SOME ISSUES ON THE DETERMINATION

1
SOME ISSUES ON THE DETERMINATION OF
BIOEQUIVALENCE FOR HIGHLY VARIABLE DRUGS Laszlo
Endrenyi University of Toronto Laszlo
Tothfalusi Semmelweis University of Budapest
FDA/CDER Advisory Committee on Pharmaceutical
Sciences Rockville, MD October 6, 2006
2
SYNOPSIS
1. What kinds of replicate designs should be
applied? 2. Should there be a constraint on
the estimated ratio of geometric means?
3
USUAL METHOD OF EVALUATIONUNSCALED AVERAGE
BIOEQUIVALENCE

• 1/BEL ? GMR ? BEL
• BEL BE limit - Usually 1.25
• GMR Ratio of geometric means
• - lgBEL ? log(GMR) ? lgBEL
• - lgBEL ? mT - mR ? lgBEL
• lgBEL logarithm of BEL
• mT, mR Estimated logarithmic means

4
AN APPROACH FOR HIGHLY VARIABLE DRUGSSCALED
AVERAGE BIOEQUIVALENCE

• Difference between logarithmic means is
  normalized by estimated variation
• - lgBELsc ? (mT - mR)/s ? lgBELsc
• A general procedure was suggested for setting BE
  limits (BELsc)
• Advantages
• - Statistical power is independent of variation
• - Statistical power is, with same sample size,
  much higher than of unscaled average BE
• - Interpretation Compare expected change due to
  switching with expected difference between
  replicate administrations
• - Interpretation Standardized effect size, as in
  clinical comparisons

5
REPLICATE STUDY DESIGNS FORAPPLYING SCALED
AVERAGE BIOEQUIVALENCE

• Difference between logarithmic means is
  normalized by estimated variation
• - lgBELsc ? (mT - mR)/s ? lgBELsc
• s Often regarded as within-subject variation of
  reference product (sWR)
• For estimating sWR a single-sequence 3 period
  design could be sufficient
• RRT

6
REPLICATE STUDY DESIGNS FORAPPLYING SCALED
AVERAGE BIOEQUIVALENCE

• An additional goal
• To compare variations of the two drug products
• sWT/sWR
• Could identify highly variable drug products
• Replications of both RR and TT are required
• For example
• RRT
• TTR

7
REPLICATE STUDY DESIGNS FORAPPLYING SCALED
AVERAGE BIOEQUIVALENCE

• Refine the additional goal
• To compare variations of the two drug products
• sWT/sWR
• within the same subject
• More effective identification of highly-variable
  drug products
• Also can identify (some) outlying observations
• Example of study design
• RTRT
• TRTR

8
REPLICATE STUDY DESIGNS FORAPPLYING SCALED
AVERAGE BIOEQUIVALENCE

• NOTE
• 3- and 4-period designs require approximately
  the same number of observations

9
SHOULD THERE BE A CONSTRAINT ON GMR?

• Concern about possibly large deviations between
  estimated logarithmic means
• i.e., about log(GMR)
• L. Benet, AAPS Workshop on Individual BE,
  1999.
• Concern about interpretation to physicians

10
SHOULD THERE BE A CONSTRAINT ON GMR?

• Larger deviation between the (logarithmic) means
  arises as a natural, direct consequence of the
  higher variability

Larger deviations occur at higher
variations They would be truncated by GMR
constraint Moreover
11
QUALITY CONTROL ORTHERAPEUTIC SURROGATE?
BioInternational 1994 Both! But Ambivalence

• QUALITY CONTROL THERAP. SURROGATE
• HIGH SENSITIVITY CLINICAL RELEVANCE
• HIGH STATISTICAL POWER
• Young, healthy (male) volunteers
  Heterogeneous study popul'n
• Single dosing Steady state
• Cmax/AUC Cmax
• Parent drug Active metabolite
• Recommendations of FDA
• Does one guidance/method satisfy both goals?

12
(No Transcript)
13
SHOULD THERE BE A CONSTRAINT ON GMR?

• Possible concern about large estimated GMR may
  possibly apply to the introduction of generic
  products,
• but NOT during the development of new drugs.
• Therefore a secondary regulatory criterion
  should NOT be generally invoked

14
CONCLUSIONS

• 3- or preferably 4-period study designs in which
  both the
• reference and test products are replicated
  (RR and TT), enable the comparison of
  within-subject variations and the identification
  of highly-variable drug products.
• A constraint of the estimated ratio of geometric
  means is
• not warranted scientifically or, generally, for
  the sake
• of public relations.