c e L e r ib L e r tu y r a Non-Cultural Methods for Aspergillosis lin n O D or I M h t C u S a E y b © Kieren A. Marr MD Professor of Medicine, Johns Hopkins School of Medicine Professor of Oncology, Sidney Kimmel Comprehensive Cancer Center Professor of Business, JH Carey School of Business Director, Transplant and Oncology Infectious Diseases Disclosures r ib L e r tu y r a • Consultant / advisory board c – Astellas, Merck, Pfizer e L • Research grantne i l – Astellas, Merck n O • Licensed technology / patent D I r o – MycoMed Technologies M h t C u S a E y b © 2 Non-Culture based Methods • Background • Available methods L e c e n i l n • Methods in development O • Remaining problems D questions, I r o M h t C u S a E y b © – Lessons learned r ib L e r tu y r a 3 Background: Non-cultural methods r ib L e r tu y r a • Detection of a biomarker • A biomarker is anything that can be used as an indicator of a particular disease state or some other physiological state of an organism L e c e n i l n O D I r – Diagnose disease risk o M h t C –S Detect theupresence of a disease a E– Surrogate endpoint to measure therapy y b © 4 Background: Diagnosis of IFI r ib L e r tu y r a • Culture of an organism is actually a biomarker • Antigen-based diagnostics have been used for many infections L e c e n i l n – PJP, Histoplasma, Coccidioides, Cryptococcus • IA: Culture sensitivity for Aspergillus in lavage and tissues is poor (<50%) – Yet.. Considered “gold standard” O D or I M h t • Technology is appropriately altering the way that C u S a aspergillosis Ewe diagnose y b © Commercially available tests • Country specific • Antigen – based assays n i l n L e c e r ib L e r tu y r a – Galactomannan Enzyme Immunoassay (GM EIA) • Serum, BAL O D I r • Blood o M h t C Molecular assays (PCR) u S a E – Blood, y compartments (BAL, tissue) b © – b-glucan • 6 Galactomannan • • • • r ib y r a L Linear core of mannan with a1,2 and a1,6 elinkages r u Platelia Aspergillus: EbA2 detects antigenic side chain of b1,5 t c galactofuronosyl (multiple epitopes) e L – Double sandwich ELISA; monoclonal IgM e n Serum i l n BAL O D or I M h t C u S a E y b © Performance of GM EIA r ib L e r tu y r a • A lot of controversy, still • Historical literature is irrelevant – Utilized inappropriate (high) cut-offs – Difficulty interpreting clinical studies • Better studies demonstrate utility, especially when integrated into clinical context, and used with other markers – Screening in high-risk patients – Adjunct to diagnosis in people with high probability of infection– BAL and serum – To measure clinical response • History with this assay has taught a lot of lessons 8 L e c e n i l n O D or I M h t C u S a E y b © Lesson #1. Do it right from the y r a r beginning b e r tu i L • Early failure to define appropriate endpoints for this test hampered progress c e L e – Cut-off shifted from 1.5 1.0 0.5 • Sensitivity / Specificity using this cut-off: 80% / 80% n i l n ROC CURVE O D or I M h t C u S a E y b © 100 Sensitivity (N=24 Pts.) 90 0.2 0.4 0.1 0.0 0.3 80 Timing (proven / probable) Cut-off 1.5 Cut-off 1.0 Cut-off 0.5 IA diagnosis -1 days -1 days -10 days Clinical onset 0 days 0 days -6 days 0.5 70 0.6 0.8 60 1.0 50 1.5 40 30 20 10 0 0 10 20 30 40 50 60 70 80 1- Specificity ( N=472 sera) 90 100 Marr et al. J Infect Dis (2004) 190(3): 641-9 Leeflang et al. Cochrane Database Syst Rev 2008 9 Lesson #2. Analyses need to account for disease • L e n i l n • • • c e O D or I M h t C u S a E y b © r ib L e r tu y r a Difficulty in assessing DISEASE vs. NO DISEASE • Presence of organism does not necessitate disease Risk Natural history of infection not well understood • When do people become infected? If culture is the gold standard, does this person really have disease? • What is a false-positive vs. false negative test? Marr and Leisenring. Clin Infect Dis 2005:41: S381 10 Statistics y r a r ib L • If timing of the test matters, antifungals effect e r u t performance, and multiple tests are taken c e per patient, then per-patient analyses are L e difficult to interpret n i l n – Per patient: one person has disease O • Case vs. control D I r • Timing of test,o and other time-dependent variables not M h accounted for appropriately t C u S a analyses E– Per test y • Need b larger datasets © 11 Per-test analyses • y r a r ib L Allow depiction of changes in sensitivity according to timing e r relative to clinical diasnotis and antifungals (fungal burden) u t c e L e n i l n O D or I M h t C u S a E y b © No antifungals With antifungals Marr et al. Clin Infect Dis 2005; 40:1762 - 9 What do you do when you have y r a sub-optimal “gold standard”? bra e r tu i L • Our tests now work better than old gold-standards (culture, histopathology) • One solution: Have a consensus committee define disease using new (accepted) tests in a syndromic definition: EORTC/MSG: possible, probable, proven – Arbitrary (opinion) – Definitions derived for treatment trials (conservative) – Definitions not validated using large databases – What happens if method is better than those 13 included? L e c e n i l n O D or I M h t C u S a E y b © Imperfect gold standards L e r ib r u – Proven / probable (informed by radiography, clinical t c e signs / symptoms, culture of respiratory tract, L histopathology and galactomannan (serum and e BAL) defines IA lin n • If sensitivity is < 100%, specificity of a new diagnostic will O be falselyD low I r o • If specificity is < 100%, specificity will be falsely low M h t C u NewSscience (statistical) solution a E y – Bayesian latent class models (calculates results b without © assuming a gold-standard) • Convention • y r a 14 “New” Developments: GM EIA • y r a r ib L Continued understandings and uses of the e r u galactomannan EIA t c e – Galactomannan (or polysaccharides that crossL react with the EbA2 antibody) is present in many e n i l sources n • Non-Aspergillus O fungi (ex. Histoplasma, Fusarium) D Food • Blood products. products (popsicles) I r o M • Antibiotics h t C u S – Biological vs. technical (test) false-positive a E y b © Nucci et al. PLoS One 2014 Guigue et al. NEJM 2013; 369(1) Martin-Rabadan et al. Clin Infect Dis 2012 55(4) 15 “New” Developments: GM EIA • r ib L e r tu y r a BAL – Early studies (HSCT): Increased sensitivity compared to culture – Cut-off 0.5 – 1.0 (depending on study / population evaluated) c e L e Pooled SEN (95% CI) Pooled SPE (95% CI) Overall analyses 13 0.90 (0.79-0.96) 0.94 (0.90-0.96) Cutoff of 0.5 for positivity 8 0.86 (0.70-0.94) Cutoff of 1 for positivity 11 Cutoff of 1.5 for positivity Cutoff of 2 for positivity Studies No. Studies n i l n Pooled PLR (95% CI) Pooled NLR (95% CI) 14.87 (8.8924.90) 0.10 (0.04-0.24) 0.89 (0.85-0.92) 7.69 (5.7510.28) 0.15 (0.07-0.35) 0.85 (0.72-0.93) 0.94 (0.89-0.97) 14.29 (8.3324.50) 0.16 (0.08-0.31) 9 0.70 (0.49-0.85) 0.96 (0.93-0.98) 18.97 (10.9332.93) 0.31 (0.17-0.57) 5 0.61 (0.38-0.80) 0.96 (0.92-0.98) 16.13 (8.0732.25) 0.40 (0.23-0.70) O D or I M h t C u S a E y b © Becker MJ, Br J Haematol 2003 Musher et al, J Clin Microbiol 2004 Maertens et al, Clin Infect Dis 2009 Guo et al, Chest 2010 16 BAL and Serum c e • Single center cohort of 210 allogeneic HSCT • All methods have independent contributions to diagnosis • BAL GMI sensitivity not impacted by antifungals L e n i l n O D or I M h t C u S a E y b © r ib L e r tu y r a Fisher et al. Transplant Infect Dis 2014 17 GM EIA: “Funny” Issues • Interpretation of result based on OD sample / OD threshold control (TC is said to be serum sample with predicted concentration of 1ng/mL) • Significant lack of reproducibility associated with different TC results from lot to lot • Sample to sample variability observed in several studies (high positive / low negative from run to run) L e c e n i l n O D or I M h t C u S a E y b © r ib L e r tu Upton et al. J Clin Microbiol 2005 Oren et al. Transplant Infect Dis 2011 Furfaro et al.; Bizzini et al Transplant Infect Dis 2012 y r a 18 “New” Developments: b-D glucan • Nature of marker: Less specific than galactomannan • Assays differ worldwide (in reagents and cut-offs) • Cut-offs established for Candida infections in neutropenic patients • Comparative GM vs. BDG study in 105 patients L e c e n i l n O D or I M h t C u S a E y b © – GM more specific; BDG more sensitive (high FP rate with bacteremia) r ib L e r tu y r a Sulahian et al. J Clin Microbiol 2014 19 PCR - Blood • Many assays, different performance, commercial availability & clinical use varies across countries • Blood: Meta-analysis – 10,000 blood, serum or plasma from 1618 patients from 16 studies – Sens 75%; Spec 87% L e c e n i l n O D or I M h t C u S a E y b © r ib L e r tu y r a Mengoli et al. Lancet Infect Dis 2009 20 PCR - BAL • Meta-analysis of17 studies with 1191 patients – Sens 91%, Spec 92% • Comparative studies vary in conclusions about performance relative to GM EIA • Commercial assays available with good data • Combination of markers may provide best results L e c e n i l n O D or I M h t C u S a E y b © r ib L e r tu y r a Sun et al. PloS One 2011 Guinea et al. PloS One 2013 Torelli et al. J Clin Microbiol 2011 Heng et al. Diag Microbiol Infect Dis 21 2014 New Tests r ib L e r tu y r a • Aspergillus-LFD: JF5 – extracellular glycoprotein • Small prospective studies on serum assay – comparable to GM EIA for screening; lower sensitivity if need confirmatory result • BAL - comparison of 4 methods in BAL (culture, GM, PCR, LFD). Sens 70 – 88%; performance best with combination • Caution: all these results are biased towards GM if that is accepted in definitions of gold-standard “proven / probable” L e c e n i l n O D or I M h t C u S a E y b © Wiederhold et al. J Clin Microbiol 2013 Held et al. Infection 2013 White et al. J Clin Microbiol 2013 Hoenigl et al. J Clin Microbiol 22 2014 In Development • Urinary GM – like antigens – MAb476 n i l n L e • Electronic nose technology • HPLC-MS/MS • PCR – MS O D or I M h t C u S a E y b © c e r ib L e r tu y r a Dufresne et al. PloS One 2012 De Heer et al. J Clin Microbiol 2013 23 Cerqueira et al. PLoS One 2014 Many Molecular Platforms • At this meeting: L e c e r ib L e r tu y r a – Several qPCRs – PCR that also detects resistance determinants – Detection of gliotoxin n i l n O D or I M h t C u S a E y b © 24 Conclusions L e c e n i l n O D isonotr going to help I – A bigger telescope you findM the next tplanet when you don’t h C u know how to use it S a –E Better understanding of disease, risks y b of analysis and methods © r ib L e r tu • Culture based diagnostics -poor • Markers are a mandatory adjunct to culture • Disease is not well understood • Better technology is not the only answer y r a 25 L e c e r ib L e r tu y r a n i l Thank you n O Dhave aogold I r Do we really standard that can support M h analyses of newttechnologies without bias? C u S a E y b © 26
© Copyright 2024