Utility of direct susceptibility testing on blood cultures: is it still worthwhile?

Journal of Medical Microbiology (2016), 65, 501 509 DOI 10.1099/jmm.0.000259 Utility of direct susceptibility testing on blood cultures: is it still worthwhile? Vidthiya Menon, Sophie Lahanas, Catherine Janto and Andie Lee Correspondence Vidthiya Menon vidthiya@hotmail.com Received 11 February 2016 Accepted 2 April 2016 Department of Microbiology and Infectious Diseases, Royal Prince Alfred Hospital, Camperdown NSW 2050, Australia Earlier targeted therapy for bacteraemia optimizes patient outcomes and reduces broad spectrum antibiotic use. Standardized susceptibility testing results are available at 36 48 h. Direct disc susceptibility testing from blood culture broth reduces time to results but the inoculum is not standardized. No studies have looked at the clinical utility of direct susceptibility results. This retrospective cohort study aimed to assess the correlation between direct and formal testing methods as well as the clinical utility of direct susceptibility results. 160 episodes of bacteraemia with paired direct and formal susceptibility testing were studied. Direct disc testing was performed on blood culture broth. Formal testing was performed on isolates, using automated broth microdilution or Etests. The rate of error was 9.0 % (95 % CI 7.0 11.6 %). In 10 cases (6.3 %, 95 % CI 3.0 11.2 %), inappropriate antibiotics were used due to direct susceptibility results, including two cases with ineffective (as opposed to too broad) antibiotics being used. Antibiotics were changed in 28.1 % of cases once direct susceptibility data was available. There was a decreased time to effective antibiotics in 9.3 % (95 % CI 5.3 15.0 %), and a decreased time to a targeted antibiotics in 14.3 % (95 % CI 9.3 20.8 %) of cases. Despite the error rate, the advantages of earlier times to effective and targeted antibiotics justifies continuing direct testing in bacteraemia episodes with Gram-negative rods. In the Gram-positive group, given the contamination rate, the availability of adjunctive PCR, and the fact that early identification of the isolate could equally influence antibiotic choices, direct susceptibility testing may no longer be warranted. INTRODUCTION Bloodstream infections are life-threatening conditions that require prompt treatment with appropriate antimicrobials. Earlier effective therapy will have positive implications for the patient (Leibovici et al., 1998; Paul et al., 2010), particularly for those who have severe infections (Kumar et al., 2006). Earlier targeted therapy will also avoid use of overly broad spectrum antibiotics and potentially limit the development of antimicrobial resistance (Joseph, et al., 2015; Rogues et al., 2007) and also have potential benefits of reducing costs (Barenfanger et al. 1999). Standardized antimicrobial susceptibility testing requires overnight growth of cultures on agar plates, which are then tested by various methods including disc diffusion and broth microdilution [Clinical and Laboratory Standards Institute (CLSI), 2014], the latter usually being performed with automated methods. This effectively means that Abbreviations: MALDI-TOF MS, matrix-assisted laser desorption/ionisation-time of flight mass spectrometry; VRE, vancomycin-resistant enterococci. antimicrobial susceptibility results will only be available 36 48 h after the initial positive blood culture is detected. Direct disc susceptibility testing is performed by most laboratories directly from blood culture broth, diluted and inoculated onto agar plates with antimicrobial discs subsequently placed on the agar. The zone sizes are then measured after overnight incubation, with presumptive susceptibility results subsequently available after only 16 24 h, effectively reducing time to results by a day. The limitation with this is a lack of standardization of the inoculum, and various studies have shown that although the correlation between direct disc testing and subsequent standardized testing is over 95 %, major and minor errors are sometimes found (Edelmann et al., 2007; Mahrer et al., 2008; Tan et al., 2008; Yu et al., 2010; Nelovkov et al., 2012). There are various techniques reported in the literature for the preparation of the inoculum for direct disc susceptibility testing. Fay & Oldfather (1979) found that the use of 0.03 ml of turbid overnight blood culture broth produced zone diameters closely approximating those after standardized methods. 000259 ã 2016 The Authors Printed in Great Britain 501

V. Menon and others The European Committee on Antimicrobial Susceptibility Testing (EUCAST), 2012 states that there are no validated methods for processing specimens to ensure the correct inoculum is used for direct susceptibility testing, and that tests should be repeated on pure cultures as needed and correlation monitored. They also state that if the inoculum is visibly light, susceptibility results should not be reported, and that results should only be utilized if the organism is identified. Positive blood culture broth has also been used to directly inoculate VITEK 2 (biomerieux) identification and susceptibility testing cards using automated broth microdilution for the latter. De Cueto et al. (2004) compared this to formal identification and susceptibility testing performed on colonies from solid agar, also using the VITEK 2 system. The organism identification rate was poor, with only 62 % concordance for Gram-negatives, and 0 % for Gram-positives. The error rate for susceptibility testing for Gram-negatives was 6.6 %, and 8.4 % for Gram-positives. The conclusion was that this method was unreliable. Although previous studies have measured the correlation between formal standardized susceptibility testing and direct disc susceptibility testing, they have not investigated the clinical impact of the direct susceptibility testing. With the recent introduction of methods such as the MALDI- TOF MS (matrix assisted laser desorption ionisation-time of flight mass spectrometry), allowing rapid identification of Gram-negative rods, and PCR-based methods for early identification of Staphylococcus aureus and methicillin resistance, it is possible that direct susceptibility testing no longer provides any additional benefit in aiding clinical decision-making. The objectives of this study were: 1. To determine the correlation between direct susceptibility testing and formal susceptibility testing on positive blood cultures. 2. To determine the impact of direct susceptibility testing on the time to appropriate antibiotics. METHODS The study was a retrospective cohort study conducted at Royal Prince Alfred Hospital, a 919-bed hospital in Sydney, Australia. The microbiology laboratory at the hospital processed 16 920 blood cultures in 2014, with 1760 (10.4 %) positive blood culture isolates identified during that year. The laboratory performs direct susceptibility testing on positive blood cultures where the pathogen resembles a Staphylococcus species on Gram stain after 3pm on weekdays or any time on weekend days [i.e. when fem/meca PCR testing, according to the method of Adams (2005) is not available], on organisms resembling streptococci/enterococci on Gram stain, and on Gram-negative rods. Formal susceptibility testing is subsequently performed by automated broth microdilution (VITEK 2 biomerieux) for staphylococci (all Staphylococcus aureus, and coagulase-negative staphylococci if deemed to reflect true bacteraemia after clinical discussion), enterococci, and Gram-negative rods. Clinically significant alpha-haemolytic streptococci are formally tested for penicillin susceptibility with an Etest for a MIC determination. Beta-haemolytic streptococci do not go on to have formal susceptibility testing given that penicillin susceptibility is generally the rule. Consecutive clinically significant positive blood cultures up to May 2014 were selected using the bacteraemia database from the Microbiology Department, including 100 episodes with staphylococci (from 10 October 2013 to 1 May 2014), 100 episodes with streptococci/enterococci (from 6 May 2013 to 1 May 2014) and 100 episodes with Gram-negative rods (from 29 December 2013 to 2 May 2014). We excluded isolates within the same bacteraemic episode (within 14 days), and polymicrobial episodes. We also excluded those isolates with no meaningful direct susceptibility results (e.g. anaerobes, Haemophilus species), and those for which the clinical notes were unavailable. Positive blood cultures were identified using the BacT/ALERT (bio- Merieux) system. Bacterial species were identified by the Bruker MALDI Biotyper, and less frequently with VITEK Identification Cards (bio- Merieux) or API strips (biomerieux) depending on the organism. Gram-negative rods also had direct identification through the Bruker MALDI Bioptyper during business hours, allowing identification directly from the positive blood culture broth rather than waiting for growth on solid agar. Direct susceptibility testing was performed by placing three drops of the positive blood culture broth for staphylococci/streptococci/enterococci (Gram-positive cocci) and one drop for Gram-negative rods, into 1.8 ml of 0.45 % saline. A sterile swab was then used to inoculate a Mueller-Hinton plate (or Mueller-Hinton +5 % Sheep s Blood plate for streptococci) by immersing the tip into the suspension and rotating for several seconds, allowing the excess to drain off inside of the tube, and then rubbing the swab over the entire surface of the plate in three directions, rotating the swab while inoculating. The same swab was then also used to inoculate a purity plate. The plates were incubated overnight in 35 37 C in O 2, or CO 2 for streptococci. Zone sizes were then recorded and interpreted according to Clinical Laboratory Services Institute (CLSI) 2014 guidelines [Clinical and Laboratory Standards Institute (CLSI), 2014]. Results were not interpreted when the inoculum was found to be too heavy or too light. Antibiotic discs used for Gram-negative rods contained the following (µg per disc unless otherwise stated): ampicillin (10), ceftriaxone (30), ciprofloxacin (5), gentamicin (10), meropenem (10), (piperacillin/tazobactam (110), or cefepime (30), and were placed in the centre of the plate. Antibiotic discs used for streptococci/enterococci contained (µg per disc unless otherwies stated): penicillin (10 U), oxacillin (1), optochin (5), erythromycin (15) or clindamycin (2) for the streptococci plate, and ampicillin (10), nitrofurantoin (300), norfloxacin (10), vancomycin (30) or mupirocin (5) for the enterococci plate. Antibiotic discs used for staphylococci were penicillin (10 U) and cefoxitin (30 µg). Quality control for disc susceptibility testing was performed at the beginning of each working week using reference strains. In addition, when a new batch of discs was received, quality control using the relevant reference strain was also performed before use. The data were then analysed to determine the correlation between the two susceptibility testing methods, and identification of errors was classified according to the method of Edelmann et al. (2007), i.e. very major errors (reporting susceptible when resistant), major errors (reporting resistant when susceptible), and minor errors (reporting susceptible/ resistant when intermediate, or reporting intermediate when susceptible/resistant). In addition, the impact of direct susceptibility testing on time to effective antibiotics, time to targeted antibiotics and use of inappropriate antibiotics was reviewed. Information extracted from the clinical notes, 502 Journal of Medical Microbiology 65

Utility of direct susceptibility testing on blood cultures medication charts, and microbiology results included the pathogen, direct susceptibility results, formal susceptibility results, discrepancies between susceptibility testing methods, initial antibiotics, antibiotic therapy changes on availability of testing results, time to effective antibiotics, time to targeted antibiotics, inappropriate antibiotics due to direct testing (ineffective, too broad and resultant toxicities), occurrence of severe sepsis or septic shock (defined as sepsis complicated by organ dysfunction or sepsis-induced hypotension persisting despite adequate fluid resuscitation, respectively, according to Levy et al., 2003), and death within 7 days and 30 days of the blood culture collection date. Effective antibiotics were defined as antibiotics to which the organism tested sensitive. Targeted antibiotics were defined as antibiotics of the narrowest spectrum that would be clinically effective against the organism based on the susceptibility testing results. Toxicity associated with the antibiotic used was determined from the clinical notes. Statistics. Descriptive statistics were used including calculation of proportions and confidence intervals. The Fisher s exact and chisquared tests were used to determined differences between proportions as appropriate. RESULTS 300 bacteraemia episodes were investigated, with 247 remaining after exclusions. This comprised 82 staphylococcal episodes (33 with paired direct disc susceptibility testing and VITEK 2 results), 88 streptococcal/enterococcal episodes (52 with paired direct and formal testing results), and 77 Gram-negative rod episodes (75 with paired direct and formal susceptibility results). Discrepancies between direct and formal susceptibility testing 9.0 % (58 of 648) of antibiotics tested showed a discrepancy between direct and formal testing. In terms of bacteraemia episodes, 46 of 160 (28.8 %, 95 % CI 21.9 36.4 %) episodes for which both direct and formal susceptibility testing were available had at least one error in the direct susceptibility results. Three of the 160 episodes had single very major errors, 10 had single major errors, 24 had single minor errors, and 9 had several errors (3 very major errors, 4 major errors and 14 minor errors were detected in the 9 episodes). The error rate as a proportion of all antibiotics tested for each organism group is shown in Fig. 1. For staphylococci, the error rate for antibiotics tested was 1.5 % (1 of 66). One of 33 bacteraemia episodes (3.0 %, 95 % CI 0.1 15.8 %) had an error in direct testing results. The one discrepant result was a major error where direct testing showed cefoxitin resistance and formal testing showed cefoxitin susceptibility (Table 1). Nine of 33 (27.3 %, 95 % CI 13.3 45.5 %) bacteraemia episodes with staphylococci had severe sepsis or septic shock, but none of these nine bacteraemia episodes had an error in direct testing results. For streptococci/enterococci, the error rate for antibiotics tested was 16 % (20 of 125) (Table 1). Seventeen of 52 bacteraemia episodes (32.7 %, 95 % CI 20.3 47.1 %) had an error in direct testing results. One case had four discrepant results, and 16 cases each with one discrepant result. Thirteen of the 20 discrepancies were found with nitrofurantoin testing, an antibiotic that is not used to treat bacteraemia in any case. Excluding nitrofurantoin from the analysis, the error rate for antibiotics tested was 7.5 % (46 of 615). Of note, one major penicillin, one major oxacillin, one minor erythromycin and one minor clindamycin error all occurred for direct susceptibility testing in a Streptococcus dysgalactiae that subsequently had formal disc testing done with the correct inoculum. Eight of 52 (15.4 %, 95 % CI 6.9 28.1 %) bacteraemia episodes with streptococci/enterococci had severe sepsis or septic shock; three of these eight bacteraemia episodes (37.5 %, 95 % CI 8.5 75.5 %) had an error in direct testing results. For Gram-negative rods, the error rate for antibiotics tested was 8.3 % (38 of 457). Twenty-eight of 75 bacteraemia episodes (37.3 %, 95 % CI 26.4 49.3 %) had an error in direct testing results. Eight cases had several errors, and 20 cases with only one error each. More absolute errors occurred in this group, with the larger panel of antimicrobials tested. Significant errors occurred with piperacillin/tazobactam (17 %) and cefepime (19 %), although these were largely minor errors. Of note the discs used for direct susceptibility testing changed several times during the study period, hence different numbers were tested for each disc in the Gram-negative group (Table 1). Twenty-nine of 75 (38.7 %, 95 % CI 27.6 50.6 %) bacteraemia episodes with Gram-negative rods had severe sepsis. Fifteen of these bacteraemia episodes (51.7 %, 95 % CI 32.5 70.6 %) had an error in direct testing results, three cases had several errors, and there were 12 cases with only one error each. Antibiotic changes when direct susceptibility results were available Fifty-seven of 203 cases (28.1 %, 95 % CI 22.0 34.8 %) had their antibiotics changed when direct susceptibility results were available (Table 2). There was no significant difference in the proportions of cases for which antibiotics were changed based on direct susceptibilities between the groups (P = 0.73). Decreased time to effective and targeted antibiotics Direct susceptibilities significantly decreased time to effective antibiotics in 15 of 160 (9.4 %, 95 % CI 5.3 15.0 %) episodes, and decreased time to targeted antibiotics in 23 of 160 (14.4 % 95 % CI 9.3 20.8 %) episodes. See Table 3 for the results for the different groups of organisms. Four of 15 (26.7 %) bacteraemia episodes with reduced time to effective antibiotics due http://jmm.microbiologyresearch.org 503

V. Menon and others 18% 16% Error rate (% of antibiotics tested) 14% 12% 10% 8% 6% 4% 2% 0% Staphylococci Streptococci / Enterococci Gram negative rods Organism group Fig. 1. Direct susceptibility error rate of antibiotics tested per organism group. to direct susceptibility results were associated with severe sepsis or septic shock, when timely appropriate antibiotics are particularly important. All of these were in the Gram-negative group. Nine of 23 (39.1 %) bacteraemia episodes with reduced time to targeted antibiotics due to direct susceptibility results were seen in patients with severe sepsis or septic shock; five of these were in the Staphylococcus group, one was in the Streptococcus/ Enterococcus group, and three were in the Gram-negative rod group. Table 1. Error rate per organism group and antibiotic tested Organism group Antibiotic Minor errors Major errors Very major errors Total errors Percentage error (95 % CI) Staphylococci Cefoxitin 0 1 0 1/33 3.0 (0.1 15.8) Penicillin 0 0 0 0/33 0 Total 0 1 0 1/66 1.5 (0.0 8.2) Streptococci/enterococci Nitrofurantoin 12 1 0 13/33 39.3 (22.9 57.9) Ampicillin 0 1 0 1/31 3.2 (0.1 16.7) Penicillin 0 2 0 2/19 10.5 (1.3 33.1) Oxacillin 0 1 0 1/2 50.0 (1.3 98.7) Erythromycin 1 0 0 1/2 50.0 (1.3 98.7) Clindamycin 1 0 0 1/4 25.0 (0.6 80.6) Vancomycin 0 1 0 1/34 2.9 (0.1 15.3) Optochin 0 0 0 0 0 Total 14 6 0 20/125 16.0 (10.1 23.6) Gram-negative rods Ciprofloxacin 4 0 2 6/74 8.1 (3.0 16.8) Piperacillin/tazobactam 9 1 2 12/69 17.4 (9.3 28.4 ) Cefepime 7 0 0 7/36 19.4 (8.2 36.0) Meropenem 0 0 1 1/74 1.4 (0.0 7.3) Ampicillin 3 3 1 7/70 10.0 (4.1 19.5) Gentamicin 0 2 0 2/72 2.8 (0.3 9.7) Ticarcillin/clavulanate 1 0 0 1/16 6.3 (0.2 30.2) Ceftriaxone 0 2 0 2/46 4.3 (0.5 14.8) Total 24 8 6 38/457 8.3 (6.0 11.2) 504 Journal of Medical Microbiology 65

Utility of direct susceptibility testing on blood cultures Table 2. Proportion of episodes of bacteraemia where there was a change of antibiotics when susceptibility results were available N/A, Not applicable. Organism group Timing of antibiotic change Direct susceptibility results available VITEK 2 susceptibility results available meca and fem PCR available for staphylococci Staphylococci 31.1 % (14/45) 13.1 % (8/61) 44.7 % (17/38) Streptococci/enterococci 24.1 % (20/83) 12.2 % (5/41) N/A Gram-negative rods No direct MALDI-TOF MS identification 32.6 % (14/43) 20.9 % (9/43) N/A Direct MALDI-TOF MS identification 28.1 % (9/32) 6.3 % (2/32) N/A There was no significant difference between time to effective antibiotics between the organism groups (P = 0.12), nor between Gram-negatives that did and did not have direct MALDI-TOF MS identification (P = 1.0). There was a significant difference between the organism groups in terms of time to targeted antibiotics (P = 0.014), with a significant decrease in time to targeted antibiotics for staphylococci over the other organisms. The availability of a direct MALDI-TOF MS identification result did not significantly reduce the time to targeted antibiotics for bacteraemia with Gram-negative rods (P = 1.0). The 30-day mortality rates were similar for bacteraemia with staphylococci (15.2 %), streptococci/enterococci (15.4 %) and Gram-negative rods (10.7 %) (P = 0.68) (Table 4). All patients who died after staphylococcal and Gram-negative bacteraemias had severe sepsis or septic shock (except one person with Gram-negative bacteraemia who died of their underlying malignancy). There were no deaths among patients with staphylococcal or Gram-negative bacteraemia who had reduced time to effective antibiotics afforded by the direct susceptibility results. In contrast, only 62.5 % (5 of 8) patients who died within 30 days of their streptococcal or enterococcal bacteraemia had severe sepsis or septic shock, and two patients died despite a decreased time to effective antibiotics facilitated by direct susceptibility results. Overall, although the 7-day mortality rates were lower among those who had decreased time to effective (6.9 % vs 0 %) and targeted (7.3 % vs 0 %) antibiotics, these differences were not statistically significant (P = 0.29 and P = 0.18, respectively). There were no differences in 30-day mortality among patients who had reduced time to effective or targeted antibiotics compared to those who did not. Inappropriate antibiotics due to direct susceptibility testing Ten of 160 (6.3 %, 95 % CI 3.0 11.2 %) cases were treated with inappropriate antibiotics due to direct susceptibility results. Among these cases, antibiotics were inappropriately Table 3. Proportion of episodes of bacteraemia with decreased time to effective and targeted antibiotics by organism group Organism group Decreased time to effective antibiotics Decreased time to targeted antibiotics Staphylococci 0 % (0/33) 36.4 % (12/33) Severe sepsis or septic shock present 0 % (0/9) 55.6 % (5/9) Streptococci 11.5 % (6/52) 11.5 % (6/52) Type of Streptococcus Alpha-haemolytic streptococci 0 % (0/10) 0 % (0/10) (excluding Streptococcus pneumoniae) Beta-haemolytic streptococci 0 % (0/6) 16.7 % (1/6) Streptococcus pneumoniae 0 % (0/4) 25 % (1/4) Enterococci 18.8 % (6/32) 12.5 % (4/32) Severe sepsis or septic shock present 0 % (0/8) 12.5 % (1/8) Gram-negative rods 12 % (9/75) 6.7 % (5/75) Direct MALDI-TOF MS identification No 11.6 % (5/43) 7.0 % (3/43) Yes 12.5 % (4/32) 6.3 % (2/32) Severe sepsis or septic shock present 13.8 % (4/29) 10.3 % (3/29) http://jmm.microbiologyresearch.org 505

V. Menon and others Table 4. 7- and 30-day mortality rates by organism group Organism group No. of deaths at 7 days (percent, 95 % CI) No. of deaths at 30 days (percent, 95 % CI) Staphylococci (n = 33) 1 (3.0 %, 0.1 15.8 %) 5 (15.2 %, 5.1 31.9 %) Reduced time to effective antibiotics Yes (n = 0) 0 0 No (n = 33) 1 (3.0 %, 0.1 15.8 %) 5 (15.2 %, 5.1 31.9 %) Reduced time to targeted antibiotics Yes (n = 12) 0 3 (25 %, 5.5 57.2 %) No (n = 21) 1 (4.8 %, 0.1 23.8 %) 2 (9.5 %, 1.2 30.4 %) Streptococci/enterococci (n = 52) 4 (7.7 %, 2.1 18.5 %) 8 (15.4 %, 6.9 28.1 %) Reduced time to effective antibiotics Yes (n = 6) 0 2 (33.3 %, 4.3 77.7 %) No (n = 46) 4 (8.7 %, 2.4 20.8 %) 6 (13.0 %, 4.9 26.3 %) Reduced time to targeted antibiotics Yes (n = 6) 0 1 (16.7 %, 0.4 64.1 %) No (n = 46) 4 (8.7 %, 2.4 20.8 %) 7 (15.2 %, 6.3 28.9 %) Gram-negative rods (n = 75) 5 (6.7 %, 2.2 14.9 %) 8 (10.7 %, 4.7 19.9 %) Reduced time to effective antibiotics Yes (n = 9) 0 0 No (n = 66) 5 (7.6 %, 2.5 16.8 %) 8 (12.1 %, 5.4 22.5 %) Reduced time to targeted antibiotics Yes (n = 5) 0 0 No (n = 70) 5 (7.1 %, 2.4 15.9 %) 8 (11.4 %, 5.1 21.3 %) broad in eight cases while they were ineffective in two cases, both of whom had Gram-negative bacteraemia. One of the two cases given ineffective antibiotics had severe sepsis; however, there were no deaths at 7 and 30 days among the 10 cases who were given inappropriate antibiotics. There were no cases of toxicities due to inappropriate antibiotics given as a result of direct testing. DISCUSSION This study found that rates of error of direct susceptibility testing on blood cultures were higher than previously reported in the literature, at 9.1 % overall. Mahrer et al. (2008) reported a 3 % error rate, comparing direct disc susceptibility testing with formal testing using the CDS method. Edelmann et al. ( 2007) reported a 5.3 % error rate with Gram-positives, and an 8.2 % error rate for Gram-negatives, using CLSI interpretations. Edelmann also found a high rate of errors with respect to beta-lactam/beta-lactamase-inhibitor combinations and Gram-negative rods, and oxacillin in Gram-positive cocci, and recommended that these not be used in direct susceptibility panels of antibiotics for the respective organisms. A significant number of discrepant results between direct and formal testing occurred due to nitrofurantoin testing in streptococci/enterococci, an antibiotic not used to treat bacteraemia that could be omitted from the panel without compromising clinical decision-making. The one case of beta-haemolytic streptococcal bacteraemia that had formal disc testing revealed four errors in the direct testing results. Tan et al. (2008) found the largest error rate with beta-haemolytic streptococci, with a 15.2 % error rate for clindamycin, 8.7 % error rate for erythromycin and 8.4 % error rate for tetracycline, casting doubts on the utility of direct susceptibility testing on streptococci. In the Gram-negative category, significant errors occurred with piperacillin/ tazobactam (17 %) and cefepime (19 %), antibiotics often used empirically in the treatment of Gram-negative bacteraemia. Given that the errors were largely minor errors, one could argue that the benefit of their inclusion in direct disc testing still remains. Although there were discrepancies seen between direct susceptibility and formal testing, this did not actually result in inappropriate prescribing in the majority of cases. There were only 10 (6.3 %) cases of inappropriate antibiotics due to direct susceptibilities, and only two of these cases had ineffective (as opposed to too broad) antibiotics and there were no cases of toxicities due to inappropriate antibiotics. Overall, antibiotics were changed once the direct susceptibility data became available in a significant proportion of bacteraemia episodes (29.0 %). This resulted in a decreased time to effective antibiotics in 9.3 % and targeted antibiotics in 14.3 % of cases. The largest decrease in time to effective antibiotic was in the Gram-negative rod group (12 %), although this was not statistically significant. There was, however, a significant decrease in time to targeted antibiotics in the Staphylococcus group (39.4 %) (P = 0.014). 506 Journal of Medical Microbiology 65

Utility of direct susceptibility testing on blood cultures Gram Gram-positive cocci in clusters Gram-positive cocci in chains Gram-negative rods Day 1 PCR for fem, meca if Mon-Fri 6am-3pm Inoculate MRSA spot plate if Friday or Saturday (when PCR unavailable the following day) PCR for DDL, vana, vanb if Mon-Fri 6am- 3pm if VRE-colonized Inoculate VRE spot plate if Friday or Saturday (when PCR unavailable the following day) Direct MALDI-TOF if Mon-Fri 6am-5pm Direct susceptibility testing with: - Ampicillin - Ceftriaxone - Ciprofloxacin - Gentamicin - Piperacillin/ Tazobactam - Cefepime - Meropenem Subculture to solid agar MALDI-TOF MS identification Day 2 Set up VITEK 2 system Set up penicillin MIC if Direct susceptibility results (biomérieux) susceptibility testing alpha-haemolytic streptococci and clinically significant meca PCR if Staphylococcus aureus Set up optochin disc testing if suspected Streptococcus pneumoniae Set up disc testing if beta -haemolytic streptococci with standard inoculum using: - Penicillin - Oxacillin - Erythromycin - Clindamycin Set up VITEK 2 system (biomérieux) susceptibility testing Set up VITEK 2 system (biomérieux) susceptibility testing for enterococcal isolates PCR for DDL, vana and vanb if Enterococcus faecium Fig. 2. Suggested flow chart for susceptibility testing for the different organism groups. A further point to consider when evaluating the utility of direct susceptibility testing is the rate of contamination of blood cultures with Gram-positive organisms. Over the period of data collection, there were 100 of 305 (32.8 %) blood cultures with Gram-positive cocci in clusters that were clinically significant, and 100/135 (74.1 %) blood cultures with Gram-positive cocci in chains that were clinically significant. For the Staphylococcus bacteraemia episodes, the 36.4 % reduction in time to targeted antibiotics would then only be applicable to 36.4 % of 32.8 % (i.e. 11.9 %) of blood cultures. For the Streptococcus/Enterococcus bacteraemia episodes, the 11.5 % reduction in time to effective antibiotics, and 11.5 % reduction in time to targeted antibiotics, would then only be applicable to 11.5 % of 74.1 % (i.e. 8.5 %) of blood cultures. http://jmm.microbiologyresearch.org 507

V. Menon and others The reduction in time to targeted antibiotics for the Staphylococcus bacteraemia episodes could also be achieved by meca PCR testing on day 2. The reduction in time to effective and targeted antibiotics for the Streptococcus/Enterococcus bacteraemia episodes could also be achieved by the identification of the isolate on day 2. For example, an identification of Enterococcus faecalis may prompt an antibiotic change to ampicillin regardless of direct susceptibility results. Similarly, an identification of Enterococcus faecium may prompt an antibiotic change to vancomycin. Although direct results would still be useful for VRE (vancomycinresistant enterococci), many of the cases of VRE bacteraemia were from patients who were known to be colonized with the organism, and as such, the antibiotics were likely altered early on to cover for VRE while awaiting formal testing. In addition, in our laboratory, patients known to be colonized with VRE have their blood culture broth sent for DDL, vana and vanb PCR testing (Adams, 2006) on the day the blood culture flags if this occurs within business hours. One advantage of direct susceptibility testing in the Gram-positive cocci in chains group is that the optochin results useful for identifying Streptococcus pneumoniae are available after 24 h. However, given the characteristic appearance of Streptococcus pneumoniae on solid agar, the patient could be presumptively treated for this while awaiting the optochin result which could be made available the following day when this organism is suspected. Another method for obtaining rapid susceptibility testing results is to directly inoculate positive blood culture broth into VITEK 2 susceptibility cards, as has been done in previous studies (De Cueto et al., 2004; Romero-Gomez et al., 2012). This method would decrease time and labour, but would potentially increase costs. Although the susceptibility testing correlation with formal testing has shown results similar to our study or better, organism identification rates are suboptimal. Direct MALDI-TOF MS for Gram-negative identification could overcome this problem. Given the limitations of the MALDI-TOF MS in identification of streptococcal and staphylococcal species from blood culture broth however, this method would largely be restricted to Gramnegative bacteraemias (Jo et al., 2016). Some laboratories are now also using very early growth (after several hours of incubation) on solid agar plates to inoculate automated MIC systems on day 1, to obtain formal susceptibility results by day 2 (Idelevich et al., 2014). While this effectively eliminates the need for direct susceptibility testing, it adds logistical issues into laboratory workflow, as the susceptibility testing can only be set up after several hours when there is sufficient growth. In addition, the clinical utility of such rapid results needs to be evaluated. A recent study assessed the clinical outcome of rapid antimicrobial susceptibility testing by using PCR to detect micro-organisms grown in the presence or absence of antibiotics after only 6 h. This randomized controlled trial was not able to show a clinical benefit compared to standard susceptibility testing, which the authors attributed to a suboptimal implementation of the rapid results (Beuving et al., 2015). Limitations of this study include its retrospective nature, and low numbers limiting the precision of the estimates. In addition, it was not always clear why antibiotics were changed at the time direct susceptibility results became available. Antibiotics may have been changed based on the identification of the isolate which may have been available at the same time as direct susceptibility results, particularly for organisms with predictable susceptibility. Antibiotics could also have been changed due to side effects, drug availability, drug interactions and allergies that have not been documented in the clinical notes. One potential disadvantage of direct susceptibility testing results is that after clinicians have looked at the direct susceptibilities, particularly if they change antibiotics, they may overlook the formal testing results even when there has been a discrepancy. This is an area that could be improved, by educating clinical staff on the difference between direct and formal testing, and also by highlighting changes in results after formal testing has been completed. This was a single centre study, and the results may not necessarily be applicable to other centres. In the era of the MALDI-TOF MS and PCR testing, our findings with regards to errors rates and clinical utility suggest that direct susceptibility testing may now only be warranted in bacteraemias due to Gram-negative rods. The higher proportion of severe sepsis in Gram-negative compared to Gram-positive bacteraemias, as well as significantly reduced time to effective and targeted antibiotics, particularly in those with severe sepsis, justifies the ongoing use of direct susceptibility testing in this group. We have devised a suggested flow chart for susceptibility testing for positive blood cultures with the different organism groups (Fig. 2). This algorithm could be adapted for use in other laboratories based on locally available methods for identification and susceptibility testing. In conclusion, direct susceptibility testing does facilitate earlier effective and targeted antibiotic use in treating bacteraemia, and our findings support its on-going use in bacteraemia due to Gram-negative rods and less so for Gram-positive bacteraemias. Our suggested protocol for susceptibility testing will decrease labour and costs, and increase accuracy of results, without a significant impact on time to effective or targeted antibiotics. REFERENCES Adams, D. N. (2005). Shortcut method for extraction of Staphylococcus aureus DNA from blood cultures and conventional cultures for use in real-time PCR assays. J Clin Microbiol 43, 2932 2933. Adams, D. N. (2006). Shortcut detection of the vanb gene cluster in enterococci by a duplex real-time PCR assay. Pathology 38, 349 352. Barenfanger, J., Drake, C. & Kacich, G. (1999). Clinical and financial benefits of rapid bacterial identification and antimicrobial susceptibility testing. J Clin Microbiol 37, 1415 1418. Beuving, J., Wolffs, P. F., Hansen, W. L., Stobberingh, E. E., Bruggeman, C. A., Kessels, A. & Verbon, A. (2015). Impact of sameday antibiotic susceptibility testing on time to appropriate antibiotic 508 Journal of Medical Microbiology 65

Utility of direct susceptibility testing on blood cultures treatment of patients with bacteraemia: A randomised controlled trial. Eur J Clin Microbiol Infect Dis 34, 831 838. Clinical and Laboratory Standards Institute (CLSI). (2014). M100- S24 Performance Standards for Antimicrobial Susceptibility Testing; Twenty-Fourth Informational Supplement. De Cueto, M., Ceballos, E., Martinez-Martinez, L., Perea, E. J. & Pascual, A. (2004). Use of positive blood cultures for direct identification and susceptibility testing with the Vitek 2 system. J Clin Microbiol 42, 3734 3738. Edelmann, A., Pietzcker, T. & Wellinghausen, N. (2007). Comparison of direct disk diffusion and standard microtitre broth dilution susceptibility testing of blood culture isolates. J Med Microbiol 56, 202 207. European Committee on Antimicrobial Susceptibility Testing (EUCAST). (2012). Direct antimicrobial susceptibility testing guidelines. http://www.eucast.org/ Fay, D. & Oldfather, J. E. (1979). Standardization of direct susceptibility test for blood cultures. J Clin Microbiol 9, 347 350. Idelevich, E. A., Schüle, I., Grünastel, B., Wüllenweber, J., Peters, G. & Becker, K. (2014). Acceleration of antimicrobial susceptibility testing of positive blood cultures by inoculation of Vitek 2 cards with briefly incubated solid medium cultures. J Clin Microbiol 52, 4058 4062. Jo, S. J., Park, K. G., Han, K., Park, D. J. & Park, Y. J. (2016). Direct identification and antimicrobial susceptibility testing of bacteria from positive blood culture bottles by matrix-assisted laser desorption/ ionization time-of-flight mass spectrometry and the vitek 2 system. Ann Lab Med 36, 117 123. Joseph, N. M.Bhanupriya, B., Shewade, D. G. & Harish, B. N,. (2015). Relationship between antimicrobial consumption and the incidence of antimicrobial resistance in Escherichia coli and Klebsiella pneumoniae isolates. J Clin Diagn Res 9, DC08 12. Kumar, A., Roberts, D., Wood, K. E., Light, B., Parrillo, J. E., Sharma, S., Suppes, R., Feinstein, D., Zanotti, S. & other authors (2006). Duration of hypotension before initiation of effective antimicrobial therapy is the critical determinant of survival in human septic shock. Crit Care Med 34, 1589 1596. Leibovici, L., Shraga, I., Drucker, M., Konigsberger, H., Samra, Z. & Pitlik, S. D. (1998). The benefit of appropriate empirical antibiotic treatment in patients with bloodstream infection. J Intern Med 244, 379 386. Levy, M. M., Fink, M. P., Marshall, J. C., Abraham, E., Angus, D., Cook, D., Cohen, J., Opal, S. M., Vincent, J. L. & Ramsay, G. (2003). 2001 SCCM/ESICM/ACCP/ATS/SIS international sepsis definitions conference. Intensive Care Med 29, 530 538. Mahrer, S., Reinbott, R. & Bell, S. (2008). Blood cultures- A retrospective evaluation of direct antibiotic susceptibility testing using positive broth cultures detected by the bact-alert system. Australian Society for Microbiology. Annual Conference, Adelaide, Australia. Nelovkov, A., Pirozkova, L., Ivanova, M. & Kolesnikova, V. (2012). Direct blood culture antimicrobial susceptibility testing in an East- Tallinn central hospital, Estonia. 22nd European Congress of Clinical Microbiology and Infectious Diseases (ECCMID). Paul, M., Shani, V., Muchtar, E., Kariv, G., Robenshtok, E. & Leibovici, L. (2010). Systematic review and meta-analysis of the efficacy of appropriate empiric antibiotic therapy for sepsis. Antimicrob Agents Chemother 54, 4851 4863. Rogues, A. M., Dumartin, C., Amadeo, B., Venier, A. G., Marty, N., Parneix, P. & Gachie, J. P. (2007). Relationship between rates of antimicrobial consumption and the incidence of antimicrobial resistance in Staphylococcus aureus and Pseudomonas aeruginosa isolates from 47 French hospitals. Infect Control Hosp Epidemiol 28, 1389 1395. Romero-Gómez, M. P., Gómez-Gil, R., Paño-Pardo, J. R. & Mingorance, J. (2012). Identification and susceptibility testing of microorganism by direct inoculation from positive blood culture bottles by combining MALDI-TOF and Vitek-2 compact is rapid and effective. J Infect 65, 513 520. Tan, T. Y., Ng, L. S. & Kwang, L. L. (2008). Evaluation of disc susceptibility tests performed directly from positive blood cultures. J Clin Pathol 61, 343 346. Yu, F., Lin, M., Lee, J., Lian, Y., Lin, C., Chen, C. & Wang, G. (2010). Comparison of antimicrobial susceptibility testing of isolates from blood cultures by direct inoculation method and PHOENIX. Journal of Biomed Lab Science 2011 23, 23 28. http://jmm.microbiologyresearch.org 509