- Email: [email protected]
Evaluation of the clinical usefulness of C. difficile toxin testing in hospitalized patients with diarrhea
Diagnostic Microbiology and Infectious Disease 36 (2000) 169 –173
Evaluation of the clinical usefulness of C. difficile toxin testing in hospitalized patients with diarrhea Ayman El-Gammala, Vincenzo Scottoa, S. Malikb, Kathleen C. Caseya,b, Ronald Codyc, David V. Alcida, Melvin P. Weinsteina,d,* a
Division of Allergy, Immunology, and Infectious Diseases, Department of Medicine, University of Medicine and Dentistry of New Jersey—Robert Wood Johnson Medical School (UMDNJ-RWJMS), 1 Robert Wood Johnson Place, New Brunswick, NJ 08903-0019, USA b Department of Medicine, Jersey Shore Medical Center, Neptune, NJ, USA c Department of Environmental and Community Medicine, UMDNJ-RWJMS, New Brunswick, NJ, USA d Department of Pathology and Laboratory Medicine, UMDNJ- RWJMS, New Brunswick, NJ, USA
Abstract Although numerous studies have evaluated the sensitivity and specificity of different assays for Clostridium difficile toxin, none has evaluated how physicians utilize these tests or respond to test results. Therefore, we assessed patient characteristics, clinical findings, and physician responses to positive and negative assay results at two university-affiliated hospitals, one of which used a cell cytotoxicity assay to test for C. difficile toxin and the other of which used an enzyme immunoassay. Two hundred one patient samples at Hospital A and 199 samples at Hospital B were assessed. Positive toxin assays were more frequent at Hospital A than at Hospital B (p ⬍ 0.001), at least in part due to the fact that patients tested at Hospital A were more likely to have fever (p ⬍ 0.001), an abnormal abdominal exam (p ⬍ 0.001), an abnormal leukocyte count (p ⬍ 0.001), and a history of prior antibiotic use (p ⬍ 0.001). Empiric therapy for C. difficile before results of the toxin assay was more common (p ⬍ 0.001) at Hospital A (83/201, 41.3%) than at Hospital B (25/199, 12.5%). Once empiric therapy was started, most physicians continued therapy despite negative test results (Hospital A, 76%; Hospital B, 69%). Patients who were treated empirically were more likely than patients not treated empirically to have positive toxin assay results and to have fever (p ⬍ 0.001), an abnormal abdominal exam (p ⫽ 0.003), or an abnormal leukocyte count (p ⬍ 0.05). Physicians seldom ordered repeat toxin assays (Hospital A, 14%; Hospital B, 10%) if the initial assay result was negative. In logistic regression analysis, predictors of a positive toxin assay were prior antibiotic therapy, an abnormal abdominal exam, residence at Hospital A, and age ⱖ 60 years. Predictors of empiric therapy were residence at Hospital A and prior antibiotic therapy. Because physicians electing to empirically treat inpatients with diarrhea rarely alter therapy based on C. difficile toxin assay results, a more cost-effective management strategy may be not to obtain a toxin assay at all in such situations. Testing should be limited to patients who have received antibiotics within the prior month and who have significant diarrhea and/or abdominal pain. © 2000 Elsevier Science Inc. All rights reserved.
1. Introduction Rising health care costs have focused attention on physician utilization of laboratory tests (Crowson et al., 1984). In patients with diarrhea illnesses, the cost-effectiveness of identifying specific pathogens has not been well studied, and it is not known how physicians use results of such testing (Hines, 1996). Clostridium difficile is the most common infectious etiology of nosocomial diarrhea, accounting for 10 to 15% of all antibiotic-associated diarrhea and causing almost all an* Corresponding author. Tel.: ⫹1-732-235-7713; fax: ⫹1-732-2357951. E-mail address: [email protected] (M.P. Weinstein).
tibiotic-associated colitis (Mitty, 1994). C. difficile associated diarrhea can be mild, but it is more often a serious illness and can have a fatal outcome despite specific antibiotic therapy (Mitty, 1994). Whereas there are several laboratory tests that aid in the diagnosis of C. difficile diarrhea, each method has limitations (Hines, 1996; Johnson, 1998). The toxin neutralization cell cytotoxicity assay for detecting cytotoxin B in stool is considered to be the gold standard laboratory test (Fekety, 1995), but it is relatively expensive and not available in many laboratories because cell culture methodology is required. Results are not available for one to two days and, as the toxin is labile, false negative results may occur because of inactivation of the toxin during transport (Fekety, 1995). In addition, false negative tissue culture results may occur
0732-8893/00/$ – see front matter © 2000 Elsevier Science Inc. All rights reserved. PII: S 0 7 3 2 - 8 8 9 3 ( 9 9 ) 0 0 1 2 9 - 7
170
Al El-Gammal et al. / Diagnostic Microbiology and Infectious Disease 36 (2000) 169 –173
due to excessive dilution of the specimen. Overall, the cell cytotoxicity assay will be positive in 10 to 20% of patients for whom testing is requested, including 15 to 25% of patients with antibiotic-associated diarrhea, 50 to 75% of patients with antibiotic-associated colitis, and 95 to 100% of patients with pseudomembranous colitis (Bartlett, 1994). Enzyme immunoassays (EIA) that detect C. difficile toxin A and/or toxin B may provide results in several hours if testing is done upon receipt in the laboratory rather than in batch mode (e.g., once daily). These assays are technically easy to perform, and are readily available commercially. However, they are less sensitive and specific than the cell cytotoxicity assay (Kelly et al., 1994). Fekety (1995) has reported that false negative results range between 5% and 33%. When hospitalized patients develop diarrhea, physicians may respond in several ways. They may order the toxin assay and initiate empiric therapy pending test results, order the toxin assay and await results before initiating therapy, or treat the patient empirically without ordering the toxin assay. Many studies have evaluated the sensitivity and specificity of different assays for C. difficile, but none has evaluated how physicians utilize the tests or respond to test results. Thus, the primary objective of this study was to assess how physicians utilized tests for C. difficile toxin. We also wished to evaluate physicians’ responses to negative test results. Were tests repeated? Was empiric therapy continued or stopped? We report herein our observations at two university-affiliated hospitals, one of which uses a commercial cell cytotoxicity assay to detect the presence of C. difficile toxin and the other of which uses an enzyme immunoassay.
2. Materials and methods Patients who were hospitalized at either of the two study sites and who had stool specimens sent for C. difficile toxin assays during a 12-month period from March 1997 through February 1998 were eligible for inclusion in the study. Data were collected by infectious disease fellows (AEG,VS) or medical residents (SM) at intervals during the study period when their duty assignments allowed. Hospital A is a 448 bed tertiary care institution that utilizes a commercially available cell cytotoxicity assay (Bartels, Issaquah, WA, USA) to detect C. difficile toxin (reported sensitivity and specificity 100% compared with conventional cell culture using W138 cells). This institution is a university medical center with house officers in all disciplines and subspecialty fellows in most disciplines; full-time academic faculty admit approximately half of the patients. Hospital B is a 476 bed community teaching hospital that utilizes an EIA (C. difficile Tox-A Test, Wampole Laboratories, Cranbury, NJ, USA) for detection of C. difficile toxin (reported sensitivity 80.7% and specificity 99.2% compared with cell culture [cell line not stated]). This hospital is a university-affiliated
Table 1 Positive C. difficile toxin assays at Hospitals A and B Hospital Aa
All tests First test only a b
Hospital Bb
p
No.
No. pos.
%
No.
No. pos.
%
201 173
42 37
20.9 21.4
199 170
15 10
7.5 5.9
⬍0.001 ⬍0.001
Cell cytotoxicity assay used. Enzyme immunoassay used.
teaching institution with residency programs in medicine, surgery, obstetrics and gynecology, and pediatrics; community physicians admit approximately 90% of all patients. The medical staffs at the two hospitals do not overlap. Patient charts were reviewed retrospectively, and data were recorded on a standardized study worksheet. The variables assessed included patient demographics, clinical findings, and physician specialty and responses to test results. Demographic data included patient age, sex, date of hospital admission, date of testing, and place of hospitalization. Clinical parameters recorded included history of antibiotic therapy in the previous 6 weeks, character of stool including presence of blood, presence of fever (ⱖ38°C), abnormal peripheral leukocyte count (⬍4000/mm3 or ⬎ 11,000/mm3), abnormal abdominal examination (tenderness, distention, hyperactive bowel sounds, or ileus), and results of C. difficile toxin assay. Physician behaviors assessed included use of empiric therapy for C. difficile before knowledge of test results and use of such therapy if test results were reported to be negative. An episode was defined by the first specimen obtained for C. difficile toxin assay or by any subsequent specimen submitted for testing ⬎7 days after the preceding specimen. If a patient had more than one episode, only the first was included for study purposes. Data were recorded on study worksheets, then tabulated in a computerized database program (Excel, Microsoft Inc, Redmond, WA, USA). Subsequently, the data were transferred to the SAS statistical analysis program (SAS, Inc., Cary, NC, USA) for significance testing and logistic regression analyses.
3. Results Two hundred one samples from 177 patients at Hospital A and 199 samples from 187 patients at Hospital B were analyzed. There were no significant differences in the age or sex distribution of patients from whom samples were obtained at the two hospitals (data not shown). Positive C. difficile toxin assays were more frequent at Hospital A than at Hospital B (Table 1) whether all tests or only the first test on any given patient was counted (p ⬍ 0.001). These differences seemed to be related in part to the clinical characteristics of the patients from whom samples
Al El-Gammal et al. / Diagnostic Microbiology and Infectious Disease 36 (2000) 169 –173
171
Table 2 Clinical characteristics of patients at Hospitals A and B who had stool specimens sent for C. difficile toxin detection
Table 3 C. difficile toxin assay positivity rates and use of empiric therapy for C. difficile diarrhea at Hospitals A and B
Clinical characteristics
Hospital A No. pos.
%
No.
No. pos.
Empiric therapy
Hospital A
No.
No. pos./total
%
No. pos./total
%
Fever Abnormal abdominal exam Bloody diarrhea Abnormal leukocyte count Prior antibiotic therapy
201 200
135 77
67.2 38.5
199 198
31 37
15.6 18.7
⬍0.001 ⬍0.001
Given Not given
22/83 20/118
26.5 16.9
4/25 11/174
16.0 6.3
201 194
16 117
8.0 60.3
74 186
5 92
6.8 49.9
NSa ⬍0.001
201
180
90.0
199
141
70.8
⬍0.001
a
Hospital B
p %
NS, p ⬎ 0.05.
were obtained for testing (Table 2). Patients at Hospital A were significantly more likely than patients at Hospital B to have fever (p ⬍ 0.001), an abnormal abdominal exam (p ⬍ 0.001), an abnormal leukocyte count (p ⬍ 0.001), and a history of prior antibiotic therapy (p ⬍ 0.001). There was no significant difference in the presence of bloody diarrhea in patients from the two hospitals (Table 2). The great majority of physicians who ordered C. difficile toxin assays at both hospitals were general internists, internal medicine subspecialists, or surgeons. At both hospitals, approximately 45% of tests were ordered by internal medicine subspecialists. However, more tests were ordered by surgeons at Hospital A (24.9% vs. 10.1% at Hospital B), whereas more tests were ordered by general internists at Hospital B (44.4% vs. 30.8% at Hospital A). These differences were statistically significant (p ⬍ 0.001). Initiation of empiric therapy for C. difficile diarrhea before receiving test results was significantly more common (p ⬍ 0.001) at Hospital A (83/201, 41.3%) than at Hospital B (25/199, 12.5%). Moreover, negative C. difficile toxin assay results had minimal influence on physicians who administered empiric therapy. At Hospital B, empiric therapy was continued in 16/21 such patients (76.2%) whereas at Hospital A empiric therapy was continued in 42/61 such patients (68.8%). The study design did not allow for determination of whether patients for whom empiric therapy was continued despite a negative assay had improved clinically on such therapy. At both hospitals, patients in whom empiric therapy was ordered had higher rates of C. difficile toxin detection than did patients not started on empiric therapy (Table 3), suggesting that physicians may have recognized that such patients were more ill and, therefore, were more inclined to begin empiric therapy. Indeed, further analysis revealed that patients who were given empiric therapy were significantly more likely to have fever (p ⬍ 0.001), an abnormal abdominal examination (p ⫽ 0.003), or an abnormal leukocyte count (p ⬍ 0.05). Physicians who did not initiate empiric therapy before C.
Hospital B
difficile toxin test results rarely began such therapy once the test results were reported to be negative. At Hospital B, only one of 163 such patients (0.6%) was started on empiric therapy; at Hospital A, 7 of 98 such patients (7.8%) were given empiric therapy. Repeat ordering of C. difficile toxin assays was uncommon in patients with a single negative test result. At Hospital A, repeat tests were sent on 14 such patients (14.3%), and at Hospital B repeat tests were ordered on 16 patients (9.8%). Logistic regression was used to determine predictors of positive C. difficile toxin assay results and use of empiric therapy. Table 4 summarizes the results of these analyses. In the full data set from both hospitals, significant predictors of a positive toxin assay were (in order of highest odds ratio) prior antibiotic administration, abnormal abdominal exam, hospitalization at Hospital A, and age ⱖ 60 years. When the data were analyzed for Hospital B only, none of the variables were significant predictors of a positive toxin assay, whereas at Hospital A both age ⱖ 60 years and an abnormal abdominal exam predicted positive test results. Significant predictors for the use of empiric therapy in the full data set were hospitalization at Hospital A and prior antimicrobial therapy. When the data were analyzed for Hospital B only, age ⱖ 60 was the only significant predictor, whereas at Table 4 Significant predictors of positive C. difficile toxin assay results and use of empiric therapy based on logistic regression analysis
Positive toxin assay
Data Set
Predictor
Odds ratio
95% Confidence interval
Full data set
Prior antibiotics Abn. abdominal exam Age ⱖ60 years Hospital A Abnormal abdominal exam Age ⱖ60 years N/Aa
3.9 3.5
1.1–13.7 1.8–6.5
2.1 2.1 4.9
1.1–4.2 1.1–4.2 2.3–10.4
2.6 N/Aa
1.2–5.8
Hospital A Prior antibiotics Prior antibiotics Age ⱖ60 years
4.2 2.8 3.2 3.0
2.5–7.0 1.3–6.2 1.0–10.1 1.0–9.3
Hospital A only
Hospital B only Empiric therapy
None Full data set Hospital A only Hospital B only
a
N/A ⫽ Information not available.
172
Al El-Gammal et al. / Diagnostic Microbiology and Infectious Disease 36 (2000) 169 –173
Hospital A, prior antimicrobial therapy was the lone significant predictor.
4. Discussion Although there have been numerous studies describing clinical aspects of C. difficile diarrhea and evaluating various C. difficile toxin assays, few if any have assessed physician utilization of toxin assays themselves or the use of empiric therapy in patients with suspected C. difficile infection. Therefore, we chose to examine this issue and to determine if possible whether physician behavior was affected by the type of toxin assay used in the two hospitals. Our observations suggest that at Hospital B toxin assays were ordered on patients who were less ill than their counterparts at Hospital A and that this finding accounted, at least in part, for the lower positivity rate for toxin assays at Hospital B. As indicated in Table 2, patients at Hospital B were significantly less likely to have fever, an abnormal abdominal exam, an abnormal leukocyte count, and a history of prior antimicrobial therapy. A clinical prediction rule for C. difficile toxin testing, recently published and validated (Katz et al., 1996; Katz et al., 1997), states that testing should be done only on hospitalized adults who have received antimicrobial therapy within the past month and who have either significant diarrhea, abdominal pain, or both. Because nearly one third of patients at Hospital B lacked a history of prior antibiotic therapy, it seems that these recommendations were less likely to be followed at Hospital B than at Hospital A. Accordingly, it may not be surprising that the positivity rates for C. difficile toxin were lower at Hospital B than even would have been expected based on the known sensitivities of the EIA method versus the cell cytotoxicity assay (Johnson, 1998, Fekety, 1995, Kelly et al., 1994). Physicians at our hospitals rarely altered their patient management in response to a negative assay for C. difficile toxin. If no therapy was given empirically, the same management almost always was continued after the negative test result was reported. Similarly, if the initial management included empiric therapy for C. difficile, that therapy was continued in approximately two thirds of patients despite the negative test result. We did not determine whether the patients given empiric therapy had improved clinically and, therefore, cannot assess whether this possibility influenced the physician decisions to continue such therapy and ignore the negative toxin assay results. Patients in whom C. difficile toxin assay results were negative rarely had repeat specimens submitted for testing. Published studies have differed on the sensitivity of a single C. difficile toxin assay. Renshaw et al. (1996) reported that the sensitivity of a single cell cytotoxicity assay was 94%, whereas Manabe et al. (1995) reported a sensitivity of only 79% from the first specimen (72% if only results from EIA
assays were counted) and 91% from two specimens. Of perhaps greater concern clinically is the fact that many physicians may not appreciate the frequency with which false negative test results occur (Fekety, 1995). Alternatively, it is possible that repeat tests were not ordered because patients were treated empirically and/or had improved clinically, negating the need for additional testing in the judgment of the treating physician. Of interest and potential importance clinically was the observation that patients in whom empiric therapy for C. difficile was initiated had higher rates of toxin assay positivity than did patients in whom empiric therapy was withheld pending test results. Patients treated empirically were more likely than their untreated counterparts to have fever, an abnormal abdominal exam, and an abnormal leukocyte count. These clinical findings may have influenced the decisions by their physicians to initiate empiric therapy. However, in the logistic regression analysis, the only statistically significant predictors of empiric therapy were hospitalization at Hospital A, prior antibiotic therapy, and, at Hospital B only, advanced age. The results of this study should be interpreted in the context of several of its limitations. A standardized method for assessing severity of illness (e.g., APACHE II score) was not utilized, individual antibiotics given to patients before the development of diarrhea were not specified, and no information was collected on patients who were given empiric therapy for C. difficile diarrhea and whose physicians chose not to obtain a C. difficile toxin test. Moreover, differences in the clinical characteristics of the patients tested and, possibly, the prevalence of C. difficile disease at our two hospitals precluded a clear determination of whether the type of toxin assay used at each institution affected physician behavior. Nevertheless, given an economic environment in which cost-effective utilization of laboratory tests is of major importance, the complex question of how test methods and characteristics (e.g., sensitivity, specificity, and predictive values) influence physician test ordering and therapeutic decisions needs further study. Finally, our results suggest that because physicians electing to treat their patients empirically rarely alter therapy based on toxin assay results, a more cost-effective management strategy may be not to obtain a toxin assay at all in such situations. Further, our observations lend support to the recommendations of Katz and colleagues (Katz et al., 1996; Katz et al., 1997) that testing be limited to hospitalized adults who have received antibiotics within the prior month and who have significant diarrhea and/or abdominal pain.
Acknowledgment The authors thank Dr. Susan Boruchoff for critical review of the manuscript.
Al El-Gammal et al. / Diagnostic Microbiology and Infectious Disease 36 (2000) 169 –173
References Bartlett, J.G. (1994). Clostridium difficile: history of its role as an enteric pathogen and the current state of knowledge about the organism. Clin Infect Dis 18(suppl 4):S265–S272. Crowson, T. W., Rich, E. C., Woolfrey, B. F., & Connelly, D. P. (1984). Overutilization of cultures of cerebrospinal fluid for mycobacteria. JAMA 251:70 –72. Fekety, R. (1995). Antibiotic-associated diarrhea and colitis. Curr Opin Infect Dis 8:391–397. Hines, J. & Nachamkin, I. (1996). Effective use of the clinical microbiology laboratory for diagnosing diarrheal diseases. Clin Infect Dis 23: 1292–1301. Johnson, S. (1998).Gerding DN Clostridium difficile-associated diarrhea. Clin Infect Dis 26:1027–1036. Katz, D. A., Bates, D. W., Rittenberg, E., Onderdonk, A., Sands, K.,
173
Barefoot, L. A., & Snydman, D. (1997). Predicting Clostridium difficile stool cytotoxin results in hospitalized patients with diarrhea. J Gen Intern Med 12:57– 62. Katz, D. A., Lynch, M. E., & Littenberg, B. (1996). Clinical prediction rules to optimize cytotoxin testing for Clostridium difficile in hospitalized patients with diarrhea. Am J Med 100:487– 495. Kelly, C. P., Pothoulakis, C., & LaMont, J. T. (1994). Clostridium difficile colitis. N Engl J Med 330:257–262. Manabe, Y. C., Vinetz, J. M., Moore, R. D., Merz, C., Charache, P., & Bartlett, J. G. (1995). Clostridium difficile colitis: an efficient clinical approach to diagnosis. Ann Intern Med 123:835– 840. Mitty, R. D. & LaMont, T. (1994). Clostridium difficile diarrhea: pathogenesis, epidemiology, and treatment. Gastroenterologist 2:61– 69. Renshaw, A. A., Stelling, J. M., & Doolittle, M.H. (1996). The lack of value of repeated Clostridium difficile cytotoxicity assays. Arch Pathol Lab Med 120:49 –52.
Evaluation of the clinical usefulness of C. difficile toxin testing in hospitalized patients with diarrhea Ayman El-Gammala, Vincenzo Scottoa, S. Malikb, Kathleen C. Caseya,b, Ronald Codyc, David V. Alcida, Melvin P. Weinsteina,d,* a
Division of Allergy, Immunology, and Infectious Diseases, Department of Medicine, University of Medicine and Dentistry of New Jersey—Robert Wood Johnson Medical School (UMDNJ-RWJMS), 1 Robert Wood Johnson Place, New Brunswick, NJ 08903-0019, USA b Department of Medicine, Jersey Shore Medical Center, Neptune, NJ, USA c Department of Environmental and Community Medicine, UMDNJ-RWJMS, New Brunswick, NJ, USA d Department of Pathology and Laboratory Medicine, UMDNJ- RWJMS, New Brunswick, NJ, USA
Abstract Although numerous studies have evaluated the sensitivity and specificity of different assays for Clostridium difficile toxin, none has evaluated how physicians utilize these tests or respond to test results. Therefore, we assessed patient characteristics, clinical findings, and physician responses to positive and negative assay results at two university-affiliated hospitals, one of which used a cell cytotoxicity assay to test for C. difficile toxin and the other of which used an enzyme immunoassay. Two hundred one patient samples at Hospital A and 199 samples at Hospital B were assessed. Positive toxin assays were more frequent at Hospital A than at Hospital B (p ⬍ 0.001), at least in part due to the fact that patients tested at Hospital A were more likely to have fever (p ⬍ 0.001), an abnormal abdominal exam (p ⬍ 0.001), an abnormal leukocyte count (p ⬍ 0.001), and a history of prior antibiotic use (p ⬍ 0.001). Empiric therapy for C. difficile before results of the toxin assay was more common (p ⬍ 0.001) at Hospital A (83/201, 41.3%) than at Hospital B (25/199, 12.5%). Once empiric therapy was started, most physicians continued therapy despite negative test results (Hospital A, 76%; Hospital B, 69%). Patients who were treated empirically were more likely than patients not treated empirically to have positive toxin assay results and to have fever (p ⬍ 0.001), an abnormal abdominal exam (p ⫽ 0.003), or an abnormal leukocyte count (p ⬍ 0.05). Physicians seldom ordered repeat toxin assays (Hospital A, 14%; Hospital B, 10%) if the initial assay result was negative. In logistic regression analysis, predictors of a positive toxin assay were prior antibiotic therapy, an abnormal abdominal exam, residence at Hospital A, and age ⱖ 60 years. Predictors of empiric therapy were residence at Hospital A and prior antibiotic therapy. Because physicians electing to empirically treat inpatients with diarrhea rarely alter therapy based on C. difficile toxin assay results, a more cost-effective management strategy may be not to obtain a toxin assay at all in such situations. Testing should be limited to patients who have received antibiotics within the prior month and who have significant diarrhea and/or abdominal pain. © 2000 Elsevier Science Inc. All rights reserved.
1. Introduction Rising health care costs have focused attention on physician utilization of laboratory tests (Crowson et al., 1984). In patients with diarrhea illnesses, the cost-effectiveness of identifying specific pathogens has not been well studied, and it is not known how physicians use results of such testing (Hines, 1996). Clostridium difficile is the most common infectious etiology of nosocomial diarrhea, accounting for 10 to 15% of all antibiotic-associated diarrhea and causing almost all an* Corresponding author. Tel.: ⫹1-732-235-7713; fax: ⫹1-732-2357951. E-mail address: [email protected] (M.P. Weinstein).
tibiotic-associated colitis (Mitty, 1994). C. difficile associated diarrhea can be mild, but it is more often a serious illness and can have a fatal outcome despite specific antibiotic therapy (Mitty, 1994). Whereas there are several laboratory tests that aid in the diagnosis of C. difficile diarrhea, each method has limitations (Hines, 1996; Johnson, 1998). The toxin neutralization cell cytotoxicity assay for detecting cytotoxin B in stool is considered to be the gold standard laboratory test (Fekety, 1995), but it is relatively expensive and not available in many laboratories because cell culture methodology is required. Results are not available for one to two days and, as the toxin is labile, false negative results may occur because of inactivation of the toxin during transport (Fekety, 1995). In addition, false negative tissue culture results may occur
0732-8893/00/$ – see front matter © 2000 Elsevier Science Inc. All rights reserved. PII: S 0 7 3 2 - 8 8 9 3 ( 9 9 ) 0 0 1 2 9 - 7
170
Al El-Gammal et al. / Diagnostic Microbiology and Infectious Disease 36 (2000) 169 –173
due to excessive dilution of the specimen. Overall, the cell cytotoxicity assay will be positive in 10 to 20% of patients for whom testing is requested, including 15 to 25% of patients with antibiotic-associated diarrhea, 50 to 75% of patients with antibiotic-associated colitis, and 95 to 100% of patients with pseudomembranous colitis (Bartlett, 1994). Enzyme immunoassays (EIA) that detect C. difficile toxin A and/or toxin B may provide results in several hours if testing is done upon receipt in the laboratory rather than in batch mode (e.g., once daily). These assays are technically easy to perform, and are readily available commercially. However, they are less sensitive and specific than the cell cytotoxicity assay (Kelly et al., 1994). Fekety (1995) has reported that false negative results range between 5% and 33%. When hospitalized patients develop diarrhea, physicians may respond in several ways. They may order the toxin assay and initiate empiric therapy pending test results, order the toxin assay and await results before initiating therapy, or treat the patient empirically without ordering the toxin assay. Many studies have evaluated the sensitivity and specificity of different assays for C. difficile, but none has evaluated how physicians utilize the tests or respond to test results. Thus, the primary objective of this study was to assess how physicians utilized tests for C. difficile toxin. We also wished to evaluate physicians’ responses to negative test results. Were tests repeated? Was empiric therapy continued or stopped? We report herein our observations at two university-affiliated hospitals, one of which uses a commercial cell cytotoxicity assay to detect the presence of C. difficile toxin and the other of which uses an enzyme immunoassay.
2. Materials and methods Patients who were hospitalized at either of the two study sites and who had stool specimens sent for C. difficile toxin assays during a 12-month period from March 1997 through February 1998 were eligible for inclusion in the study. Data were collected by infectious disease fellows (AEG,VS) or medical residents (SM) at intervals during the study period when their duty assignments allowed. Hospital A is a 448 bed tertiary care institution that utilizes a commercially available cell cytotoxicity assay (Bartels, Issaquah, WA, USA) to detect C. difficile toxin (reported sensitivity and specificity 100% compared with conventional cell culture using W138 cells). This institution is a university medical center with house officers in all disciplines and subspecialty fellows in most disciplines; full-time academic faculty admit approximately half of the patients. Hospital B is a 476 bed community teaching hospital that utilizes an EIA (C. difficile Tox-A Test, Wampole Laboratories, Cranbury, NJ, USA) for detection of C. difficile toxin (reported sensitivity 80.7% and specificity 99.2% compared with cell culture [cell line not stated]). This hospital is a university-affiliated
Table 1 Positive C. difficile toxin assays at Hospitals A and B Hospital Aa
All tests First test only a b
Hospital Bb
p
No.
No. pos.
%
No.
No. pos.
%
201 173
42 37
20.9 21.4
199 170
15 10
7.5 5.9
⬍0.001 ⬍0.001
Cell cytotoxicity assay used. Enzyme immunoassay used.
teaching institution with residency programs in medicine, surgery, obstetrics and gynecology, and pediatrics; community physicians admit approximately 90% of all patients. The medical staffs at the two hospitals do not overlap. Patient charts were reviewed retrospectively, and data were recorded on a standardized study worksheet. The variables assessed included patient demographics, clinical findings, and physician specialty and responses to test results. Demographic data included patient age, sex, date of hospital admission, date of testing, and place of hospitalization. Clinical parameters recorded included history of antibiotic therapy in the previous 6 weeks, character of stool including presence of blood, presence of fever (ⱖ38°C), abnormal peripheral leukocyte count (⬍4000/mm3 or ⬎ 11,000/mm3), abnormal abdominal examination (tenderness, distention, hyperactive bowel sounds, or ileus), and results of C. difficile toxin assay. Physician behaviors assessed included use of empiric therapy for C. difficile before knowledge of test results and use of such therapy if test results were reported to be negative. An episode was defined by the first specimen obtained for C. difficile toxin assay or by any subsequent specimen submitted for testing ⬎7 days after the preceding specimen. If a patient had more than one episode, only the first was included for study purposes. Data were recorded on study worksheets, then tabulated in a computerized database program (Excel, Microsoft Inc, Redmond, WA, USA). Subsequently, the data were transferred to the SAS statistical analysis program (SAS, Inc., Cary, NC, USA) for significance testing and logistic regression analyses.
3. Results Two hundred one samples from 177 patients at Hospital A and 199 samples from 187 patients at Hospital B were analyzed. There were no significant differences in the age or sex distribution of patients from whom samples were obtained at the two hospitals (data not shown). Positive C. difficile toxin assays were more frequent at Hospital A than at Hospital B (Table 1) whether all tests or only the first test on any given patient was counted (p ⬍ 0.001). These differences seemed to be related in part to the clinical characteristics of the patients from whom samples
Al El-Gammal et al. / Diagnostic Microbiology and Infectious Disease 36 (2000) 169 –173
171
Table 2 Clinical characteristics of patients at Hospitals A and B who had stool specimens sent for C. difficile toxin detection
Table 3 C. difficile toxin assay positivity rates and use of empiric therapy for C. difficile diarrhea at Hospitals A and B
Clinical characteristics
Hospital A No. pos.
%
No.
No. pos.
Empiric therapy
Hospital A
No.
No. pos./total
%
No. pos./total
%
Fever Abnormal abdominal exam Bloody diarrhea Abnormal leukocyte count Prior antibiotic therapy
201 200
135 77
67.2 38.5
199 198
31 37
15.6 18.7
⬍0.001 ⬍0.001
Given Not given
22/83 20/118
26.5 16.9
4/25 11/174
16.0 6.3
201 194
16 117
8.0 60.3
74 186
5 92
6.8 49.9
NSa ⬍0.001
201
180
90.0
199
141
70.8
⬍0.001
a
Hospital B
p %
NS, p ⬎ 0.05.
were obtained for testing (Table 2). Patients at Hospital A were significantly more likely than patients at Hospital B to have fever (p ⬍ 0.001), an abnormal abdominal exam (p ⬍ 0.001), an abnormal leukocyte count (p ⬍ 0.001), and a history of prior antibiotic therapy (p ⬍ 0.001). There was no significant difference in the presence of bloody diarrhea in patients from the two hospitals (Table 2). The great majority of physicians who ordered C. difficile toxin assays at both hospitals were general internists, internal medicine subspecialists, or surgeons. At both hospitals, approximately 45% of tests were ordered by internal medicine subspecialists. However, more tests were ordered by surgeons at Hospital A (24.9% vs. 10.1% at Hospital B), whereas more tests were ordered by general internists at Hospital B (44.4% vs. 30.8% at Hospital A). These differences were statistically significant (p ⬍ 0.001). Initiation of empiric therapy for C. difficile diarrhea before receiving test results was significantly more common (p ⬍ 0.001) at Hospital A (83/201, 41.3%) than at Hospital B (25/199, 12.5%). Moreover, negative C. difficile toxin assay results had minimal influence on physicians who administered empiric therapy. At Hospital B, empiric therapy was continued in 16/21 such patients (76.2%) whereas at Hospital A empiric therapy was continued in 42/61 such patients (68.8%). The study design did not allow for determination of whether patients for whom empiric therapy was continued despite a negative assay had improved clinically on such therapy. At both hospitals, patients in whom empiric therapy was ordered had higher rates of C. difficile toxin detection than did patients not started on empiric therapy (Table 3), suggesting that physicians may have recognized that such patients were more ill and, therefore, were more inclined to begin empiric therapy. Indeed, further analysis revealed that patients who were given empiric therapy were significantly more likely to have fever (p ⬍ 0.001), an abnormal abdominal examination (p ⫽ 0.003), or an abnormal leukocyte count (p ⬍ 0.05). Physicians who did not initiate empiric therapy before C.
Hospital B
difficile toxin test results rarely began such therapy once the test results were reported to be negative. At Hospital B, only one of 163 such patients (0.6%) was started on empiric therapy; at Hospital A, 7 of 98 such patients (7.8%) were given empiric therapy. Repeat ordering of C. difficile toxin assays was uncommon in patients with a single negative test result. At Hospital A, repeat tests were sent on 14 such patients (14.3%), and at Hospital B repeat tests were ordered on 16 patients (9.8%). Logistic regression was used to determine predictors of positive C. difficile toxin assay results and use of empiric therapy. Table 4 summarizes the results of these analyses. In the full data set from both hospitals, significant predictors of a positive toxin assay were (in order of highest odds ratio) prior antibiotic administration, abnormal abdominal exam, hospitalization at Hospital A, and age ⱖ 60 years. When the data were analyzed for Hospital B only, none of the variables were significant predictors of a positive toxin assay, whereas at Hospital A both age ⱖ 60 years and an abnormal abdominal exam predicted positive test results. Significant predictors for the use of empiric therapy in the full data set were hospitalization at Hospital A and prior antimicrobial therapy. When the data were analyzed for Hospital B only, age ⱖ 60 was the only significant predictor, whereas at Table 4 Significant predictors of positive C. difficile toxin assay results and use of empiric therapy based on logistic regression analysis
Positive toxin assay
Data Set
Predictor
Odds ratio
95% Confidence interval
Full data set
Prior antibiotics Abn. abdominal exam Age ⱖ60 years Hospital A Abnormal abdominal exam Age ⱖ60 years N/Aa
3.9 3.5
1.1–13.7 1.8–6.5
2.1 2.1 4.9
1.1–4.2 1.1–4.2 2.3–10.4
2.6 N/Aa
1.2–5.8
Hospital A Prior antibiotics Prior antibiotics Age ⱖ60 years
4.2 2.8 3.2 3.0
2.5–7.0 1.3–6.2 1.0–10.1 1.0–9.3
Hospital A only
Hospital B only Empiric therapy
None Full data set Hospital A only Hospital B only
a
N/A ⫽ Information not available.
172
Al El-Gammal et al. / Diagnostic Microbiology and Infectious Disease 36 (2000) 169 –173
Hospital A, prior antimicrobial therapy was the lone significant predictor.
4. Discussion Although there have been numerous studies describing clinical aspects of C. difficile diarrhea and evaluating various C. difficile toxin assays, few if any have assessed physician utilization of toxin assays themselves or the use of empiric therapy in patients with suspected C. difficile infection. Therefore, we chose to examine this issue and to determine if possible whether physician behavior was affected by the type of toxin assay used in the two hospitals. Our observations suggest that at Hospital B toxin assays were ordered on patients who were less ill than their counterparts at Hospital A and that this finding accounted, at least in part, for the lower positivity rate for toxin assays at Hospital B. As indicated in Table 2, patients at Hospital B were significantly less likely to have fever, an abnormal abdominal exam, an abnormal leukocyte count, and a history of prior antimicrobial therapy. A clinical prediction rule for C. difficile toxin testing, recently published and validated (Katz et al., 1996; Katz et al., 1997), states that testing should be done only on hospitalized adults who have received antimicrobial therapy within the past month and who have either significant diarrhea, abdominal pain, or both. Because nearly one third of patients at Hospital B lacked a history of prior antibiotic therapy, it seems that these recommendations were less likely to be followed at Hospital B than at Hospital A. Accordingly, it may not be surprising that the positivity rates for C. difficile toxin were lower at Hospital B than even would have been expected based on the known sensitivities of the EIA method versus the cell cytotoxicity assay (Johnson, 1998, Fekety, 1995, Kelly et al., 1994). Physicians at our hospitals rarely altered their patient management in response to a negative assay for C. difficile toxin. If no therapy was given empirically, the same management almost always was continued after the negative test result was reported. Similarly, if the initial management included empiric therapy for C. difficile, that therapy was continued in approximately two thirds of patients despite the negative test result. We did not determine whether the patients given empiric therapy had improved clinically and, therefore, cannot assess whether this possibility influenced the physician decisions to continue such therapy and ignore the negative toxin assay results. Patients in whom C. difficile toxin assay results were negative rarely had repeat specimens submitted for testing. Published studies have differed on the sensitivity of a single C. difficile toxin assay. Renshaw et al. (1996) reported that the sensitivity of a single cell cytotoxicity assay was 94%, whereas Manabe et al. (1995) reported a sensitivity of only 79% from the first specimen (72% if only results from EIA
assays were counted) and 91% from two specimens. Of perhaps greater concern clinically is the fact that many physicians may not appreciate the frequency with which false negative test results occur (Fekety, 1995). Alternatively, it is possible that repeat tests were not ordered because patients were treated empirically and/or had improved clinically, negating the need for additional testing in the judgment of the treating physician. Of interest and potential importance clinically was the observation that patients in whom empiric therapy for C. difficile was initiated had higher rates of toxin assay positivity than did patients in whom empiric therapy was withheld pending test results. Patients treated empirically were more likely than their untreated counterparts to have fever, an abnormal abdominal exam, and an abnormal leukocyte count. These clinical findings may have influenced the decisions by their physicians to initiate empiric therapy. However, in the logistic regression analysis, the only statistically significant predictors of empiric therapy were hospitalization at Hospital A, prior antibiotic therapy, and, at Hospital B only, advanced age. The results of this study should be interpreted in the context of several of its limitations. A standardized method for assessing severity of illness (e.g., APACHE II score) was not utilized, individual antibiotics given to patients before the development of diarrhea were not specified, and no information was collected on patients who were given empiric therapy for C. difficile diarrhea and whose physicians chose not to obtain a C. difficile toxin test. Moreover, differences in the clinical characteristics of the patients tested and, possibly, the prevalence of C. difficile disease at our two hospitals precluded a clear determination of whether the type of toxin assay used at each institution affected physician behavior. Nevertheless, given an economic environment in which cost-effective utilization of laboratory tests is of major importance, the complex question of how test methods and characteristics (e.g., sensitivity, specificity, and predictive values) influence physician test ordering and therapeutic decisions needs further study. Finally, our results suggest that because physicians electing to treat their patients empirically rarely alter therapy based on toxin assay results, a more cost-effective management strategy may be not to obtain a toxin assay at all in such situations. Further, our observations lend support to the recommendations of Katz and colleagues (Katz et al., 1996; Katz et al., 1997) that testing be limited to hospitalized adults who have received antibiotics within the prior month and who have significant diarrhea and/or abdominal pain.
Acknowledgment The authors thank Dr. Susan Boruchoff for critical review of the manuscript.
Al El-Gammal et al. / Diagnostic Microbiology and Infectious Disease 36 (2000) 169 –173
References Bartlett, J.G. (1994). Clostridium difficile: history of its role as an enteric pathogen and the current state of knowledge about the organism. Clin Infect Dis 18(suppl 4):S265–S272. Crowson, T. W., Rich, E. C., Woolfrey, B. F., & Connelly, D. P. (1984). Overutilization of cultures of cerebrospinal fluid for mycobacteria. JAMA 251:70 –72. Fekety, R. (1995). Antibiotic-associated diarrhea and colitis. Curr Opin Infect Dis 8:391–397. Hines, J. & Nachamkin, I. (1996). Effective use of the clinical microbiology laboratory for diagnosing diarrheal diseases. Clin Infect Dis 23: 1292–1301. Johnson, S. (1998).Gerding DN Clostridium difficile-associated diarrhea. Clin Infect Dis 26:1027–1036. Katz, D. A., Bates, D. W., Rittenberg, E., Onderdonk, A., Sands, K.,
173
Barefoot, L. A., & Snydman, D. (1997). Predicting Clostridium difficile stool cytotoxin results in hospitalized patients with diarrhea. J Gen Intern Med 12:57– 62. Katz, D. A., Lynch, M. E., & Littenberg, B. (1996). Clinical prediction rules to optimize cytotoxin testing for Clostridium difficile in hospitalized patients with diarrhea. Am J Med 100:487– 495. Kelly, C. P., Pothoulakis, C., & LaMont, J. T. (1994). Clostridium difficile colitis. N Engl J Med 330:257–262. Manabe, Y. C., Vinetz, J. M., Moore, R. D., Merz, C., Charache, P., & Bartlett, J. G. (1995). Clostridium difficile colitis: an efficient clinical approach to diagnosis. Ann Intern Med 123:835– 840. Mitty, R. D. & LaMont, T. (1994). Clostridium difficile diarrhea: pathogenesis, epidemiology, and treatment. Gastroenterologist 2:61– 69. Renshaw, A. A., Stelling, J. M., & Doolittle, M.H. (1996). The lack of value of repeated Clostridium difficile cytotoxicity assays. Arch Pathol Lab Med 120:49 –52.