This document discusses bioequivalence standards for highly variable drug products. It begins by defining highly variable drug substances and products as those with intra-subject variabilities greater than 30%. It notes the sources of high variability can include administration conditions, physiological factors, and technical aspects. The usual standards for passing bioequivalence require average AUC and Cmax values to fall within 80-125% intervals. However, these criteria may be impossible to meet for highly variable drugs even with large sample sizes. The document therefore proposes alternative approaches for demonstrating bioequivalence of highly variable drugs, including replicate study designs and reference scaled average bioequivalence criteria. It provides examples and discusses some issues that can arise with these alternative approaches.

1. Bioequivalence of Highly
Variable Drug Products
Dr. Bhaswat S. Chakraborty

2. 2

3. Highly Variable Drug Products
• Definition: BIO-international '92 [2001] "Drugs which exhibit intra
"Drugs which exhibit intra-subject variabilities >30 % (CV from
ANOVA) are to be classified as highly variable …"
• Essential differentiation
• Highly variable drug substances, e.g. statins
• Highly variable drug products, e.g. enteric coated
• Sources of (high) variability
• Administration conditions, interactions with food
• Physiological factors (GE, transit, first-pass, ...), technical aspects, e.g.
bioanalytical procedures
3

4. 4

5. Usual Standards for Passing ABE
• AUC: 90% CI limits 80-125%
• Cmax: 90% CI limits 80-125%
• Data generated in a 2x2 crossover study
• Criteria applied to drugs of low and high variability
5

6. TwCounter-intuitive to the Concept of BE
• Two formulations with a
large difference in
Means: Bioequivalent
(if variances are low)
• Two formulations with a
small difference in
Means: Not
Bioequivalent (if
variances are high).
6

7. Reference Test
PI PII PI PII
A better generic product gets penalized for high within-
subject & within product variability of the reference! 7

8. • For HVDs and HVDPs, it may be almost impossible to show
BE with a reasonable sample size.
• The common 2×2 cross-over design over assumes
Independent Identically Distributions (IID), which may not
hold
• If e.g., the variability of the reference is higher than the one
of the test, one obtains a high common (pooled) variance and
the test will be penalized for the ‘bad’ reference (previous
slide)
Impact of High Variability
8

9. • Produces medical dilemma (Switchability for NTRs, Prescribability for
nth generic
• Ignores distribution of Cmax and AUC
• Within subject variation is not accurate
• Ignores correlated variances and subject-by-formulation interaction
• One criteria irrespective of inherent patterns of product, drug or
patient variations
• Although rare, but may not be therapeutic equivalent
Other Limitations of a 2x2 Crossover Study
9

10. HVDs & HVDPs are usually safe and of wide
therapeutic range
Concentration
Time
10

11. Power to show BE with 40
subjects for CVintra 30–50%
μT/μR 0.95, CVintra 30%
→ power 0.816
μT/μR 1.00, CVintra 45%
→ power 0.476
μT/μR 0.95, CVintra 50%
→ n=98 (power 0.803)
11

12. US FDA ANDAs: 2003 – 2005 Example
(1010 studies, 180 drugs)
• 31% (57/180) highly variable (CV ≥30%)
• of these HVDs/HVDPs,
• 60% due to PK (e.g., first pass metabolism)
• 20% formulation performance
• 20% unclear
12

13. • Reduce human experimentation (number of participants) in BE
studies
• Prohibitive size of BE studies for some HVDs means no generic
is available – many patients go untreated
• Changing criteria to reduce number of participants in BE studies
on HVDs can be accomplished without compromising
safety/efficacy
• 80 – 125% BE criteria not universally implemented worldwide
Why a Different Set of Passing Criteria Needed
for HVDs & HVDPs?
13

14. Approaches to Solution
• US-FDA:
• In favour of replicate design approach
• Rejection of multiple dosing as less discriminative
• Individual BE:
• “Prescribability" vs. “switchability/interchangeability”
• S*F interaction – what does it mean therapeutically?
• Concept on trials for years, than dismissed
• Reference scaled procedure
• Widening of acceptance criteria due to scaling
• Based on Reference product related variability
14

15. Highly VariableDrugs
• Includes many therapeutic classes
• Includes both newer and older products
• Potential savings to patients in the billions of dollars if generics
are approved
• Examples: atorvastatin, esomeprazole, pantoprazole,
clarithromycin, paroxetine (CR), risedronate, metaxalone,
itraconazole, balsalazide, acitretin, verapamil, atovaquone,
disulfiram, erythromycin, sulfasalazine, many delayed release and
modified release products etc.
15

16. Fed BE Studies
• Confidence interval criteria now required for BE studies under
fed conditions
• General paucity of information on variability under fed
conditions
• Some drugs show much more variability under fed conditions
than fasting conditions, making them HVDs (e.g., esomeprazole,
pantoprazole, tizanidine)
• May be more HVDs than generally appreciated
16

17. Hierarchy of Designs
• The more ‘sophisticated’ a design is, the more information can be extracted.
• Hierarchy of designs:
• Variances which can be estimated:
Full replicate (TRTR | RTRT or TRT | RTR)
Partial replicate (TRR | RTR | RRT)
Standard 2×2 cross-over (RT | RT)
Parallel (R | T)
Full replicate: Total variance + within subjects (reference, test)
Partial replicate: Total variance + within subjects (reference)
Standard 2×2 cross-over: Total variance + incorrect within subject
Parallel: Total variance
17

18. Design of 4-period, Replicate Studies
Subjects
Sequence 1
Sequence 2
T
R
PI
W
A
S
H
O
U
T
1
Randomizaion
PII PIII PIV
W
A
S
H
O
U
T
2
W
A
S
H
O
U
T
3
R
RR
TT
T
18

19. Replicate Designs
• Each subject is randomly assigned to sequences, where at least one of the
treatments is administered at least twice
• Not only the global within-subject variability, but also the within-subject variability
within product can be estimated
• Smaller subject numbers compared to a standard 2×2×2 design – but
outweighed by an increased number of periods
• Two-sequence three-period
TRT RTR
• Two-sequence four-period (>2-sequence does not have any particular
advantage)
TRTR RTRT
19

20. Conduct of Replicate Studies
• Generally dosing, environmental control, blood sampling scheme and
duration, diet, rest and sample preparation for bioanalysis are all the same as
those for 2-period, crossover studies
• Avoid first-order carryover (from preceding formulation) & direct-by-
carryover (from current and preceding formulation) effects
• Unlikely when the study is single dose, drug is not endogenous, washout is adequate, and
the results meet all the criteria
• If conducted in groups, for logistical reasons, ANOVA model should take the
period effect of multiple groups into account
• Use all data; if outliers are detected, make sure that they don’t indicate product
failure or strong subject-formulation interaction
20

21. Evaluation of BE: Replicate Studies
• Any replicate design can be evaluated according to ‘classical’
(unscaled) Average Bioequivalence (ABE)
• ABE mandatory if scaling not allowed FDA: sWR <0.294 (CVWR
<30%); different models depend on design (e.g., SAS Proc MIXED
for full replicate and SAS Proc GLM for partial replicate)
• EMA: CVWR ≤30%; all fixed effects model according to 2011’s
Q&A-document preferred (e.g., SAS Proc GLM)
• Even if scaling is not intended, replicate design give more
information about formulation(s)
21

22. Sample size and Ethics: Replicate Studies
• 4-period replicate designs:
• Sample size = ~½ of 2×2 study’s sample size
• 3-period replicate designs:
• Sample size = ~¾ of 2×2 study’s sample size
• Number of treatments (and biosamples)
• Same asconventional 2×2 cross-over
• Allow for a safety margin – expect a higher number of drop-outs
due to additional period(s).
• More time commitment from subjects; Consider increased blood
loss; improved Bioanalytical method required
22

23. 23
3-period Replicate 4-period Replicate
SAMPLE SIZES

24. Unscaled & Scaled ABE from Replicate Studies
• Common to EMA and FDA
• ABE:
• Scaled ABE model:
• Regulatory switching condition θS is derived from the regulatory
standardized variation σ0 (proportionality between acceptance limits
in ln-scale and σW in the highly variable region)
24

25. Reference Scaling
• A general objective in assessing BE is to compare the log-transformed BA
measure after administration of the T and R products
• An expected squared distance between the T and R formulations to the
expected squared distance between two administrations of the R
formulation
• An acceptable T formulation is one where the T-R distance is not
substantially greater than the R-R distance
• When comparison of T happens in central tendencies and also to the
reference variance, this is referred to as scaling to the reference
variability
25

26. Reference Scaled BE Criteria
• Highly Variable Drugs / Drug Products with CVWR >30 %
• USA: Recommended in API specific guidances; scaling for AUC
and/or Cmax acceptable,
• GMR 0.80 – 1.25; ≥24 subjects
• EU: Widening of acceptance range (only Cmax ) to maximum of
69.84% – 143.19%), GMR 0.80 – 1.25; Demonstration that CVWR
>30% is not caused by outliers; justification that the widened
acceptance range is clinically irrelevant.
26

27. Reference Scaled BE Criteria: USA & EMA
• There is a difference between EMA and FDA scaling approaches
• US FDA: Regulatory regulatory switching condition θS is set to 0.893, which
would translate into
• RSABE is allowed only if CVWR ≥ 30% (sWR ≥ 0.294), which explains to the
discontinuity at 30%
• EMA: Regulatory regulatory switching condition θS avoids discontinuity
27

28. Example 1: Data set 1
• RTRT | TRTR full replicate, 77 subjects, imbalanced, incomplete
• FDA: sWR 0.446 ≥ 0.294 → apply RSABE (CVWR 46.96%)
• a. critbound -0.0921 ≤ 0 and
• b. PE 115.46% ⊂ 80.00–125.00%
• EMA: CVWR 46.96% → apply ABEL (> 30%)
• Scaled Acceptance Range: 71.23–140.40%
• Method A: 90% CI 107.11–124.89% ⊂ AR; PE 115.66%
• Method B: 90% CI 107.17–124.97% ⊂ AR; PE 115.73%
PE = Point estimate; AR = Acceptance range28

29. Example 2: Data set 2
• TRR | RTR | RRT partial replicate, 24 subjects, balanced, complete
• FDA: sWR 0.114 < 0.294 → apply ABE (CVWR 11.43%)
• 90% CI 97.05–107.76 ⊂ AR (CVintra 11.55%)
• EMA: CVWR 11.17% → apply ABE (≤ 30%)
• Method A: 90% CI 97.32–107.46% ⊂ AR; PE 102.26%
• Method B: 90% CI 97.32–107.46% ⊂ AR; PE 102.26%
• A/B: CVintra 11.86%
PE = Point estimate; AR = Acceptance range29

30. Canadian BE Criteria for HVDPs
• The 90% confidence interval of the relative mean AUC of the test to reference
product should be within the following limits:
• 80.0%-125.0%, if sWR ≤0.294 (i.e., CV ≤30.0%);
• [exp(-0.76sWR) × 100.0%]-[exp(0.76sWR) × 100.0%] if 0.294 <sWR ≤0.534 (i.e., 30.0% <CV
≤57.40%); or,
• 66.7%-150.0%, if sWR >0.534 (i.e., CV >57.4%).
• The relative mean AUC of the test to reference product should be within 80.0%
and 125.0% inclusive;
• The relative mean maximum concentration (Cmax) of the test to reference
product should be between 80.0% and 125.0% inclusive.
30

31. Analysis by SAS Proc Mixed
31

32. Example 3: Inverika Data Set;
Two Alverine Formulations; Intra-subject CV ~35%; n = 48
32

33. Individual Bioequivalence (IBE) Metric
2 2 2 2
2 2
0
( ) ( )
max( , )
T R D WT WR
I
WR W
    

 
   

2
2
0
(ln1.25)
I
W





Where
Where
µT = mean of the test product
µR = mean of the reference product
σD
2 = variability due to the subject-by-formulation interaction
σWT
2 = within-subject variability for the test product
σWR
2 = within-subject variability for the reference product
σW0
2 = specified constant within-subject variability
33

34. Population Bioequivalence (PBE) Metric
Where
µT = mean of the test product
µR = mean of the reference product
σTT
2 = total variability (within- and between-subject) of the test product
σTR
2 = total variability (within- and between-subject) of the reference
product
σ0
2 = specified constant total variance
≤θP
34

35. Example 3: Inverika Data Set;
Two Alverine Formulations; Intra-subject CV ~35%; n = 48
35

36. Issues with RSABE
• Advantages
• Sometimes fewer subjects can be used to demonstrate BE for a HVD
• Concerns
• Borderline drugs
• Submission of unscaled and reference-scaled BE statistics for same product
• What if T variability > R variability
• Unacceptably high or low T/R mean ratios
• Number of study subjects
36

37. Questions?
Comments?
drb.chakraborty@cadilapharma.co.in
37

38. Thank You Very
Much
38