructose, and egg yolk agar (CCFA, Difco Laboratories, Inc., Detroit, MI) were inoculated with stool and incubated at 36°C in an anaerobic atmosphere for 24-48 hr. The presence of suspicious colonies were confirmed by colony fluorescence and biochemical and chromatographic techniques (George et al., 1979).
Stool Filtrate Preparation for Toxin Assays The procedure used for stool filtrate preparation is that of Chang and coworkers (1979). Stool filtrates were refrigerated (4°C) 0-24 hr before processing.
Antigen and Antibody Preparations
Clostridium difficile antitoxins prepared in rabbit serum and goat serum were used for toxin neutralization in the ELISA and cytotoxicity assays, respectively. Affinitypurified antibodies against C. difficile toxin A and B were used in the ELISA for the specific detection of these toxins in stool filtrates. A culture filtrate prepared from C. difficile stock strain 10463 (ATCC) was used as the positive control in the ELISA and both cytotoxicity assays. The above materials were prepared and supplied by DM Lyerly and TD Wilkins (Department of Anaerobic Microbiology, Virginia Polytechnic Institute and State University of Blacksburg, Blacksburg, VA).
Comparison of Tests for C. difficile
63
Tube Cytotoxicity and Neutralization Assays Monolayers of MRC-5 cells (Flow Laboratories, Rockville, MD) were prepared in 16 × 125 mm culture tubes and maintained in Eagles minimal essential medium (MEM) containing 10% fetal bovine serum (FBS), penicillin, streptomycin, and gentamycin until specimen inoculation. One MRC-5 monolayer was inoculated per patient specimen with 0.2 ml of stool filtrate. An uninoculated MRC-5 monolayer (negative control) and a monolayer inoculated with the positive control filtrate was included with each batch of patient specimens. All tubes were incubated at 36°C in room air on a fiat culture rack overnight. The monolayers were read twice daily for cytopathic effect (CPE) characteristic of that produced by C. difficile toxins. If CPE was detected, a neutralization procedure was done to prove the specificity of the CPE by using a 1:100 dilution of C. difficile antitoxin. The neutralization procedure involved the incubation of equal amounts (0.5 ml) of stool filtrate with antitoxin at 36°C for 30 min in a test vial, followed by the inoculation of two monolayers with the filtrate/antitoxin combination (0.1 ml) and one monolayer with the original stool filtrate. A positive neutralization control (C. difficile toxin plus antitoxin) and a negative neutralization control (C. difficile toxin) was included with each batch of a specimen's neutralization assays. The neutralization test was considered positive when the CPE was eliminated in the presence of C. difficile antitoxin.
Microtiter Cytotoxicity and Neutralization Assay Chinese hamster ovary cells (CHO-KI cells, ATCC, Rockville, MD) were prepared in microtiter plates (Dynatech, Alexandria, VA) as previously described (Ehrich et al., 1980) with two modifications: the cell medium was Eagle's MEM containing 2% FBS, and 1 ml each of sodium bicarbonate and Tris buffer, and the microtiter wells were seeded with 5,000 CHO-K1 ceils per well. Undiluted and serial 10-fold dilutions (to 1:1000) of patient stool filtrate (0.02 ml) were added to separate microtiter wells and then incubated 18 hr at 36°C in a CO2 (5%) atmosphere. A neutralization assay for each patient specimen was included in the same microtiter plate. Cytotoxicity was determined to be positive if the monolayers showed 100% cell death.
Enzyme-Linked Immunosorbent Assay for C. difficile Toxin A and B The ELISA method for C. difficile toxin A and B detection was previously described by Lyerly et al. (1983).
Clinical Evaluation The case histories of 113 adult patients whose physicians had submitted stool samples for routine laboratory diagnosis of C. difficile-related diarrhea were systematically reviewed. The reviewers were unaware of the ELISA and MCTXA results, although stool culture and tube cytotoxicity test results were available to the patients' clinicians and, thus, were documented in the medical records. Only the first clinical episode of diarrhea in a patient and the corresponding test results were studied. Relapse or follow-up cultures were not included in this study. The 113 case summaries were evaluated according to the presence or absence of four major clinical criteria (George et al., 1982): 1) A history of antibiotic use in the 2 months before the stool specimen was obtained. 2) Clinically significant diarrhea: defined as a change in usual bowel habit with three or more loose stools per day for 2 or more days in a row.
64
R.C. Walker et al.
3) Treatment w i t h and response to accepted t h e r a p y for C. difficile enteric disease (vancomycin, metronidazole, or bacitracin). Response was defined as a reduction of 50% in the n u m b e r of stools per day w i t h i n 2 days of starting therapy. Cases that r e l a p s e d after t h e r a p y were s t o p p e d were also considered as having had a positive and specific response to therapy. 4) Absence of an alternative diagnosis to e x p l a i n the patient's diarrhea, that is, active inflammatory bowel disease, other stool pathogens, use of laxatives, recent gastrointestinal surgery, or resolving bowel obstruction. Cases were classified in terms of the l i k e l i h o o d of a clinical diagnosis of CDRED as: very likely, probable, possible, and doubtful, d e p e n d i n g on the fulfillment of these criteria. The presence or absence of p s e u d o m e m b r a n e s or abnormal sigmoidoscoptic e x a m i n a t i o n were not i n c l u d e d in the major criteria because so few patients had this e x a m i n a t i o n performed. RESULTS
Laboratory Results In the 113 s p e c i m e n s tested, the C. difficile culture was positive in 27 (24%); the tube assay for cytotoxicity was positive in 22 (19%); the MCTXA was positive in 16 (14%); the toxin A ELISA was positive in 14 (12%); and the toxin B ELISA was positive in 54 (47%) cases.
Clinical Classification of Disease Probability Using the above criteria, 14 cases were classified as very likely having C. difficilerelated disease, four were classified as probable cases, nine as possible cases, and 86 as doubtful cases (Table 1.) The 14 very likely cases met all four of the clinical criteria. The four probable cases had significant diarrhea r e s p o n d i n g to specific C. difficile therapy, but had other c o n d i t i o n s that also could have caused diarrhea. One of the probable cases had
TABLE 1. Clinical Likelihood of
Clostridium difficile-Related Enteric Disease Clinical criteria fulfilled
Clinical likelihood
No.
Antibiotic
Significant
of disease
of cases
exposure °
diarrhea b
14
+
+
+
+
4 g
+ +
+ +
+ 0
+
Very likely Probable Possible
Doubtful (total of 86)
21 11 52
Therapeutic
trialc
No other possible
cause d
+
+
-(2)
+
+ + -
+ +/-
0 / - (2) 0 / - (1) 0/-(2)
+/+/-
°Antibiotic e x p o s u r e in 2 m o n t h s before s p e c i m e n w a s collected. One probable case h a d chronic diarrhea a n d r e s p o n d e d p r o m p t l y to oral v a n c o m y c i n but h a d no record of recent antibiotic exposure. bBased on acute onset, frequency, quality, a n d d u r a t i o n of diarrhea. ~+ indicates p r o m p t response to specific t h e r a p y (e.g., oral v a n c o m y c i n ) for CDRED, 0 indicates no treatment, - indicates treatment given but no response (no. of cases). d + indicates no other possible causes for diarrhea were present.
Comparison of Tests for
C. difficile
65
chronic diarrhea for several years that responded promptly to specific C. difficile therapy, although antibiotic exposure before the onset of symptoms was not specified in the case record. All of the n i n e possible cases had recent antibiotic exposure, significant diarrhea, and no other c o n d i t i o n evident besides C. difficile enterocolitis that could have caused diarrhea. However, diarrhea in each case resolved spontaneously without specific therapy for CDRED. In 84 doubtful cases, no more than two clinical criteria were fulfilled. Another two doubtful cases fulfilled three criteria, but had persistent diarrhea despite specific therapy for CDRED. Only 34 doubtful cases had recent antibiotic exposure.
Correlation of Laboratory Results With Clinical Classification Of the 14 very likely cases of CDRED, the C. difficile culture was positive i n 12 (86%); the tube cytotoxicity assay was positive in 10 (71%); the MCTXA was positive in 12 (86%); the toxin A ELISA was positive in n i n e (64%); and the toxin B ELISA was positive in 11 (79%) (Table 2). All four of the probable cases had positive C. difficile cultures and toxin B ELISAs. Tube cytotoxicity, MCTXA, and toxin A ELISAs were each positive in two of these four cases. Among the n i n e possible cases, there was one case with a positive C. difficile culture and five cases with positive toxin B ELISAs. The other tests were negative in all n i n e cases. Among the 34 doubtful cases that had recent antibiotic exposure, the culture was positive in 10 (29%); the tube cytotoxicity assay was positive in n i n e (26%); the MCTXA was positive i n two (6%); the toxin A ELISA was positive in n o n e (0%); and the toxin B ELISA was positive in 14 (41%). None of the 52 patients who had not received antibiotics had a positive stool culture or MCTXA. One had a positive tube cytotoxicity test, whereas three had a positive toxin A and 20 had a positive toxin B ELISA.
Sensitivity and Specificity of Laboratory Tests The sensitivity of laboratory tests was based on the n u m b e r of positive results (true positives) in that clinical group of cases where tests should have been positive (true positives plus false negatives (Table 3). Based on the 18 cases with very likely or
TABLE 2. Clinical Correlation of Clostridium difficile Test Results Diagnostic test no. (%) positive Clinical likelihood of disease
No. of cases
Very likely Probable Possible Doubtful
14 4 9 86
12 4 1 10
34
10 (29)
9 (26)
2 (5.9)
113
27 (24)
22 (19)
16 (14)
Antibiotic recipients only° All cases
C. difficile culture (86) (100) (11) (12)
Tube cytotoxicity 10 2 0 10
(71) (50) (0) (12)
Microtiter cytotoxicity 12 2 0 2
(86) (50) (0) (2.3)
Toxin A ELISA 9 2 0 3
(64) (50) (0) (3.5)
Toxin B ELISA 11 4 5 34
(79) (100) (55) (40)
0 (0)
14 (41)
14 (12)
54 (47)
°Fifty-twoof 86 doubtful cases had no recorded history of antibiotic exposure in the 2 months before the date the stool specimenwas collected.
66
R.C. Walker et al.
TABLE 3. Comparison of Sensitivity and Specificity of Clostridium difficile Tests Using Clinical Likelihood of Disease as Reference Diagnostic
Clinical likelihood of diseasea
Test
Sensitivity
Specificity
Microtiter cytotoxicity C. difficile culture Tube cytotoxicity Toxin A ELISA Toxin B ELISA
78 89 67 61 83
95 74 79 100 56
aBased on 61 antibioticrecipients.Sensitivityis percentpositiveamong18 caseswith significantdiarrhea and promptresponseto specifictherapyfor CDRED(very likely and probable categories);specificityis percent negative among 43 cases with low clinical probabilityof CDRED(passible and doubtful categories). Note that patients who did not receive antimicrobialsare excluded.
probable CDRED, the sensitivities for each test were: C. difficile culture, 89%; tube cytotoxicity, 67%; MCTXA, 78%; toxin A ELISA, 61%; and toxin B ELISA, 83%. The specificity of laboratory tests was based on the n u m b e r of negative results (true negatives) in that clinical group of cases where tests should have been negative (true negatives plus false positives (Table 4). Based on all 86 cases in w h o m CDRED was doubtful and the n i n e possible cases, the specificity of each test was: C. difficile culture, 88%; tube cytotoxicity, 90%; MCTXA, 98%; toxin A ELISA, 97%; and toxin B ELISA, 60%. However, if one selects only the 34 antibiotic recipients in w h o m CDRED was doubtful (considering those who did not receive antimicrobials as inappropriate candidates for testing) and combines these with the n i n e possible cases, the specificity of each test was: culture, 74%; tube cytotoxicity, 79%; MCTXA, 95%; toxin A ELISA, 100%; and toxin B ELISA, 56%. Based on all 113 cases studied, the predictive values of positive and negative results and efficiencies, respectively, for each test were (%): culture 59.2, 97.6, and 88.4; tube cytotoxicity 54.5, 93.4, and 85.8; MCTXA 87.5, 95.8, 94.7; ELISA for toxin A 78.5, 92.9, and 91.1; and ELISA for toxin B 27.7, 94.9, and 62.8. In this population, MCTXA was the most efficient test (Table 4).
TABLE 4. Comparative Predictive Values and Efficiencies° of Diagnostic Tests for Clostridium difficile-Related Enteric Disease in Clinical Patients ° Predictive value (%)
C. difficile culture Tube cytotoxicity Microtiter cytotoxicity ELISA A ELISA B
Test positive
Test negative
Efficiency (%)
59.2 54.5 87.5 78.5
97.6 93.4 95.8 92.9 94.9
88.4 85.8 94.7 91.1 62.8
27.7
°Galen RS, GambinoSR (1975) Beyond Normality: the Predictive Value and Efficiency of Medical Diagnosis. New York: John Wiley& Sons. hAll clinical patients, n = 113.
Comparison of Tests for C.
difficile
67
Quantitative Results of Enzyme-Linked Immunosorbent Assay for Toxin A and Toxin B The ELISA assays for toxin A and toxin B were determined positive when the optical density ratio (ODR) (optical density of sample divided by optical density of control) for the sample tested was >2.00. Of 18 very likely and probable cases of CDRED, 11 were positive (flue positives) for toxin A. The range of ODR values in this group was 2.15-6.35 with a mean of 4.0 -- 1.39 (SD). Of 86 doubtful cases of CDRED, three were positive (false-positives) for toxin A. The range of ODR values in this group was 2.11-2.60 with a mean of 2.31 -* 0.25 (SD). There was no difference in the range of or mean ODR values for toxin B among the very likely, probable, possible, and doubtful cases of CDRED.
Quantitative Results of Microtiter Cytotoxicity Assay The MCTXA was positive if cytotoxicity occurred at a dilution of I>1:10. Cytotoxicity occurring only in undiluted samples was considered as negative. Of 18 very likely or probable cases of CDRED, 14 had positive (true-positive) microtiter cytotoxicity results. None had cytotoxicity in the undiluted samples only. Four were positive at a 1:10 dilution, six at 1:100, and four at 1:1000. Of 86 doubtful cases, two had positive (false-positive) MCTXA results both at a 1:10 dilution. Twelve other doubtful cases had cytotoxicity occurring in the undiluted sample only. In those specimens where both the MCTXA and an ELISA toxin assay were positive, the ODR for the toxin A ELISA showed a greater tendency to increase with increasing MCTXA titers than did the ODR values for toxin B.
Proctosigmoidoscopic Examination Proctosigmoidoscopic examination was not done routinely in patients whose stools were submitted for C. difficile testing. Among the 18 patients with very likely or probable CDRED, 10 had proctoscopic examinations; in eight (80%) of these, abnormalities consistent with CDRED were noted. None, however had pseudomembranes. Forty-three of the 86 doubtful cases had proctoscopic exams; 20 of these yielded diagnoses other than CDRED, and the remaining 26 were normal.
Test Results in Hospitalized Patients Receiving Antibiotics With no Diarrhea An additional 19 patients who were hospitalized and were receiving antibiotics but had no diarrhea had formed stools submitted for C. difficile testing. One patient developed significant diarrhea within 2 days after testing; all C. difficile tests were positive in this patient. The other 18 patients remained asymptomatic; in this group there were two positive C. difficile cultures, one positive toxin A ELISA, and two positive toxin B ELISAs, but no positive cytotoxicity tests. DISCUSSION The purification of C. difficile toxin A and toxin B has permitted the development of new immunoassays to detect these toxins in stool specimens. In this study we compared toxin A and toxin B ELISA with standard tests using fresh diarrheal specimens. The diagnostic accuracy of each test was based on a blinded clinical assessment of disease probability. In this sense this study differs from some prior evaluations of laboratory tests for the diagnosis of CDRED, which simply compared the results with an arbitrary "standard" (usually a cytotoxicity assay) without considering
68
R.C. Walker et al.
the clinical status of the patients. It is apparent from this study and prior studies (George et al., 1982) that there is a wide spectrum of disease related to C. difficile ranging from pseudomembranous colitis to simple diarrhea, and that there is no single laboratory test that is absolutely diagnostic of CDRED. Moreover, other than pseudomembranous colitis, none of the clinical findings is pathognomonic of the disease. Determination of the most accurate tests, therefore, requires comparison of all laboratory results with a systematic, detailed review of patient medical records and compilation of the most important clinical features relative to the diagnosis of CDRED. Such a review revealed to us that, indeed, the toxin A ELISA was actually more specific than the MCTXA (100% versus 95%), although the latter was more sensitive than the toxin A assay (78% versus 61%). In fact, clinical observations led to the setting of the MCTXA positive "breakpoint" at a 1:10 dilution of stool filtrate when it was observed in a preliminary study that numerous positive results in undiluted specimens were not associated with clinical evidence of CDRED. These observations, therefore, helped develop criteria resulting in a high degree of specificity for the MCTXA. Additionally, the clinical evaluation allowed us to eliminate from the specificity calculations those patients who were found not to have received antimicrobials and therefore would be inappropriate for CDRED laboratory studies. In particular, the specificity of the culture fell from 88% to 74% after elimination of these inappropriate specimens. This relatively low specificity is an important consideration in view of the rather high sensitivity of the culture. It has been previously shown that C. difficile can be cultured from some patients who have received no antimicrobials as well as those with no diarrhea who have received antimicrobials (George et al., 1982). This method and the toxin B ELISA had the lowest specificities of those tested (74% and 56%, respectively). Overall, the MCTXA was the most efficient diagnostic test considering both sensitivity and specificity. The high degree of specificity of the toxin A ELISA in this study is consistent with its definition as an enterotoxin and the likelihood that this toxin plays a more important role in the pathogenesis of human CDRED than does the cytotoxin (toxin B) (Lyerly et al., 1982). The toxin B ELISA achieved the lowest specificity of any of the procedures tested in this study (56%). This finding is somewhat curious in view of the good performance of the MCTXA, but probably is the result of nonspecific binding of the antitoxin B antibody used in this assay. The use of affinity-absorbed toxin B mixed with formalin-stabilized whole cells of nontoxigenic C. difficile in the preparation of antiserum to toxin B may account for the higher specificity (100%) of the toxin B ELISA reported by Laughon et al. (1984). If one calculates the sensitivity of the ELISA for toxin A using the MCTXA as a standard, a value of 69% is obtained (based on 16 positive results in antibiotic recipients). This compares favorably with the sensitivity of 59% reported by Lyerly et al. (1983) using the same assay. The toxin A ELISA developed by Laughon et al. (1984) detected 91% of cytotoxin-positive specimens. The addition of fecal calf serum to stool diluent to prevent proteolytic desorption of antibody from the solid phase of the ELISA may account for this increased sensitivity, and also appears to enhance specificity as well (Hanvanich et al., 1985). CONCLUSION The comparison of results of several laboratory tests for CDRED with clinical evidence of disease has allowed us to more accurately determine the sensitivity and specificity of these tests. We believe this approach is more appropriate than simply comparing our results with an arbitrary standard test. The toxin A ELISA was the most specific
C o m p a r i s o n of Tests for C. difficile
69
test (100%), but was less sensitive than the MCTXA, which, in terms of both sensitivity and specificity (78% and 95%), is currently the most efficient laboratory test to correlate w i t h clinical findings in d e t e r m i n i n g the l i k e l i h o o d of the presence of CDRED. I m p r o v e m e n t in the sensitivity of and simplification in the m e t h o d o l o g y of assays for toxin A m a y well lead to w i d e r clinical usage in the future. REFERENCES
Chang TW, Lauerman M, Bartlett SG (1979) Cytotoxicity assay in antibiotic-associated colitis. J Infect Dis 140:765. Ehrich M, Van Tassell RL, Libby JM, Wilkins TD (1980) Production of Clostridium dif~cile antitoxin. Infect Immun 28:1041. George WL, Sutter VL, Citron D, Finegold S (1979) Selective and differential medium for isolation of C|ostridium difficile. J Clin Microbiol 9:214. George WL, Rolfe RD, Finegold SM (1982) Clostridium difficile and its cytotoxin in feces of patients with antimicrobial agent-associated diarrhea and miscellaneous conditions. J C]in Microbiol 15:1049. Hanvanich M, Viscidi R, Laughon BE, Bartlett JG, Yolken RH (1985) Stool desorbing activity: a possible cause of false-positive reactions in competitive enzyme immunoassays. J C/in Microbiol 21:184. Laughon BE, Viscidi RP, Gdovin SL, Yocken RH, Bartlett JG (1984) Enzyme immunoassays for detection of Clostridium difficile toxins A and B in fecal specimens. J Infect Dis 149:781. Lyerly DM, Lockwood DE, Richardson SH, Wikins TD (1982) Biological activities of toxins A and B of Clostridium difficile. Infect Immun 35:1147. Lyerly DM, Sullivan NM, Wilkins TD (1983) Enzyme-linked immunosorbent assay for Clostridium di~cile toxin A. J Clin Microbiol 17:72. Sullivan NM, Pellet S, Wilkins TD (1982) Purification and characterization of toxins A and B of Clostridium difficile. Infect Immun 35:1032. Taylor NS, Thorne GM, Bartlett SG (1981) Comparison of two toxins produced by Clostridium difficile. Infect Immun 34:1036. Tedesco FJ (1979) Antibiotic-associated pseudomembranous colitis with negative proctosigmoidoscopy examination. Gastroenterology 77:295. Varki NM, Aquino TJ (1982) Isolation of Clostridium difficile from hospitalized patients without antibiotic-associated diarrhea or colitis, l Clin Microbiol 16:659. Wu TC, Fung JC (1983) Evaluation of the usefulness of counterimmunoelectrophoresis for diagnosis of Clostridium difficile-associated colitis in clinical specimens, l Clin Microbial 17:610.
61
Comparison of Culture, Cytotoxicity Assays, and Enzyme-Linked Immunosorbent Assay for Toxin A and Toxin B in the Diagnosis of Clostridium difficile-Related Enteric Disease Randall C. Walker, Peter J. Ruane, Jon E. Rosenblatt, David M. Lyerly, Curt A. Gleaves, Thomas F. Smith, Phillip F. Pierce Jr., and Tracy D. Wilkins
Clostridium difficile culture, test tube, and micretiter cytotoxicity assays, and enzyme-linked immunosorbent assays (ELISAs) for toxin A and toxin B, were simultaneously performed on 113 fresh diarrheal steal specimens randomly selected from those submitted to our clinical laboratory for routine C. difficile testing. The performance of these tests in diagnosing C. difficile-related enteric disease (CDBED) was based on a clinical assessment of the likelihood of CDRED as determined by a systematic review of case histories blinded from the test results. Among 61 antibiotic recipients, both the microtiter cytotoxicity assay and the toxin A ELISA were highly specific for CDRED (95% and 100%, respectively). Specificities for the ether procedures were much lower (tube cytotoxicity assay, 79%; culture, 74%; and toxin B ELISA, 56%). The high sensitivities of the culture (89%) and toxin B ELISA (83%) were somewhat negated by their low specificities. The only test that was both specific and had acceptable sensitivity (78%) was the microtiter cytotoxicity assay. This study indicates that ELISAs for detection of C. difffieile toxins are net as reliable as the cytotoxicity assay in the laboratory diagnosis of CDRED, and that clinical correlation is essential in the evaluation of any new test for CDRED. INTRODUCTION The e x a m i n a t i o n of fecal specimens from large n u m b e r s of patients suspected of having Clostridium difficile-related enteric disease (CDRED) has become a major task for m a n y clinical microbiology laboratories. There is a need for a rapid, simple, and efficient test for C. difficile that might replace culture of stool specimens and tissue culture cytotoxicity assay. Clostridium difficile produces two toxins: toxin A, which is primarily an enterotoxin (Taylor et al., 1981} and toxin B, w h i c h is 1000 times more cytotoxic than toxin A and is responsible for the cytotoxicity observed in tissue culture assays
From the Divisions of Clinical Microbiology and Infectious Diseases, Mayo Clinic and Mayo Foundation Rochester, MN and Department of Anaerobic Microbiology, Virginia Polytechnic Institute and State University, Blacksburg, VA. Address reprint requests to: Jan E. Rosenblatt, M.D., Divisions of Clinical Microbiology and Infectious Diseases, Mayo Clinic and Mayo Foundation, Rochester, MN 55905. Received June 18, 1985; revised and accepted February 20, 1986. © 1986 Elsevier Science PublishingCo., Inc. 52 VanderbiltAvenue,New York, NY 10017
0732-8893/86/$03.50
62
R.C. Walker et al.
(Sullivan et al., 1982). The exact role of each toxin in producing CDRED has not been determined. The isolation and purification of these toxins has led to the development of enzyme-linked immunosorbent assays (ELISAs) using antisera directed at the purified toxins (Lyerly et al., 1983; Laughon et al., 1984). A problem in evaluating the clinical significance of newer tests for CDRED is the relationship between laboratory results and the clinical diagnosis (George et al., 1982). Culture and cytotoxicity assays can be falsely positive in patients who do not have enteric disease (George et al., 1982; Varki et al., 1982). While test results in patients with antibiotic-associated pseudomembranous colitis (PMC) can be used to determine its sensitivity, PMC is uncommon and C. difficile has been shown to cause enteric disease other than proctosigmoidoscopic-positive PMC (George et al., 1982; Tedesco, 1979). Therefore, clinical correlation is important in the evaluation of any new test. We performed C. difficile culture, two tissue culture cytotoxicity assays (tube and microtiter methods), and ELISAs directed against purified toxin A and toxin B on flesh diarrheal specimens randomly selected from clinical specimens submitted for routine testing. The patients' histories were systematically reviewed and classified in order to compare test results on the basis of the clinical likelihood of CDRED.
MATERIALS AND METHODS Stool Specimens During a 5-month time period, diarrheal stool samples from adult patients submitted for C. difficile testing were randomly selected for C. difficile toxin determination by ELISA and a microtiter cytotoxicity assay (MCTXA) method, in addition to routine bacteriologic culture and tube cytotoxicity assay. Besides the requirement that stools be diarrheal, there was no prior screening or selection of samples on the basis of any clinical data.
Stool Culture Upon arrival in the microbiology laboratory, culture plates containing cycloserine, cefoxitin, fructose, and egg yolk agar (CCFA, Difco Laboratories, Inc., Detroit, MI) were inoculated with stool and incubated at 36°C in an anaerobic atmosphere for 24-48 hr. The presence of suspicious colonies were confirmed by colony fluorescence and biochemical and chromatographic techniques (George et al., 1979).
Stool Filtrate Preparation for Toxin Assays The procedure used for stool filtrate preparation is that of Chang and coworkers (1979). Stool filtrates were refrigerated (4°C) 0-24 hr before processing.
Antigen and Antibody Preparations
Clostridium difficile antitoxins prepared in rabbit serum and goat serum were used for toxin neutralization in the ELISA and cytotoxicity assays, respectively. Affinitypurified antibodies against C. difficile toxin A and B were used in the ELISA for the specific detection of these toxins in stool filtrates. A culture filtrate prepared from C. difficile stock strain 10463 (ATCC) was used as the positive control in the ELISA and both cytotoxicity assays. The above materials were prepared and supplied by DM Lyerly and TD Wilkins (Department of Anaerobic Microbiology, Virginia Polytechnic Institute and State University of Blacksburg, Blacksburg, VA).
Comparison of Tests for C. difficile
63
Tube Cytotoxicity and Neutralization Assays Monolayers of MRC-5 cells (Flow Laboratories, Rockville, MD) were prepared in 16 × 125 mm culture tubes and maintained in Eagles minimal essential medium (MEM) containing 10% fetal bovine serum (FBS), penicillin, streptomycin, and gentamycin until specimen inoculation. One MRC-5 monolayer was inoculated per patient specimen with 0.2 ml of stool filtrate. An uninoculated MRC-5 monolayer (negative control) and a monolayer inoculated with the positive control filtrate was included with each batch of patient specimens. All tubes were incubated at 36°C in room air on a fiat culture rack overnight. The monolayers were read twice daily for cytopathic effect (CPE) characteristic of that produced by C. difficile toxins. If CPE was detected, a neutralization procedure was done to prove the specificity of the CPE by using a 1:100 dilution of C. difficile antitoxin. The neutralization procedure involved the incubation of equal amounts (0.5 ml) of stool filtrate with antitoxin at 36°C for 30 min in a test vial, followed by the inoculation of two monolayers with the filtrate/antitoxin combination (0.1 ml) and one monolayer with the original stool filtrate. A positive neutralization control (C. difficile toxin plus antitoxin) and a negative neutralization control (C. difficile toxin) was included with each batch of a specimen's neutralization assays. The neutralization test was considered positive when the CPE was eliminated in the presence of C. difficile antitoxin.
Microtiter Cytotoxicity and Neutralization Assay Chinese hamster ovary cells (CHO-KI cells, ATCC, Rockville, MD) were prepared in microtiter plates (Dynatech, Alexandria, VA) as previously described (Ehrich et al., 1980) with two modifications: the cell medium was Eagle's MEM containing 2% FBS, and 1 ml each of sodium bicarbonate and Tris buffer, and the microtiter wells were seeded with 5,000 CHO-K1 ceils per well. Undiluted and serial 10-fold dilutions (to 1:1000) of patient stool filtrate (0.02 ml) were added to separate microtiter wells and then incubated 18 hr at 36°C in a CO2 (5%) atmosphere. A neutralization assay for each patient specimen was included in the same microtiter plate. Cytotoxicity was determined to be positive if the monolayers showed 100% cell death.
Enzyme-Linked Immunosorbent Assay for C. difficile Toxin A and B The ELISA method for C. difficile toxin A and B detection was previously described by Lyerly et al. (1983).
Clinical Evaluation The case histories of 113 adult patients whose physicians had submitted stool samples for routine laboratory diagnosis of C. difficile-related diarrhea were systematically reviewed. The reviewers were unaware of the ELISA and MCTXA results, although stool culture and tube cytotoxicity test results were available to the patients' clinicians and, thus, were documented in the medical records. Only the first clinical episode of diarrhea in a patient and the corresponding test results were studied. Relapse or follow-up cultures were not included in this study. The 113 case summaries were evaluated according to the presence or absence of four major clinical criteria (George et al., 1982): 1) A history of antibiotic use in the 2 months before the stool specimen was obtained. 2) Clinically significant diarrhea: defined as a change in usual bowel habit with three or more loose stools per day for 2 or more days in a row.
64
R.C. Walker et al.
3) Treatment w i t h and response to accepted t h e r a p y for C. difficile enteric disease (vancomycin, metronidazole, or bacitracin). Response was defined as a reduction of 50% in the n u m b e r of stools per day w i t h i n 2 days of starting therapy. Cases that r e l a p s e d after t h e r a p y were s t o p p e d were also considered as having had a positive and specific response to therapy. 4) Absence of an alternative diagnosis to e x p l a i n the patient's diarrhea, that is, active inflammatory bowel disease, other stool pathogens, use of laxatives, recent gastrointestinal surgery, or resolving bowel obstruction. Cases were classified in terms of the l i k e l i h o o d of a clinical diagnosis of CDRED as: very likely, probable, possible, and doubtful, d e p e n d i n g on the fulfillment of these criteria. The presence or absence of p s e u d o m e m b r a n e s or abnormal sigmoidoscoptic e x a m i n a t i o n were not i n c l u d e d in the major criteria because so few patients had this e x a m i n a t i o n performed. RESULTS
Laboratory Results In the 113 s p e c i m e n s tested, the C. difficile culture was positive in 27 (24%); the tube assay for cytotoxicity was positive in 22 (19%); the MCTXA was positive in 16 (14%); the toxin A ELISA was positive in 14 (12%); and the toxin B ELISA was positive in 54 (47%) cases.
Clinical Classification of Disease Probability Using the above criteria, 14 cases were classified as very likely having C. difficilerelated disease, four were classified as probable cases, nine as possible cases, and 86 as doubtful cases (Table 1.) The 14 very likely cases met all four of the clinical criteria. The four probable cases had significant diarrhea r e s p o n d i n g to specific C. difficile therapy, but had other c o n d i t i o n s that also could have caused diarrhea. One of the probable cases had
TABLE 1. Clinical Likelihood of
Clostridium difficile-Related Enteric Disease Clinical criteria fulfilled
Clinical likelihood
No.
Antibiotic
Significant
of disease
of cases
exposure °
diarrhea b
14
+
+
+
+
4 g
+ +
+ +
+ 0
+
Very likely Probable Possible
Doubtful (total of 86)
21 11 52
Therapeutic
trialc
No other possible
cause d
+
+
-(2)
+
+ + -
+ +/-
0 / - (2) 0 / - (1) 0/-(2)
+/+/-
°Antibiotic e x p o s u r e in 2 m o n t h s before s p e c i m e n w a s collected. One probable case h a d chronic diarrhea a n d r e s p o n d e d p r o m p t l y to oral v a n c o m y c i n but h a d no record of recent antibiotic exposure. bBased on acute onset, frequency, quality, a n d d u r a t i o n of diarrhea. ~+ indicates p r o m p t response to specific t h e r a p y (e.g., oral v a n c o m y c i n ) for CDRED, 0 indicates no treatment, - indicates treatment given but no response (no. of cases). d + indicates no other possible causes for diarrhea were present.
Comparison of Tests for
C. difficile
65
chronic diarrhea for several years that responded promptly to specific C. difficile therapy, although antibiotic exposure before the onset of symptoms was not specified in the case record. All of the n i n e possible cases had recent antibiotic exposure, significant diarrhea, and no other c o n d i t i o n evident besides C. difficile enterocolitis that could have caused diarrhea. However, diarrhea in each case resolved spontaneously without specific therapy for CDRED. In 84 doubtful cases, no more than two clinical criteria were fulfilled. Another two doubtful cases fulfilled three criteria, but had persistent diarrhea despite specific therapy for CDRED. Only 34 doubtful cases had recent antibiotic exposure.
Correlation of Laboratory Results With Clinical Classification Of the 14 very likely cases of CDRED, the C. difficile culture was positive i n 12 (86%); the tube cytotoxicity assay was positive in 10 (71%); the MCTXA was positive in 12 (86%); the toxin A ELISA was positive in n i n e (64%); and the toxin B ELISA was positive in 11 (79%) (Table 2). All four of the probable cases had positive C. difficile cultures and toxin B ELISAs. Tube cytotoxicity, MCTXA, and toxin A ELISAs were each positive in two of these four cases. Among the n i n e possible cases, there was one case with a positive C. difficile culture and five cases with positive toxin B ELISAs. The other tests were negative in all n i n e cases. Among the 34 doubtful cases that had recent antibiotic exposure, the culture was positive in 10 (29%); the tube cytotoxicity assay was positive in n i n e (26%); the MCTXA was positive i n two (6%); the toxin A ELISA was positive in n o n e (0%); and the toxin B ELISA was positive in 14 (41%). None of the 52 patients who had not received antibiotics had a positive stool culture or MCTXA. One had a positive tube cytotoxicity test, whereas three had a positive toxin A and 20 had a positive toxin B ELISA.
Sensitivity and Specificity of Laboratory Tests The sensitivity of laboratory tests was based on the n u m b e r of positive results (true positives) in that clinical group of cases where tests should have been positive (true positives plus false negatives (Table 3). Based on the 18 cases with very likely or
TABLE 2. Clinical Correlation of Clostridium difficile Test Results Diagnostic test no. (%) positive Clinical likelihood of disease
No. of cases
Very likely Probable Possible Doubtful
14 4 9 86
12 4 1 10
34
10 (29)
9 (26)
2 (5.9)
113
27 (24)
22 (19)
16 (14)
Antibiotic recipients only° All cases
C. difficile culture (86) (100) (11) (12)
Tube cytotoxicity 10 2 0 10
(71) (50) (0) (12)
Microtiter cytotoxicity 12 2 0 2
(86) (50) (0) (2.3)
Toxin A ELISA 9 2 0 3
(64) (50) (0) (3.5)
Toxin B ELISA 11 4 5 34
(79) (100) (55) (40)
0 (0)
14 (41)
14 (12)
54 (47)
°Fifty-twoof 86 doubtful cases had no recorded history of antibiotic exposure in the 2 months before the date the stool specimenwas collected.
66
R.C. Walker et al.
TABLE 3. Comparison of Sensitivity and Specificity of Clostridium difficile Tests Using Clinical Likelihood of Disease as Reference Diagnostic
Clinical likelihood of diseasea
Test
Sensitivity
Specificity
Microtiter cytotoxicity C. difficile culture Tube cytotoxicity Toxin A ELISA Toxin B ELISA
78 89 67 61 83
95 74 79 100 56
aBased on 61 antibioticrecipients.Sensitivityis percentpositiveamong18 caseswith significantdiarrhea and promptresponseto specifictherapyfor CDRED(very likely and probable categories);specificityis percent negative among 43 cases with low clinical probabilityof CDRED(passible and doubtful categories). Note that patients who did not receive antimicrobialsare excluded.
probable CDRED, the sensitivities for each test were: C. difficile culture, 89%; tube cytotoxicity, 67%; MCTXA, 78%; toxin A ELISA, 61%; and toxin B ELISA, 83%. The specificity of laboratory tests was based on the n u m b e r of negative results (true negatives) in that clinical group of cases where tests should have been negative (true negatives plus false positives (Table 4). Based on all 86 cases in w h o m CDRED was doubtful and the n i n e possible cases, the specificity of each test was: C. difficile culture, 88%; tube cytotoxicity, 90%; MCTXA, 98%; toxin A ELISA, 97%; and toxin B ELISA, 60%. However, if one selects only the 34 antibiotic recipients in w h o m CDRED was doubtful (considering those who did not receive antimicrobials as inappropriate candidates for testing) and combines these with the n i n e possible cases, the specificity of each test was: culture, 74%; tube cytotoxicity, 79%; MCTXA, 95%; toxin A ELISA, 100%; and toxin B ELISA, 56%. Based on all 113 cases studied, the predictive values of positive and negative results and efficiencies, respectively, for each test were (%): culture 59.2, 97.6, and 88.4; tube cytotoxicity 54.5, 93.4, and 85.8; MCTXA 87.5, 95.8, 94.7; ELISA for toxin A 78.5, 92.9, and 91.1; and ELISA for toxin B 27.7, 94.9, and 62.8. In this population, MCTXA was the most efficient test (Table 4).
TABLE 4. Comparative Predictive Values and Efficiencies° of Diagnostic Tests for Clostridium difficile-Related Enteric Disease in Clinical Patients ° Predictive value (%)
C. difficile culture Tube cytotoxicity Microtiter cytotoxicity ELISA A ELISA B
Test positive
Test negative
Efficiency (%)
59.2 54.5 87.5 78.5
97.6 93.4 95.8 92.9 94.9
88.4 85.8 94.7 91.1 62.8
27.7
°Galen RS, GambinoSR (1975) Beyond Normality: the Predictive Value and Efficiency of Medical Diagnosis. New York: John Wiley& Sons. hAll clinical patients, n = 113.
Comparison of Tests for C.
difficile
67
Quantitative Results of Enzyme-Linked Immunosorbent Assay for Toxin A and Toxin B The ELISA assays for toxin A and toxin B were determined positive when the optical density ratio (ODR) (optical density of sample divided by optical density of control) for the sample tested was >2.00. Of 18 very likely and probable cases of CDRED, 11 were positive (flue positives) for toxin A. The range of ODR values in this group was 2.15-6.35 with a mean of 4.0 -- 1.39 (SD). Of 86 doubtful cases of CDRED, three were positive (false-positives) for toxin A. The range of ODR values in this group was 2.11-2.60 with a mean of 2.31 -* 0.25 (SD). There was no difference in the range of or mean ODR values for toxin B among the very likely, probable, possible, and doubtful cases of CDRED.
Quantitative Results of Microtiter Cytotoxicity Assay The MCTXA was positive if cytotoxicity occurred at a dilution of I>1:10. Cytotoxicity occurring only in undiluted samples was considered as negative. Of 18 very likely or probable cases of CDRED, 14 had positive (true-positive) microtiter cytotoxicity results. None had cytotoxicity in the undiluted samples only. Four were positive at a 1:10 dilution, six at 1:100, and four at 1:1000. Of 86 doubtful cases, two had positive (false-positive) MCTXA results both at a 1:10 dilution. Twelve other doubtful cases had cytotoxicity occurring in the undiluted sample only. In those specimens where both the MCTXA and an ELISA toxin assay were positive, the ODR for the toxin A ELISA showed a greater tendency to increase with increasing MCTXA titers than did the ODR values for toxin B.
Proctosigmoidoscopic Examination Proctosigmoidoscopic examination was not done routinely in patients whose stools were submitted for C. difficile testing. Among the 18 patients with very likely or probable CDRED, 10 had proctoscopic examinations; in eight (80%) of these, abnormalities consistent with CDRED were noted. None, however had pseudomembranes. Forty-three of the 86 doubtful cases had proctoscopic exams; 20 of these yielded diagnoses other than CDRED, and the remaining 26 were normal.
Test Results in Hospitalized Patients Receiving Antibiotics With no Diarrhea An additional 19 patients who were hospitalized and were receiving antibiotics but had no diarrhea had formed stools submitted for C. difficile testing. One patient developed significant diarrhea within 2 days after testing; all C. difficile tests were positive in this patient. The other 18 patients remained asymptomatic; in this group there were two positive C. difficile cultures, one positive toxin A ELISA, and two positive toxin B ELISAs, but no positive cytotoxicity tests. DISCUSSION The purification of C. difficile toxin A and toxin B has permitted the development of new immunoassays to detect these toxins in stool specimens. In this study we compared toxin A and toxin B ELISA with standard tests using fresh diarrheal specimens. The diagnostic accuracy of each test was based on a blinded clinical assessment of disease probability. In this sense this study differs from some prior evaluations of laboratory tests for the diagnosis of CDRED, which simply compared the results with an arbitrary "standard" (usually a cytotoxicity assay) without considering
68
R.C. Walker et al.
the clinical status of the patients. It is apparent from this study and prior studies (George et al., 1982) that there is a wide spectrum of disease related to C. difficile ranging from pseudomembranous colitis to simple diarrhea, and that there is no single laboratory test that is absolutely diagnostic of CDRED. Moreover, other than pseudomembranous colitis, none of the clinical findings is pathognomonic of the disease. Determination of the most accurate tests, therefore, requires comparison of all laboratory results with a systematic, detailed review of patient medical records and compilation of the most important clinical features relative to the diagnosis of CDRED. Such a review revealed to us that, indeed, the toxin A ELISA was actually more specific than the MCTXA (100% versus 95%), although the latter was more sensitive than the toxin A assay (78% versus 61%). In fact, clinical observations led to the setting of the MCTXA positive "breakpoint" at a 1:10 dilution of stool filtrate when it was observed in a preliminary study that numerous positive results in undiluted specimens were not associated with clinical evidence of CDRED. These observations, therefore, helped develop criteria resulting in a high degree of specificity for the MCTXA. Additionally, the clinical evaluation allowed us to eliminate from the specificity calculations those patients who were found not to have received antimicrobials and therefore would be inappropriate for CDRED laboratory studies. In particular, the specificity of the culture fell from 88% to 74% after elimination of these inappropriate specimens. This relatively low specificity is an important consideration in view of the rather high sensitivity of the culture. It has been previously shown that C. difficile can be cultured from some patients who have received no antimicrobials as well as those with no diarrhea who have received antimicrobials (George et al., 1982). This method and the toxin B ELISA had the lowest specificities of those tested (74% and 56%, respectively). Overall, the MCTXA was the most efficient diagnostic test considering both sensitivity and specificity. The high degree of specificity of the toxin A ELISA in this study is consistent with its definition as an enterotoxin and the likelihood that this toxin plays a more important role in the pathogenesis of human CDRED than does the cytotoxin (toxin B) (Lyerly et al., 1982). The toxin B ELISA achieved the lowest specificity of any of the procedures tested in this study (56%). This finding is somewhat curious in view of the good performance of the MCTXA, but probably is the result of nonspecific binding of the antitoxin B antibody used in this assay. The use of affinity-absorbed toxin B mixed with formalin-stabilized whole cells of nontoxigenic C. difficile in the preparation of antiserum to toxin B may account for the higher specificity (100%) of the toxin B ELISA reported by Laughon et al. (1984). If one calculates the sensitivity of the ELISA for toxin A using the MCTXA as a standard, a value of 69% is obtained (based on 16 positive results in antibiotic recipients). This compares favorably with the sensitivity of 59% reported by Lyerly et al. (1983) using the same assay. The toxin A ELISA developed by Laughon et al. (1984) detected 91% of cytotoxin-positive specimens. The addition of fecal calf serum to stool diluent to prevent proteolytic desorption of antibody from the solid phase of the ELISA may account for this increased sensitivity, and also appears to enhance specificity as well (Hanvanich et al., 1985). CONCLUSION The comparison of results of several laboratory tests for CDRED with clinical evidence of disease has allowed us to more accurately determine the sensitivity and specificity of these tests. We believe this approach is more appropriate than simply comparing our results with an arbitrary standard test. The toxin A ELISA was the most specific
C o m p a r i s o n of Tests for C. difficile
69
test (100%), but was less sensitive than the MCTXA, which, in terms of both sensitivity and specificity (78% and 95%), is currently the most efficient laboratory test to correlate w i t h clinical findings in d e t e r m i n i n g the l i k e l i h o o d of the presence of CDRED. I m p r o v e m e n t in the sensitivity of and simplification in the m e t h o d o l o g y of assays for toxin A m a y well lead to w i d e r clinical usage in the future. REFERENCES
Chang TW, Lauerman M, Bartlett SG (1979) Cytotoxicity assay in antibiotic-associated colitis. J Infect Dis 140:765. Ehrich M, Van Tassell RL, Libby JM, Wilkins TD (1980) Production of Clostridium dif~cile antitoxin. Infect Immun 28:1041. George WL, Sutter VL, Citron D, Finegold S (1979) Selective and differential medium for isolation of C|ostridium difficile. J Clin Microbiol 9:214. George WL, Rolfe RD, Finegold SM (1982) Clostridium difficile and its cytotoxin in feces of patients with antimicrobial agent-associated diarrhea and miscellaneous conditions. J C]in Microbiol 15:1049. Hanvanich M, Viscidi R, Laughon BE, Bartlett JG, Yolken RH (1985) Stool desorbing activity: a possible cause of false-positive reactions in competitive enzyme immunoassays. J C/in Microbiol 21:184. Laughon BE, Viscidi RP, Gdovin SL, Yocken RH, Bartlett JG (1984) Enzyme immunoassays for detection of Clostridium difficile toxins A and B in fecal specimens. J Infect Dis 149:781. Lyerly DM, Lockwood DE, Richardson SH, Wikins TD (1982) Biological activities of toxins A and B of Clostridium difficile. Infect Immun 35:1147. Lyerly DM, Sullivan NM, Wilkins TD (1983) Enzyme-linked immunosorbent assay for Clostridium di~cile toxin A. J Clin Microbiol 17:72. Sullivan NM, Pellet S, Wilkins TD (1982) Purification and characterization of toxins A and B of Clostridium difficile. Infect Immun 35:1032. Taylor NS, Thorne GM, Bartlett SG (1981) Comparison of two toxins produced by Clostridium difficile. Infect Immun 34:1036. Tedesco FJ (1979) Antibiotic-associated pseudomembranous colitis with negative proctosigmoidoscopy examination. Gastroenterology 77:295. Varki NM, Aquino TJ (1982) Isolation of Clostridium difficile from hospitalized patients without antibiotic-associated diarrhea or colitis, l Clin Microbiol 16:659. Wu TC, Fung JC (1983) Evaluation of the usefulness of counterimmunoelectrophoresis for diagnosis of Clostridium difficile-associated colitis in clinical specimens, l Clin Microbial 17:610.