Background and Plan of Analysis

ENTEROCOCCI Background and Plan of Analysis UR-11 (2017) was sent to API participants as a simulated urine culture for recognition of a significant pathogen colony count, to perform the identification of the cultured organism, and to test the organism by the routinely used antimicrobial susceptibility testing (AST) method/product. The sample contained Enterococcus faecium, a species often observed to be multidrug-resistant (MDR), thus requiring accurate antimicrobial susceptibility testing to guide infection treatments. This particular strain was MDR, carrying resistances to penicillins, fluoroquinolones, macrolides, tetracyclines, linezolid (an oxazolidinone), rifampin, and vancomycin, and high-level resistance to streptomycin among the aminoglycosides. In contrast to sample UR-01 (2017) tested earlier this year which was a pan-susceptible quality control Escherichia coli (ATCC 25922), 1 this enterococcal strain was selected to analyze survey results for recognition of true resistance to key therapeutic agents (categorical [CA] and quantitative accuracy [QA]) e.g., rates of serious false-susceptible error results. Participating laboratories were requested to return a separate form with additional disk diffusion (DD) zone diameter or MIC results. Those forms with quantitative/categorical results listed by drug, were tallied by breakpoint category and method or system utilized. 2-6 The results from UR-01 were compared to UR-11 to contrast rates of false-susceptible and false-resistance for AST methods/drugs occurring in contemporary clinical microbiology practice. The overall categorical response statistics (Table 1) included grading for 15 antimicrobials having >40 results for DD and/or MIC methods. Among these over 1,000 participating laboratories, 320 forwarded quantitative AST results (Tables 2-6). The interpretive breakpoint category criteria 4-6 most utilized by the API participants were from the Clinical and Laboratory Standards Institute (CLSI) 4 at 68.5%; see Table 2. This was true across all four test methods/systems, with USA-FDA criteria used by 19.2% of participants. Non-USA or unlisted/unknown criteria were reported by 12.3% of API survey respondents. 5,6 These results were very similar to those reported for UR-01 (2017). 1 Examination of results while preparing Table 2 also indicated that method/system utilization favors MIC methods and automation. Only 2.5% of laboratories used the DD method, and the use rank order of MICbased systems was: MicroScan (52.2%) > Vitek 2 (41.6%) > BD Phoenix (3.4%). This distribution of product application when testing an enterococcus was slightly different than that observed for UR-01 (2017), where 91.5% of participants used MIC systems (Vitek 2 > MicroScan) for testing the E. coli strain. 1

Table 1. Categorical agreement (CA) for DD and MIC methods 1,2 when testing UR-11 (2017), a MDR VRE E. faecium. a Expected No. tests (% CA) for: Antimicrobial Agent category a DD method MIC method Ampicillin Resistant 47 (100.0) 888 (99.1) Ciprofloxacin Resistant 51 (100.0) 746 (99.2) Daptomycin Susceptible 0 (--) 175 (93.2) Erythromycin Resistant 3 (100.0) 77 (100.0) Gentamicin b Susceptible 14 (78.6) 221 (97.3) Levofloxacin Resistant 25 (100.0) 806 (99.6) Linezolid Resistant 7 (42.9) 515 (93.6) Nitrofurantoin Susceptible-Intermediate 55 (76.4) 846 (98.1) Norfloxacin Resistant 10 (90.0) 36 (100.0) Penicillin Resistant 20 (100.0) 662 (99.4) Quinupristin-Dalfopristin Susceptible 1 (100.0) 158 (98.7) Rifampin Resistant-Intermediate 0 (--) 132 (90.9) Streptomycin b Resistant 0 (--) 174 (99.4) Tetracycline Resistant 32 (90.6) 792 (97.2) Vancomycin Resistant 30 (100.0) 946 (99.5) a. Correct categorical interpretation was determined by the reference MIC result, using the M07-A10, M100-S27 and USA-FDA breakpoint criteria or participant consensus (gentamicin, norfloxacin, penicillin, rifampin, streptomycin). b. Susceptibility indicates potential synergistic activity in combination with cell wall active agents. Table 2. API Survey subscriber use of antimicrobial susceptibility breakpoint organization criteria for sample UR-11 (2017), by listed AST method or commercial product applied (320 responses). a AST breakpoint AST method used by API participant (%): organizations Disk Diffusion (DD) BD Phoenix MicroScan Vitek 2 CLSI 87.5 66.7 65.5 70.5 FDA 0.0 11.1 17.9 22.5 Other Not listed 12.5 22.2 16.6 6.2 Listed 0.0 0.0 0.0 0.8 a. 320 laboratories provided quantitative results, e.g., 14.5% (DD method) to 33.0% (MIC method) of all participants; see Table 1. One response from a Sensititre user named CLSI as the category criteria source.

Categorical and Quantitative Accuracy Analyses Overall Categorical Accuracy (CA) for DD and MIC Methods Table 1 lists the CA of both major AST testing groups (DD and commercial MIC devices) for the 15 most tested antimicrobial agents, regardless of appropriateness for the infection site listed. These tested drugs had >40 reported categorical results, with the greatest number of responses for vancomycin (976), ampicillin (943), nitrofurantoin (901), levofloxacin (831), tetracycline (824), and ciprofloxacin (801). Only nitrofurantoin among the six most reported agents was marginally active (MIC, 32-64 µg/ml); significant resistances (linezolid, rifampin, etc.) of epidemiologic importance were reported by less than half of the laboratories. Graded overall CA for DD method was lower than that of the MIC systems (Table 1) at 91.9% versus 97.7%. Lowest CA values were observed for linezolid (42.9%), nitrofurantoin (76.4%) and high-level gentamicin resistance (78.6%). The errors for linezolid may be caused by inappropriate endpoint (zone edge reading) interpretation; see references 4 and 5 for reading instructions. False-susceptible errors (called very major errors) were most common for tetracycline HCl (nearly 10%) among DD test users. Among MIC test results, the most serious and significant errors were (Table 1): 1. Daptomycin false-resistant or false-nonsusceptible reports at 6.9% of results; 2. Linezolid false-susceptible results at 6.4%; and 3. Rifampin false-susceptible values at 9.1%. CA and QA Analyses of the DD Method 3,4 Table 3 lists the DD zones of inhibition for the seven drugs most reported on the supplemental quantitative result form. CA among these tallied antimicrobials was low at 90.3% (91.1% overall; Table 1). The QA results (% of zones within the correct category 4 ) was similar at 91.2%. The antimicrobials having the highest rates of QA results out of range were gentamicin HLAR and nitrofurantoin (28.6% with zones 14 mm or resistant). Sample sizes for DD QA were considered small (34), in contrast to sample UR-01 (2017) 1 for a Gram-negative bacillus.

Table 3. Categorical agreement (CA) for the disk diffusion (DD) method including the reported zone diameters (mode and range) compared to the all participant CA values for sample UR-11, a MDR VRE- E. faecium (8 reporting sites). Antimicrobial CA % Quantitative subset results (mm): agent Expected category a All Subset b Modal zone (%) Range c Ampicillin Resistant 100.0 100.0 6 (100.0) 6 Ciprofloxacin Resistant 100.0 100.0 6 (87.5) 6-15 Gentamicin Susceptible 78.6 33.3 6 (66.7) 6-25 Linezolid Resistant 42.9 100.0 15 (100.0) 15 Nitrofurantoin Susceptible-Intermediate 76.4 85.7 15 (28.6) 12-23 Tetracycline Resistant 90.6 80.0 6 (50.0) 6-14 Vancomycin Resistant 100.0 100.0 6 (75.0) 6-14 a. Graded categorical response for all participants in this challenge. b. Subset of DD users that reported category and quantitative (mm zone diameter) results. c. Acceptable categorical results would be zone diameters (mm) as follows: ampicillin ( 16), ciprofloxacin ( 15), gentamicin ( 10 mm; 120-µg disk content), linezolid ( 20), nitrofurantoin ( 15), tetracycline ( 14) and vancomycin ( 14). CA and QA Analyses of the BD Phoenix MIC Device Three commercial MIC devices were tabulated separately (Tables 4-6), each having enough results to calculate both CA and QA (compared to correct MIC category ranges) rates. 3 Table 4 lists the BD Phoenix results (3.4% of MIC reporting sites; 2.8% for UR-01) for the ten antimicrobial agents that allowed comparisons of the most reported MICs 1) to the breakpoint categorical criteria, 2) to the MIC range for the correct category, and 3) to calculate the CA and QA. The CA rates for BD Phoenix calculated across these 10 listed antimicrobial agents was similar to the all MIC systems CA reported in Table 1, except for three drugs (daptomycin, linezolid and quinupristindalfopristin [Q-D]). All three daptomycin MIC results forwarded for analysis were >4 µg/ml, or not susceptible, and represented a major testing error. Three of nine linezolid MIC results were either 2 or 4 µg/ml, and represented a false-intermediate or false-susceptible error. Finally, the only Q-D MIC reported (2 µg/ml) was a minor, false-intermediate result.

Table 4. Categorical agreement (CA) for the BD Phoenix commercial MIC system subset including MIC (µg/ml) quantitative results (mode and range) compared to the all participant MIC and CA values for sample UR-11, a MDR VRE-E. faecium (11 participating sites). Antimicrobial agent Quantitative subset MIC Expected CA % results (µg/ml): category a All Subset b Modal MIC (%) Range Acceptable MICs (%) Ampicillin Resistant 99.1 100.0 >8 (100.0) >8 16 (100.0) Daptomycin Susceptible 93.2 0.0 >4 (100.0) >4 4 (0.0) c Gentamicin Susceptible 97.3 100.0 500 (100.0) 500 500 (100.0) Levofloxacin Resistant 99.6 100.0 >4 (100.0) >4 8 (100.0) Linezolid Resistant 93.6 70.0 >4 (66.7) 2->4 8 (66.7) d Penicillin Resistant 99.4 100.0 >8 (100.0) >8 16 (100.0) Quinupristin-Dalfopristin Susceptible 98.7 0.0 2 (100.0) 2 1 (0.0) d Streptomycin Resistant 99.4 100.0 >2000 (100.0) > 2000 >1000 (100.0) Tetracycline Resistant 97.2 100.0 >8 (100.0) >8 16 (100.0) Vancomycin Resistant 99.5 100.0 >16 (100.0) >16 32 (100.0) a. Graded categorical response for all participants in this challenge. b. Subset of BD Phoenix system users that reported category and quantitative (MIC) results. c. All reported MIC values for daptomycin were in error (false-resistant or non-susceptible). d. Poor CA and quantitative performance. CA and QA Analyses of the MicroScan MIC System Table 5 lists results from the MicroScan device (used by 52.2% of MIC reporting sites) for 12 tabulated antimicrobials. When the MicroScan CA rates were compared to the all MIC method CA rates (90.9-99.6%) several concerns were noted. The MicroScan subset (Table 5) showed CA rates of <90.0% for daptomycin (86.7%) and gentamicin (84.2%); and the QA were 89.8% and 100.0%, respectively. The reference laboratory daptomycin MIC (4 µg/ml) was at the CLSI susceptible breakpoint concentration as was the MicroScan modal value, thus minimizing the method-based concern due to normal MIC variation. Other potential MicroScan device concerns were the linezolid false-susceptible results found by 3.6% of participants (QA also at 3.6%); however, the all MIC method CA shows greater numbers of errors (6.4%) produced by other systems or methods. Similarly, rifampin false-susceptible results (MIC, 1 µg/ml) were observed by 4.3% of MicroScan participants but 9.1% of all MIC participants. Conversely,the MicroScan CA rate for tetracycline HCl (94.1%) was worse than the all MIC method rate (97.2%). For this MIC product interpretive testing errors were generally uncommon for this MDR Gram-positive pathogen, most often encountered as false resistances for daptomycin and falsely susceptible range MICs for linezolid, rifampin and the tetracyclines.

Table 5. Categorical agreement (CA) for the MicroScan commercial MIC system subset including MIC (µg/ml) quantitative results (mode and range) compared to the all participant MIC and CA values for sample UR-11, a MDR VRE-E. faecium (167 participating sites). Antimicrobial agent Quantitative subset MIC Expected CA % results (µg/ml): category a All Subset b Modal MIC (%) Range Acceptable MICs (%) Ampicillin Resistant 99.1 100.0 >8 (98.2) >8 - >16 16 (100.0) Ciprofloxacin Resistant 99.2 100.0 >2 (96.9) >2 - >16 4 (100.0) Daptomycin Susceptible 93.2 86.7 c 4 (79.7) 0.5 - >4 4 (89.8) c Gentamicin Susceptible 97.3 84.2 500 (73.9) 4-500 500 (100.0) Levofloxacin Resistant 99.6 100.0 >4 (100.0) >4 8 (100.0) Linezolid Resistant 93.6 96.4 d >4 (96.4) 1 - >4 8 (96.4) d Nitrofurantoin Susceptible- Intermediate 98.1 99.3 32 (91.0) 32-64 64 (100.0) Penicillin Resistant 99.4 100.0 >8 (100.0) >8 16 (100.0) Rifampin Resistant- Intermediate 90.9 95.7 d >2 (54.3) 1 - >2 2 (95.7) d Streptomycin Resistant 99.4 100.0 >1000 (100.0) >1000 >1000 (100.0) Tetracycline Resistant 97.2 94.1 >8 (90.4) 4 - >16 16 (97.1) Vancomycin Resistant 99.5 99.3 >16 (97.2) 16-32 32 (99.3) a. Graded categorical response for all participants in this challenge. b. Subset of MicroScan MIC system users that reported category and quantitative (MIC) results. c. False-resistant values of >4 µg/ml reported by >10% of laboratories. This error rate was greater than the all laboratory average. d. False-susceptible reports were observed for two significant treatment agents at unacceptable rates of 3.6-4.3%. CA and QA Analyses of the Vitek 2 MIC System Table 6 presents results from the Vitek 2 device (used by 41.6% of MIC reporting sites) for 12 tabulated antimicrobial agents. MIC results from this product are "calculated" from testing several drug concentrations and reported for a validated MIC range for each agent. The most reported calculated MIC values are found in Table 6, and were usually at the highest extreme of the calculated MIC range. 4-6 Where MIC results could be evaluated for QA, 75.0% (daptomycin) to 100.0% (six agents) of MIC values were within the acceptable category concentration range. Most of the Vitek 2 CA results (75-100.0%) were similar to the all MIC device CAs of 93.2-99.6%. For these enterococcus susceptibility tests, the most compromised Vitek 2 MIC results, like other systems, were for daptomycin (CA and QA, 83.3 and 75.0%) and linezolid (CA and QA, 91.1 and 89.3%). The false-resistant results for Vitek 2 showed inconsistent MIC values, but from a small sample size (only

eight responses). Furthermore, some participants stated these were supplemental tests, possibly from agar diffusion techniques (DD or Etest). Linezolid Vitek 2 MIC results had 51 of 56 values at 8 µg/ml, with MICs as low as 2 µg/ml. Participants should examine their routine QC results for these two drugs for possible trends toward falsely-high (daptomycin) or falsely-low (linezolid) MICs. Most other Vitek 2 MIC and categorical results for these reported antimicrobials had accuracy rates (QA or CA) superior to the all MIC method CA (Tables 1 and 6). Table 6. Categorical agreement (CA) for the Vitek 2 commercial MIC system subset including MIC (µg/ml) quantitative results (mode and range) compared to the all participant MIC and CA values for sample UR- 11, a MDR VRE-E. faecium (133 participating sites). Antimicrobial agent Quantitative subset MIC Expected CA % results (µg/ml): category a All Subset b Modal MIC (%) Range Acceptable MICs (%) Ampicillin Resistant 99.1 98.4 32 (97.6) 2 - >256 16 (97.6) Ciprofloxacin Resistant 99.2 100.0 8 (100.0) 8 4 (100.0) Daptomycin Susceptible 93.2 83.3 c 4 (50.0) 3-16 4 (75.0) c Gentamicin Susceptible 97.3 90.9 500 (100.0) 500 500 (100.0) Levofloxacin Resistant 99.6 100.0 8 (100.0) 8 8 (100.0) Linezolid Resistant 93.6 91.1 d 8 (89.3) 2-8 8 (89.3) d Nitrofurantoin Susceptible- Intermediate 98.1 99.2 64 (67.2) 16-128 64 (99.2) Penicillin Resistant 99.4 100.0 64 (100.0) 64 16 (100.0) Quinupristin- Dalfopristin Susceptible 98.7 100.0 1 (83.0) 0.5-32 1 (96.4) Streptomycin Resistant 99.4 75.0 >2000 (100.0) > 2000 >1000 (100.0) Tetracycline Resistant 97.2 99.2 16 (99.2) 0.12-16 16 (99.2) Vancomycin Resistant 99.5 100.0 32 (99.2) 32 - >256 32 (100.0) a. Graded categorical response for all participants in this challenge. Small sample sizes for tigecycline (susceptible at 0.12 µg/ml; two responses), doxycycline (intermediate at 8 µg/ml; one), and fosfomycin (susceptible at 48 µg/ml; one) were received, but not presented in this table. b. Subset of Vitek 2 MIC system users that reported category and quantitative (MIC) results. c. Generally poor performance, but some participants stated they used Etest and other products (reported as Vitek 2). d. Poor performance with 8.9 to 10.7% categorical or quantitative false-susceptible or false-intermediate errors.

Major Findings and Points of Concern from Analysis of this Sample (UR-11, 2017) Major Findings for E. faecium UTI Sample Over 1,000 laboratories participated in this survey, approximately 35% forwarding additional quantitative results (MICs and zone diameters). Commercial MIC-based devices (MicroScan > Vitek 2 > BD Phoenix) dominated AST use at over 97%. Interpretive breakpoint criteria applied to susceptibility test results were: CLSI (68.5%) > USA- FDA (19.2%) > others, i.e., consistent with sample UR-01 (2017); see Table 2. CA rate results for the evaluated drugs were (UR-01/UR-11): DD (97.8/91.9%) and all commercial MIC devices (99.5/97.7%). QA Findings (% of results within acceptable MIC or DD zone category ranges) For DD method the QA was at 91.2%, with the highest "out-of-acceptable category range" rate for nitrofurantoin. For BD Phoenix device the QA was only 84.1%, compromised by daptomycin and linezolid results among a small number of users. For MicroScan products the QA was 98.3%. For Vitek 2 system the QA was 98.7%. Points of Greatest Concern DD QA was below 90% overall on this sample and for UR-01 (2017). We observed in UR-01 (2017) that commercial MIC method products rarely test MIC endpoints in a dilution range where the E. coli ATCC 25922 or other QC organisms can be used as an effective strain. This fact limits precise use of MIC results to guide clinical therapy (dosing adjustments and TDM) via evolving "Antibiotic Stewardship" programs. Commercial MIC device testing of MDR strains appear to produce high CA & QA, but worrisome occurrences of false susceptibility (linezolid) and false resistance (daptomycin) have been observed. Furthermore, the antimicrobials recently approved by the USA-FDA (ceftaroline, ceftazidime-avibactam, ceftolozane-tazobactam, dalbavancin, delafloxacin, meropenemvaborbactam, oritavancin, tedizolid, telavancin, etc.) have not been translated into the devices which would be valuable for treatment of MDR strains.

As with UR-01 (2017), some results forwarded by participating laboratories imply that category breakpoint interpretations are not current with regulatory or other breakpoint publications. 3-5 Recommendations Assure availability of diagnostic products that provide category interpretations of antimicrobials, old and recently approved, for therapies of MDR Gram-positive and -negative pathogens. Re-evaluate any test procedure where unacceptable categorical grades were received, especially for daptomycin and linezolid among Gram-positive-targeted compounds. Re-evaluate QC procedures to assure routine testing of appropriate strains that provide measures of quantitative precision for each antimicrobial, not just a CA metric. Assure that current category interpretive breakpoints 4-6 are being applied locally and/or via utilized commercial AST product software. References 1. Jones RN. "Contemporary Assessment of Antimicrobial Susceptibility Testing (AST) Accuracy: Methods Used, Categorical Performance, and Quantitative Precision (Sample UR-01, 2017)." Traverse City, MI. American Proficiency Institute, 2017. 2. Clinical and Laboratory Standards Institute M07-A10. Methods for dilution antimicrobial susceptibility tests for bacteria that grow aerobically; approved standard. 10th ed. Wayne, PA. CLSI, 2015. 3. Clinical and Laboratory Standards Institute M02-A12. Performance standards for antimicrobial disk susceptibility tests. 12th ed. Wayne, PA. CLSI, 2015. 4. Clinical and Laboratory Standards Institute M100-S27. Performance standards for antimicrobial susceptibility testing: 27th informational supplement. Wayne, PA. CLSI, 2017. 5. EUCAST. Breakpoint tables for interpretation of MIC's and zone diameters. Version 7.1, March 2017. Available at http://www.eucast.org/fileadmin/src/media/pdfs/eucast_files/breakpoint_tables/v_7.1_breakpoint _Tables.pdf. Accessed March 2017. 6. USCAST. Breakpoint tables for interpretations of MICs and Zone Diameters, Version 3.0, January 2017. Available at http://www.uscast.org/breakpoints.html. Accessed March 2017.