Models of Drug Behavior: The Basis of TCI

1
Models of Drug BehaviorThe Basis of TCI
Steven L. Shafer, M.D. Palo Alto VA Health Care
System Stanford University School of
Medicine University of California at San Francisco
2
Please Fasten Safety BeltsPrior to Take Off
3
Basic PK and PD Relationships
4
Simple Pharmacokinetic Model Volume of
Distribution
5
Simple Pharmacokinetic Model Clearance
6
More complex PK ModelMulti-compartment
7
More complex PK ModelMulti-compartment
100
Rapid
10
Concentration
Intermediate
Slow
1
0
120
240
360
480
600
Minutes since bolus injection
8
Awake EEG
Gregg K, Varvel JR, Shafer SL. J Pharmacokinet
Biopharm 20, 611-635, 1992
9
Profound Opioid EEG Effect
Gregg K, Varvel JR, Shafer SL. J Pharmacokinet
Biopharm 20, 611-635, 1992
10
EEG Time Course with Fentanyl
Scott J, Ponganis KV, Stanski DR. Anesthesiology
62234-241, 1985
11
EEG Time Course with Alfentanil
Scott J, Ponganis KV, Stanski DR. Anesthesiology
62234-241, 1985
12
Extended PK/PD Concept The Effect Site
13
PK/PD IntegrationEffect site concentrations
over time
14
Application of Drug ModelsTarget Controlled
Delivery
15
Fentanyl Target 1 ng/ml
16
Fentanyl TCI
17
Fentanyl TCIPlasma Target
18
Fentanyl TCIEffect Site Target
19
Start with bolus drug kinetics
20
Give by computer controlled infusion
21
Test refined kinetics
22
Propofol PK/PD in the ICU
23
Target Controlled Lidocaine

• Used at Stanford Pain Clinical for patients with
  neuropathic pain.

24
CSF Targeted Epidural Clonidine
Eisenach et al, Anesthesiology 1995 8333-47
25
TCI and Biological Variability
26
TCI Variability

• Biological variability exists
• TCI devices cannot increase biological
  variability
• The maximum variability in concentration is
  observed with a single bolus dose
• The variability with TCI is bounded by the
  variability following bolus dose

27
TCI Variability is Bounded
28
TCI Variability is Bounded

• TCI variability cannot be greater than that
  following bolus injection.
• Based solely on linear systems theory with no
  other assumptions about underlying PK model.
• Has implications for regulatory approval of TCI
  devices.

29
Bolus Simulation 50-60 CV
30
Infusion Simulation 25 CV
31
TCI Simulation 25 CV
32
Simulation CVs
33
Bolus Variability 31 CV
34
TCI Variability 11 CV
35
TCI Variability 18 CV
36
TCI Can Reduce Variability

• TCI removes time as a confounding variable
  between the device setting and the patient
  response
• TCI can incorporate patient covariates to
  individualize drug dosing
• Weight, height, gender, ethnicity
• Disease state
• Drug interactions
• Pharmacogenetics

37
Regulatory Implications

• TCI devices cannot have variability greater than
  that seen with bolus injection
• TCI devices can reduce variability by reducing
  the influence of time and patient covariates,
  including pharmacogenetics
• If drugs are approved for use by bolus injection,
  then there should be minimal regulatory burden
  for to approve drugs for administration by TCI.

38
Regulatory Implications

• TCI devices give
• approved drugs
• by approved routes
• for approved indications
• at doses that conform to the package insert
• There should be minimal regulatory burden for TCI
  devices if the drug, route, indication, and doses
  are already approved for bolus injection

39
Are Drug Models Predictiveof Drug Effect?
40
The Aspect Data Base Evaluation

• Patient trials (movement)
• Thiopental
• Propofol
• Fentanyl/Alfentanil/Sufentanil
• Isoflurane
• Nitrous Oxide
• Volunteer trials (recall, sedation, eyelash)
• Propofol
• Isoflurane
• Alfentanil
• Midazolam

41
The Aspect Data Base Evaluation

• Aspect Investigators
• Peter Sebel (Emory)
• Peter Glass (Duke)
• Carl Rosow (Harvard/MGH)
• Lee Kearse (Harvard/MGH)
• Marc Bloom (University of Pittsburgh)
• Ira Rampil (University of California, San
  Francisco)
• Randy Cork (University of Arizona)
• Mark Jopling (Ohio State University)
• N. Ty Smith (University of California, San Diego)
• Paul White (University of Texas at Dallas)

42
Recall vs. BIS(unstimulated)
43
Predictors of Movement
Measure
Pk
0.74
Blood propofol
0.76
Effect-site propofol
Bispectral Index
0.86
Relative delta power
0.79
Relative beta power
0.83
95 SEF (Hz)
0.81
Median Frequency (Hz)
0.8
Leslie et al, Anesthesiology 8452-63, 1996
44
Sedation, BIS, and Propofol
Glass et al, Anesthesiology 86836-847, 1997
45
Conscious/Unconscious Prediction (Pk)
Target
Measured
Agent (n)
BIS
Concentration
Concentration
Propofol (399)
0.976 0.006
0.936 0.010
0.937 0.013
Isoflurane (70)
0.959 0.021
0.965 0.015
0.967 0.016
Midazolam (50)
0.885 0.047
0.859 0.045
0.886 0.048
Significantly different from Pk value for Target
Concentration (p lt 0.001),
and Measured concentration (p lt 0.01)
Glass et al, Anesthesiology 86836-847, 1997
46
PK for AAI, BIS, and Predicted Propofol
Concentrations(when combined with remifentanil)
Struys et al, Anesthesiology 99802-812
47
Propofol-Remifentanil Interaction(loss of
response to laryngoscopy)
Measured and predicted have nearly identical
likelihood values
Propofol (?g/ml)
Predicted (TCI)
Measured
Bouillon et al, Anesthesiology 2004
48
Are drug models predictive?

• Mathematical models of drug behavior
  incorporating effect site concentrations and drug
  interactions predict anesthetic drug effect
  (e.g., loss of response to stimulation) as well
  as
• Measured concentrations
• BIS
• AAI
• I am not aware of any counter examples.

49
Models of Drug Interaction
50
Basic Concentration vs Response Relationship
1
0.8
0.6
50 Probability
Probability of no response
0.4
C
0.2
50
0
0.1
1
10
100
Drug concentration
51
Basis of Response Surface A Sigmoid in Every
Slice
52
How a response surface relates to an isobole
53
Simple Additivity
54
Synergy
55
Hierarchical Model of Drug Interaction
Hypnotics
Opioids,N2O
Conscious,Responsive
Cortex
AmbientStimuli
Unconscious,Unresponsive
SystemicOpioids
Pain projection
Midbrain, Thalamus
Severe
N2O
to cortex
None
Spinal
Local
Opioids
Anesthetics
Pain projection
Severe
to midbrain
Peripheral nerves, Spinal cord
None
Pain
56
Hierarchical Model of Drug Interaction
AmbientStimuli
Afferent Stimuli
Pain projection
to cortex
Pain
Pain
57
Propofol-RemifentanilInteraction Surface
Laryngoscopy
Bouillon et al, Anesthesiology 2004
58
Propofol-RemifentanilInteraction Surface
Laryngoscopy
Bouillon et al, Anesthesiology 2004
59
Propofol-RemifentanilInteraction Surface BIS
Bouillon et al, Anesthesiology 2004
60
Propofol-RemifentanilInteraction Surface BIS
Bouillon et al, Anesthesiology 2004
61
Display IncorporatingDrug Interactions
From Westenskow, Kern, and Egan
62
Display IncorporatingDrug Interactions
From Westenskow, Kern, and Egan
63
The next step in closed loopbringing it all
together
64
Therapeutic Objective

• Maximize Efficacy
• Minimize Toxicity
• Requires models of efficacy and toxicity
• Requires an objective function to balance lack of
  efficacy with risk of toxicity

65
Balancing Act
High
High
Efficacy
Safety
Low
Low
66
Cost Function
67
General Anesthesia Cost Function
68
GA Cost Functionfor Target BIS 65
69
GA Cost Functionfor Target BIS 60
70
Dynamic Ventilatory DepressionBouillon Model
Ventilation
Effect Site Drug Concentration
Depresses
Increases CO2
Stimulates
Carbon Dioxide
71
Model of Ventilatory DepressionRemifentanil 70
µg bolus
72
Model of Ventilatory DepressionRemifentanil 12
µg/min infusion
73
Models of Toxicity

• Several available models of awareness vs.
  propofol and remifentanil
• Models of ventilatory depression with propofol,
  remifentanil, although not with both.
• Should be incorporated into any closed-loop PCA
  system with rapidly acting opioids

74
Bringing it all together
Population PK/PD Model of Therapeutic Effect
Population PK/PD Model of Toxic Effect
Predict
Update
Predict
Update
Measure Therapeutic Effect
Measure Toxic Effect
Calculate Dose to Minimize Cost Function
Dose Drug
75
(No Transcript)