CAP Laboratory Improvement Programs. Performance Accuracy of Antibacterial and Antifungal Susceptibility Test Methods

CAP Laboratory Improvement Programs Performance Accuracy of Antibacterial and Antifungal usceptibility Test Methods Report From the College of American Pathologists Microbiology urveys Program (001 003) Michael A. Pfaller, MD; Ronald N. Jones, MD; for the Microbiology Resource Committee, College of American Pathologists Context. The College of American Pathologists Microbiology urveys Program provides external proficiency samples that monitor the performance of nearly 3000 laboratories that perform and report antimicrobial susceptibility tests. Objective. To summarize results obtained with bacterial and yeast challenge samples (001 through 003). Design. One organism every 4 months was tested by surveys participants against antibacterials/antifungals by routinely used methods. Reports were graded by interpretive category (susceptible, intermediate, resistant) based on an 80% consensus of referees/participants. Results. The most common antibacterial test methods/ systems were Vitek (38% 43%), Microcan (39% 43%), and the disk diffusion test (14% 15%), although Etest was most used for fastidious species. YeastOne was the dominant antifungal test (50% 55%). Antifungal results demonstrated continuous, improved accuracy (83% 88%), highest for YeastOne (96%) and broth microdilution (95%) methods. Antibacterial test accuracy was consistently greater than % against gram-positive organism challenges and greater than 98% against gram-negative challenges. For gram-negative strains with well-characterized resistance mechanisms, the accuracy by method was disk diffusion greater than broth microdilution greater than automated systems. Major problems identified were (1) Haemophilus influenzae control ranges require re-evaluation, () overuse of -lactamase tests, (3) errors among Enterococcus faecium against penicillins (Vitek, Microcan), (4) false-susceptible results with trimethoprim/sulfamethoxazole against coagulase-negative staphylococci (Micro- can), (5) macrolide false-susceptibility for -hemolytic streptococcus (Microcan), (6) flawed reporting for antimicrobials not active at the infection site, (7) use of outdated interpretive criteria, and (8) failure to follow Clinical and Laboratory tandards Institute testing/reporting recommendations. Conclusions. usceptibility tests were generally performing satisfactorily as measured by the surveys, but serious errors were identified with some drug/organism combinations that may require action by the Clinical and Laboratory tandards Institute and/or the Food and Drug Administration. (Arch Pathol Lab Med. 006;130:767 778) The participation of clinical laboratories in proficiency testing programs is considered an important step in ensuring quality and standardization in the performance of antibacterial and antifungal susceptibility testing. 1 4 The College of American Pathologists (CAP) urveys Program for clinical microbiology represents one of the largest comprehensive external proficiency testing programs in the world. An important component of the CAP Bacteriology and Mycology urveys has been the graded antibacterial/antifungal susceptibility testing challenges (3 Accepted for publication December 7, 005. From the Departments of Pathology and Epidemiology, Roy J. and Lucille A. Carver College of Medicine and College of Public Health, niversity of Iowa, Iowa City (Dr Pfaller); and JMI Laboratories, North Liberty, Iowa, and Tufts niversity chool of Medicine, Boston, Mass (Dr Jones). The authors have no relevant financial interest in the products or companies described in this article. Reprints: Michael A. Pfaller, MD, Medical Microbiology Division, C606 GH, Department of Pathology, niversity of Iowa College of Medicine, Iowa City, IA 54 (e-mail: michael-pfaller@uiowa.edu). organisms per year). Previous reports by this program dating from 198 have documented consistent high-quality performance overall by participating laboratories,,5,6 but periodic methods or commercial product deficiencies detected by the CAP have led to technical or methods modifications by the Clinical and Laboratory tandards Institute (CLI, formerly the National Committee for Clinical Laboratory tandards [NCCL]) 7 9 and/or product changes by various commercial systems. 5,10 15 The results of the 001 003 CAP antibacterial and antifungal surveys are summarized in this presentation including discussion of the most used susceptibility testing methods or systems. MATERIAL AND METHOD In 001 003, the CAP Microbiology urveys (D-series) had 685 to 9 subscribing laboratories reporting data from unknown challenge organisms for routine susceptibility testing. The organisms included 5 gram-positive species (Enterococcus faecium, Enterococcus faecalis, taphylococcus epidermidis, treptococcus dysgalactiae, andtreptococcus pneumoniae) and 4 gram-negative organisms (Acinetobacter baumannii, Klebsiella pneumoniae, Haemoph- Arch Pathol Lab Med Vol 130, June 006 Antimicrobial usceptibility Proficiency Testing Pfaller & Jones 767

ilus influenzae, and erratia marcescens). Each specimen was to be processed by identification and susceptibility testing methods routinely used in the participating laboratory, with the reporting of only those antimicrobial agents considered appropriate for the clinical settings stated on the CAP survey. For example, the clinical infection settings included meningitis caused by pneumoniae, bloodstream infection caused by K pneumoniae, and an A baumannii lower respiratory tract infection diagnosed by bronchoalveolar lavage. In these specific settings, the reporting of antimicrobials that achieved clinically adequate concentrations only in the urine (eg, cinoxacin, nitrofurantoin, norfloxacin, trimethoprim) was considered unacceptable performance. Acceptable graded categorical results were those achieved by 90% or more (001 and 00) or 80% or more (003) of referees and participants allowing grading of all samples. nacceptable results included reporting of inappropriate antimicrobial agents as described previously, reporting of agents for which interpretive criteria are lacking, and noncompliance with CLI guidelines. Reported results judged to be false-susceptible (false-) based on the graded consensus category were considered very major errors and those that were false-resistant (false-r) were considered major errors. Likewise in 001 003, the CAP Mycology urvey (F-series) had 11 to 179 subscribing laboratories reporting data that were sent 3 unknown yeast isolates each year as antifungal susceptibility challenges of which 63 to 113 reported antifungal susceptibility results. The susceptibility test results were monitored, but grading of responses was not initiated until 004. As with the antibacterials, consensus susceptibility categorical responses from 80% or more of referees and participants will guide the grading process in the future. For purposes of this report, only the quantitative (minimum inhibitory concentration [MIC]) results were tabulated and discussed. RELT AND COMMENT Antibacterial urvey The graded categorical accuracy for the gram-positive and gram-negative challenge strains tested by disk diffusion and MIC methods are shown in Tables 1 and 5, respectively. Gram-Positive trains. The overall results for the gram-positive challenge strains were generally quite good (Table 1). Among 16 graded organism drug pairs (56 disk and 70 MIC), only 3 pairs, all involving pneumoniae, had an accuracy level of less than 90%, 1 for the disk test and for the MIC test methods. Despite these mostly favorable results, some specific and recurring problems merit further discussion and comment. Performance issues with each of the challenge strains are addressed sequentially. Challenge strain D-03 (001) was an E faecium that was isolated from a simulated blood culture. This E faecium strain was a well-characterized clinical isolate that was resistant to vancomycin (vancomycin-resistant enterococci). In addition, this vancomycin-resistant enterococci strain had several unique characteristics that may have been confusing to some survey participants, including (1) a discord in susceptibility between ampicillin (susceptible) and penicillin (resistant); () moderate susceptibility to fluoroquinolones; and (3) generalized susceptibility to chloramphenicol, rifampin, and aminoglycoside synergy (gentamicin and streptomycin) but resistant to quinupristin/dalfopristin. Overall, the survey participants performance on this challenge was outstanding and the recognition of the unusual resistance features of this isolate (ie, resistant to both quinupristin/dalfopristin and vancomycin) was detected at rates of 98.4% to % among disk diffusion test users and 99.% to % for those using MIC methods (Table 1). The discord between ampicillin and penicillin susceptibility categories may be explained by this strain s penicillin-binding protein target affinity differences for the agents and was problematic for some of the commercial susceptibility testing systems (Table ). Vitek- (bio- Mérieux, Hazelwood, Mo) was the most likely system to produce a major error with ampicillin (.% of 46 reported results), and the Microcan ystem (Dade Behring, acramento, Calif) was the most likely to report a very major error with penicillin (54.7% of 844 reported results). As resistance to previous drugs of choice becomes more common among the enterococci (eg, ampicillin, vancomycin aminoglycosides), clinical microbiology laboratories must use reliable methods to test new agents. Delays in the introduction of newer agents such as the streptogramin combination (quinupristin/dalfopristin) and an oxazolidinone (linezolid) into commercial systems have occurred. In the present Microbiology urvey, participants reported results for these agents using a variety of methods or commercial systems including disk diffusion, Microcan, Etest (AB BIODIK, olna, weden), Vitek, PACO (Becton Dickinson, parks, Md), in-house broth microdilution and macrodilution, ensititre, and agar dilution methods. Major (false-r) errors were recorded for linezolid from Microcan (1 report; 3.% of results) and from the disk diffusion method (3 reports; 3.9%, data not shown). Linezolid-resistant gram-positive cocci are rare and only a few enterococci with resistance have been encountered, each emerging during long-term therapy usually associated with infected indwelling devices. 16,17 Likewise, although acquired resistance to quinupristin/dalfopristin can occur during therapy of infections resulting from E faecium, 18 the finding of primary (initial) resistance to this agent is rare. 19 This challenge isolate produces a streptogramin acetyl-transferase enzyme (vatd) that inactivates the drug. The vatd-positive strains of E faecium usually have MIC results for quinupristin/dalfopristin of 8 g/ml or greater and are easily detected by disk or MIC methods (Table 1). When streptogramin resistance has been documented in the nited tates or elsewhere, the principal pattern is a low-grade resistance (MIC of or 4 g/ml) without evidence of inactivating enzymes by molecular methods or microbiologic assays. 0 Challenge strain D-15 (001) was a multidrug-resistant (MDR) bloodstream isolate of epidermidis. This MDR strain had several susceptibility pattern characteristics that were considered to offer important challenges to existing test methods and systems: (1) resistant to oxacillin, and thus by consensus standards 7 resistant to all other -lactams regardless of in vitro testing results; () MDR with additional resistance to ciprofloxacin, macrolides, and trimethoprim/sulfamethoxazole; and (3) susceptible to glycopeptides (vancomycin and teicoplanin), aminoglycosides, chloramphenicol, clindamycin, rifampin, and the tetracyclines. The accumulated evidence regarding oxacillin resistance in coagulase-negative staphylococci is that those strains possessing the meca resistant gene and its product (PBP A) may have reference MIC values that are g/ ml or less (breakpoint for aureus), but usually g/ ml or greater. 1 5 These findings resulted in a decision by the CLI AT subcommittee to modify the oxacillin MIC breakpoint for susceptibility from g/ml or less to g/ml or less for coagulase-negative staphylococci (ie, resistant at g/ml). 6 The D-15 (001) challenge strain was characterized in these previous studies as containing 768 Arch Pathol Lab Med Vol 130, June 006 Antimicrobial usceptibility Proficiency Testing Pfaller & Jones

Table 1. Categorical Accuracy of Antimicrobial usceptibility Test Methods for College of American Pathologists urveys Gram-Positive Challenge trains in 001 003 (D-03, Enterococcus faecium [001]; D-15, taphylococcus epidermidis [001]; D-03, Enterococcus faecalis ATCC 91 [00]; D-15, treptococcus dysgalactiae [00]; and D-10, pneumoniae [003])* Antimicrobial Agent E faecium E faecalis Disk Diffusion epidermidis dysgalactiae Amikacin (16) (4) Amoxicillin/ clavulanate (94) Ampicillin 98.5 94.7 (9) (61) (131) (171) Ampicillin/ sulbactam (6) Azithromycin (19) Cefazolin 98. (161) (54) Cefepime (19) Cefotaxime. (6) (7) Ceftriaxone 93.0 (41) (00) Cefuroxime (30) Chloramphenicol 90.9 98.3 (175) (33) (58) (178) Ciprofloxacin 94.3 (75) (54) (10) (9) Clindamycin 94.9 91.4 (34) (373) (514) Erythromycin 99.3 99. (144) (59) (384) (608) Gatifloxacin 91.7 (4) Gentamicin (54) (6) (87) (30) Imipenem (0) Levofloxacin 98.0.7 93.7.9 (49) (43) (17) (85) Linezolid 94.8 (77) (1) (3) Ofloxacin 93.8 (16) (3) Oxacillin 93.4 (470) (6) Penicillin.5.7 96.9 (01) (174) (390) (418) Quinupristin/ 98.4 dalfopristin (17) (17) Rifampin (111) (11) (94) (5) treptomycin 90.0 (3) (19) Tetracycline 94.3 98.8.1 (177) (87) (48) (06) Trimethoprim/ 98.7 sulfamethoxazole (8) (305) (35) Trovafloxacin (5) Vancomycin 94.3 98. (380) (314) (435) (617) % Accuracy (No. of Results) pneumoniae (0) (5) () 95. (14) (7) (138) (9).1 (104).7 (811) 89.6 (48) 90.4 (83) (11) (53) E faecium 95.6 (1701) 94.1 (51) 98.4 (863) (65) E faecalis (1761) (31) 99. (747) (454) MIC Methods epidermidis 94.5 (17) 96. (664) (737) 96.1 (646). (1437) 95.9 (1) 91. (137) 93.3 (60) (383) 98.0 (1465) 98.6 (347).8 (155) 98.5 (1435) 98. (1356) (13) 91.4 (45) 99. 96.4 (866) (89) (113) 93.6 (31).1 (70) 98.5 (633) 99. (1795) (1808) (061) 99. (18) 99.0 (581) (39) (178) 94. 93.9 (78) (668) (168) (906) (1551) (1537) 99.0 (96) (63) (534) (441) (703) dysgalactiae (577) 98. (54) 98.1 (5) 96.5 (85) (50) (668) 94.1 (406) 90.3 (31) (773) 91.6 (740) (51) (481) (63) (67) (1158) 98.1 (414) (31) (853) pneumoniae (1087) (1603) 89.9 (69) 93. (0) 94.4 (90) 95.0 (161) (50) (161) 89.4 (193) 91.3 (16) 93.9 (165) (1358) * % Accuracy indicates good performance; MIC, minimum inhibitory concentration;, ungraded because of too few participant reports or not achieving 80% consensus. nderlined value is the lowest graded accuracy for that challenge and method. Gentamicin and streptomycin results for entertococci indicate tests for the detection of high-level resistance negating potential synergistic bactericidal action between ammoglycosides and cell-wall active agents. No entry indicates not applicable. Arch Pathol Lab Med Vol 130, June 006 Antimicrobial usceptibility Proficiency Testing Pfaller & Jones 769

Table. Method or ystem Results Varying From the Consensus Categorical Determination for Ampicillin (usceptible) and Penicillin (Resistant); 5 Responses From Enterococcus faecium D-03 (001) Antimicrobial Variant Result Method % Error Ampicillin Resistant Etest Microcan Vitek Vitek- Penicillin usceptible Etest Microcan Vitek Vitek- 6.8 3.4.. 9.1 54.7.3 8.7 Table 3. Accuracy of Detecting Resistance to Trimethoprim/ulfamethoxazole for taphylococcus epidermidis D-15 (001) by Various Test ystems/ Methods ( 5 Responses) Method or ystem Microcan Vitek All methods No. of Reports by Category usceptible Intermediate Resistant % Error* 33 9 64 0 479 705 115 3.7 4. 1.9 * Correct results indicate resistant. Excludes Vitek-, which had 40.0% error but only 5 responses. the meca gene and with oxacillin MIC results in the range of to g/ml. 6 urvey participants performed well with this challenge strain (Table 1), especially for oxacillin using commercial systems such as Microcan (1177 reports; % resistant) and Vitek (106 reports; 98.3% resistant). However, several issues were observed that require comment: 1. Laboratories continuing to use outdated oxacillin interpretive criteria: Examination of the reported zone diameters and associated interpretations for oxacillin shows that at least 4.3% of disk diffusion test users (18 sites) still are using the older (1998) breakpoint criteria resulting in a report of false- (very major error). These laboratories should acquire the most recent M100-15 document 9 and use zone diameter breakpoints of 17 mm or less for resistant and 18 mm or greater for susceptible. Alternatively, these laboratories may use a 30- g cefoxitin disk and breakpoints of 4 mm or less (resistant) and 5 mm or greater (susceptible). 9. Reporting errors of false- (very major error) for lactams when the oxacillin (or cefoxitin) test result is resistant: Those laboratories reporting a susceptible result for a penicillin, cephalosporin, carbapenem, -lactamase inhibitor combination (eg, amoxicillin/clavulanate), or monobactam were clearly not complying with the CLI standards for testing and reporting results for oxacillinresistant staphylococci. 9 Among the reports from disk test users (439 total results), 189 (4.5%) were false- including results for penicillins, carbapenems, 4 -lactamase inhibitor combinations, and 13 cephalosporins (Table 1). The 4 agents most often reported in error as susceptible were amoxicillin/clavulanate (34 reports), ceftriaxone (19 reports), ampicillin/sulbactam (14 reports), and cefuroxime (10 reports). In contrast, only 0.6% of MIC test participant results (158/5 054 responses) were false- -lactam values. Two thirds of those very major errors were generated by users of the Microcan system. Laboratories must be reminded that the reporting criteria for all -lactams and staphylococci should be based on the oxacillin test results and not on the individual drug results. 9 3. ystem-associated false- results for trimethoprim/ sulfamethoxazole (Table 3): Participants reporting disk diffusion results for trimethoprim/sulfamethoxazole correctly classified the organism as resistant (98.7% of 305 results). In contrast, only 81.7% of 1537 participants using MIC methods reported the correct resistant category interpretation. Analysis of the responses from various test methods (Table 3) demonstrated that very major (false-) errors were more likely to be generated by Microcan users (3.7% error, 88% of all false- MIC results) compared with all other systems (4.% 6.1% error). Resolution of this product-specific testing problem requires attention from the manufacturers. 4. se of -lactamase detection tests: A significant number of laboratories (16 and 1077 laboratories using disk diffusion and MIC methods, respectively) continue to report -lactamase test results from various test methods. uch reports for staphylococcal isolates are not recommended by the CLI 9 or by the CAP Microbiology Resource Committee for strains having clearly resistant test results for penicillin. Among the 139 total responses, only 18 (1.5%) were false-negative for this penicillinaseproducing isolate, the majority of which (78%) were produced by the nitrocephin method. By comparison with previous challenges, these reported surveys results show considerable progress toward reducing testing and reporting errors when processing staphylococcal isolates.,5,6,1 Attention to the issues noted previously could further improve the quality of susceptibility testing, reduce errors, and provide more reliable commercial products. Challenge strain D-03 (00) was a blood culture isolate of E faecalis. This E faecalis strain represents the CLI quality control (QC) strain, ATCC 91, which has been used for nearly 3 decades to provide quality assurance for CLI quantitative dilution methods. 7,9 The graded performance accuracy by MIC test users ranged from 99.% (ciprofloxacin, penicillin) to 100% (trovafloxacin) (Table 1). When the potential for synergy with aminoglycosides was assessed, the test accuracy was 93.9% (positive synergy) for streptomycin and 98.% (positive synergy) for gentamicin. Because this was a QC strain, we were able to compare the reported MIC results with the ranges expected for this organism. 9 The percentage of MIC results within published ranges were ampicillin (93.0%), chloramphenicol (96.9%), ciprofloxacin (%), erythromycin (95.8%), levofloxacin (%), penicillin (96.9%), rifampin (%), tetracycline (.5%), and vancomycin (%). A high degree of accuracy was also demonstrated by the disk diffusion participants. The graded accuracy ranged from 90.0% (300- g streptomycin disk for synergy) to % (ampicillin susceptibility). Of concern with this challenge was the reporting of drugs not recommended by the CLI for enterococci: cephalosporins (57 occurrences), clindamycin (34 occurrences), and trimethoprim/sulfamethoxazole (8 occurrences). 9 Furthermore, even with warnings on the CAP urveys reporting form, 8 participants reported nitrofurantoin results for this bloodstream infection. everal new potent agents are available for the treatment of enterococcal bacteremia, especially for isolates that are MDR. Among these, quinupristin/dalfopristin has no significant 770 Arch Pathol Lab Med Vol 130, June 006 Antimicrobial usceptibility Proficiency Testing Pfaller & Jones

Table 4. Error Rates for Erythromycin and Clindamycin by Testing Method When Applied to erogroup G -Hemolytic treptococcus (D-15, 00; 5 Responses) Antimicrobial/ Method (No. of Reports) Erythromycin Etest (118) Microcan (453) PACO (38) Vitek (40) Clindamycin Etest (64) Microcan (55) PACO (35) Vitek (38) No. (%) of Errors by Category usceptible 0 (0.0) 43 (9.5) 1 (.6) 4 (10.0) 3 (4.7) 177 (3.1) 3 (8.6) 6 (68.4) Intermediate 0 (0.0) (0.4) (5.3) (5.0) 1 (1.6) 8 (1.4) 1 (.9) (5.3) Total % Error 0.0 9.9 7.9 15.0 6.3 33.5 11.4 73.7 activity against E faecalis or this QC strain 18,19,7 ; however, of 17 participants providing data on this agent reported a susceptible or intermediate result. Linezolid susceptibility was reported by 1 participants, each quantitatively correct with a corresponding susceptible or intermediate category report. 8 The CLI recommends the use of a nitrocephin-based -lactamase test to detect ampicillin or penicillin resistance due to -lactamase production among blood and cerebrospinal fluid (CF) isolates of enterococci. 9 A total of 1370 participants reported -lactamase test results, 195 of which were determined by the chromogenic cephalosporin methods. False-positive test results were reported by 13 participants, 5 generated from acidometric methods and 6 were from chromogenic methods. pecimen D-15 (00) was a simulated blood culture sample containing a serogroup G -hemolytic streptococcus ( dysgalactiae). This organism exhibited a broadly susceptible antibiogram but with significant resistance to tetracyclines and also to macrolide-lincosamide-streptogramin B class agents by a methylase. 9 Performance was generally good for both disk diffusion and MIC methods participants (Table 1). One of the most significant reporting errors in this challenge was the reporting of ciprofloxacin and trimethoprim/sulfamethoxazole disk and MIC category results for this streptococcus when no interpretive criteria are published in the CLI tables. 9 This macrolide-lincosamide-streptogramin B resistant strain had various levels of reported clindamycin very major (false- ) errors for each method. The mean zone diameter for clindamycin was only 9.9 mm (D 5.), and very major (false-) or minor (false-intermediate) errors for this test were only 8.6%. In contrast, the rate of clindamycin-susceptible (very major errors) results was nearly 30% for all reported MIC methods (Table 4). Extremely high very major (false-) rates were detected for the Microcan (3.1%) and Vitek (68.4%) methods (Table 4). Likewise, erythromycin very major errors were dominated by Microcan test results, in which 9.9% of these participants incorrectly reported erroneous susceptible or intermediate results (Table 4). sers of these commercial systems should be aware of the high possibility of testing errors when performing susceptibility tests on these streptococci. ome reporting errors were observed for the disk diffusion test users as follows: (1) urinary tract infection specific agents reported (3 occurrences for nitrofurantoin, norfloxacin, and trimethoprim); () reporting of agents not having published CLI interpretive criteria including amikacin (4 occurrences), carbenicillin (1 occurrence), cefamandole (1 occurrence), cefixime (1 occurrence), cefotetan (1 occurrence), ceftazidime (4 occurrences), ciprofloxacin (9 occurrences), doxycycline ( occurrences), gentamicin (30 occurrences), netilmicin (1 occurrence), oxacillin (6 occurrences), piperacillin tazobactam (3 occurrences), rifampin (5 occurrences), spectinomycin (1 occurrence), teicoplanin (1 occurrence), ticarcillin clavulanate (1 occurrence), tobramycin (4 occurrences), trimethoprim/sulfamethoxazole (35 occurrences), lomefloxacin ( occurrences), and moxifloxacin (1 occurrence); and (3) false- clarithromycin ( occurrences) on erythromycin-resistant strains. Among the MIC test users, the reports of major concern were similar to those of the disk diffusion test participants: (1) urinary tract infection specific drugs reported ( occurrences), () reporting of agents without in vitro breakpoint criteria (143 occurrences among 3 drugs), and (3) false- clarithromycin ( occurrences) and azithromycin (1 occurrence) reports. The clinical setting for this organism was selected to establish the rate of susceptibility testing of -hemolytic streptococci from invasive infections. Considering that penicillin is usually the drug of choice for these infections, 30 the CLI guidelines 9 indicate that susceptibility testing of penicillin and other -lactams approved by the Food and Drug Administration for the treatment of -hemolytic streptococci is not necessary for clinical purposes and need not be done routinely, because as with vancomycin, resistant strains have not been recognized or reported. However, this patient was said to have a penicillin allergy and several alternative agents might be considered including cephalosporins, glycopeptides, fluoroquinolones, macrolides, or clindamycin. Thus, antimicrobial susceptibility testing was indicated and a total of 1775 participants (68.7% of laboratories) reported 1 or more susceptibility test results for this sample. The macrolide-lincosamide-streptogramin B resistance expressed by this organism was a major finding that would negate the use of this class. Laboratories should realize that susceptibility and resistance of dysgalactiae to azithromycin, clarithromycin, and dirithromycin can be predicted by testing erythromycin. 9 Likewise, a streptococcal isolate that is susceptible to penicillin may be considered susceptible to other -lactams and need not be tested against those agents. 9 Finally, macrolide-resistant isolates of -hemolytic streptococci may have constitutive or inducible resistance to clindamycin. Inducible clindamycin resistance can be detected by placing a - g clindamycin disk 1 mm from the edge of a 15- g erythromycin disk ( D test ). Following incubation, organisms that do not show flattening of the clindamycin zone should be reported as clindamycin-. Organisms that show flattening of the clindamycin zone adjacent to the erythromycin disk (a positive D test) demonstrate inducible clindamycin resistance and should be reported as such (ie, resistant 9 ). pecimen D-10 (003) was a mock CF specimen from a patient with meningitis due to pneumoniae. This isolate expressed constitutive resistance to macrolides and clindamycin, as well as additional resistances to -lactams, trimethoprim/sulfamethoxazole, and tetracyclines (Table 1). The overall graded performance accuracy for the disk diffusion tests was 98.% of 033 results and for the MIC methods was 94.0% of 433 results. A total of 811 participants used the 1- g oxacillin disk screening test for pen- Arch Pathol Lab Med Vol 130, June 006 Antimicrobial usceptibility Proficiency Testing Pfaller & Jones 771

icillin resistance and.7% reported the intended result (possibly resistant, MIC testing required for confirmation). All but 0.9% of MIC tests users obtained a non-susceptible result for penicillin (89.4% resistant; 9.7% intermediate), whereas.3% of oxacillin disk screens were erroneously susceptible (Table 1). Cerebrospinal fluid isolates of pneumoniae found to be nonsusceptible by the oxacillin disk screen (zone diameter, 19 mm) should be subjected to quantitative MIC testing against penicillin, cefotaxime, ceftriaxone, meropenem, and vancomycin. 9 Ideally, MIC testing should be performed primarily on all isolates from CF. The participant MIC results for cefotaxime, ceftriaxone, and meropenem were not gradable because of lack of consensus and variable reporting of either intermediate or resistance categories; however, the strain had a clearly resistant MIC of 4 g/ml for penicillin. Again, it must be stressed that disk diffusion and/or MIC testing should not be performed with drugs that lack CLI-approved interpretive criteria. erious very major (false-) errors may result from this practice, causing increased risk of inappropriate antimicrobial therapy and patient morbidity and mortality. As such, disk test results for amoxicillin, ampicillin, cefepime, cefotaxime, ceftriaxone, cefuroxime, imipenem, and meropenem are unreliable and should not be reported. 9 In addition, routine reporting of agents administered by the oral route only, first- or second-generation cephalosporins (except cefuroxime sodium), clindamycin, macrolides, tetracyclines, and fluoroquinolones on pneumococci (or any species of bacteria) isolated from CF could be dangerously misleading for patient care as these agents do not reliably achieve therapeutic concentrations in the CF and are inappropriate for treating such an infection. 6 Evidence of such dangerous reporting was seen in this challenge when results were reported for clindamycin (16 responses), erythromycin (99 responses), gatifloxacin (59 responses), levofloxacin (65 responses), and tetracycline (09 responses). Furthermore, because no interpretive criteria have been published for disk diffusion or MIC testing of pneumoniae by the CLI, 9 results would be unacceptable for several other agents including aminoglycosides, ceftazidime, teicoplanin, and ticarcillin/ clavulanate. Taking all of the caveats into consideration, unacceptable or inappropriate reports comprised 561 (36% of all results) disk test responses and 787 (10.1%) MIC responses. These are unacceptably high rates of clinically important errors in testing and reporting, and every effort should be made to correct these dangerous practices. Gram-Negative trains. For gram-negative organisms, Table 5 summarizes the test accuracy for the 4 challenge strains. The number of graded organism drug pairs was markedly increased compared with that of the grampositive challenge strains (Table 1) and the overall accuracy was also improved. Although both disk diffusion and MIC methods performed very well with overall rates of accuracy exceeding 98%, there were some important testing and reporting issues that bear discussing. Challenge strain D-09 (001) was an isolate of Abaumannii/haemolyticus obtained from a simulated bronchoalveolar lavage specimen. This MDR strain has several resistant characteristics that posed challenges to surveys participants: (1) susceptible to imipenem, meropenem, and polymyxins (B and E); () marginal inhibition by tetracyclines and by some -lactamase inhibitor combinations such as ampicillin/sulbactam, piperacillin/tazobactam, and ticarcillin/clavulanate; and (3) frank resistance to all remaining routinely tested compounds. These types of MDR strains, usually represented by A baumannii, are being reported at greater frequency worldwide. 31 34 Therapy can be difficult, but most investigators confirm the conclusions of Wisplinghoff et al 34 that carbapenems (imipenem and meropenem), amikacin, tobramycin, and tetracyclines remain most active and clinically useful. tandardized and reference susceptibility tests 9 appear adequate to detect antimicrobial resistances and to guide therapy for Acinetobacter spp infections. Both disk diffusion and MIC participants correctly detemined the susceptibility of this isolate to the carbapenems (91% 100%) and resistance to virtually all other agents. The borderline nature of the susceptibility to the -lactamase inhibitor combinations was evident in the lack of consensus as to the category response of this organism (Table 5). Challenge strain D-09 (00) was an isolate of H influenzae from a simulated blood culture. This H influenzae strain (ATCC 4947) represents one of the CLI Haemophilus QC strains and has been used for more than a decade to provide quality assurance for CLI MIC and disk diffusion methods. 9 This organism is well characterized and has an altered penicillin-binding protein target that renders resistance to ampicillin, ampicillin/sulbactam, amoxicillin/ clavulanate, and some older orally administered cephalosporins. 9,35 Most laboratories continue to use the disk diffusion test (81.7% of 1303 laboratories) to determine the susceptibility of H influenzae to ampicillin and other therapeutic agents (Table 5). This trend has not varied significantly since the initial H influenzae challenges in 1990 when 80.9% of participants used the disk diffusion method. When MIC values were determined for the most often tested drug (ampicillin), the methods used were Etest (58.7%) greater than ensititre (7.0%; TREK Diagnostic ystems, Cleveland, Ohio) greater than broth microdilution prepared in-house (5.8%) greater than Microcan (5.4%; not approved by the Food and Drug Administration). Ten years earlier, Microcan was the most used system (56.7%). 36 The performance accuracy by the MIC method users ranged from 9.5% (tetracycline) to % (4 agents) among graded drugs. Many agents were ungraded because their MIC values were near the breakpoint MIC, a finding observed in all prior surveys for this QC strain. For the disk diffusion test, the accuracy for the 13 graded drugs ranged from 9.8% (meropenem) to % (3 agents). When graded responses were available for both disk diffusion and MIC methods (11 antimicrobials), the accuracy rates were.5% and.7%, respectively (Table 5). Because this was a QC strain, we could compare the reported MIC or zone diameter against CLI published QC ranges. 9 Table 6 lists the modal MIC and mean zone diameter for selected agents. A chronic problem of reported zone diameters for the H influenzae QC strain at the lower limits (small zones) of published ranges was observed in this challenge as it was in 1993. The best examples of this phenomenon are the cephalosporins (cefotaxime, ceftazidime, and ceftriaxone), chloramphenicol, and the fluoroquinolones (ciprofloxacin and levofloxacin). Haemophilus Test Medium QC ranges for several agents may require re-evaluation using the latest CLI M3-A guidelines 37 and contemporary Haemophilus Test Medium lots. Most MIC results for this strain were within control ranges when the dilution schedules permitted. ome problems may exist with azithromycin (high) and rifampin (low) (Table 6). The most frequent reporting error for this 77 Arch Pathol Lab Med Vol 130, June 006 Antimicrobial usceptibility Proficiency Testing Pfaller & Jones

Table 5. Categorical Accuracy of Antimicrobial usceptibility Test Methods for College of American Pathologists urveys Gram-Negative Challenge trains in 001 003 (D-09, Acinetobacter baumannin [001]; D-09, Haemophilus influenzae ATCC 4947 [00]; D-03, erratia marcescens [003]; and D-17, Klebsiella pneumoniae [003])* A Antimicrobial Agent baumannii Amikacin (9) Amoxicillin/ clavulanate (47) Ampicillin (156) Ampicillin/ sulbactam (104) Azithromycin ( 50) Aztreonam 99. (19) Carbenicillin (4) Cefaclor Disk Diffusion (No. of Results) H influenzae (14) (116) (1) (115) (58) (173) Cefazolin (108) Cefepime 99.1 (106) Cefixime 96.6 (58) Cefoperazone (1) Cefotaxime 96.7 (86) (47) Cefotetan (46) Cefoxitin () Ceftazidime. 93.9 (81) (114) Ceftizoxime (5) Ceftriaxone 96.4 (177) (780) Cefuroxime (40) (396) Cephalothin (50) Chloramphenicol 96.8 (34) (759) Ciprofloxacin 99.3 (330) (77) Gentamicin (3) Imipenem 98. (8) (109) Levofloxacin (117) (01) Meropenem 90.9 9.8 () () Mezlocillin (5) Ofloxacin (1) Piperacillin (11) Piperacillin/ 98. tazobactam (108) Tetracycline 96.4 (7) (50) Ticarcillin (36) Tobramycin.3 (60) Trimethoprim/ 99. sulfamethoxazole (53) (689) marcescens (66) (10) (305) 96.7 (90) (40) 90.0 (40) 90.0 (199) (48).3 (74) 98.3 (6) (31).8 (90) 98.1 (157) 96.6 (87) (116) 98.5 (01) (89).4 (76) 99.3 (135) (16) 98.4 (63) 98.5 (68) (9) 85.7 (14).8 (90) 98.7 (307) K pneumoniae (141) (8) (38) 99.0 (99) (86) (174) (73) (99) 88.0 (75) (51) (165) (170) (75) (87) (14) (175) (70) (164) (106) (9) (65) 99.1 (106) 90.9 (66) (137) (190) A baumannii (1568) (178) (695) (5) MIC Methods (No. of Results) H influenzae marcescens 99.0 (618) 98.3 (833) (065) (713) K pneumoniae 9 (1435) (531) (1848) (1070) Arch Pathol Lab Med Vol 130, June 006 Antimicrobial usceptibility Proficiency Testing Pfaller & Jones 773 98.8 (960) 95.5 (44) (677) 96.8 (65) (105) 94.9 (97) (34) 99.0 (96) (04) 90. (51) (1345) 99.3 (148) (88) 99.1 (105) 99.3 (1986) (45) 99. (03) (1011) (39) (88) (106) (194) (464) (95) 93.5 (153) (1854) 96.1 (151) (31) (61) (4) (7) 9.9 (14) 9.9 (14) (10) 99. (16) (11) (194) (68) (144).0 (33) 95.0 (0) (5) (4) 9.5 (40) 99. (18) (34) 87.5 (179) (1908) 99.3 (405) 95.5 (487) (508) (08) 94.8 (808). (131) (775) (451) (158) (415) 93.5 (673) (140) 98.1 (157) 93.8 (551) 9.8 (547) (456) (71) (760) (07) (745) (1780) (64) (686) 91.6 (904) 83.8 (84) (1195) (198) (561) (47) (55) (1405) (1) (1857) (148) 99. (4) (5) (1004) 98.4 (380) (1307) (19) * MIC indicates minimum inhibitory concentration;, ungraded because of too few participant reports or not achieving 80% consensus. nderlined value is the lowest graded accuracy for that challenge and method. No entry indicates not applicable.

Table 6. Comparisons of urveys Participant Modal Minimum Inhibitory Concentrations (MICs) and Mean Zone Diameters Compared With Clinical and Laboratory tandards Institute Quality Control Ranges for Haemophilus influenzae ATCC 4947, D-09 (00; 100 Total Responses)* Antimicrobial Agent (No. Tested, Disk/MIC) Ampicillin (1049/54) Ampicillin/sulbactam (110/3) Azithromycin (104/7) Cefotaxime (394/14) Ceftazidime (104/10) Ceftriaxone (707/189) Chloramphenicol (700/138) Ciprofloxacin (5/56) Levofloxacin (183/38) Imipenem (10/18) Meropenem (85/40) Tetracycline (7/37) Trimethoprim/sulfamethoxazole (5/119) Disk Diffusion Zone, mm Mean 16.6 (18.1) 16.6 (18.) 18.1 31.1 (31.3) 8.9 (8.6) 3 (31.4) 3.3 (31.8) 33.3 (33.0) 3.3.0 (3.4).9 15.5 (16.3) 7.4 (6.5) NCCL Range 13 1 14 13 1 31 39 7 35 31 39 31 40 34 4 3 40 1 9 0 8 14 4 3 Mode 4 4 4 0.1 1 8 0.1 MIC, g/ml NCCL Range 8 8 1 4 0.1 0.1 1 0.06 1 0.004 0.03 0.008 0.03...... 4 3 0.03 * NCCL indicates National Committee for Clinical Laboratory tandards. Zone diameter (mm) in parentheses was the mean zone from 1993 D-05 sample, indicating the long-term problem with the H influenzae QC ranges. Table 7. Accuracy of Various -Lactamase Tests When Tested Against Haemophilus influenzae ATCC 4947 (D-09, 00; 5 Responses) Method (No.) Chromogenic cephalosporins Nitrocefin (95) Cefinase- (154) Acidometric (08) Iodometric (31) % Correct by usceptibility ser Group (No. of Reports) Disk Diffusion (1911) 98.0 96.5 94.9 95. * MIC indicates minimum inhibitory concentration. MIC (760)* 84.7 90.0 challenge was the reporting of fluoroquinolone results (not appropriate for a patient of this age [4 years]) by both MIC (39 reports) and disk diffusion test users (530 reports). Although this isolate was from a blood culture, the clinical setting was suggestive of meningitis. To this end, participant laboratories should be aware of CLI recommendations indicating that only results of testing with ampicillin, one of the third-generation cephalosporins, chloramphenicol, and meropenem should be reported routinely with CF isolates of H influenzae. 9 Finally, a total of 701 lactamase test results were reported, 1911 from disk diffusion test users and 760 from those laboratories using a MIC method (Table 7). This organism is -lactamase negative, but false-positive results were noted with all methods. The rank order of test accuracy remains unchanged over the last decade: chromogenic cephalosporins greater than iodometric greater than acidometric tests. Challenge D-17 (003) was a K pneumoniae isolate from a simulated bloodstream infection. The patient was a - year-old with leukemia. This isolate was an extendedspectrum -lactamase (EBL) producing strain of K pneumoniae. The EBL produced by this organism is a CTX-Mtype enzyme of the Bush group be and Ambler class A. 38,39 This EBL has high affinity for many penicillins and cephalosporins (less so for ceftazidime and cephamycins than for cefotaxime or ceftriaxone) and is inhibited by clavulanate. 38,40 The performance accuracy of the disk diffusion method among 5 graded antimicrobials ranged from 88.0% (cefotetan) to % (15 drugs) with an average approaching 99%. All 7 antimicrobials achieved consensus levels for grading MIC results with a range from 83.8% (cefoxitin) to % (chloramphenicol, gatifloxacin, and ticarcillin/clavulanate) and an overall average accuracy of 98.4%. The recognition of the EBL was easily accomplished by participants in this challenge because of the high-affinity hydrolysis of key screening drugs such as cefotaxime (modal MIC, 3 g/ml and median zone diameter at 6 mm), ceftriaxone ( 64 g/ml and 6 mm), ceftazidime ( 3 g/ml and 6 mm), and aztreonam ( 16 g/ml and 6 mm). Cefpodoxime, often recommended as a sensitive screen for EBLs, 9 was not widely used by surveys participants (8 disk test results [6 mm median zone diameter] and 7 MIC values [ 4 g/ ml MIC mode]). Although reference MIC and disk diffusion tests clearly produced results for both cephamycins (cefoxitin and cefotetan) that were at or below the CLI breakpoints for susceptibility, a significant number of laboratories erroneously changed susceptible results to resistant for this EBL-producing strain. These agents are not hydrolyzed efficiently by EBLs and have been used with clinical success against such EBL-producing strains. 41 Notably, the MIC methods used had a great impact on the reporting of resistance to the cephamycins. Major errors (false-r results) for cefotetan and cefoxitin, respectively, were reported by 5.8% and 7.4% of Microcan users, by 10.9% and 5.4% of Vitek users, and by 8.9% and 77.8% of Vitek- users. Consistent reporting programs are needed across test methods and commercial systems. The reporting of actual susceptible MIC results with discordant resistant interpretations by surveys participants using these commercial products underscores this problem. Participants using the disk diffusion method underestimated the level of susceptibility (major errors) to amikacin (19.1%) compared with those using a MIC method (9%). Among the 136 MIC participants reporting an intermediate or resistant result for amikacin, 85.7% were Microcan users. Challenge strain D-03 (003) was an isolate of marcescens obtained in the setting of a urinary tract infection. 774 Arch Pathol Lab Med Vol 130, June 006 Antimicrobial usceptibility Proficiency Testing Pfaller & Jones

This organism produced an unusual -lactamase, ME- 1. 38,4 45 This enzyme is classified as a carbapenemase and hydrolyses ertapenem, imipenem, and meropenem as well as aztreonam. 4,45 It is one of several Bush group f lactamases including ME-1 to ME-3, IMI-1, KPC-1 to KPC-3, GE-, and NmcA. 45 49 Expression of this enzyme may be either chromosomal- or plasmid-mediated, and many advanced-generation cephalosporins (cefepime, cefotaxime, ceftazidime, ceftriaxone) remain active, as do selected -lactamase inhibitor combinations. 4,49 Of concern in this challenge was the variable results reported for aztreonam (ungraded, resistance detected by only 0% of disk test users and 11% of MIC participants), a substrate for the ME-1 enzyme, and the lower accuracy for the imipenem MIC tests (93.5% vs.4% for disk tests). The imipenem false-intermediate (MIC, 8 g/ml) and false- (MIC, 4 g/ml; very major errors) results were most often reported from Vitek (37 occurrences; 13.% of all reports by this system). False- results (very major errors) for meropenem were also reported in more than 50% of tests reported from Vitek and Vitek- systems. There are no specific tests available for detection of Bush group fproducing strains that have recently become epidemic in the New York City area. Laboratories must be vigilant for the unique antibiograms of these strains characterized by resistance to carbapenems and aztreonam but susceptibility to the newer extended-spectrum cephalosporins. Antifungal urvey Prior to the publication of CLI document M7-A, 50 little was known of the proficiency of clinical laboratories in performing antifungal susceptibility testing aside from those laboratories actively engaged in CLI-conducted studies. Following the publication in 19 of the CLI M7-A reference method for testing yeasts, 50 the CAP initiated a proficiency program for antifungal susceptibility testing. ince that time, the number of participants has increased from approximately 50 to more than 100 laboratories. During the period discussed in this presentation the average number of participants has increased from 68 in 001 (range, 63 78), to 78 in 00 (range, 69 86), and in 003 (range, 88 113). Likewise, the performance of laboratories has steadily improved (Tables 8 through 10). A total of 9 yeast challenge samples were tested for susceptibility to antifungal agents by CAP urveys subscribers between 001 and 003 (Table 8). The challenge strains were all well characterized CLI QC or reference strains with published reference ranges for each antifungal agent. 51 The results were quantitatively compared with those ranges as shown in Table 8. The performance varied somewhat depending on the challenge strain and antifungal agent; however, there was a steady trend toward improved performance during the 3-year period with each of the antifungal agents (Table 9). Overall, the quantitative accuracy improved from 83% in 001 to 88% in 003 (Table 9). The lowest overall performance was seen with amphotericin B (74%), and the highest accuracy was observed with fluconazole (90%) and ketoconazole (96%) (Table 9). A set of supplemental questions was included with each challenge to ascertain the type of methodology used, compliance with CLI guidelines, numbers of tests performed annually, and QC and reporting issues. The most significant observations from the questions portion of the survey are (1) the vast majority (76.6%) of laboratories use broth microdilution methods (55.3% YeastOne [TREK Diagnostic ystems] and 1.3% other broth microdilution) conforming to CLI guidelines 51 ; () commercial MIC tests (YeastOne and Etest) dominate the methods (55.3% YeastOne and 16.0% Etest); (3) RPMI 1640 is the preferred medium; (4) MIC endpoints are most often read colorimetrically (YeastOne); (5) the QC strains are those recommended by the CLI; (6) annual numbers of antifungal tests have increased with one third of laboratories performing more than 100 tests per year; (7) most laboratories (83%) perform testing on isolates from blood and normally sterile body fluids; (8) most laboratories report both the MIC and interpretive category (5%), whereas 7.9% only report MIC values and the category interpretations are generally those of the CLI or the test manufacturer; and (9) the frequency of QC testing is generally at the time of each test occurrence (69.6%) or no less than weekly (6.1%). Table 10 lists the quantitative accuracy of those MIC methods that were in compliance with the CLI guidelines 51 for the three 003 challenge strains. Agreement with the CLI reference ranges was highest with the colormetric YeastOne and the in-house broth microdilution methods. Good performance was generally noted with Etest, whereas the broth macrodilution method provided inferior accuracy. These data are encouraging and show both an increased level of standardization and a concomitant improvement in the level of performance of the participating laboratories. With the introduction of several new antifungal agents, it is likely that antifungal susceptibility testing will assume an increasingly important clinical role. 5 These new agents are rapidly becoming incorporated into commercial testing panels, 53 the performance of which will be monitored and graded in future CAP urveys. In summary, antimicrobial and antifungal test accuracy remains high during the period from 001 to 003 as monitored by the CAP Microbiology urveys Program. However, several serious organism-method very major and major errors were detected during the 001 003 interval, most of which involved automated methods (see Tables through 4). Perhaps the most serious and persistent problem was the continued practice of laboratories testing and reporting results for antimicrobial agents for which there are no CLI interpretive criteria. This has been detected most often for challenges with pneumoniae but has been noted with other organisms as well. imilarly, reporting of results for agents that are not recommended by the CLI and other consensus groups for the treatment of serious (eg, CN) infections is another dangerous reporting practice that has been noted and must be corrected. The most commonly used tests have changed markedly since the initial published reports of CAP urveys in the early 1980s. 5,6 Disk diffusion testing, although very accurate and cost-effective, has declined to less than 15% of participating laboratories, replaced by commercial, automated methods having a convenient laboratory information system interface. As noted in Tables through 4 and discussed in the text of this article, these automated and semiautomated systems do make mistakes and thus cannot be treated as some infallible black box. Laboratory workers must be on the alert for false- and false-r results. One of the important roles of proficiency testing programs is to detect these system-based errors and report them in such a way that the responsible manufacturers are encouraged to take prompt corrective action. Improvement in antimicrobial testing methodology is evidenced by the develop- Arch Pathol Lab Med Vol 130, June 006 Antimicrobial usceptibility Proficiency Testing Pfaller & Jones 775