Antifungal susceptibility testing: Which method and when?

Antifungal susceptibility testing: Which method and when? Maiken Cavling Arendrup mad@ssi.dk SSI & Juan Luis Rodriguez Tudela jlrtudela@isciii.es ISCIII

Agenda Summary of current standards and selected commercial methods BPs and BP setting process CLSI EUCAST Indications for AFST Conclusion

In vitro susceptibility testing Broth-dilution Macro-format Micro-format CLSI Reference method EUCAST Reference method Odds method SensiTitre Disk / Etest diffusion Vitek Agar dilution

Yeast standards Free at EUCAST web page!!! 55 $ (non member price including supplement)

CLSI M7-A><A><A3 & EUCAST M7-A M7-A M7-A3 EUCAST Glucose /Inoculum size.%,.5-.5 x 3 %,.5-.5 x 5 Plates & Reading Round bottom & Visual Flat & Spec. Incubation time h - h h h End point % inhib. 5% inhib. 5% inhib. 5% inhib C. parapsilosis ATCC 9 C. krusei ATCC 65 - µg/ml.5- / - µg/ml.5- µg/ml.5- µg/ml 6-6 µg/ml -6 / 6- µg/ml -6 µg/ml 6-6 µg/ml

Antifungal S, R breakpoints for Candida Breakpoints given as: S: X; R: >Y Proposed Breakpoints- not yet approved!! CLSI EUCAST (specific for) Amphotericin B ; > Anidulafungin.3; >.3 (C. alb).6; >.6 (C. g,k,t) Caspofungin - Micafungin - Fluconazole ; >3 ; > (C. a,t,p) Itraconazole.5; >.5 - Posaconazole -.6; >.6 (C. a,t,p) Voriconazole ; >.5; >.5 (C. a,t,p) EUCAST BP: Clin Microbiol Infect : 93-5 & : 95 97

Vitek, Sensititre, Etest & Ref methods Cuenca-Estrella M JCM In press

Aspergillus Standards Free at EUCAST web page!!! $ (Non member price)

Differences between CLSI M3-A & E. Dis 9. CLSI M-3-A EUCAST-AFST E. DEF 9. Glucose Shape of well Inoculum (cfu/ml) Incubation (h) Reading Endpoint. % Round bottom.5 -.5 x Visual optically clear & MEC for echinocandins % Flat bottom -5 x 5 Visual optically clear & MEC for echinocandins

A. fumigatus ECOFFs & BP suggestions ECOFF / Susceptible Intermediate Resistant Amphotericin B µg/ml - - Itraconazole µg/ml > Voriconazole µg/ml > Posaconazole.5 µg/ml.5 >.5 Rodriguez-Tudela AAC, Verweij Drug Res Updates 9, Pfaller JCM 9

o Voriconazole: 93.3 to % agreement (within 3 dilutions) o independent of the E test incubation time (for four of five species at both incubation times). o Amphotericin B, itraconazole, and posaconazole, o E test results were more dependent on the incubation time for certain species. o Posaconazole and A. fumigatus, o highest agreement at h (95.%) (versus 9% at h)

Agenda Summary of current standards and selected commercial methods BPs and BP setting process CLSI EUCAST Indications for AFST Conclusion

CLSI vs EUCAST breakpoint establishing procedure CLSI EUCAST MIC distributions PK/PD MIC/Clinical outcome relationships Cumulative % per species From animal models or humans Rule 9-6 Several data sets Epidemiological cutoffs per species Target validated by means of Montecarlo simulations Data mining BP never higher than ECOFF unless supported by clinical data

Fluconazole

3 5 6 7 9.6.3.6.5.5.5 6 3 6 56 5 C. tropicalis 3 5 6 7.6.3.6.5.5.5 6 3 6 56 5 C. albicans 6 6.6.3.6.5.5.5 6 3 6 56 5 C. glabrata 5 5 5 3 35.6.3.6.5.5.5 6 3 6 56 5 C. krusei 6.6.3.6.5.5.5 6 3 6 56 5 C. parapsilosis

Summary of epidemiological cut offs values for this bug drug combination EUCAST methodology C. albicans < mg/l C. parapsilosis < mg/l C. tropicalis < mg/l C. glabrata < 3 mg/l C. krusei < mg/l

Candidaemia. 6 patients & 6 episodes Oropharyngeal candidiasis. HIV+ patients & 3 episodes A total of 5 episodes of Candida infection were available for analysis

Distribution between successes and failures of 5 episodes

EUCAST Clinical Data MIC in mg/l > mg/d No cure/total % response <.5 /33 93. / / / 66.7 9/37.3 >6 3/6.7

Strategy to analyze results by data mining Different models have been used: Two decision trees: J and CART A classification rule: One R A probabilistic classifier: Naive Bayes A logistic regression model: Simple logistic Performance of the tests were assessed by means of: Sensitivity and specificity False positive rate Area under ROC curve Matthews Correlation Coefficient (MCC) To check if different algorithms produce similar answers Cuesta I. et al. AAC. 9. 53:99-5.

EUCAST clinical data Classifiers MIC value for failures in mg/l Sensitivity in % False positive rate in % Area under ROC curve J > 7.6 CART > 7.9 one R > 7.9 Naïve Bayes Simple Logistic > 9 3.9 > 9 3.9

CART for all set: 5 cases Predicting failures Node Class = Success LOG_MIC <=.5 Class Cases % Failure 39.5 Success 56 6.5 W = 5. N = 5 MIC= mg/l Relative error:. ROC area:.9 Sensitivity: 9% False positive rate: % LOG_MIC <=.5 Terminal Node Class = Success Class Cases % Failure 3.3 Success 9.7 W = 57. N = 57 LOG_MIC >.5 Terminal Node Class = Failure Class Cases % Failure 9. Success.9 W =. N =

Clinical BPs for fluconazole have been validated by data mining Susceptible Intermediate Resistant Apply to C. albicans, C. parapsilosis and C. tropicalis But not to EUCAST < mg/l mg/l > mg/l C. krusei which is inherently resistant C. glabrata because BPs divide wild type populations & only 9 cases for analysis

Fluconazole Anidulafungin Caspofungin Micafungin mg/ mg mg/ mg 7 mg/5 mg mg 5 mg C. glabrata 5 56 67 6

% of species causing fungal infections Specie Candidemia N (%) OPC N (%) All N (%) Candida albicans 73 3 5 (79.) Candida parapsilosis 7 --- 7 (.) Candida tropicalis --- (.6) Candida glabrata 9 --- 9 (3.) Other yeast* 5 --- 5 (.9) All 6 (.) 3 (5.) 5 Other yeast comprise an isolate of the following species: C. krusei, C. guilliermondii, C. kefyr, C. dubliniensis and Geotrichum capitatum

One C. glabrata susceptible (MIC = mg/l) to fluconazole according EUCAST BPs S I R < > 5 repetitions in different days

One C. glabrata susceptible (MIC = mg/l) to fluconazole according EUCAST BPs 5 S < S I R < > 5 5 5 repetitions in different days

One C. glabrata susceptible (MIC = mg/l) to fluconazole according EUCAST BPs 5 S < I S I R 5 < > 5 5 repetitions in different days

One C. glabrata susceptible (MIC = mg/l) to fluconazole according EUCAST BPs 5 S < I R > S I R 5 < > 5 5 repetitions in different days

6 6 BPs divide wild-type population of C. glabrata S < I R >.6.5.5.5 6 3 6 56 5.3.6

6 6 BPs divide wild-type population of C. glabrata S < 3.6.5.5.5 6 3 6 56 5.3.6

BPs divide wild-type population We do not have of C. glabrata clinical data 6 showing MICs < S < 3 3 mg/l respond to fluconazole 6.6.3.6.5.5.5 6 3 6 56 5

35 3 5 5 5 BPs identify C. krusei as intrinsically resistant to FZ S < I R >.5.5.5 6 3 6 56 5.6.3.6

Voriconazole CLSI vs EUCAST

Step : Voriconazole MIC distribution MIC distribution Pfaller JCM 6

Step : MIC In vivo outcome Pfaller JCM 6

Step Logistic analysis MIC-outcome 9 cases BP suggestion R > μg/ml S.5 /.5 / μg/ml?? BP for resistance (response <6%)

CART analysis of studies # 63, 6, 39/6 & 3/66 Node Class = yes MIC <=.9 Class Cases % no 7. yes 79 7.9 W = 9. N = 9 ROC area for VZ:.6 MIC <=.9 Terminal Node Class = yes Class Cases % no 36.6 yes 3 7. W = 67. N = 67 MIC >.9 Terminal Node Class = no Class Cases % no 3.5 yes 5.5 W =. N =

Step 3: PK/PD data Mouse model Isolates: C. albicans MIC.7.5 μg/ml Human PK/PD AUC ~ μg*h/ml supports a susceptibility BP of μg/ml Andes AAC 3

Montecarlo simulation for a PD target

Choice of susceptibility breakpoint Arguments for S μg/ml No. of isolates 35 3 5 5 Candida spp C. tropicalis C. parapsilosis C. krusei C. glabrata C. albicans 5...3.6.3.5.5 MIC distributions Not bisect C. glabrata population includes 99% of isolates Outcome <6% response if MIC > μg/ml for average not for 95% population PK/PD Serum levels frequently μg/ml AUC/MIC will be. MIC (μg/ml)

European concerns Outcome C. glabrata 55% versus 7-3% for C. albicans, C. tropicalis and C. parapsilosis Logistic regression for CLSI BPs based on average population and not the lower 95% CI Montecarlo simulations: Target AUC/MIC ~ covers 99% population at a MIC of.5 mg/l for iv MIC of.5 for oral EUCAST data mining does not show any statistical support for in vitro in vivo outcome

EUCAST approach for Voriconazole ECOFFs as breakpoint C. glabrata and C. krusei No breakpoints, Insufficient data Voriconazole Rationale for the EUCAST clinical breakpoints, version May. MIC wild type distributions (based on MIC-values determined with EUCAST, CLSI and Etest methodology) Species.... 6. 3. 6. 5. 5.5 6 3 6 ECOFF (mg/l)* Candida albicans 5 996 53 9 53 5 7 53 7 6 63 3 7 6.5 Candida glabrata 5 6 63 9 53 7 3 9 6 3 9 69 6 Candida krusei 3 9 5 9 9 3 Candida parapsilosis 6 7 3 6 5 9 7 3 77 3 6 3.5 Candida tropicalis 33 6 93 73 7 3 5 3 6 3.5

Echinocandins BPs

Step MIC distributions & ECOFFs Anidulafungin. MIC wild type distributions (based on MIC-values determined with EUCAST and Etest methodology*) Species....6.3.6.5.5.5 6 3 6 Tot ECOFF (mg/l)** Candida albicans Full distributions 36 77 6 6 95 Truncated at.3 6 3 63.6-.3 Truncated at.6 33 5 Candida glabrata Full distributions 55 3 6 9 6 6 39 Truncated at.3 33 37.3-.6 Truncated at.6 3 9 6 Candida krusei Full distributions 7 3 6 Truncated at.3 5 7 65.6 Truncated at.6 Candida parapsilosis 3 6 36 7 7 96 3 7 9 Candida tropicalis Full distributions 9 3 Truncated at.3.3-.6 Truncated at.6 Candida guilliermondii 5 7 6 33

Step : MIC - outcome data No such data using EUCAST! Anidulafungin superior to fluconazole C. albicans (35 patients) (% >< 6%) P<.5 C. tropicalis ( patients) (93% >< 5%) P<.5 C. glabrata (3 patients) (56% >< 5%) Candida spp. (7 patients) (75% >< 67% ) Anidulafungin inferior to fluconazole C. parapsilosis (3 patients) (6% >< 3%) Good targets for anidulafungin Poor target for anidulafungin Reboli N Engl J Med 7

Step : MIC distribution fks mutants C. albicans C. glabrata A N I No. of isolates 9 7 6 5 3...3.6.3.5.5 MIC (μg/ml) C. albicans (S) C. albicans (R) 6 No. of isolates 9 7 6 5 3...3.6.3.5.5 MIC (μg/ml) C. krusei C. tropicalis C. glabrata (S) C. glabrata (R) 6 MICs for fks mutants C. albicans >.3 C. glabrata >.6 C. krusei >.3/ >.55 C. tropicalis >.6 No. of isolates 6 C. krusei (S) C. krusei (R) No. of isolates 9 7 6 5 3...3.6.3.5.5 6...3.6.3.5.5 6 C. tropicalis (S) C. tropicalis (R) MIC (μg/ml) MIC (μg/ml) Arendrup AAC

Step 3: No PK/PD ECOFFs as BP Clinical breakpoints C. albicans : S.3, R >.3 mg/l C. glabrata : S.6, R >.6 mg/l C. tropicalis : S.6, R >.6 mg/l C. krusei : S.6, R >.6 mg/l C. parapsilosis & C. guillermondii : no Breakpoints (not good targets) A N I No. of isolates 9 7 6 5 3 C. albicans C. glabrata C. krusei C. parapsilosis C. tropicalis...3.6.3.5.5 C. albicans (S) C. albicans (R) 6 No. of isolates 9 7 6 5 3...3.6.3.5.5 C. glabrata (S) C. glabrata (R) 6 No. of isolates 6...3.6.3.5.5 C. krusei (S) C. krusei (R) No. of isolates 6 7 6 5 3...3.6.3.5.5 C. parapsilosis (S) 6 No. of isolates 9 7 6 5 3...3.6.3.5.5 C. tropicalis (S) C. tropicalis (R) 6 MIC (μg/ml) MIC (μg/ml) MIC (μg/ml) MIC (μg/ml) MIC (μg/ml) Arendrup et al AAC and EUCAST-AFST

Mis-classifications using CLSI BP CLSI BP: Susceptible μg/ml fks hot spot mutants classified as S Anidulafungin Caspofungin Micafungin VMEs CLSI 9.% 6.7% 9.9% Arendrup et al AAC

Etest and Anidulafungin 9 7 6 5 3 9 7 6 5 3 <.....5.3.6.5.5.5 6 >3 No. of isolates <.....5.3.6.5.5.5 6 >3 No. of isolates A N I C. albicans C. glabrata EUCAST draft BP CLSI BP MIC (μg/ml) MIC (μg/ml) C. krusei C. tropicalis C. parapsilosis 9 7 6 5 3 9 7 6 5 3 <.....5.3.6.5.5.5 6 >3 <.....5.3.6.5.5.5 6 >3 No. of isolates No. of isolates 9 7 6 5 3 <.....5.3.6.5.5.5 6 >3 No. of isolates MIC (μg/ml) MIC (μg/ml) MIC (μg/ml) Arendrup et al AAC and EUCAST-AFST

Agenda Summary of current standards and selected commercial methods BPs and BP setting process CLSI EUCAST Indications for AFST Conclusion

When? Guiding therapy Invasive mycoses Immunocompromised patients Failure cases Breakthrough isolates preceeding AF therapy Surveillance etc. Epidemiologic studies Uncommon species New compounds Cross resistance issues Breakpoint development process

How? Standardized methodology EUCAST CLSI Commercial method validated against reference methodologies ASTY,Etest, Sensititre, Vitek Careful selection of appropriate breakpoints

Conclusion Why do we differ?? CLSI BPs species with highest MIC drives the BP BPs have not been adjusted to change in methodology Risk of misclassifying R isolates as S. Very major errors!!!! EUCAST BPs Conservative approach Clinical correlation has to have statistical support if data arise showing isolates with resistance mechanisms are good targets BPs have to be raised My recommendation: Use EUCAST BPs

Antifungal S, R breakpoints for Candida Breakpoints given as: S: X; R: >Y Proposed Breakpoints- not yet approved!! CLSI EUCAST (specific for) NRMM (9) Ampho ; > ; > Anidula.3; >.3 (C. alb).6; >.6 (C. g,k,t).5 DK.5 SE Caspo - DK, NO SE, FI Mica - - Fluco ; >3 ; > (C. a,t,p) ; > Itra.5; >.5 -.5; >.5 Posa -.6; >.6 (C. a,t,p).5 DK, SE Vori ; >.5; >.5 (C. a,t,p).5 EUCAST BP: Clin Microbiol Infect : 93-5 & : 95 97

Thank you for your attention STATENS SERUM INSTITUT