Bioequivalence and Bioavailability Forum

Main page Policy/Terms of Use Abbreviations Latest Posts

 Log-in |  Register |  Search

Back to the forum  Query: 2017-11-22 10:25 CET (UTC+1h)
 
DaLoe
Junior

Spain,
2017-10-19 10:50

Posting: # 17899
Views: 492
 

 Outlier PK Study – Non Normal Distribution [Regulatives / Guidelines]

Good Morning

Recently we have conducted a Fasting PK study (single croosover). The statistical report is as follow:

Dependent   Ratio_%Ref_ CI_90_Lower CI_90_Upper CV%
Ln(Cmax)    118.34      103.66      135.11      37%
Ln(AUClast) 103.72       97.87      109.92      15%


As can be seen, the upper limit of the confidence interval do not meet PK criteria.

One of the subject is clearly an outlayer according Lund test. Removing the subject of the study, the results would be as follow:

Dependent   Ratio_%Ref_ CI_90_Lower CI_90_Upper CV%
Ln(Cmax)    112.3365    101.6649    124.1284    29.36%
Ln(AUClast) 102.4207     96.91779   108.2361    14.98%


resulting in meeting PK criteria, borderline, but meets criteria.

The thing is in my understanding it is not accepted to exlude a subject after the bioanalysis. How can we argue an outlier?

On the other hand, we have demostrate that the distribution of the RLD in case of CMax it is not normal. If we apply the arithmetic means (not transformed) to our subjects, the result is:

Obs PK   Mean_Ratio Lower_90 Upper_90 BE_Limits Result
 1  AUCT 101.989    96.703   107.274  80-120    BE Untrans.
 2  Cmax  93.634    83.252   104.016  80-120    BE Untrans


meeting perfectly the BE criteria (80 - 120%) even including the outlier.

Due the distribution is not normal, neither in RLD nor in Test, but extremely not normal in RLD, when we normalize the sequence the results show we are not meeting PK criteria because of the upper limit of the confidence interval of the CMax.

Obs PK   Mean_Ratio Lower_90 Upper_90 BE_Limits Result
 1  AUCT 101.989    96.703   107.274  80-120    BE untransformed
 2  Cmax  93.634    83.252   104.016  80-120    BE untransformed


It is possible to arg the distribution RLD is non mormal to file the PK report with untransformed results?

Thanks for your help


Edit: Please don’t shout. Tabulators changed to spaces and BBcoded; see also this post #6. [Helmut]
jag009
Hero

NJ,
2017-10-20 06:33

@ DaLoe
Posting: # 17900
Views: 394
 

 Outlier PK Study – Non Normal Distribution

HI

» The thing is in my understanding it is not accepted to exlude a subject after the bioanalysis. How can we argue an outlier?

Check previous post about what to do with outliers. I don't know the current view of the regulatory agency (FDA). I have heard that they don't accept re-dosing studies anymore (rumors).

» It is possible to arg the distribution RLD is non mormal to file the PK report with untransformed results?

1) Please revisit why PK parameters (AUC, Cmax) need to be logs transformed.
2) My answer to your question is no. Others might have different views.

J
Helmut
Hero
Homepage
Vienna, Austria,
2017-10-23 10:25

@ DaLoe
Posting: # 17906
Views: 282
 

 Normal Distribution: Japan only

Hi DaLoe,

unless you submit the study in Japan and the analysis was planned based on untransformed data, no way (see the Japanese Q&A document, Q-32 to Q-34).
Note that you would have calculate the PE and its CI in the original domain (i.e., as differences) and the acceptance range for untransformed data would be –0.20xR to +0.20xR.
Please give complete information in the future. Since you are speaking about “RLD”: Is the study for FDA-submission?

» Recently we have conducted a Fasting PK study (single croosover). The statistical report is as follow:
» Dependent   Ratio_%Ref_ CI_90_Lower CI_90_Upper CV%
» Ln(Cmax)    118.34      103.66      135.11      37%
» Ln(AUClast) 103.72       97.87      109.92      15%
»
» As can be seen, the upper limit of the confidence interval do not meet PK criteria.

I’m struggling to reproduce your CVs. Was the sample size for Cmax 42 and for AUC 38? If yes, why do they differ?

» One of the subject is clearly an outlayer according Lund test. Removing the subject of the study, the results would be as follow …

Outlier tests are generally not acceptable (even if stated in the SAP).

» … resulting in meeting PK criteria, borderline, but meets criteria.

BTW, your calculation is wrong. If you are interested in the ratio of untransformed data, use Fieller’s confidence interval.

» The thing is in my understanding it is not accepted to exlude a subject after the bioanalysis.

Correct. Only exception (EMA): Exclusion if AUC of the suspected outlier <5% of the geometric mean of subjects with “normal” AUCs. Applicable to the reference only. IMHO, no way for the FDA.

» How can we argue an outlier?

Bad luck. Smells of cherry picking. :cherry picking:

» On the other hand, we have demostrate that the distribution of the RLD in case of CMax it is not normal.

» Due the distribution is not normal, neither in RLD nor in Test, but extremely not normal in RLD, …

The distributions of R and T are not relevant in BE. Only the model’s residuals are. It is possible that both R and T have skewed distributions and the residuals are still normal.

» … when we normalize the sequence …

What?

» … the results show we are not meeting PK criteria because of the upper limit of the confidence interval of the CMax.

This table is exactly like the one before. What are you trying to tell us?

» It is possible to arg the distribution RLD is non mormal to file the PK report with untransformed results?

For the FDA, no. For the EMA, very unlikely. In Japan, maybe.

[image]Regards,
Helmut Schütz 
[image]

The quality of responses received is directly proportional to the quality of the question asked. ☼
Science Quotes
Back to the forum Activity
 Thread view
Bioequivalence and Bioavailability Forum | Admin contact
17,494 Posts in 3,746 Threads, 1,086 registered users;
34 users online (0 registered, 34 guests).

The whole purpose of education is
to turn mirrors into windows.    Sydney J. Harris

The BIOEQUIVALENCE / BIOAVAILABILITY FORUM is hosted by
BEBAC Ing. Helmut Schütz
XHTML/CSS RSS Feed