Download Translational Modeling Tools for Combination Drug Development

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Transla'onal Modeling Tools for Combina'on Drug Development Tawanda Gumbo, MD Office of Global Health & Department of Medicine University of Texas Southwestern Medical Center Dallas, Texas Siyambalapitiyage C, Norton C, Musuka S, Srivasatava S, Sherman C
Preclinical model •  Add PK system for one drug •  Disease pathology •  Cell type and metabolic status (PD readout) •  Add PK system for second, third, and 4th drug Experiment QuanFtaFve output: PK readout PD read: # of cells (live& dead), #resistant cells Molecular marker readout (RNA, protein, etc) Modeling and simulaFon Clinical comparison QuanFtaFve PredicFons for Drug CombinaFons (n=12 exp.) Assessment of accuracy in forecasFng clinical efficacy Hollow fiber model of TB example Hollow fiber PK system: in academia for GNR and GPC (Blaser, Stone & Zinner) Funding from philanthropy & intramural funds Hollow fiber model of TB: 2001-­‐2003 Funding from NIH Funding from BMGF 1,2, 3 drug Funding from industry studies (2006 -­‐2012) QuanFtaFon of predicFve accuracy of HFS-­‐TB for Efficacy Cri'cal Path/CPTR Regulatory bodies: FDA/EMA AddiFon of toxicity module Hollow fiber model of TB (HFM-­‐TB) Mtb has complex metabolic phenotypes relevant to therapeu'cs, and thus the HFM-­‐TB is actually several models Log-­‐phase growth Mtb in ambient air Semi-­‐dormant Mtb at low pH NR Mtb under hypoxia Intracellular Mtb: macrophages Intracellular Mtb: neutrophils MOXIFLOXACIN in the hollow fiber •  We wished to examine the relaFonship between MOXI bactericidal acFvity and the emergence of resistance •  We simulated the human PK profile of a 100 to 1600 mg daily MOXI regimens •  Mtb cultures were sampled during the experiment in order to determine the effect of therapy on the total microbial populaFon as well as the MOXI-­‐resistant populaFon Gumbo T, et al. J Infect Dis. 2004 ;190(9):1642-­‐51. Gumbo T et al. J Infect Dis. 2004;190:1642-­‐1651 © 2004 by the Infectious Diseases Society of America
MOXIFLOXACIN AGAINST M. TUBERCULOSIS
Simulation for Dose Selection
•  Given that MOXI AUC/MIC raFos of >53 (total drug 106) prevented resistance amplificaFon, we wished to examine the probability of afaining this exposure across a range of dosing regimens •  We conducted a 10,000 subject Monte Carlo simulaFon –  UFlized a populaFon pharmacokineFc model, and –  UFlized the MIC distribuFon for MOXI against Mtb Dose (mg/day)
Probability of Target
Attainment
400
53%
600
87%
800
93%
Gumbo T, et al. J Infect Dis. 2004 ;190(9):1642-­‐51. • 
Study design
Three drug therapy in HFS (INH, RIF, PZA), with 3 different t1/2 in HFS •  Bactericidal effect: drug suscep'ble •  Sterilizing effect: drug suscep'ble •  Bactericidal effect: pre-­‐seeded with INH resistant (katG 315) and RIF resistant (rpoB S531L) of 0.5% total propor'on each •  Different paderns of non adherence examined: random forgeeng, start-­‐stop, start-­‐stop-­‐start-­‐stop •  Dura'on of therapy: 28 & 56 days Srivastava S, et al. J. Infect. Dis. 2011; 204:1951-9.
DOTS & ADHERENCE •  DOTS an organizing principle for delivery of combina'on therapy in B •  How much non-­‐adherence leads to acquired drug resistance? •  How much non-­‐adherence leads to therapy failure? Srivastava S, et al. J. Infect. Dis. 2011; 204:1951-9.
Drug Resistance
— There was no emergence of resistance to
INH or RIF
Srivastava S, et al. J. Infect. Dis. 2011; 204:1951-9.
Modeling
& simulation assumptions
•  PK and PK variability from Western Cape, South Africa,
• 
• 
• 
• 
patients (studies by Wilkins & McIlleron)
Resistance to INH and RIF arises as Poisson type
event: 100% biofitness
Patients 100% adherent to INH, RIF and PZA
Resistance rates constrained to those observed with
monotherapy in clinical studies in the 1960s and 1970s
How many patients on the REGIMEN will be
effectively on monotherapy due to PK variability and
in what proportion will MDR-TB arise?
Srivastava S, et al. J. Infect. Dis. 2011; 204:1951-9.
External validation of model: sputum conversion rates in 10,000 patients
Sputum conversion rate predicted = 56% of pa'ents Sputum conversion rate from prospec've clinical studies in Western Cape= 51-­‐63% Srivastava S, et al. J. Infect. Dis. 2011; 204:1951-9.
Many (simulated)
patients had 1-2 of the 3
drugs at very low concentration throughout,
leading to monotherapy of the remaining drug
Drug resistance predicted to arise in 0.68% of
all pts on therapy in first 2 months despite
100% adherence
Srivastava S, et al. J. Infect. Dis. 2011; 204:1951-9.
Pooled risk differences for defaulFng in paFents on directly observed therapy compared to self-­‐administered therapy. Pasipanodya J G , and Gumbo T. Clin Infect Dis. 2013;57:21-­‐31 © The Author 2013. Published by Oxford University Press on behalf of the Infectious Diseases
Society of America.
Pooled risk differences for microbiologic failure in paFents on directly observed therapy compared to self-­‐administered therapy.
Pasipanodya J G , and Gumbo T. Clin Infect Dis. 2013;57:21-­‐31 © The Author 2013. Published by Oxford University Press on behalf of the Infectious Diseases
Society of America.
Pooled risk difference for relapse on directly observed therapy compared to self-­‐administered therapy. Pasipanodya J G , and Gumbo T. Clin Infect Dis. 2013;57:21-­‐31 © The Author 2013. Published by Oxford University Press on behalf of the Infectious Diseases
Society of America.
Effect of directly observed therapy vs self-­‐
administered therapy on acquired drug resistance.
Pasipanodya J G , and Gumbo T. Clin Infect Dis. 2013;57:21-­‐31 © The Author 2013. Published by Oxford University Press on behalf of the Infectious Diseases
Society of America.
ProspecFve combinaFon therapy studies for validaFon •  STUDY: Cape Town •  142 paFents, hospitalized for 2 months: 100% adherence •  Examined full scale PKs at 8 weeks –  compartmental PKs for each pa'ent for each drug •  2 month sputum conversion, 'me to sputum conversion •  6 (4%) pa'ents became either nonadherent or absconded in the last 4 months of therapy. •  Followed up to 2 years; outcomes in pa'ents who were non-­‐adherent ascertained. Pasipanodya JG, McIlleron H, Burger A, Wash P, Smith P, Gumbo T. J Infect Dis. 2013;infdis.jit352 ClassificaFon and Regression-­‐tree analysis: Variables predicFve of poor long-­‐term outcome in 142 paFents. Virtually the same opFmal drug exposures for efficacy as idenFfied in HFS-­‐TB studies between 2007 and 2010! Pasipanodya J G et al. J Infect Dis. 2013;infdis.jit352 © The Author 2013. Published by Oxford University Press on behalf of the Infectious Diseases
Society of America. All rights reserved. For Permissions, please e-mail:
[email protected].
Low Pyrazinamide or low Rifampin •  Relapse: •  OR:
None low
Both low 0/64 17/60 51.90 (95% CI, 3.04–886) •  0.7% developed acquired drug resistance (0.68% forecast from HFS-­‐TB plus M&S) Pasipanodya J G et al. J Infect Dis. 2013;infdis.jit352 © The Author 2013. Published by Oxford University Press on behalf of the Infectious Diseases
Society of America. All rights reserved. For Permissions, please e-mail:
[email protected].
U'lity of approach •  Allow examina'on of drugs in combina'on at different doses for a synergy plot; ranking regimens or comparison to current standard of care •  Predicts drug concentra'ons in humans in combina'on therapy associated with op'mal outcome •  Using these results in M&S to select op'mal doses for combina'on therapy •  Predic'ng what propor'on of pa'ents will achieve efficacy and what % will have drug resistance •  Then design more targeted clinical trial in terms of doses used, propor'on of pa'ents likely to respond etc CPTR PCS-­‐WG AIM To iden'fy the accuracy of the HFM-­‐TB, and Monte Carlo simula'ons based on HFM-­‐TB output, in forecas'ng : (a) therapeu'c concepts and hypotheses (b) quan'ta've therapeu'c parameters DEFINITION OF FORECASTING HFM-­‐TB publica'on or conference presenta'on followed 6 months later by clinical findings HFM-­‐TB
± MCS
>6 months Clinical study publica'on “In forecasting information is transferred from time 1 to a future time 2 ”
0% accuracy or 100% error 100% accuracy or 0% error What is quan'ta've predic've accuracy? Examples •  Type #1: Direct output of the HFM-­‐TB –  In Year 2009, the HFM-­‐TB predicted pyrazinamide AUC0-­‐24 /MIC of 209 at site of infec'on was op'mal for sterilizing effect (translates to 11.7 in serum/plasma) •  Type #2: HFM-­‐TB output followed by Monte Carlo simula'ons. –  MIC resistance breakpoints (Year 2009) RIF
INH Standard/accepted( mg/L):
Predic'on (mg/L):
PZA 1.0
2.0/0.2
100 0.0625 0.0313/0.125 25-­‐50 PredicFve accuracy is how well these predicted quanFFes were to what was found in future in clinical studies: how close to the bullseye. HFM-­‐TB quan'ta've predic'ons •  Eight prospec've studies that examined 14 parameters/quan'ta've predic'ons: –  Quality of evidence for 7 studies: 1 or 2 –  8th was 3 –  used together with sample size in weigh'ng •  PredicFve accuracy = 94.4% (CI, 84.3-­‐99.9%) •  Bias =1.8% (CI,-­‐13.7 to 6.2%) – crossed zero. Next stage •  Toxicity with 3D architecture human “organoids” •  Completed 4 week study that looked at both efficacy and hepatotoxicity using human doses in a dose ranging study •  Same M&S approach to be used