THERAPEUTIC DRUG MONITORING

1
THERAPEUTIC DRUG MONITORING WHAT YOU CAN DO
WITH IT TODAYTarget goal-oriented, model-based
TDM

• Roger Jelliffe,
• Laboratory of Applied Pharmacokinetics
• USC Keck School of Medicine
• Los Angeles CA

2
Important points

• Must use models of drug behavior, and software.
• Clinician must set target goal(s) Must SEE the
  pt!
• Assay error pattern no LOQ!
• Better TDM strategiesnot just troughs
• Nonparametric pop PK/PD models
• Multiple Model dosage design
• Nonparametric Bayesian posteriors
• IMM Bayesian posteriors
• Three illustrative cases

3
Determining the assay error pattern
4
Fisher Information

• Fisher Info x/Varx
• So need to know, or have a good estimate, of the
  SD of every serum level
• Var SD2
• Weightx 1/Varx

5
Assay CVSD

• Concentration

6
Determining the Assay SD polynomial

• Measure blank, low, medium, high, and very high
  samples in at least quadruplicate.
• Get mean SD for each sample.
• Fit a polynomial to the mean and SD data.
• SD A0C0 A1C1 A2C2 A3C3
• Then can weight each measurement by the
  reciprocal of its variance (Fisher Info)
• No lower detectable limit for PK work!

7
(No Transcript)
8
Better TDM strategiesnot just troughs

• Optimal Times to get Serum Samples
• D - optimal experimental Design

9
When and why to get Serum Samples

• To achieve target goals precisely.
• To get good AUC values for individual patients.
• And good AUCs from population models.
• Many drug regimens target desired AUCs.
• The AUCs MUST be reliable!

10
Example Methotrexate

• Most samples obtained to see if patient needs
  leucovorin rescue, not for good PK models.
• A real problem!
• How reliable are target AUCs when samples not
  obtained before 6, 8, 12, 24 hours?
• How reliable are patient AUCs when only trough
  samples are drawn?

11
(No Transcript)
12
(No Transcript)
13
Monitoring Lidocaine Optimally DArgenio, 1981

• Loading infusion over 1 min, then 1.45 mg/min.
  10 subjects.
• Conventional strategy 8 measurements, at 5,
  10, 30, 60, 120, 180, 360, and 720 min into the
  regimen.
• D Optimal strategy 4 samples, each in
  duplicate, at 1, 10, 70, and 720 min.
• Assay SD 0.0078 0.1236 x conc.

14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
Good TDM strategies

• Can check achievement of target goals
• Can get good parameter and AUC values for
  individual patient PK models
• Can make good population models from TDM data
• Permit reliable target AUC values
• Permit reliable target goals for individual
  patients.

18
What is the IDEAL Pop Model?

• The correct structural PK/PD Model.
• The collection of each subjects exactly
  known parameter values for that model.
• Therefore, multiple individual models, up to
  one for each subject.
• Usual statistical summaries can also be
  obtained, but usually will lose info.
• How best approach this ideal?

19
A Parametric Population Model Joint Density
20
A Population Model, as made by Breugel
21
An NPML Population Model, as made by Mallet
22
Nonparametric Population Models

• Get the entire ML distribution, a Discrete Joint
  Density one param set per subject, its prob.
• Shape of distribution not determined by some
  equation, only by the data itself.
• Multiple indiv models, up to one model per
  subject.
• Can discover, locate, unsuspected subpopulations.
• The multiple models permit multiple predictions.
• Can predict precision of goal achievement by a
  dosage regimen.

23
The Separation Principle

• Whenever you separate the process of controlling
  the behavior of a system into
• First, getting the best single point parameter
  estimates, and then
• Second, using those point values to control the
  system to achieve target goals,
• The control is usually done suboptimally.
• No performance criterion is optimized.

24
The Way around the Separation Principle

• Use a pop model with discrete multiple models -
  an NP pop model, for example.
• Give a candidate regimen to each model.
• Predict each result.
• Compute weighted squared error of failure to
  hit target goal at target time.
• Find the regimen having the minimal weighted
  squared error. This is multiple model
  (MM) dosage design.

25
Lido Regimen based on Param means Predicted
response of full lido pop model
26
MM lido regimenPredicted response of full lido
pop model
27
Three illustrative cases

• A patient on gentamicin -changing renal function
• A patient on tobramycin changing clinical
  status
• A patient on Digoxin converting A Fib