VSL
☆    

India,
2015-04-20 23:46
(3264 d 21:15 ago)

Posting: # 14721
Views: 12,865
 

 Study Results [Study As­sess­ment]

Dear All,
Please explain what to do if following results obtained in BE-Study:
  1. Study is passed in CI but mean T/R ratio is not within the 0.80-1.25.
  2. Study is passed in CI but there is significant p-value for treatment/period/sequence
Shall we submit the study to regulatory agencies?
Regards...
VSL


Edit: Category changed. [Helmut]
ElMaestro
★★★

Denmark,
2015-04-21 01:47
(3264 d 19:14 ago)

@ VSL
Posting: # 14722
Views: 11,456
 

 Study Results

Hi VSL,

❝ Please explain what to do if following results obtained in BE-Study:

❝ 1. Study is passed in CI but mean T/R ratio is not within the 0.80-1.25.


This does not correspond to the world I know. Can you show the example?
If I am not mistaken HRotter or some other guru once mentioned on this forum that a PE outside the CI is theoretically possible; from the equation for the 90% BE CI -which I hope is generally accepted- I think it is evident that the PE is not outside the CI.

❝ 2. Study is passed in CI but there is significant p-value for treatment/period/sequence


The term is nuisance effect. This forum has discusssed it many times, try and search for it. Basically, it should not be too much of a worry. Those p-values correspond to a null hypothesis you are not really interested in when testing for BE.
Pre-dose levels of the analyte is, however, more tangible ground for concern.

Pass or fail!
ElMaestro
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2015-04-21 02:39
(3264 d 18:22 ago)

@ ElMaestro
Posting: # 14724
Views: 11,449
 

 upper CL < PE

Hi ElMaestro,

❝ If I am not mistaken HRotter or some other guru once mentioned on this forum that a PE outside the CI is theoretically possible;


Thomas Jaki. When he tried to explain that to me I got mental hickups. But was other stuff: The one-sided CI of a CV with an α >0.5 – which is nuts.

library(PowerTOST)
CV    <- 0.2
df    <- 22
alpha <- 0.55
CL    <- CVCL(CV=CV, df=df, side="upper", alpha=alpha)
cat("CV      :", CV, "\nCL of CV:", CL[1], "to", CL[2], "\n")

# CV      : 0.2
# CL of CV: 0 to 0.1992335


Implemented according to Chow/Liu… Beyond me.

❝ from the equation for the 90% BE CI -which I hope is generally accepted- I think it is evident that the PE is not outside the CI.


Rightyright. Not possible.

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2015-04-21 02:24
(3264 d 18:37 ago)

@ VSL
Posting: # 14723
Views: 11,233
 

 PE restriction for HVDs/HVDPs; testing effects

Hi VSL,

❝ 1. Study is passed in CI but mean T/R ratio is not within the 0.80-1.25.


ElMaestro’s answer is correct – if the acceptance range (AR) is 0.8–1.25 (unscaled ABE). If you applied reference-scaling (in FDA’s or EMA’s flavor) such a result is possible indeed, especially if:
  • CVWR ≫ CVWT.
  • Large “true” deviation of T from R.
  • … or both.
The modified acceptance range depends only on the variability of the reference. If the test’s variability is comparatively small, the CI might be relatively narrow. That would mean a formulation could pass with a GMR far away from 1 (even <0.8 or >1.25). Statistically that would still make sense, but for political reasons the restriction of the PE of 0.8–1.25 was introduced. Actually for CVWR ≥50% the scaled AR (and the CI inclusion rule) is of little importance at all. It is mostly the PE which counts. Tons of papers.

In the future please give complete information (acceptance range, CI, CVs, regulatory agency).

❝ 2. Study is passed in CI but there is significant p-value for treatment/period/sequence

❝ Shall we submit the study to regulatory agencies?


Ahem – did your read the guidelines?

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
jag009
★★★

NJ,
2015-04-21 18:49
(3264 d 02:11 ago)

@ Helmut
Posting: # 14728
Views: 11,155
 

 PE restriction for HVDs/HVDPs; testing effects

?

You guys are messing with me... If the T/R Ratio is <0.8 or >1.25, how can you generate a 90 CI that is within 80.00 – 125.00 since the 90% CI consists of (PE ± values)?

John
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2015-04-21 19:12
(3264 d 01:49 ago)

@ jag009
Posting: # 14729
Views: 11,312
 

 PE restriction for HVDs/HVDPs; testing effects

Hi John,

❝ You guys are messing with me... If the T/R Ratio is <0.8 or >1.25, how can you generate a 90 CI that is within 80.00 – 125.00 since the 90% CI consists of (PE ± values)?


If we are dealing with unscaled ABE, ElMaestro’s answer (and your doubt) is correct: No way.

VSL’s post is ambiguous:

❝ Study is passed in CI but mean T/R ratio is not within the 0.80-1.25.


If the study was reference-scaled, the (implied) BE limits may be much wider (my answer).
Example in FDA’s flavor: CVWR 60%, implied limits 60.96–164.04%, CI 97.00–162.00%; would pass (CI within limits), but fail due to the PE of 125.36 >125.00%.

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
jag009
★★★

NJ,
2015-04-22 17:47
(3263 d 03:13 ago)

@ Helmut
Posting: # 14732
Views: 10,941
 

 PE restriction for HVDs/HVDPs; testing effects

Hi Helmut,

❝ VSL’s post is ambiguous:


And guess what, he still hasn't replied :-|

But if he was talking about ref scaled, then why would he bring up 90 CI? Strange, actually scary...

John
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2015-04-22 18:41
(3263 d 02:20 ago)

@ jag009
Posting: # 14733
Views: 11,050
 

 PE restriction for HVDs/HVDPs

Hi John,

❝ ❝ VSL’s post is ambiguous:


❝ And guess what, he still hasn't replied :-|


❝ But if he was talking about ref scaled, then why would he bring up 90 CI? Strange, actually scary...


Duno. That’s why I asked for more information. He/she didn’t mention a 90% CI, only told us

❝ Study is passed in CI


Maybe he/she meant the upper 95% CL of FDA’s approach (≤0)? BTW, since for the EMA scaling is capped at CVWR 50% passing with a 90% CI and failing on the PE is very, very unlikely.

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
Averroes
☆    

Spain,
2017-05-19 13:53
(2505 d 07:07 ago)

@ Helmut
Posting: # 17372
Views: 6,268
 

 GMR outside the limits when using ABEL

Dear Helmut,

Recently we performed a full replicate (4-period) study using ABEL for EMA and actually the results shows AUC within 80-125, Cmax within expanded CI but unfortunately GMR fall outside 80-125.

I found in the forum that the GMR constrant is there for political reasons more than for scienfic or statistical reasons. Do you think we could still have an opportunity with European regulatory agencies?

Thanks,
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2017-05-19 16:07
(2505 d 04:54 ago)

@ Averroes
Posting: # 17373
Views: 6,361
 

 Regulations ≠ Science

Hi Averroes,

❝ […] the GMR constrant is there for political reasons more than for scienfic or statistical reasons.


Exactly. Suggested by Les Benet.

❝ Do you think we could still have an opportunity with European regulatory agencies?


I’m always happy to challenge regulatory practices. But in this case the chances to succeed are close to nil. IMHO, even trying is a complete waste of time. Alfredo will never, ever buy it.

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
Averroes
☆    

Spain,
2017-05-22 22:51
(2501 d 22:10 ago)

@ Helmut
Posting: # 17389
Views: 6,131
 

 Regulations ≠ Science

❝ I’m always happy to challenge regulatory practices. But in this case the chances to succeed are close to nil. IMHO, even trying is a complete waste of time. Alfredo will never, ever buy it.


Thanks Helmut!!
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,636 registered users;
80 visitors (0 registered, 80 guests [including 13 identified bots]).
Forum time: 20:01 CET (Europe/Vienna)

Nothing shows a lack of mathematical education more
than an overly precise calculation.    Carl Friedrich Gauß

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5