Prequalification programme: Priority essential medicines

1
Prequalification programmePriority essential
medicines
Training programme on pharmaceutical quality,
good manufacture practice and bioequivalence with
a focus on TB products. Jiaxing Peoples
Republic of China 5 9 November 2007
2
Training Workshop on Evaluation of quality and
interchangeability of medicinal products.

• Statistical considerations for BE
• Presenter Drs. J. Welink
• Senior pharmacokineticist
• Medicines Evaluation Board, NL
• WHO adviser
• E-mail j.welink_at_cbg-meb.nl

3
Bioequivalence
Bioequivalence bioavailability with
pre-defined criteria for the rate and extent of
absorption!!
4
Bioequivalence
The primary concern in bioequivalence
assessment is to limit the risk of a false
declaration of equivalence. Statistical analysis
of the bioequivalence trial should demonstrate
that the clinically significant difference in
bioavailability is unlikely..
WHO working document multisource (generic)
pharmaceutical products Guidelines on
registration requirements to establish
interchangeability, Nov. 2005
5
Bioequivalence
2 pharmaceutical products
Test
Reference
Bioequivalent??
6
Bioequivalence
Important PK parameters
Cmax the observed maximum concentration of a
drug ?? measure of the rate of absorption
AUC area under the concentration-time curve ??
measure of the extent of absorption
tmax time at which Cmax is observed ?? measure
of the rate of absorption
7
Statistical considerations
How similar is similar?
8
Statistical considerations
9
Statistical considerations
Statistical test should take into account

• The consumer (patient) risk of erroneously
  accepting bioequivalence (primary concern health
  authorities)

• Minimize the producer (pharmaceutical company)
  risk of erroneously rejecting bioequivalence

• Choice
• two one-side test procedure
• confidence interval ratio T/R 100 (1-2?)
• ? set at 5 (90 CI)

10
Statistical considerations
11
Statistical considerations
Consumer Risk

• The risk of declaring two product BE when theyre
  not is called the consumer risk

• In statistical terms, this is a Type I error
• The risk of rejecting the null hypothesis when
  its true

• The consumer risk is set at 5

12
Statistical considerations
Producer Risk

• The risk of declaring two products NOT BE when
  they truly are BE is called the producer risk

• In statistical terms, this is a Type II error
• The risk of accepting the null hypothesis when
  its false

13
Statistical considerations
The risks are related

• If the consumer risk is reduced, the producer
  risk increases

• In statistical terms, if you lower the acceptable
  risk of making a Type I error, the risk of making
  a Type II error increases

14
Statistical considerations
Average Bioequivalence
two drug products are bioequivalent on the
average when the (1-2a) confidence interval
around the Geometric Mean Ratio falls entirely
within 80-125 (regulatory control of specified
limit)
15
Statistical considerations
Some International Criteria
16
Statistical considerations
17
Statistical considerations
Least Square Means from ANOVA
t-statistic with 0.05 in one tail
Standard Error
18
Statistical considerations
BE Limits

• The concept of the ?20 difference is the basis
  of BE limits (goal posts)

• If the concentration dependent data were linear,
  the BE limits would be 80-120

• On the log scale, the BE limits are 80-125

• The 90CI must fit entirely within specified BE
  limits e.g. 80-125

19
Statistical considerations
Variables..

• Log transformation
• For all concentration dependent pharmacokinetic
  variables (AUC and Cmax)

• Analysis of log-transformed data by means of
  ANOVA (analysis of variance)
• includes usually formulation, period, sequence or
  carry-over, and subject factors
• parametric test (normal theory)

20
Statistical considerations

• The sources of variance in the model are
• Product
• Period
• Sequence
• Subject (Sequence)
• Residual variance

These account for all the inter-subject variabilit
y
This estimates Intra-subject variability
21
Statistical considerations

• The width of the 90CI depends on
• The magnitude of the WSV (ANOVA-CV (residual
  variance))
• The number of subjects in the BE study

• The bigger the WSV, the wider the CI

• If the WSV is high, more subjects are needed to
  give statistical power compared with when the WSV
  is low

22
Statistical considerations
ANOVA CV
23
Statistical considerations
24
Statistical considerations
why log-transformation
25
Statistical considerations
Why parametric testing and not non-parametric
applicant after log transformation not normal
distributed!

• based upon test for normality, however these are
  insensitive and it concerns a small study

• normally after log transformation AUC and Cmax
  are normal distributed

• reason for non-normality should be explained

26
Example
Statistical analysis
Case testing.

• - number of subjects 24
• used for statistical analysis 24
• non-parametric testing
• reason non normal distribution
• calculated 90 CI AUCinf 0.98 1.23
• Cmax 0.99 1.24
• Conclusion Bioequivalent!

27
Example
Statistical analysis
Case testing.
28
Example
Statistical analysis
Case testing.

• - number of subjects 24
• used for statistical analysis 24
• parametric testing !!!!!!!!!!
• reason detection of an outlier considered not
  acceptable
• calculated 90 CI AUC0-t 0.98 1.23
• Cmax 1.01 1.38
• Conclusion Not bioequivalent!
• non parametric testing considered not
  acceptable!

29
Outliers

• Outliers
• Definition
• aberant/irregular values (e.g. no plasma
  concentration, late high concentrations.)

30
Outliers

• Outliers
• Explanation
• vomiting?
• non-compliant volunteers?
• bioanalytical failure?
• individual pharmacokinetics?
• protocol violations?

31
Outliers

• Outliers
• Handling
• pharmacokinetic data can only be excluded based
  on non-statistical reasons that have been defined
  previously in the protocol.
• Exclusion of data can never be accepted on the
  basis of statistical analysis or for
  pharmacokinetic reasons alone, because it is
  impossible to distinguish between formulation
  effects and pharmacokinetic effects.
• Results of statistical analyses with and without
  the excluded subjects should be provided.
  (excerpt from QA Doc Ref EMEA/CHMP/EWP/40326/200
  6)

32
Carry-over

• carry-over
• Definition

33
Carry-over

• differential carry-over
• is coming from the drug given at the previous
  period
• in a 2x2 crossover it occurs if the carry-over
  is not the same for the sequence TR and RT
• a non-differential carry-over translates into a
  period effect
• a poor randomization can be wrongly interpreted
  as a carry-over
• no valid study with a significant carry-over

34
Carry-over

• to avoid a carry-over
• the wash-out period should be sufficiently long
  (appr. 5 x t1/2)

35
sequence effect

• a sequence effect may occur
• if the two groups of volunteers (TR and RT) are
  really different (unlikely by correct selection)
• due to a different carry-over effect from the
  formulations (may be visible by non-zero samples
  at the second period)
• due to a formulation-period interaction
  (minimized by the same standardized conditions
  for both periods)

36
End