Four Levels of In-Vitro-In-Vivo Correlation

Four levels of in-vitro-in-vivo correlation and their regulatory relevance

1. FOUR LEVELS OF IVIVC
AND THEIR EFFECTIVE
IMPLEMENTATION TO
DEVELOP QUALITY
DRUGS
Dr. Bhaswat S. Chakraborty
Sr. VP & Chair, R&D Core Committee
Cadila Pharmaceuticals Ltd.
Former Senior Clinical Reviewer, TPD (Canadian FDA)
Presented at the IVIVC & BABE SUMMIT 2015
Holiday Inn, Mumbai, Nov. 23, 2015
1

2. CONTENT GUIDELINES
 LEVEL A
Correlation of the entire in vitro and in-vivo profiles pertaining to
regulatory relevance
LEVEL B
The principles of statistical moment analysis
LEVEL C
One or more PK parameters correlated with amount of drug dissolved at
several time point of dissolution profile
LEVEL D
Semi quantitative; rank order correlation
Case study
Conclusion
2

3. UNDERSTANDING
CORRELATION
Correlation:
Strength of associative relationship between two variables
Broad class of statistical relationships involving dependence
3

4. 4
Four sets of data with the same correlation of
0.816

5. DEFINITION OF IVIVC
IVIVC is the predictive, mathematical models relating
an in-vitro property such as dissolution and an
in-vivo response, e.g., amount of drug absorbed, thus
allowing an evaluation of the QC specifications, change
in process, site, formulation and application for a
biowaiver etc. –US FDA
Establishment of a rational relationship between a
biological property, or a parameter derived from a
biological property produced by a dosage form, and a
physicochemical property or characteristic of the
same dosage form. – USP
5Valid in-vitro and in-vivo methods valid IVIVC

6. BIOPHARMACEUTICS
CLASSIFICATION SYSTEM (BCS) &
IVIVC EXPECTATIONS
Amidon et al. (1995), Pharm Res, 12, 413-420
6

7. SYSTEMIC DRUG ABSORPTION:
CARBAMAZEPINE CR
15
N STABLE ISOTOPE
STUDY
Wilding et al. Br J Clin Pharmac (1991), 32, 573-579
7

8. CUMULATIVE
IN-VIVO
ABSORPTION
OF CBZ FROM
THE OROS
SYSTEM IN
INDIVIDUAL
SUBJECTS
COMPARED
WITH
CUMULATIVE
IN VITRO
RELEASE
Wilding et al. Br J Clin Pharmac (1991), 32, 573-579
8

9. BCS Class PK Data IVIVR
API –
Physicochemi-
cal Properties
Scale factor
Dosage Form
Properties
Biorelevent
Dissolution
Computer Modeling Using Convolution including Transporters, PK Models,
and PK Parameters, API properties or Drug Release Data
IVIVC
1
2
3
GENERAL APPROACH TO DEVELOP
IVIVC
Wang et al (2009) Diss Tech, 8, 6-12
9

10. IVIVC CORRELATION LEVELS
1.Level A: highest level; point to point relationship
between in-vitro dissolution rate and in-vivo input
rate of the drug from the dosage form
2.Level B: uses statistical moments; MDT
vitro
of the
product is compared to either MRT or MDT
vivo

3.Level C: one dissolution time point (t
50%
, t
90
%..) is
compared to one mean PK parameter (AUC, t
max
or C
max
)
4.Multiple Level C: relates one or several PK
parameters (AUC, C
max ..
) to the amount of drug
dissolved in-vitro at several time points
5.Level D: is a rank order and qualitative analysis.
10

11. LEVEL A: POINT TO POINT
CORRELATION
% Drug absorbed calculated by means of model dependent
techniques such as Wagner-Nelson or Loo-Riegelman or by
model-independent deconvolution
These techniques utilize all of the dissolution and plasma
level data
Purpose of Level A corr. is to define a direct relationship such
that measurement of in-vitro dissolution rate is a surrogate
for in-vivo performance
change in manufacturing site, method of manufacture, raw
material supplies, minor formulation modification, product
strength using the same formulation can be justified
excellent quality control procedure since it is predictive of the
dosage form’s in-vivo performance
Of highest regulatory value
11
Sakore S.and Chakraborty B. (2011). J Bioequiv Availab, S3: 1-12.

12. LEVEL B: UTILIZING
STATISTICAL MOVEMENT
As mentioned above, it uses statistical moments;
MDT
vitro
of the product is compared to either MRT or
MDT
vivo
Level B correlation uses entire in-vitro & in-vivo
data, yet it is not a point-to-point corr., since
number of different in-vivo curves will produce
similar MRT values
A level B correlation does not uniquely reflect the
actual in-vivo plasma level curves
Alone is not enough to justify SUPAC, biowaiver etc.
12
Sakore S.and Chakraborty B. (2011). J Bioequiv Availab, S3: 1-12.

13. LEVEL C: ONE DISSOLUTION PT. TO
ONE PK PARAMETER
Level C relates one dissolution time point (t50%,
t90%, etc.) to one mean PK param. e.g., AUC, tmax
or Cmax
This is a weak level of correlation as only partial
relationship between absorption and dissolution
Does not reflect the complete shape plasma-conc.
time curve, defining performance of a drug in-vivo
In the early stages of formulation development Level
C correlations can be useful when pilot formulations
are being selected
Biowaiver is generally not possible
13
Sakore S.and Chakraborty B. (2011). J Bioequiv Availab, S3: 1-12.

14. MULTIPLE LEVEL C
Relationship between Cmax, AUC, or any PK
parameters and amount of drug dissolved at several
time points of dissolution profile
It may be used to justify a biowaivers provided that
the correlation has been established over the entire
dissolution profile with one or more PK parameters
Multiple Level C correlation should be based on at
least three dissolution time points covering the early,
middle, and late stages of the dissolution profile
Level A is sometimes likely when multiple level C is
achieved at each time point at the same parameter
thus effect on the in-vivo performance of any change in
dissolution can be assessed
14
Sakore S.and Chakraborty B. (2011). J Bioequiv Availab, S3: 1-12.

15. LEVEL D: RANK ORDER &
QUALITATIVE
It is not a formal correlation but it is a semi
quantitative (qualitative analysis) and rank order
correlation
Not considered useful for regulatory purpose but can
be serves as an aid in the development of a
formulation or processing procedure
15
Sakore S.and Chakraborty B. (2011). J Bioequiv Availab, S3: 1-12.

16. Level A – point-point; first
deconvolution to get in-vivo
%drug absorbed, then
compare with %dissolved
Level B – Statistical
moments; MRT or MDT in-
vivo vs. MDT in vitro
Level C – single point; PK
parameter vs. %dissolved
Level A
Level B
Level ALevel C
Malinowski and Marroum, Encyclopedia of Contr. Drug Deliv.
16

17. OVERALL IVIVC DEVELOPMENT FOR
MR FORMULATIONS
For Market
Retig et al. Diss Tech, Feb. 2008, 6-8
17

18. CASE STUDY TO DEVELOP
A LEVEL A CORRELATION
18

19. THE PRODUCT
A new type of prolonged release Hydrocodone formulation
based on Egalet® technology
Three tablet GMP batches (A to C) developed, all containing
20mg of Hydrocodone as tartrate salt.
They differed solely by the mass of the final tablet,
corresponding to different diameters and lengths of the
tablet (6, 7.5 or 9 mm, respectively) and adjusted by an
increase of excipients’ mass
An IR tablet of 10 mg of Hydrocodone in combination with
325 mg of Paracetamol was also included in the clinical
study as a reference and in order to perform deconvolution
19
Hemmingsen PH et al.(2011). Pharmaceutics,3:73-87

20. DISSOLUTION STUDIES
Pharmacopoeial media: phosphate buffer pH 6.8; USP
Apparatus 2 paddle method (Vankel VK7025 coupled to a
Varian Cary 50 UV-visible spectrophotometer); dissolution
vol: 1,000 mL; paddle speed: 50 rpm: temp.: 37 °C.
20
Hemmingsen PH et al.(2011). Pharmaceutics,3:73-87

21. IN-VIVO BIOAVAILABILITY
STUDIES
4-arm, single dose, randomized cross-over study comparing three
test tablets to the reference
Plasma samples (0 - 42 hr) were measured by a validated HPLC-
MS method
C
max
, T
max
& AUC were calculated, bioequivalence parameters under
consideration were Cmax and AUC
T
max
was not analyzed as prolonged release formulation is involved
Absorption kinetics were calculated using a deconvolution
technique using the IR reference tablet as response function
Deconvolution allows isolating the input (« absorption ») function as a
function of the observed concentration for the studied tablet and for
the IR reference tablet
This input function reflected the in-vivo release observed after
administration of the PR test tablets
Simulations of the curves from the theoretic input were
performed using convolution
21
Hemmingsen PH et al.(2011). Pharmaceutics,3:73-87

22. IN-VIVO BIOAVAILABILITY
STUDIES
22
Hemmingsen PH et al.(2011). Pharmaceutics,3:73-87

23. IN-VIVO RELEASE &
ABSORPTION OF 3 TABLETS
23
Hemmingsen PH et al.(2011). Pharmaceutics,3:73-87

24. IVIVC
The clinical study was designed to support a level A
correlation
In a linear correlation, the in-vitro dissolution and in-vivo
input curves may be directly super-imposable or may be
made to be super-imposable by the use of a scaling factor
e.g., if the dissolution is faster than the in vivo input rate then the
two curves are not super-imposable
in this case a time scaling may be applied on the in vitro data for each
%absorbed, the corresponding time in-vitro using a Levy’s plot
Model predictability was estimated internally by
comparison of prediction errors on Cmax & AUC derived
from mean observed and predicted in vivo data obtained by
convolution
Regulatory guidelines state prediction errors for Cmax and
AUC should not exceed 10 % as a mean and none greater
than 15%
24

25. DISSOLUTION VS. IN-VIVO
ABSORPTION
25
TIME SCALING: LEVY’S PLOT
Hemmingsen PH et al.
(2011).
Pharmaceutics,3:73-87

26. IVIVC OF ALL FORMULATIONS
USING A COMMON NON-LINEAR
TIME SCALE
26
Hemmingsen PH et al.(2011). Pharmaceutics,3:73-87

27. PREDICTABILITY BASED ON
AUC AND C
MAX
27
Based on IVIVC & in vitro data, input kinetics were back calculated and then,
based on this input function, a convolution was performed to simulate the in
vivo plasma concentration curve
The predictability was good and in accordance with the FDA recommendation
(5) with a mean error of −0.32% and −6.63% on Cmax and AUCinf, respectively,
no case being greater than +10%
Hemmingsen PH et al.(2011). Pharmaceutics,3:73-87

28. CALCULATION OF
DISSOLUTION LIMITS
One application of IVIVC is to predict bioavailability and to set dissolution
limits. The residual error from ANOVA for Cmax and AUC and modelized
absorption to a multi zero order absorption and a few more steps dissolution
limits for bioequivalence.
28

29. IN-VIVO PREDICTION OF
FORMULATION B TO ENSURE
BIOEQUIVALENCE
Based on the dissolution limits and on the modelization of the
absorption, the in-vivo curves were simulated in three conditions
corresponding to the modelized absorption
29
Hemmingsen PH et al.(2011). Pharmaceutics,3:73-87

30. CONCLUSIONS
Biorelevant and reliable dissolution profiles can predict the in-
vivo absorption of drugs from CR formulations
Batches with similar dissolution will be BE and dissimilar
dissolution will be non-BE
Level A (point-to-point) is most useful for regulatory purposes;
multiple level C is also acceptable for regulatory purposes
At least 3 lots (desirable, fast and slow) must be established with
IVIVC and proper reference
Time scaling and modeling of in-vitro & in-vivo parameters must
be accurate and validated
Predictability should be high
IVIVC is useful in
QbD, SUPAC and biowaivers…
30

31. Thank You Very Much
Acknowledge: Raji Nair
31