Download presentation_5-22-2015-10-57-4.counting procedure and dose

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Large sample dose content
uniformity test: parametric and
nonparametric (counting)
Meiyu Shen, PhD
Collaborators: Xiaoyu Dong, Ph.D., Yi Tsong, PhD
Office of Biostatistics, CDER, FDA
* This presentation contains opinions of the authors that do not
represent the official position of U.S. Food and Drug Administration
1
Outline
• Purpose of uniformity of dosage unit
• Harmonized USP dose content uniformity test with a small
sample
• Large n dose content uniformity test
– EU methods
• Option 1: Parametric method
• Option 2: Nonparametric method (Counting method)
– Two one-sided tolerance interval method
• Comparison between the EU method and the two one-sided
tolerance interval method
• Conclusion
2
Uniformity of dosage unit
• The purpose of uniformity of dosage unit
– The degree of uniformity in the amount of the
drug substance among dosage units.
• Demonstrated by one of the follows
– Dose content uniformity (focus here)
• based on the assay of the individual content of drug
substance(s) in a number of dosage units
– Weight variation
3
Indifferent zone
• M:
M  98.5% if X  98.5%.;
M  101.5% if X  101.5%;
M  X otherwise.
•
M  X  K  s  15
83.5%  X  K  s  X  K  s  113.5% if X  98.5%.;
86.5%  X  K  s  X  K  s  116.5% if X  101.5%;
K  s  15 otherwise.
4
Harmonized USP DCU for small n
Step 1, 10 tablets
Step 2, additional 20 tablets
M  X  K  s  15%
No outside (0.75 M, 1.25M)
K  2.4
M  X  K  s  15%
No outside (0.75 M, 1.25M)
K  2.0
No
Yes
Pass
Fail
Yes
5
Pass
EU option 1 for large n≥100
Take n tablets, {Xi}, i=1,2,…,n
M  X  K  s  15%
n
&  I  X i  1  0.25M   c 2.
i 1
Here, K  Z 1 
2
n  1/ n
,   0.84,   0.91.
 2 n  1
No
Fail
Yes
Pass
6
EU option 2 for large n≥100
Take n tablets, {Xi}, i=1,2,…,n
n
 I X
i 1
i
 1  0.15  c1
n
&  I  X i  1  0.25  c 2.
i 1
No
Fail
Yes
Pass
7
EU Option 2 acceptable number of
individual units c1 outside (1±0.15) and c2
outsides (1±0.25)
n
c1
c2
100
3
0
123
3
0
150
3
0
176
5
1
196
6
1
….
…
…
8
PTIT_matchUSP90
Take n tablets, {Xi}, i=1,2,…,n
100%  X  K  s  15%
1  Pn 
 X  Ks
Here, K : Pr 
f x :  ,  dx 
  1
2 
 
2
1
f x :  ,   
e  x    2 2
2 


No
Fail
Yes
Pass
9
EU option 1 and PTIT_matchUSP90
EU Option 1
• Two-sided tolerance
interval
PTIT_matchUSP90
• Two one-sided tolerance
interval
– Control probability within
(85,115)%
• Two-sided hypothesis
H 0 : Pr L  X  U   P
H 1 : Pr L  X  U   P
P
– Control probability each tail
outside (85,115)%
•
Two one-sided hypothesis
H 0 : Pr  X  L   1  Pn  / 2  Pr X  U   1  Pn  / 2
H 1 : Pr  X  L   1  Pn  2  Pr  X  U   1  Pn  2
(1-P)/2
10
EU option 1 and PTIT_matchUSP90
EU Option 1
• Formula for K
– K  Z 1
2
n  1/ n
 2 n  1
– Confidence level:   0.84
– Center Coverage:   0.91
PTIT_matchUSP90
• Formula for K
–
K  tn 1,1 (Z1 p ( n ) n ) / n
2
– Confidence level: 1-α=0.95
– Each tail probability: (1-p(n))/2
– For n=30, p=82.04%,
11
K values of EU Option 1 and
PTIT_matchUSP90
Sample size
PTIT_matchUSP90
EU Option 1
100
2.0475
2.15
1000
2.2321
2.27
12
Normal: on target product, mean=100%
13
Normal: off target product, mean=102%
14
Mixed normal: on target overall mean
15
Mixed normal: off target overall mean
16
Bias of EU Option 1
17
Special distribution
• Assume the individual tablet dose content is
distributed as a uniform distribution in the range
from 85% to 115% with 97% probability and a
value 84% with 3% probability.
– The probability of passing USP harmonized DCU is
3.72% for a sample size of 30 tablets.
– Comparison of EU Option 2 and PTIT_matchUSP90
in next table
• EU Option 2 has 45.5% probability to pass the DCU test
when n=300.
• the PTIT_macthUSP90 has zero passing probability for
n≥100.
18
EU Option 2 and PTIT_matchUSP90
Xi: a uniform distribution in the range from
85% to 115% with 97% probability and a value
84% with 3% probability
Sample size, n
100
150
200
300
500
1000
Acceptance probability
EU option 2 PTIT_USPmatch90
0.6458
0
0.5276
0
0.6047
0
0.455
0
0.3509
0
0.2075
0
19
EU Option 2 and PTIT_matchUSP90
for 2 special cases
Case
Sample size, n Passing probability
EU Option 2 PTIT_matchUSP90
Xi is 100 with 97%
probability, and 50
with 3%
probability.
Xi is 90 with 97%
probability, and 50
with 3%
probability.
100
150
0.641
0.523
0.196
0.172
100
150
0.648
0.530
0.045
0.012
20
Conclusion
• A large difference in acceptance probability between EU
option 1 and PTIT_matchUSP90 when the batch mean is
off-target.
– Larger passing probability for EU Option 1 than PTIT_matchUSP90
• No much difference in acceptance probability between EU
option 1 and PTIT_matchUSP90 when the batch mean is
on-target.
• Bias of EU Option 1
– EU Option 1 has higher probability of passing the off-target product
than that of passing the on-target mean product for a given coverage
within (85%, 115%)
21
Conclusion (continued)
• EU Option 2
– Issue with a large variability for a mixture of 97% probability of
distributing uniformly with (85%, 115%) and 3% probability of being
84%) using a sample of 200
• 60% probability to pass EU Option 2
• 0% probability to pass PTIT_matchUSP90
• 3% probability to pass USP harmonized
– Issue with a location shift of the mean product
• The same probability to pass the EU Option 2 for 97% population with 100%
content and 97% population with 90% content.
• Off target product: 97% population with 90% content using a sample of 150.
– >50% probability to pass the EU Option 2
– About 1% probability to pass PTIT_matchUSP90
22
References
•
•
•
•
USP Pharmacopoeia 2015
European Pharmacopoeia 7.7
European Pharmacopoeia 8.1
Meiyu Shen, Yi Tsong, Xiaoyu Dong, Statistical
Properties of Large Sample Tests for Dose Content
Uniformity, Therapeutic Innovation & Regulatory
Science, 2014, Vol. 48(5) 613-622
• Meiyu Shen and Yi Tsong, Bias Of The United States
Pharmacopeia Harmonized Test For Dose Content
Uniformity, United States Pharmacopeia forum, January
2011
23
24
Related documents