Experiences in applied clinical pharmacometrics: challenges, recommendations, and research opportunities. Ron Keizer

Transcription

1 Experiences in applied clinical pharmacometrics: challenges, recommendations, and research opportunities Ron Keizer

2 2 USC*PACK JPKD TCIworks NextDose EzeCHiel TDMx RxKinetics TDMS2000 MW/Pharm BestDose DoseMe Optimum Dosing Strategies Precise PK Baysient InsightRX Tucuxi MyPKFit AutoKinetics free commercial unknown / other * start/end dates are approximate Sources: Fuchs et al. Clin Pharmacokinet

3 3 Introduction EMR integration Clinical analytics Continuous learning

4 4 challenges to successfuladoption 1. user interface (UI/UX) 2. education / support 3. integration with hospital systems (EHR) 4. funding 5. prove cost/benefit 6. regulatory 7. science ACoP preconference 2018

5 5 themes of this talk 1. model selection 2. individual fit 3. between-occasion variability 4. beyond exposure

6 1 model selection

7 7 model selection How do I know this model works for my patients?

8 8 model selection even when right age class, assumptions: trial population == new population parameter distribution covariate effects error magnitude no bias data collection / analysis drug administration drug assay creatinine assay etc

9 9 model selection: retrospective evaluation Pull data from EMR demographics + dosing + TDM Implement candidate models Perform predictive checks population-level / individual level a priori / a posteriori

10 10 model selection: a priori evaluation Manuscript in preparation. Collaboration with Radboud Applied Pharmacometrics Group (R ter Heine, E Svensson, R Aarnoutse, R Bruggeman)

11 11 model selection: a posteriori evaluation subject 1 subject 2 Similar functionality available in proseval (PsN) Manuscript in preparation. Collaboration with Radboud Applied Pharmacometrics Group (R ter Heine, E Svensson, R Aarnoutse, R Bruggeman)

12 model selection: retrospective evaluation 12

13 13 model selection: retrospective evaluation goal = fit for purpose

14 2 individual fit

15 15 individual fit Why is the fit for this patient off?

16 16 individual fit: outlier subject 125 fit to TDM data 100 concentration ipred pred time

17 17 individual fit: outlier subject distribution of η fit to TDM data concentration ipred pred time

18 18 individual fit: parametric prior-adjustment distribution of η fit to TDM data concentration ipred pred time

19 19 individual fit Apply with care! overfitting inter-occasion variability regression to the mean

20 20 20 individual fit: non-parametric prior-adjustment Jelliffe and Neely. Individualized Drug Therapy for Patients CL V

21 21 individual fit: non-parametric prior-adjustment V CL Jelliffe and Neely. Individualized Drug Therapy for Patients

22 22 Individual fit: model updating Implement PK model refine model use in practice e.g. with flattened priors for extreme subjects collect data

23 3 inter-occassion variability

24 24 Inter-occasion variability I saw this patient last month, can we use the knowledge learned from his previous visit?

25 25 inter-occasion variability Use of individual estimates specific to a previous occasion lead to reduced predictive power in forecasting future exposure 1 12 clearance (L/hr) occasion 1. Abrantes J et al. PAGE 2017

26 26 inter-occasion variability first issue what is occasion? visit treatment cycle 1 day arbitrary n days often not defined specifically in original paper not always matching clinical practice

27 27 inter-occasion variability: models including IOV % of 18 6 of 32 9 of 45 0 neonates children adult

28 28 inter-occasion variability long-term data is common subject 2

29 29 inter-occasion variability when no IOV reported: ignore IOV, but weigh data with time

30 30 inter-occasion variability when no IOV reported: ignore IOV, but weigh data with time

31 31 Inter-occasion variability I saw this patient last month, can we use the knowledge learned from his previous visit?

32 4 beyond exposure

33 33 Exposure-outcome relationships C min should be mg/l to be effective If C min >20 then 5x higher nephrotoxicity subjective, qualitative, usually ROC-based 1 population-dependent 1) A-K Hamberg and RJ Keizer. Ther Drug Monit Jun;39(3):303

34 34 Exposure-outcome relationships binary decision rules % nephrotoxicity vancomycin trough concentration (mg/l)

35 35 Exposure-outcome relationships binary decision rules % nephrotoxicity vancomycin trough concentration (mg/l)

36 36 Exposure-outcome relationships continuous link with outcome / toxicity % nephrotoxicity vancomycin trough concentration (mg/l)

37 37 Exposure-outcome relationships Instead of binary exposure rules: Example table regimen AUC 24 C trough efficacy toxicity 1000 mg q % 5 % 1500 mg q % 13 % 2000 mg q % 30 % 1000 mg q % 20 % allows individualization on PD, toxicity, outcome, as well as on PK

38 38 Exposure-outcome relationships Exposure target attainment Pharmacological aim Improved outcome / Reduced toxicity Medical aim Reduced costs Financial aim

39 39 Acknowledgements for our collaborations and discussions along the presented topics Sirj Goswami Ranvir Mangat Adam Frymoyer Shabnam Gaskari Jonathan Faldasz Janel Long-Boyle Rada Savic Rob ter Heine Roger Bruggeman Elin Svensson Krista van Steeg Anna-Karin Hamberg Joao Abrantes Elisabet Nielsen Mats Karlsson