1. Individualization of Dosage Regimens
Michael Neely, MD, MSc, FCP, FAAP
Associate Professor and Clinical Scholar of Pediatrics
Children’s Hospital of Los Angeles
University of Southern California
Laboratory of Applied Pharmacokinetics and Bioinformatics

2. Disclosures
Co-founder and chief scientific advisor to Applied Pharmacometrics, Inc. who is negotiating to license BestDose for commercial distribution. Compensation: $0.
Director of the University of Southern California Laboratory of Applied Pharmacokinetics (LAPK), developer of BestDose.
I will be discussing off-label dosing.

3. Does the regimen matter?

4. How to tailor the dose

5. References for previous slide
1. Buijk SES, Mouton JWJ, Gyssens ICI, Verbrugh HAH, Bruining HAH. Experience with a once-daily dosing program of aminoglycosides in critically ill patients. Intensive Care Med 2002; 28:936–942.
2. Avent ML, Teoh J, Lees J, Eckert KA, Kirkpatrick CM. Comparing 3 methods of monitoring gentamicin concentrations in patients with febrile neutropenia. Ther Drug Monit 2011; 33:592–601.
3. Elligsen M, Walker SAN, Walker SE, Simor A. Optimizing initial vancomycin dosing in burn patients. Burns : journal of the International Society for Burn Injuries 2011; 37:406–414.
4. Kullar R, Leonard SN, Davis SL, et al. Validation of the Effectiveness of a Vancomycin Nomogram in Achieving Target Trough Concentrations of 15–20 mg/L Suggested by the Vancomycin Consensus Guidelines. Pharmacother 2011; 31:441–448.
5. Pea F, Bertolissi M, Di Silvestre A, Poz D, Giordano F, Furlanut M. TDM coupled with Bayesian forecasting should be considered an invaluable tool for optimizing vancomycin daily exposure in unstable critically ill patients. Int J Antimicrob Agents 2002; 20:326–332.
6. Thomson AH, Campbell KC, Kelman AW. Evaluation of Six Gentamicin Nomograms Using a Bayesian Parameter Estimation Program. Ther Drug Monit 1990; 12:258.
7. Mohan M, Batty KT, Cooper JA, Wojnar-Horton RE, Ilett KF. Comparison of gentamicin dose estimates derived from manual calculations, the Australian ‘Therapeutic Guidelines: Antibiotic’ nomogram and the SeBA-GEN and DoseCalc software programs. British J Clin Pharm 2004; 58:521–527.

6. Bayes’ Theorem Current Prior Posterior
Formal method of incorporating prior (past) experience to bear on current circumstances.
Assume that you’re presented with three coins, two of them fair and the other a counterfeit that always lands heads. If you randomly pick one of the three coins, the probability that it’s the counterfeit is 1 in 3. This is the prior probability of the hypothesis that the coin is counterfeit. Now after picking the coin, you flip it three times and observe that it lands heads each time. Seeing this new evidence that your chosen coin has landed heads three times in a row, you want to know the revised posterior probability that it is the counterfeit. The answer to this question, found using Bayes’ theorem, is 4 in 5. You thus revise your probability estimate of the coin’s being counterfeit upward from 1 in 3 to 4 in 5.
Prior
Posterior

7. Voriconazole prior experience
Age (years)
N=141
2-6 31
>6-12 33
>12-18
>18 56
Hope W. Population pharmacokinetics of voriconazole in adults. Antimicrob Agents Chemother 2012; 56:526–531.
Friberg LE, Ravva P, Karlsson MO, Liu P. Integrated population pharmacokinetic analysis of voriconazole in children, adolescents, and adults. Antimicrob Agents Chemother 2012; 56:3032–3042.

8. Prior experience Median (range) 22 (1- 75) samples/subje ct
All but 1 subject ≥7 samples

9. *Allometrically scaled
A model of the data
Oral Dose
IV Dose FA
A model is just a series of equations that describe the time course of the concentrations, with or without effects, in the body after any given dose. Those equations have variables, and there is a distribution of values for those variables in a sample study population that represents the larger pool.
KCP Ka
Vc*
KPC
Michaelis-Menten:
f (Vmax*, Km,Hill,C)
*Allometrically scaled

10. A model of the data

11. A model of the data
You can see clusters in the non-parametric distribution
Neely MN, van Guilder MG, Yamada WM, Schumitzky A, Jelliffe RW. Accurate Detection of Outliers and Subpopulations With Pmetrics, a Nonparametric and Parametric Pharmacometric Modeling and Simulation Package for R. Ther Drug Monit 2012; 34:467–476.

12. A model of the data

13. Using the model How does Bayes’ theorem apply to optimal dosing?
Each patient is not an island? They exist in the universe of other patients who have been given the drug. We can capture that past experience in drug model.

14. Multiple Models

15. Multiple Models

17. Patient 1 7 year old girl with bone marrow transplant for AML
Weight 28.2 kg
Started on oral voriconazole for prophylaxis
Enrolled in local study of the ontogeny of voriconazole pharmacokinetics*
*NIH-NICHD R01 HD (PI: Neely)

18. All the Data
Used BestDose to calculate the dose required to achieve a target that was an actual measured concentration, and compared the calculated dose with the actual dose that resulted in the measured concentration.
Colored lines
Black line
Date
Time
Dose
Units
%Err 1
5/21/13
20:30
76.8 mg
-48.8% 2
5/22/13
8:30
157.14
4.8%

19. Optimal Sampling*
Note that we know less about the patient, so there are more possible curves.
*Neely MN , Bayard DS, Hope WW . Multiple Model Optimal (MMopt) Sampling for First Dose Oral Voriconazole TDM in Children, IATDMCT,Salt Lake City, September 22-26, 2013
Date
Time
Dose
Units
%Err 1
5/21/13
20:30
177.30 mg
18.2% 2
5/22/13
8:30
177.15

20. Optimal Sampling, Fixed Dose
Date
Time
Conc
Pred
%Err 1
5/22/13
7:00
1.72
1.56
-9.3% 2
19:00
1.48
-14.9%

21. Initial Dosing Loading: 520 mg (~19 mg/kg) x 1
Maintenance: 200 mg (~7 mg/kg) q12 h
Date
Time
Conc
Pred
%Err 1
5/22/13
7:00
1.72
1.71
-0.6% 2
19:00
1.60
-7.0% 3
5/23/13
1.64
-4.7% 4
1.69
-1.7% 5
5/24/13
1.76
2.3%

22. Initial Dosing, No Load Loading: none
Maintenance: 200 mg (~7 mg/kg) q12 h
Date
Time
Conc
Pred
%Err 1
5/22/13
7:00
1.72
0.4
-76.7% 2
19:00
0.68
-60.4% 3
5/23/13
0.89
-48.2% 4
1.06
-38.4% 5
5/24/13
1.21
-30.0%

23. Patient 2
13 month old bone marrow transplant patient on long-term oral voriconazole for pulmonary nodules with +Galactomannan
Weight 8.76 kg
Voriconazole trough concentration repeatedly <0.5 mg/L. Dose gradually increased to 17 mg/kg/dose PO tid (51 mg/kg/day).
A 75 kg adult at this dose would be getting mg/day compared to usual dose of 400 mg/day.

24. AUC
AUC ref
MIC breakpoint
67.8
40.6 (69%)1
73.0 (45%)2
1.14 3
Vfend package insert mean (CV%) adult AUC 200 mg PO q12h, (accessed 9/10/13)
Vfend package insert mean (CV%) adult AUC 300 mg PO q12h
Based on mean protein binding of 58% in Vfend package insert and target fAUC:MIC of ≥25, (Andes D, Marchillo K, Stamstad T, Conklin R. In vivo pharmacokinetics and pharmacodynamics of a new triazole, voriconazole, in a murine candidiasis model. Antimicrob Agents Chemother 2003; 47:3165–3169).

25. MM and Vori in Adults
Hope W, van Guilder M, Donnelly JP, et al. Software for dosage individualization of voriconazole for immunocompromised patients. Antimicrob Agents Chemother 2013; 57:1888–1894.

26. MM and HIV ARV
A - efavirenz, lower dose; B - nelfinavir, higher than adult dose in a 10 year old; C - fos-amprenavir, alternative dosing regimen to manage toxicity; D - atazanavir, extrapolating from adult dosing
Neely M, Jelliffe R. Practical therapeutic drug management in HIV-infected patients: use of population pharmacokinetic models supplemented by individualized Bayesian dose optimization. J Clin Pharmacol 2008; 48:1081–1091.

27. MM and Vancomycin
77 adults getting vancomycin. Overall bias excellent, precision moderate - half of predictions <23% error. Error worst in obese patients and those with unstable renal function.
However, the precision of fixed dosing expressed as abs(100*(min - meanmin)/min)) in vori dataset was 92% - 4-fold worse! Plus among 372 trough concentrations sent in a 3 month period at our hospital, precision was 77% fold worse than BestDose. Among 58 children with vancomycin troughs, precision was 46% - 2-fold higher than BestDose.
Nunn MO, Corallo CE, Aubron C, Poole S, Dooley MJ, Cheng AC. Vancomycin dosing: assessment of time to therapeutic concentration and predictive accuracy of pharmacokinetic modeling software. The Annals of Pharmacotherapy 2011; 45:757–763.

28. MM and Piperacillin
Excluding patient 3 changed bias to -7% and precision to 28%
Based on paper from Buck, precision of standard dosing in hospitalized patients is about 60%
Bias
Precision
6.7%
37.5%
Felton TW, Roberts JA, Lodise TP, VanGuilder M, Boselli E, Neely MN, Hope WW. Software for Dosage Individualisation of Piperacillin/Tazobactam for critically ill patients, poster A-482, ICAAC 2013, Denver, CO.

29. MM summary Done To Do Generalizable
Multiple models ensure maximally precise and accurate dosing
Does not require overly complex covariate models
Runs in seconds
More validation
Smoothly integrate into workflow

30. Optimization Software
BestDose - Laboratory of Applied Pharmacokinetics, University of Southern California, Los Angeles, CA (
DoseMe* - DoseMe Pty Ltd, Brisbane, Australia (
TCIWorks - Universities of Queensland, Australia and Otago, New Zealand (
MWPharm* - Medimatics, Maastricht, Netherlands ( w/home) MM
MAP Bayesian
*Paid subscription

31. Acknowledgements
Support from NIH-NICHD R01 HD and NIH-NIGMS R01 GM068968
LAPK - Mike van Guilder, Alan Schumitzky, David Bayard, Jay Bartroff, Alona Chubatiuk, Tatiana Tatarinova, Walter Yamada, Roger Jelliffe
University of Liverpool - William Hope
University of Manchester - Timothy Felton