Metformin: Mechanistic Absorption Modeling and IVIVC Development

Metformin: Mechanistic Absorption
Modeling and IVIVC Development
Maziar Kakhi*, Ph.D.
FDA
Silver Spring, MD 20993
Viera Lukacova, Ph.D.
Simulations Plus
Lancaster, CA 93534
[email protected]
[email protected]
AAPS Annual Meeting and Exposition
Orlando FL, October 25-29, 2015
*
Disclaimer: The views expressed in this presentation are those of the author and do not reflect the opinion nor
the policy of the FDA.
1
Presentation Overview
• Metformin – general considerations
• Validation study database
• Metformin absorption model
• IVIVC development/validation
• Conclusions
2
Metformin – General Considerations
• Used in the treatment of type 2 diabetes mellitus.
• Highly soluble.
• Incompletely absorbed with an oral bioavailability ranging from 40 to
60%.
• Undergoes site-dependent absorption primarily in the upper part of
the intestine.
• Carrier-mediated transport known to play active role in drug
absorption and disposition.
• Poses a challenge for classical level A in vitro – in vivo correlation
(IVIVC) approaches.
3
Validation Study Database
Data were only considered for pharmacokinetics of metformin administered in healthy human subjects.
Study
Tucker et al. Br. J. Clin. Pharmac
(1981), 12, 235-246
Pentikäinen et al. Eur. J. Clin.
Pharmacol. (1979), 16, 195-202
Formulation
Route of
Administration
Dose
Fed /
Fasted
Avg / Ind
Cp-t Data
infusion (15 min)
Intravenous
250 mg
-
Average (N=4)
IR tablet
(Glucophage)
Oral
500 &1500 mg
Fed
Average (N=4)
Infusion (5 min)
Intravenous
IR tablet
Oral
4.4 μCi/500 mg
Fasted
Individual (N=3)
500 mg
Fasted
Bolus, solution
BMS clinical data study 1
(April-July 1996)
Proximal SI
Distal SI
Proximal SI
Infusion (4 hrs)
Distal SI
Average (N=8)
Average (N=4)
Average (N=10)
500 mg
Fasted
Colon
Average (N=6)
N= 1 or 2
Crossover




IR tablet ('minitabs')
Oral
8x50 mg / 30 min
Fasted
Average (N=16)

BMS clinical data study 2
(November 1996)
IR tablet
(Glucophage)
ER capsule (fast,
medium, slow)
Oral
1000 mg
Fed &
Fasted
Individual (N=8)

BMS clinical data study 3
(March-April 2010)
IR tablet
Oral
500 mg
1000 mg
Fed
Individual (N=27)
Individual (N=26)

4
Metformin absorption model
• Metformin in solution behaves as an hydrophilic cation.
 Limited passive transcellular absorption.
• Passive paracellular absorption mainly in upper GI tract: molecule size vs.
decreasing paracellular pore radius.
• Paracellular absorption is not assumed to be saturable.
• Metformin is primarily a substrate for organic cation transporters OCT1 and
OCT3 along basolateral and apical membrane of intestinal epithelia.
• OCT1 influxes metformin from portal vein into enteroctye; OCT3 influxes
metformin from gut lumen into enterocyte.
5
Metformin absorption model – contd.
• Carrier-mediated transport modeled using Michealis-Menten kinetics.
• Disposition of metformin modeled using compartmental mammillary PK
model.
• Major route of elimination is secretion in urine.
6
Sponge Effect Hypothesis1
∆𝒕
• Metformin dose undergoes
predominantly paracellular absorption.
• Uptake across apical (AP) membrane
occurs by means of OCT3-transporter.
1
Assumed OCT-like transporter effluxes
metformin into lumen for further
paracellular absorption or re-uptake
into enterocytes downstream.
Proctor WR, Bourdet DL, Thakker DR. 2008. Drug Metabolism and Disposition 36(8):1650-1658.
7
Modeling Strategy
Investigation of the effect of:
•
Dosing locations for site administration studies.
Labeled
site
Intestinal compartment where the drug was administered in the simulation
Proximal
SI
Model 1
Middle of
Duodenum
Distal SI
Jejunum2
Colon
Ascending
colon
Model 2
Model 3A
Model 3B
Model 4
Jejunum1
Jejunum1
Jejunum1
Jejunum1
Beginning of
Ileum1
Ascending
colon
Middle of
Ileum1
Ascending
colon
Middle of
Ileum1
Ascending
colon
End of Ileum1
Ascending
colon
• ASF models of Gastro Plus v6.8 (Opt logD 6.1 SA/V, Theoretical SA/V, and
Theoretical).
• Inclusion/exclusion of colonic OCT expression level.
• Intestinal electrical potential gradient: 58,102 (default) mV.
8
Modeling Strategy - cont.
• Fit CLr, Total Peff, and Vmax values for intestinal transporters for each set of
model options (administration sites, ASF model, EPG, presence of OCT in
colon).
• Models screened for:
• Fitted ratio Peff,Trans/Peff,Para across site administration studies.
• Quality of prediction of Cp-t observations from PO-administration studies.
9
Absorption Model Validation – Tucker IV/PO
• Higher STT means initially less drug
available
for
absorption,
but
the
absorption will continue for longer.
• Tucker et al. Br. J Clin Pharmac. (1981)
12:235-246 report Cmax of 3.1 µg/ml for
1500 mg.
10
Absorption Model Validation – Pentikäinen IV/PO
• Little observed variability
clearance between 3 subjects.
in
renal
• Simulations based on default STT of
0.25h.
11
Absorption Model Validation – BMS Study 3
•
STT fixed at 1h for both simulations.
•
CLr and Vmax scaling factors fitted against both IR doses.
12
ER Formulation – BMS Study 3 / In Vitro Data
f2 Analysis
Fast-Medium
22
Fast-Slow
17
Medium-Slow
35
Apparatus
USP 2
Medium
Phosphate buffer, pH 6.8
Volume
1000 ml
Paddle speed
100 rpm
Average
(N=6)
• Tmax,Fed < Tmax,Fasted
• Cmax,Fasted-Fast > Cmax,Fasted-IR
• Fasted IR & Fast formulation Cp-t curves exhibit
similar initial rate of increase.
• Balan et al. 2001. JPS. 90(8):1176-1185: use of
numerical deconvolution failed for fasted data
set.
• ‘Extended’ convolution method required to
achieve (internally) predictive level A correlation.
13
Procedure for Mechanistic Deconvolution
• Mechanistic deconvolution performed for fed and fasted arm of BMS clinical
data study 2.
• In vivo dissolution profile assumed to behave as double Weibull function.
• Initial estimate for in vivo dissolution profile obtained from in vitro data.
• STT, Vmax scaling factors and CLr fitted to average and individual Cp-t
profiles for IR formulation.
• STT additionally fitted to individual Cp-t profiles for ER formulations to study
effect on resulting IVIVC.
14
IVIVC – Levy Plot for Fasted State
STT rate limiting
step for absorption
of fast formulation?
Correlation function
defined to be ≥ 0
•
•
•
STT fitted to individual Cp-t profiles for ER formulations.
IVIVC constructed using Medium and Slow formulations.
Quadratic function had lowest AIC.
15
IVIVC – Validation Fasted State
Individual deconvolution, STT fitted to Ind ER profiles
Study
STT=1.69 h
STT=1.63 h
Fasted
STT=2.58 h
Average deconvolution, STT fitted to Avg ER profiles
Study
Fasted
Formulation
Parameter
AUC(0-t) ng h/ml
Slow Int
Cmax ng/ml
AUC(0-t) ng h/ml
Medium Int
Cmax ng/ml
AUC(0-t) ng h/ml
Avg Abs %PE Int
Cmax ng/ml
AUC(0-t) ng h/ml
Fast Ext
Cmax ng/ml
%PE
3.9
-23.0
-10.1
-22.6
7.0
22.8
-9.1
-16.9
Formulation
Parameter
AUC(0-t) ng h/ml
Slow Int
Cmax ng/ml
AUC(0-t) ng h/ml
Medium Int
Cmax ng/ml
AUC(0-t) ng h/ml
Avg Abs %PE Int
Cmax ng/ml
AUC(0-t) ng h/ml
Fast Ext
Cmax ng/ml
%PE
19.2
4.5
12.4
-10.3
15.8
7.4
5.7
-14.0
Individual deconvolution, STT fitted to IR profile
Study
Fasted
Formulation
Parameter
AUC(0-t) ng h/ml
Slow Int
Cmax ng/ml
AUC(0-t) ng h/ml
Medium Int
Cmax ng/ml
AUC(0-t) ng h/ml
Avg Abs %PE Int
Cmax ng/ml
AUC(0-t) ng h/ml
Fast Ext
Cmax ng/ml
%PE
10.5
7.0
16.0
-7.9
13.2
7.4
1.7
-19.6
16
IVIVC – Levy Plot for Fed State
•
•
•
STT fitted to individual Cp-t profiles for ER formulations.
IVIVC constructed using all formulations.
Power law function had lowest AIC.
17
IVIVC – Validation Fed State
Individual deconvolution, STT fitted to Ind ER profiles
STT=1.18 h
Study
STT=0.95 h
Fed
STT=1.05 h
Average deconvolution, STT fitted to Avg ER profiles
Study
Fed
Formulation
Parameter
AUC(0-t) ng h/ml
Slow Int
Cmax ng/ml
AUC(0-t) ng h/ml
Medium Int
Cmax ng/ml
AUC(0-t) ng h/ml
Fast Int
Cmax ng/ml
AUC(0-t) ng h/ml
Avg Abs %PE Int
Cmax ng/ml
%PE
-2.3
-23.2
-0.6
-11.0
-8.5
-12.9
3.8
15.7
Formulation
Parameter
AUC(0-t) ng h/ml
Slow Int
Cmax ng/ml
AUC(0-t) ng h/ml
Medium Int
Cmax ng/ml
AUC(0-t) ng h/ml
Fast Int
Cmax ng/ml
AUC(0-t) ng h/ml
Avg Abs %PE Int
Cmax ng/ml
%PE
1.0
-30.3
7.2
-7.6
-6.1
-10.5
4.8
16.1
Individual deconvolution, STT fitted to IR profile
Study
Fed
Formulation
Parameter
AUC(0-t) ng h/ml
Slow Int
Cmax ng/ml
AUC(0-t) ng h/ml
Medium Int
Cmax ng/ml
AUC(0-t) ng h/ml
Fast Int
Cmax ng/ml
AUC(0-t) ng h/ml
Avg Abs %PE Int
Cmax ng/ml
%PE
-16.3
-44.0
-3.2
-17.8
-9.2
-6.5
9.6
22.8
18
IVIVC – Validation Comparison
Study: Fasted State
Formulation
Slow Int
Balan et al.
Basic Convolution
Medium Int
Fast Int
Slow Int
Balan et al.
Extended Convolution
Medium Int
Fast Int
Slow Int
Phoenix WinNonlin
Numerical Deconvolution
Medium Int
Fast Int
Parameter
AUC(0-22) ng h/ml
Cmax ng/ml
AUC(0-22) ng h/ml
Cmax ng/ml
AUC(0-22) ng h/ml
Cmax ng/ml
%PE
102.7
50.4
56.1
9.9
17.1
3.9
AUC(0-22) ng h/ml
Cmax ng/ml
AUC(0-22) ng h/ml
Cmax ng/ml
AUC(0-22) ng h/ml
Cmax ng/ml
-1.6
-7.3
1.4
-10.8
-2.4
-11.0
AUC(0-24) ng h/ml
Cmax ng/ml
AUC(0-24) ng h/ml
Cmax ng/ml
AUC(0-24) ng h/ml
Cmax ng/ml
46.8
-22.9
26.9
-41.5
-0.3
-35.8
19
IVIVC – Validation Comparison
• Mechanistic model predicted the
changes in % of drug entering portal
vein across the three formulations
• Similar trends were previously fitted
using the extended convolution
based approach
20
Conclusions
• Developed level A IVIVC is currently not adequately predictive but
even the preliminary IVIVC based on mechanistic model:
• performs better than numerical deconvolution and basic one-
stage direct convolution methods.
• does not require adjustment of bioavailable fraction to dose, as
required in extended convolution.
• Model predictions show that stomach transition time plays a critical
role in absorption behavior of metformin.
• Mechanistic absorption model predicts Cmax roughly proportional to
dose.
21
Acknowledgments
• Peter Timmins Ph.D., Drug Product Science and Technology,
Bristol-Myers Squibb, Moreton, United Kingdom
• Jonathan Brown, Ph.D., Drug Product Science and Technology,
Bristol-Myers Squibb, Moreton, United Kingdom
• John Crison, Ph.D., formerly Drug Product Science and
Technology, Bristol-Myers Squibb, New Brunswick, NJ, USA
22
Back Up slides
23
Fraction(s) of Drug Absorbed
Fa
F
FDp
Absorption
Dose
Portal
vein
D
A
Liver
SC Bioavailability
PV
Gut
wall
To faeces
Metabolism
Metabolism
* Modified from van de Waterbeemd, H, and Gifford, E. ADMET In Silico Modelling:
Towards Prediction Paradise? Nat. Rev. Drug Disc. 2003, 2:192-204
24
ACAT Model – Gastro Plus v8.6
25
Permeability
Efflux
Transporter
Influx
Transporter
Passive
Diffusion
Blood
Paracellular
Transport
Tight
Junction
26
Physicochemical Properties
Parameter
Value
Source
LogP
-1.26 (-1.432)
ADMET Predictor
LogD
-2.27 (-3.37@pH 4.03)
ADMET Predictor
pKa
2.67 & 12.04 (2.8 &11.52)
ADMET Predictor
Molar mass
129.17 g/mol4
Solubility
≥ 165 mg/ml (300 mg/ml5, 500 mg/ml6)
Total Peff (jejunem)
Diffusion coefficient
1.18 x10-5 cm/s (3.26x10-5 cm/s in rat
 1.20 x10-5 cm/s in human7)
1.14 x10-5 cm2/s
ADMET Predictor
Optimized8
ADMET Predictor
2
Graham et al. 2011. Clin Pharmacokinet. 50(2):81-98.
BMS internal data.
4 Dose set to metformin free base in all simulation studies
5 Desai et al. 2014. J. Pharm. Sci. 103:920-926
6 Bretnall, Clarke. 1998. Brittain HG, ed. Analytical Profiles of Drug Substances and Excipients, Vol. 25. Academic Press, pp. 243-293.
7 Song et al. 2006. World J Gastroenterol. 12(25): 4064-4070
8 Optimized based on BMS site administration/absorption data. ADMET Predictor calculates 0.66 x 10 -4 cm/s.
3
27
PK and Physiological Properties
Parameter
Value
Source
% drug unbound in protein
100 %
Tucker et al. 1981
Blood/plasma conc. ratio
0.559
Tucker et al. 1981
Central compt. VC
Fitted to IV data
Tucker et al. 1981
Pentikäinen et al. 1979
Renal Clearance
Fitted to IV/IR data
Multiple studies
Efflux/Influx Vmax
Fitted to IR data
Multiple studies
Influx Km (Basolateral, OCT1)
292 mg/l
UCSF-FDA TransPortal10
Influx Km (Apical, OCT3)
202 mg/l
UCSF-FDA TransPortal
Efflux Km (Apical)
1655 mg/l
Optimized
9
From data of Tucker et al. B/P ratio was calculated and determined to be very low. Therefore, no RBC binding assumed within first few
hours suggesting that for normal hematocrit B/P ratio is 0.55.
10 http://dbts.ucsf.edu/fdatransportal/compounds/metformin/
28
Transporter (Initial) Estimates
29
Tucker et al. Br. J Clin Pharmac. (1981) 12:235-246
30
Tucker et al. Br. J Clin Pharmac. (1981) 12:235-246
31
Mechanistic Deconvolution: In Vivo Release
Fasted
Fed
32