9. • Innovator Pharmaceutical Product ( Safety and
efficacy)
• A generic product should not be a comparator as long as
an innovator product is available.
• Selection should be made at the national level by the
drug regulatory agency
– National Innovator
– WHO comparator product ( quality-safety-efficacy and
has reference to manufacturing site)
– ICH or associated country comparator product
Comparator Product
10. In The Case that Innovator Product cannot be
identified
• Important Criteria for Selection
– Product is in the WHO list
– Approval in an ICH – Associate Country- Pre-qualified by WHO
– Extensive documented use in clinical trials
reported in peer-reviewed scientific journals
– Long unproblematic post-market surveillance (“well selected
comparator”)
A product approved based on comparison with
A non domestic comparator product may not be
interchangeable with currently marketed
domestic products
20. Therapeutic Equivalence can be assured when
the multisource product is:
pharmaceutically equivalent and
bioequivalent.
TE = PE + BE
Therapeutic Equivalence of Multisource
Product
The concept of interchangeability applies to:
1. - the dosage form and
2. - the indications and instruction for use.
25. Origin of ABE
• A survey of physicians suggested that for
most drugs, a difference of up to 20% in
dose between two treatments would have
no clinical significance
26. Average Bioequivalence
• two drug products are Bioequivalent ‘on
the average’ when the (1-2α) confidence
interval around the Geometric Mean Ratio
falls entirely within 80-125% (regulatory
control of specified limit)
27. AVERAGE BIOEQUIVALENCE
Comapre the population average response of the log-transformed Bioavailability
Parameters after administration of the Test and Reference products.
Test
Reference
The same Mean different
Variances ? What to do?
{ } Pf RT −≤∆≤ΘΘ≤∆ 1)(Pr 2,1
Confidence Interval
Which BA metrics and which distribution
parameters must meet criteria
The width of the interval
The assigned assurance probability
Average Response
test within 80 -125% 25.1)(8.0 LnLn RT ≤−≤ µµ
80 125
90 111NTI
digoxin, phenytoin, warfarin,
theophylline, lithium
67 150
Who decides the goal post?
Clinical Judgement
CMS
Variability of Reference Product
Population vs Individual Dose
-Response curves
28. Some International Criteria
Country/Region AUC 90% CI
Criteria
Cmax 90% CI
Criteria
Canada (most drugs) 80 – 125% none
(point estimate only)
Europe (some drugs) 80 – 125% 75 – 133%
South Africa (most drugs) 80 – 125% 75 – 133% (or broader
if justified)
Japan (some drugs) 80 – 125% Some drugs wider than
80 – 125%
Worldwide (WHO) 80 – 125% “acceptance range for
Cmax may be wider
than for AUC”
29.
30. )( 05.0,
exp.100%90 abdfbA SEtLSMLSM
CIGeometric
±−
=
Least Square
Means from ANOVA
t-statistic with
0.05 in one
tail
Standard
Error
31.
32.
33.
34.
35.
36. Limitations of 2-Period Designs
• The intra subject variance associated with the Test
and Ref products may not be the same
• A poor pharmaceutical product may have inflated
intrasubject variance because of high within
formulation variability
• The residual variance in 2-period designs averages the
intrasubject variance of the two products
– The Test and Ref intrasubject variance cannot be
separated
37.
38. REPLICATED CROSSOVER DESIGNS FOR TWO
FORMULATIONS
OPTIMAL FOR CARRYOVER ESTIMATION
PERIOD
SEQUENCE 1 2 1 2 3 1 2 3 4
1 T T T R R T T R R
2 R R R T T R R T T
3 T R T R R T
4 R T R T T R
SWITCHBACK DESIGNS
SEQUENCE 1 2 3 1 2 3 4
1 T R T A B A B
2 R T R B A B A
39. Replicate Designs
• Yields information on the Intrasubject
Variance
• Ideally, intrasubject variance of the Test
product should be similar to the
intrasubject variance of the Reference
product
40.
41. What do we learn from ANOVA Analysis
• The sources of variance in the model are
– Product
– Period
– Sequence
– Subject (Sequence)
– Residual variance
These account
for all the inter-subject
variability
This estimates
Intra-subject
variability
Source: Modified from K. Midha
42. ‘Fixed Effects” in ANOVA
• Product
• Period
• Sequence
• Subject nested within sequence is usually
significant (f-test) because of large
variability between subjects
These fixed
effects usually are not
significant in the f-test
Source: Modified from K. Midha
43. The Residual Variance (SW
2
)
• Sources of Variability
– Intra-subject variance in Pharmacokinetics
– Analytical variability
– Subject by formulation interaction
– Unexplained random variation
WSVCVANOVA
VarianceResidualCVANOVA
≈−
×=− %100
Source: Modified from K. Midha
44. µT, obs = 24.7 ng/ml
µR, obs = 23.7 ng/ml
v = 22
t0.95(v) = 1.7171
s = 5.693
n = 24
s*sqrt 2/n = 1.543
24.7 – 23.7 +/- 1.717 (1.643) ng/ml
1 +/- 2.82 ng/ml
-1.82 ng/ml; 3.82 ng/ml
The lower CI limit = 23.7 – 1.82 / 23.7
= 92.3 %
The Upper CI limit = 23.7 + 3.83 / 23.7
= 116%
N
StobsRobsT
2
)(95.0,, νµµ ±−
Example using ANOVA results
45.
46.
47.
48. The ‘ANOVA-CV’
• The ANOVA-CV which is easily
calculated from the residual variance is
an estimate of WSV
)(
%100
WSVeectVariancWithinSubjCVANOVA
VarianceResidualCVANOVA
≈−
×=−
49. Variability
• It is well known the Between Subject
Variance (BSV) can be very high
– Biological variation
– Within Subject Variance (WSV) contributes to
BSV
• WSV can also be high e.g. highly variable
drugs and highly variable drug products
• Drugs with an ANOVA-CV ≥ 30% are
defined as ‘highly variable drugs’
<number>
Over the years, BE limits of 80-120% (80-125% on the log scale) have served us well with drugs that do not exhibit complicated behaviour such as:
-a narrow therapeutic range
-toxicity associated with Cmax
-non-linear kinetics
<number>
<number>
The problems in dealing with HVDs and HVDPs is outside the scope of this presentation.
<number>
<number>
<number>
<number>
<number>
The above equation works OK at low WSVs. At higher WSVs the following equation gives a better answer.
<number>
Even today, I suspect many physicians have no idea the magnitude BSV can assume, let alone WSV > 30%!
<number>
<number>
The usual SE is calculated as follows:
In the Two One-Sided Test, the denominator is the number of subjects in each sequence when n1 = n2
<number>
The consumer risk is set at 0.05 (5%) by means of the t-statistic with 0.05 in one tail.