Pharmacokinetic and pharmacodynamic modelling in drug development

1
Pharmacokinetic and pharmacodynamic modelling in
drug development

• David Giltinan
• Genentech Inc.
• Conference on Applied Statistics in Ireland
• Killarney May, 2006

2
(No Transcript)
3
Outline

• Relevance of PK/PD behaviour to choosing a dose
  regimen general remarks
• Case study a Phase II clinical study of an
  inhibitor of platelet aggregation
• Pharmacokinetic analyis and results
• Pharmacodynamic modeling and results
• Implications for dosing
• General conclusions

4
Target concentration paradigm

• Basic idea guiding dose selection for a drug to
  work, one must reach (and maintain) a certain
  minimum concentration at the site of action
• For practical reasons (feasibility of sampling)
  aim to keep concentration of the drug in the
  bloodstream within the appropriate target range
  for efficacy, the therapeutic window
• Pharmacokinetics (PK) what the body does to
  the drug
• Pharmacodynamics (PD) what the drug does to
  the body

5
(No Transcript)
6
(No Transcript)
7
Inter-subject variation in pharmacokinetics

• Patients may have very different absorption,
  distribution, or elimination characteristics
• Thus, attained plasma concentration profiles may
  differ considerably among patients following the
  same dosing regimen
• Identify patient characteristics such as sex,
  age, weight, renal function that have a
  systematic effect on PK behavior, and adjust
  dosing accordingly
• If there is substantial inter-subject variability
  in kinetic behavior that cannot be controlled,
  and if the therapeutic window is narrow, some
  monitoring of attained concentrations, with
  subsequent individualization of dosing, may be
  needed

8
Inter-subject variation in pharmacodynamics

• Inter-subject variability in the
  concentration-effect relationship reflects
  differential susceptibility to the drug (e.g.
  different subjects have different EC50s)
• Such variability, if it exists, may stem from
  specific patient characteristics, or overall
  clinical status
• If inter-subject variation is substantial, use of
  a patient-specific therapeutic window may be
  appropriate. That is, dosing may need to target
  different concentration ranges for different
  patients

9
Inter-subject variation in pharmacodynamics

• In this situation (target concentration range
  varies by subject), one may need to explore the
  dependence of the concentration-effect
  relationship on patient characteristics and
  adjust dosing accordingly
• If inter-patient differences in responsiveness to
  the drug cannot be predicted or controlled, e.g.
  in terms of known subject characteristics, it may
  be necessary to titrate dosing based on some
  measure of effect
• Coumadin (warfarin) is one drug which is dosed in
  this manner (dose is adjusted to maintain a
  specified prothrombin time). For newer candidate
  drugs, the impact of this scenario on development
  is likely to be lethal

10
The Clotting Cascade

• The process by which blood clots in case of
  injury is both extremely complex and highly
  regulated
• There is a need for tight control
• Too slow bleed to death from a papercut
• Too quick myocardial infarct or embolism in
  store
• Successful pharmacologic intervention is possible
  in several coagulation disorders, though rarely
  straightforward

11
IIbIIIa inhibitors

• It is well known that aspirin has a protective
  effect against coronary events such as M.I. or
  stroke
• Aspirin blocks one pathway by which platelets
  aggregate
• IIbIIIa is a particular glycoprotein found on the
  surface of platelets which facilitates their
  aggregation (platelet velcro)
• This suggests that a drug which binds to IIbIIIa
  may have a clinical effect, mediated through
  blockage of platelet aggreagation
• Reopro, an injectable monoclonal antiibody, is
  one therapy with this method of action
• Search for a super-aspirin, oral agent in this
  class

12
Phase II trial design

• Novel, small-molecule, IIbIIIa antagonist, under
  study for possible chronic administration
• Patient population recent acute coronary event
• 28-day treatment period
• Initial randomization to one of four dose groups,
  targeted to span a range of inhibition of
  platelet aggregation
• Ability to add/delete dose groups, based on
  aggregation results for first 7 subjects in each
  group
• Once-a-day dosing groups dropped from
  consideration early on more twice-daily dose
  groups added as study progressed

13
Phase II trial design (continued)

• Intensive PK/PD evaluations on day 1 and day 28
  (6 to 9 timepoints each day)
• PK serial measurements of free and total drug
  concentration
• PD serial measurements of ADP-induced platelet
  aggregation at timepoints roughly concurrent with
  PK sampling (expressed as inhibition relative
  to the subjects baseline)
• About 100 patients in this PK/PD portion
  subsequently 250 subjects added to selected
  dose groups to augment safety data
• Main safety variable incidence of bleeding
  events

14
Pharmacokinetic analysis - objectives

• Characterise the typical PK behavior in the
  dose groups studied
• Elucidate relationship between exposure and dose
• Quantify inter-subject variability in PK behavior
• Identify subject characteristics which are
  predictive for differences in PK response

15
Results of PK Analysis

• Kinetics do not support once-daily dosing
• There is a clear dose-response in (average)
  kinetic behavior
• Relatively high degree of inter-subject
  variability
• Model-free and model-based conclusions were
  similar (sampling scheme was intensive enough to
  give excellent estimates of measures of exposure
  without necessarily invoking a parametric model
  to describe the concentration-time profiles)

16
(No Transcript)
17
(No Transcript)
18
Results of PK Analysis (continued)

• Covariates with explanatory power for measures of
  exposure such as AUC or Cmax
• Weight (exposure ? as weight ?)
• Renal function (exposure ? as , e.g. , GFR ?)
• Other covariates (age, sex, other measures of
  renal function) probably reflect correlation with
  these two
• Substantial unexplained inter-subject variation
  remains
• Covariates reduce, e.g. CV, of dose-adjusted AUC
  from 40 to 27

19
(No Transcript)
20
Pharmacodynamic analysis - objectives

• Characterise the concentration-response
  relationship
• Identify dose groups achieving inhibition of
  platelet aggregation in the target range (20 to
  80) with an acceptable safety profile
• Quantify variability in pharmacodynamic behavior
• Identify subject characteristics which are
  predictive for differences in PD response

21
Pharmacodynamic analysis results (ctd.)

• These data provide a textbook illustration of the
  need to account appropriately for the repeated
  measures character of the data when fitting the
  PD model
• Naïve fitting approaches, which fail to
  accommodate both the within-subject and
  between-subject variability, give misleading
  parameter estimates, underestimating the slope
  parameter in particular
• Fitting techniques based on an underlying
  (nonlinear) mixed model, which do accommodate
  both levels of variability, provide better
  parameter estimates and a better fit overall
• Estimated Hill coefficient is close to 2,
  indicating a very steep concentration-response
  curve

22
(No Transcript)
23
Pharmacodynamic analysis results (ctd.)

• High degree of variability across subjects in the
  estimated EC50 values
• The combined plot of inhibition versus
  concentration has very high potential to mislead
  if subject ID is suppressed
• Why? Because the overall visual impression of the
  combined data plot is that of a more gradual
  dependence of inhibition on dose than is actually
  the case
• Inspecting the individual-subject profiles
  reveals a series of extremely steep response
  curves, anchored at different EC50 values for
  different subjects
• The same concentration which induces complete
  inhibition in one subject may not be enough to
  generate any response at all in a more resistant
  subject

24
(No Transcript)
25
(No Transcript)
26
(No Transcript)
27
Pharmacodynamic analysis - results

• PD model fit well in the mid-range of inhibition,
  but not at the extremes
• Values close to 100 under-predicted model
  unable to account for subjects with inhibition
  values less than 0
• Potential explanation for negative inhibition
  values discontinuation of baseline medications?

28
(No Transcript)
29
Pharmacodynamic analysis - results

• The high inter-subject variability in EC50 values
  is not good news
• No identifiable subject covariates which
  accounted for the apparent differences in
  responsiveness candidate variables such as
  clinical status, platelet count, receptor
  density, concomitant medication usage were not
  predictive
• The single variable which appeared to contribute
  most to inter-subject variation in EC50s was
  study site
• Depressingly, this more likely reflects
  differences in assay conduct across sites than
  genuine differences among subjects in their
  responsiveness to the drug

30
(No Transcript)
31
Pharmacodynamic analysis - conclusions

• Although the PD model fit is not without
  problems, the substantial inter-subject
  differences in concentration-effect profiles are
  quite real, and large enough to complicate dosing
  enormously
• Based on the excessive (poorly understood and
  thus uncontrollable) pharmacodynamic variability,
  Genentech concluded that it was not possible to
  identify a workable dosing strategy, and passed
  on its option to proceed with joint development
  in Phase III
• Roche went ahead with a Phase III study, using a
  dosing scheme that adjusted for weight and renal
  function, and allowed for within-subject dose
  reduction in the event of toxicity
• The Phase III trial was unsuccessful
• Personal view (DG) this was hardly a surprise,
  as the dosing strategy addressed only PK
  variability, with no provision for dealing with
  the more serious issue of PD variation. The Phase
  III trial was a completely avoidable, extremely
  expensive, mistake. Interesting insights into
  team dynamics, collaboration with partners etc.
  etc.

32
(No Transcript)
33
General conclusions

• If the target concentration window is wide
  enough, life can be straightforward, requiring no
  fine-tuning of the dosing regimen
• If not, then substantial inter-subject PK
  variability can be problematic, particularly if
  it cannot be explained in terms of readily
  available subject covariate information.
  Concentration monitoring can provide a remedy,
  but may not be consistent with an acceptable
  product profile
• High inter-subject PD variability which cannot be
  attributed to simple subject-specific covariates
  is even more problematic and is likely to
  constitute a terminal hurdle to successful
  development of a candidate drug
• Even though a drugs characteristics may be
  satisfactory, in some average sense, this may be
  of little use in a practical sense, as there is
  no such thing as an average patient

34
(No Transcript)