Laboratory testing of stool samples for toxigenic Clostridium

Transcription

1 Comparison of Analytical and Clinical Performance of Three Methods for Detection of Clostridium difficile P. Rocco LaSala, MD; Annika M. Svensson, MD, PhD; Amin A. Mohammad, PhD; Peter L. Perrotta, MD N Context. Diagnostic laboratory testing for Clostridium difficile infection has undergone considerable and rapid evolution during the last decade. The ideal detection method(s), which should exhibit high analytical and clinical sensitivity and specificity, remains undefined. Objective. We sought to evaluate the analytical and clinical performance characteristics of three methods for the laboratory detection of C difficile. Design. This study used 114 consecutive stool samples to compare three methods of C difficile detection: an enzyme immunoassay (EIA) for toxins A/B, a lateral flow membrane immunoassay for glutamate dehydrogenase (GDH), and a qualitative real-time polymerase chain reaction (PCR) assay. Medical records of all patients having $1 positive test result were reviewed to estimate the clinical likelihood of C difficile infection. Results. Based upon laboratory result consensus values, analytical sensitivity was significantly higher for GDH (94%) and PCR (94%) assays than for toxin EIA (25%). Analytical specificity was significantly higher for PCR (100%) and EIA (100%) than for GDH assay (93%). In contrast, assay performance based upon clinical probability of C difficile infection suggested lower discriminatory power (ie, clinical specificity) of the more analytically sensitive methods. Conclusions. Higher rates of C difficile detection will be realized upon implementation of GDH assay and/or real-time PCR based testing algorithms than by testing with EIA alone. Further study is required to elucidate potential downstream costs for higher detection rates. (Arch Pathol Lab Med. 2012;136: ; doi: / arpa oa) Laboratory testing of stool samples for toxigenic Clostridium difficile infection (CDI) has evolved considerably during the past 3 decades. Cytotoxicity to fibroblast monolayers, a key feature that helped to implicate the organism as an etiological factor in colitis, remained the in vitro diagnostic gold standard for many years. 1 Due largely to its complexity and long turnaround time, the cytotoxicity assay was superseded by various toxin enzyme immunoassay (EIA) formulations. While those assays demonstrate favorable performance characteristics compared to one another or to the cytotoxicity assay, 2 recent evidence suggests their sensitivity is unacceptably low compared to techniques such as anaerobic toxigenic culture assay. 3 As a result, alternative diagnostic methods have been proposed, including the use of toxigenic culture itself, two-step algorithms incorporating glutamate dehydrogenase (GDH) common antigen assays, and more recently, nucleic acid amplification tests (eg, polymerase chain reaction [PCR]). The purpose of this study was to evaluate in parallel a few of these methods (ie, EIA, real-time PCR, and GDH assay) in order to Accepted for publication July 26, From the Clinical Laboratories, Department of Pathology, Robert C. Byrd Health Science Center, West Virginia University Hospitals, Morgantown. Dr Mohammad is currently with The Texas A&M University System Health Science Center, Temple. The authors have no relevant financial interest in the products or companies described in this article. Reprints: P. Rocco LaSala, MD, Department of Pathology, West Virginia University, WVUH Clinical Labs, Box 8009, One Medical Center Dr, Morgantown, WV ( plasala@hsc.wvu.edu). assess their analytical and logistical performances and the potential impact of each method on patient management at a 500-bed regional academic medical center. MATERIALS AND METHODS Between February 18, 2010, and March 19, 2010, consecutive, nonformed stool samples submitted Monday through Friday to West Virginia University Hospital (WVUH) Clinical Laboratories, Morgantown, West Virginia, for C difficile toxin testing by routine EIA (Wampole C difficile Tox A/B; Inverness Medical Diagnostics, Princeton, New Jersey) were also tested in parallel with the GeneXpert C difficile real-time PCR assay (Cepheid Diagnostics, Sunnyvale, California) and a rapid lateral flow immunochromatographic GDH assay (C difficile Quik Chek Complete; Techlab, Blacksburg, VA). All three assays were performed according to the manufacturers instructions and included appropriate controls. All discrepant stool samples with sufficient residual material maintained at 220uC were sent for supplemental testing at two outside laboratories using different commercially available PCR assays (GeneXpert C difficile/epi assay [Cepheid Diagnostics, Sunnyvale, California] and Gene- Ohm Cdiff assay [BD Diagnostics, Franklin Lakes, New Jersey]). Control strains for all testing included toxigenic C difficile American Type Culture Collection (ATCC) strain , Clostridium perfringens ATCC strain 13124, and Clostridium sordelli ATCC strain True positive results were defined as those found to be positive by two or more in-house or reference laboratory molecular assays. Following approval by local institutional review board, a retrospective review of electronic medical records was performed by a single author (P.R.L.) who was blinded to study findings, with the exception of EIA results, which had already been reported in patients charts. The following parameters were assessed for purposes of CDI categorization: (1) exposure to Arch Pathol Lab Med Vol 136, May 2012 Laboratory Detection of C difficile LaSala et al 527

2 Results distribution is shown for 24 positive stool samples submitted for C difficile testing by polymerase chain reaction (PCR) (GeneXpert; Cepheid Diagnostics, Sunnyvale, California), glutamate dehydrogenase (GDH) common antigen (Quik Chek Complete; Techlab, Blacksburg, Virginia), and toxin enzyme immunoassay (EIA) (Wampole C difficile Tox A/B; Inverness Medical Diagnostics, Princeton, New Jersey). antibiotics within 6 weeks preceding testing; (2) documented diarrhea (of any quantity or duration) within 7 days of specimen collection; (3) leukocytosis ( white blood cells [WBC]/mL) within 1 day of specimen collection; (4) a history of CDI (with or without laboratory confirmation); (5) prior or subsequent treatment for CDI; and (6) $2 toxin EIA tests ordered on different days. Probability of CDI was defined as unlikely, possible, or likely with fulfillment of #1, 2, or $3 of these parameters, respectively; if $2 parameters could not be determined from available records, probability of CDI was categorized as unknown. RESULTS During the study period, WVUH clinical laboratories received and processed 114 consecutive unformed stool samples that were included for analysis. Among the total, 87 samples were found to be negative by all three assays. Three additional samples were unresolved by PCR (ie, duplicate invalid results) and negative by both of the other methods. The remaining 24 samples demonstrated $1 positive test result (Figure). Only 3 of 24 samples (12%) were positive by all three methods. The GDH assay detected analyte in 22 of 24 samples, of which 7 samples (33%) were positive only by GDH assay. The PCR assay detected analyte in 16 of 24 samples, of which only 2 samples (12%) were positive exclusively by PCR. The toxin EIA detected analyte in only 4 of 24 samples, none of which was positive by EIA alone. For discrepant samples, residual frozen material was forwarded to two reference laboratories for supplementary PCR testing, which confirmed initial GeneXpert PCR results in all specimens tested. Under the assumption that samples unavailable for referral testing would have been similarly confirmed, and excluding 1 sample found positive only by in-house PCR, true positive and true negative values were assigned based on assay consensus results (ie, $2 tests had positive results) (Table 1). Using these data, sensitivity, specificity, and positive and negative predictive values (PPVs and NPVs, respectively) were calculated for each detection method (Table 2). A chart review was conducted for all 22 patients who had $1 positive test result (1 patient had 3 samples with positive results) (Table 3). The number of positive results was the same for both males and females. Mean age was 50 years old, and the cohort included 4 pediatric-aged patients, 2 of whom were less than 1 year of age. Less than half the patients with available WBC counts had evidence of leukocytosis, and nearly one-fourth (all outpatients) had no concomitant orders for blood counts. Of 18 patients with available treatment records, all had prior exposure to antibiotics within the preceding 42 days. Only 7 of 18 patients received directed therapy for CDI based on EIA results and/or clinical impression, while an additional 4 patients received intravenous vancomycin and/or metronidazole therapy for another indication (eg, intraabdominal abscess). More than one-third (7 of 18) of patients with available treatment records received no directed or incidental therapy for C difficile activity (ie, vancomycin or metronidazole). Clinical and laboratory parameters were then compared among patient subsets according to the estimated likelihood of clinical CDI (Table 4). As expected, features supporting a diagnosis of CDI were more prevalent among those patients with a high likelihood of diagnosis than those with a possible or unlikely diagnosis. Analysis of individual or collective C difficile laboratory results revealed little appreciable distinction among the three diagnostic likelihood categories (Table 4). That is, while patients with high disease probability had results that were uniformly or frequently positive by GDH and PCR assays (suggesting high clinical sensitivity), results from these same methods were also positive in a large proportion of patients with lower estimated likelihood of CDI, based on our clinical model (suggesting low clinical specificity). Although EIA demonstrated good clinical specificity (ie, few positive results among patients with low likelihood of disease), it was not especially useful for detecting patients with high disease probability (suggesting low clinical sensitivity). Furthermore, true positive results, as defined by the laboratory test consensus standard ($2 tests with positive results), showed little association with any particular diagnostic likelihood group (Table 4). COMMENT It has recently become apparent that widespread use of EIA by clinical laboratories for detecting C difficile exotoxin(s) may be insufficient for establishing a diagnosis of CDI. Several reports have documented poor analytical sensitivity of toxin immunoassays in comparison to toxigenic culture and/or various PCR-based assays The present study supports those findings. Coupled with the fact that many laboratories still use the relatively low-cost toxin EIA as the primary method for laboratory diagnosis, 11,12 medical providers may frequently face the decision whether to initiate therapy in the absence of laboratory confirmation or to repeat laboratory tests. The former choice is recommended only for cases of severe or complicated disease 13 and is not consistent with recent case definitions. 14,15 The latter approach is hampered by 528 Arch Pathol Lab Med Vol 136, May 2012 Laboratory Detection of C difficile LaSala et al

3 Table 1. Results by Assay of 24 Specimens Having $1 Positive Result for C difficile Sample No. GDH PCR Toxin EIA Reference Laboratory Result a True Positive b Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y N N N N N N N ND c Y Abbreviations: EIA, enzyme immunoassay; GDH, glutamate dehydrogenase; ND, not determined; PCR, polymerase chain reaction. a Reference laboratory methods included GeneXpert C difficile/epi assay (Cepheid Diagnostics, Sunnyvale, California) and GeneOhm Cdiff assay (BD Diagnostics, Franklin Lakes, New Jersey). Testing was performed with stool aliquots stored and shipped at 220uC. b As determined by laboratory test result consensus. c Sample was unavailable for supplemental testing; result status was not determined. successive decreases in PPV with each sequential test, as illustrated by Peterson and Robicsek. 16 Improvement of immunoassay formats that detect C difficile-specific GDH marked a potential diagnostic advancement with minimal increases in assay cost compared to that of toxin EIAs. Ticehurst and colleagues 17 were the first to report high NPVs for that test, and many studies have upheld their conclusion that it can sufficiently exclude carriage of C difficile Clinical specificity and PPV of the GDH assay are negatively impacted, however, by the test s inherent reactivity against both toxigenic and nontoxigenic strains, as corroborated by the present study (Tables 2 and 4). Hence, GDH-positive samples require follow-up testing to determine whether they harbor organisms capable of toxin production. The need for this second-step test to exhibit reasonable sensitivity and turnaround time, in addition to specificity, restricts the utility of toxigenic culture, toxin EIA, and cell cytotoxicity assays in this role. Recent investigations have focused on real-time PCR as a candidate for GDH screen confirmation. Using the same GDH assay described here, Sharp et al 21 and Swindells et al 8 reported excellent performance characteristics of a two-step GDH-PCR algorithm. They observed,100% sensitivity of the GDH screen component compared with toxigenic culture and/or PCR, as well as excellent correlation between second-step PCR testing and toxigenic culture results. Others have noted somewhat lower sensitivity (,85% 90%) for GDH screening assays than for universal PCR and/or toxigenic culture. 4,5,22 Interestingly, one report highlighted different performance characteristics for GDH algorithms depending upon bacterial ribotype. 9 It remains unclear whether differences in published values of GDH sensitivity are due to regional strain variation, variable performance characteristics of molecular assays, differences in reference comparator methods (eg, culture techniques), or perhaps other causes; and debate persists regarding the benefits of two-step algorithms incorporating GDH assay (eg, offering significantly decreased costs) versus universal PCR testing. 23 The present study substantiates the reported high sensitivity values for GDH (94%) and PCR (94%), as well as their excellent NPVs (99%). The present study further demonstrates that a substantial increase in the rate of C difficile detection would be realized upon implementation of universal PCR or a GDH-PCR algorithm in comparison to testing by toxin EIA alone. This investigation also incorporated a retrospective clinical assessment that revealed several interesting findings. Foremost among the findings was the absence of any documentation of diarrhea in more than one-third of patients whose records were accessible. In their prospective study, Peterson and colleagues 15 found a similarly high proportion (39%) of potential CDI episodes that did not meet criteria for diarrhea (ie, $3 loose stools in 1 day). The primary limitation of the current approach was its retrospective nature, inasmuch as the absence of documented diarrhea cannot be interpreted definitively as a documented absence; yet the lack of mention altogether in such a large proportion of cases is noteworthy given the fact that diarrhea is the most common underlying feature of CDI. 1 Assuming that our clinical model is accurate despite other limitations (eg, inherent bias), inclusion of patients with low pretest probability could potentially account for the poor correlation observed between analytical and clinical performances of the laboratory assays. Table 4 suggests that assays with high clinical Table 2. Performance Characteristics of Three Laboratory Assays for Detection of C difficile Using Assay Consensus as the Reference Method Characteristic GDH Toxin EIA PCR a % of sensitivity (95% CI) 94 (70 99) b 25 (10 50) b 94 (70 99) b % of specificity (95% CI) 93 (87 97) c 100 (95 100) c 100 (95 100) c NPV % (95% CI) 99 (94 100) 89 (82 94) 99 (94 100) PPV % (95% CI) 69 (47 84) 100 (45 100) 100 (77 100) Abbreviations: CI, confidence interval; EIA, enzyme immunoassay; GDH, glutamate dehydrogenase; NPV, negative predictive value; PCR, polymerase chain reaction; PPV, positive predictive value. a Calculations excluded 3 indeterminate PCR results. b Significant differences (P,.001) were found between GDH and EIA and between PCR and EIA results. c Significant differences (P 5.01) were found between EIA and GDH and between PCR and GDH assay results. Arch Pathol Lab Med Vol 136, May 2012 Laboratory Detection of C difficile LaSala et al 529

4 Table 3. Overall Clinical Features of 22 Patients Testing Positive by $1 Laboratory Assay for Detection of C difficile Parameter Number % of total Male:Female ratio 11:11 Patient age (years) # Patients with history of antibiotic 18/18 patients a 100 exposure Described nature of diarrhea Severe and/or persistent 9 41 Documented but not categorized 4 18 Not recorded in medical record 7 32 Outpatient clinic note not 2 9 available WBC count (WBC/mL) Count not performed 5 23 Mean number of toxin EIAs 2 (range, 1 5) History of CDI 2 9 Treatment for CDI Targeted oral therapy 7 32 Nontargeted intravenous therapy 4 18 Unknown a 4 18 None 7 32 Probability of CDI Likely 9 41 Possible 5 23 Unlikely 4 18 Unknown 4 18 Abbreviations: CDI, C difficile infection; EIA, enzyme immunoassay; WBC, white blood cell. a Four outpatients had insufficient antibiotic prescription data to determine past exposure or treatment for CDI. sensitivity (ie, few negative results among patients with high likelihood of disease) may exhibit reduced clinical specificity (ie, many positive results among patients with low likelihood of disease). Conversely, high clinical specificity seemed to be achieved at the cost of reduced clinical sensitivity. Interestingly, among the 18 patients in this study with available treatment records and $1 positive C difficile test result, 7 patients received no therapy for CDI, including 1 likely and 3 probable cases (based on likelihood estimates) and 3 true positives (based upon laboratory consensus results). Given that antibiotic-associated diarrhea may subside following discontinuation of the inciting antibiotic alone, 24 this observation may not be especially remarkable. Had a more analytically sensitive assay been in place during the study period, however, many of these patients would likely have received directed antimicrobial therapy for CDI. Further investigation of the direct and indirect economic, epidemiological, and clinical consequences of higher detection rates and resulting impact on therapeutic interventions is warranted. Expansion of our observations, particularly under circumstances where pretest probabilities can be firmly assessed, will be required to determine the overall benefits and costs of implementing more analytically sensitive assays for C difficile detection. Chart review also highlighted instances of potential misuse of laboratory resources. Two patients with a history of CDI and prior positive laboratory results reported 6 and 19 days earlier, respectively, had $1 positive test result. Recent consensus reports point out that a test for cure is not indicated 13 and that an additional positive [laboratory] result... performed on a specimen collected #2 weeks after the last specimen that tested positive represents continuation of the same CDI case. 14 Additionally, 2 patients with positive results were noted to be,1 year of age. Given the high rates of C difficile colonization documented during infancy and early childhood, 25 the clinical specificity of CDI diagnostic assays applied to this population has been called into question. 26 Table 4. Clinical and Laboratory Features of 22 Patients with $1 Laboratory Test Result Positive for C difficile Listed by CDI Probability Parameter No. of Patients in the Likelihood Group/Total Number of Patients a Likely (n = 9) Possible (n = 5) Unlikely (n = 4) Unknown (n = 4) Clinical feature Documented diarrhea 9/9 1/4 0/4 2/2 Leukocytosis 6/8 2/4 0/4 0/1 Treated for CDI 6/7 2/5 b 2/4 b 1/2 History of CDI 2/9 0/5 0/4 0/4 Mean no. of EIAs performed Positive test result GDH 9/9 4/5 4/4 3/4 Toxin EIA 3/9 0/5 0/4 1/4 PCR 7/9 2/5 4/4 1/4 Combined laboratory results 3 positive test results 3/9 0/5 0/4 0/4 $2 positive test results 7/9 1/5 4/4 1/4 Abbreviations: CDI, C difficile infection; EIA, enzyme immunoassay; GDH, glutamate dehydrogenase; PCR, polymerase chain reaction. a For some clinical parameters, the sum of denominators does not total 22 patients because data were unavailable in some patient records. b Four patients who received nondirected therapy. 530 Arch Pathol Lab Med Vol 136, May 2012 Laboratory Detection of C difficile LaSala et al

5 This study suggests that both GeneXpert C difficile PCR and the Quik Chek GDH immunoassay demonstrate acceptably high analytical and clinical sensitivity values for diagnosis of CDI at our institution. Analytical specificity of PCR is also high, suggesting that it could serve as an acceptable confirmatory test for positive GDH screening. However, the clinical specificity of both of these assays may be reduced, particularly in scenarios where pretest probability is low and/or asymptomatic colonization is high. Finally, diagnostic reliance on toxin EIA appears to be limited by its poor sensitivity. References 1. Bartlett JG, Gerding DN. Clinical recognition and diagnosis of Clostridium difficile infection. Clin Infect Dis. 2008;46(suppl 1):S12 S Musher DM, Manhas A, Jain P, et al. Detection of Clostridium difficile toxin: comparison of enzyme immunoassay results with results obtained by cytotoxicity assay. J Clin Microbiol. 2007;45(8): Crobach MJ, Dekkers OM, Wilcox MH, Kuijper MJ. European Society of Clinical Microbiology and Infectious Diseases (ESCMID): data review and recommendations for diagnosing Clostridium difficile-infection (CDI). Clin Microbiol Infect. 2009;15(12): Eastwood K, Else P, Charlett A, Wilcox M. Comparison of nine commercially available Clostridium difficile toxin detection assays, a real-time PCR assay for C. difficile tcdb, and a glutamate dehydrogenase detection assay to cytotoxin testing and cytotoxigenic culture methods. J Clin Microbiol. 2009; 47(10): Larson AM, Fung A, Fang FC. Evaluation of tcdb real-time PCR in a threestep diagnostic algorithm for detection of toxigenic Clostridium difficile. J Clin Microbiol. 2010;48(1): Sloan LM, Duresko BJ, Gustafson DR, Rosenblatt JE. Comparison of realtime PCR for detection of the tcdc gene with four toxin immunoassays and culture in diagnosis of Clostridium difficile infection. J Clin Microbiol. 2008; 46(6): Stamper PD, Babiker W, Alcabasa R, et al. Comparison of a commercial real-time PCR assay for tcdb detection to a cell culture cytotoxicity assay and toxigenic culture for direct detection of toxin-producing Clostridium difficile in clinical samples. J Clin Microbiol. 2009;47(2): Swindells J, Brenwald N, Reading N, Oppenheim B. Evaluation of diagnostic tests for Clostridium difficile infection. J Clin Microbiol. 2010;48(2): Tenover FC, Novak-Weekley S, Woods CW, et al. Impact of strain types on detection of toxigenic Clostridium difficile: comparison of molecular diagnostic and enzyme immunoassay approaches. J Clin Microbiol. 2010;48(10): Wren MWD, Sivapalan M, Kinosn R, Shetty NP. Laboratory diagnosis of Clostridium difficile infection. An evaluation of tests for faecal toxin, glutamate dehydrogenase, lactoferrin and toxigenic culture in the diagnostic laboratory. Br J Biomed Sci. 2009;66(1): Cooke RPD, Collins J, Galloway A, Holland D, Trigg G. Audit of the laboratory diagnosis of Clostridium difficile infection in the UK. J Hosp Infect. 2010;75(4): Microbiology Resource Committee, Proficiency Test Survey, Bacteriology D-A, Northfield, IL: College of American Pathologists; Cohen SH, Gerding DN, Johnson S, et al. Clinical practice guidelines for Clostridium difficile infection in adults: 2010 update by the Society for Healthcare Epidemiology of America (SHEA) and the Infectious Diseases Society of America (IDSA). Infect Control Hosp Epidemiol. 2010;31(5): McDonald LC, Coignard B, Dubberke E, et al. Recommendations for surveillance of Clostridium difficile-associated disease. Infect Control Hosp Epidemiol. 2007;28(2): Peterson LR, Manson RU, Paule SM, et al. Detection of toxigenic Clostridium difficile in stool samples by real-time polymerase chain reaction for the diagnosis of C. difficile-associated diarrhea. Clin Infect Dis. 2007;45(9): Peterson LR, Robicsek A. Does my patient have Clostridium difficile infection? Ann Intern Med. 2009;151(3): Ticehurst JR, Aird DZ, Dam LM, Borek AP, Hargrove JT, Carroll KC. Effective detection of toxigenic Clostridium difficile by a two-step algorithm including tests for antigen and cytotoxin. J Clin Microbiol. 2006;44(3): Gilligan PH. Is a two-step glutamate dehyrogenase antigen-cytotoxicity neutralization assay algorithm superior to the premier toxin A and B enzyme immunoassay for laboratory detection of Clostridium difficile? J Clin Microbiol. 2008;46(4): Quinn CD, Sefers SE, Babiker W, et al. Diff Quik Chek complete enzyme immunoassay provides a reliable first-line method for detection of Clostridium difficile in stool specimens. J Clin Microbiol. 2010;48(2): Reller ME, Lema CA, Perl TM, et al. Yield of stool culture with isolate toxin testing versus a two-step algorithm including stool toxin testing for detection of toxigenic Clostridium difficile. J Clin Microbiol. 2007;45(11): Sharp SE, Ruden LO, Pohl JC, Hatcher PA, Jayne LM, Ivie WM. Evaluation of the C.Diff Quik Chek Complete Assay, a new glutamate dehydrogenase and A/B toxin combination lateral flow assay for use in rapid, simple diagnosis of Clostridium difficile disease. J Clin Microbiol. 2010;48(6): Novak-Weekley SM, Marlowe EM, Miller JM, et al. Clostridium difficile testing in the clinical laboratory by use of multiple testing algorithms. J Clin Microbiol. 2010;48(3): Wilcox MH, Planche T, Fang FC, Gilligan P. What is the current role of algorithmic approaches for diagnosis of Clostridium difficile infection? J Clin Microbiol. 2010;48(12): Bartlett JG. Clinical practice. Antibiotic-associated diarrhea. N Engl J Med. 2002;346(5): Jangi S, Lamont JT. Asymptomatic colonization by Clostridium difficile in infants: implications for disease in later life. J Pediatr Gastroenterol Nutr. 2010; 51(1): Bryant K, McDonald LC. Clostridium difficile infections in children. Pediatr Infect Dis J. 2009;28(2): Arch Pathol Lab Med Vol 136, May 2012 Laboratory Detection of C difficile LaSala et al 531