Ken Peh
★    

Malaysia,
2014-07-20 21:01
(3539 d 02:48 ago)

Posting: # 13276
Views: 7,381
 

 Revalidate the study [Study As­sess­ment]

Dear Members,

We ran a BE study on 24 subjects. Unfortunately, on the day of study, two subjects dropped out and 7 subjects worked night shift and could only reported to the trial site after midnight. We were told that the 7 subjects could not be considered as these subjects could not be proven to have fasted. The study passed with 15 subjects and CVintrasubject of AUC and Cmax were below 10%.

We were asked to revalidate the study since the power of study should be 24 subjects. How do we revalidate the study ? Can we use the posthoc CV to show that the power is more than 80% even with 15 subjects ?

Highly appreciate your comments.

Thank you.

Regards,
Ken


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

Denmark,
2014-07-20 22:04
(3539 d 01:44 ago)

(edited by ElMaestro on 2014-07-20 23:08)
@ Ken Peh
Posting: # 13277
Views: 7,073
 

 Revalidate the study

Hi Ken Peh,

❝ We ran a BE study on 24 subjects. Unfortunately, on the day of study, two subjects dropped out and 7 subjects worked night shift and could only reported to the trial site after midnight. We were told that the 7 subjects could not be considered as these subjects could not be proven to have fasted. The study passed with 15 subjects and CVintrasubject of AUC and Cmax were below 10%.


Being lucky is not a crime.

❝ We were asked to revalidate the study since the power of study should be 24 subjects. How do we revalidate the study ? Can we use the posthoc CV to show that the power is more than 80% even with 15 subjects ?


I do not know what it means to revalidate a study. Probably someone has misunderstood something.
24 subjects dosed, 15 evaluated of which 7 are likely due to bad planning on your part. Fix it before doing further trials. Make sure to discuss night shifts etc with volunteers. You really do not wish to enroll people who are not available (willing to make themselves available) due to mundane issues such as work habits.
Anyways, the study passed the product is declared BE even with 15 volunteers. Well done. If anyone asks about power, they are wrong. But they could very justifiedly ask you to embark on preventive actions.

Consider: You roll a dice 9 times and get a total of 43 eyes. Afterwards you ask questions. What's the chance of this outcome? How can you get such an outcome when the dice is fair, can you revalidate the dice?

Pass or fail!
ElMaestro
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2014-07-21 00:33
(3538 d 23:16 ago)

@ Ken Peh
Posting: # 13278
Views: 7,910
 

 α ≠ β

Hi Ken,

I agree with ElMaestro.

❝ We were told […]


❝ We were asked […]



By whom? By the sponsor? By a regulator? I give some points below, but if I would know who asked these questions I could give you a more detailed answer.

❝ […] revalidate the study since the power of study should be 24 subjects.


Was this term used in English or is it a translation? Sounds like plain nonsense to me. You planned the study according to information / assumptions for 24 subjects. If all (‼) assumptions turn out to be exactly (‼) correct and you don’t have a single (‼) drop-out then you have a chance of passing BE, which is exactly the power you aimed at. Below possible other combinations:

assumed in     observed in    producer’s   power
study design   actual study   risk (β)     (1–β)
────────────────────────────────────────────────
T/R ratio      closer to 1        ⇓          ⇑  
               away from 1        ⇑          ⇓  
CV             lower              ⇓          ⇑  
               higher             ⇑          ⇓  
# drop-outs    lower              ⇓          ⇑  
               higher             ⇑          ⇓  


The patient’s risk (α = probability of the type I error) – this is the only one regulators should care about! – is ≤5% in any case – independent from power (1–β, where β = probability of the type II error, aka producer’s risk).

❝ How do we revalidate the study ?


Forget it. The study passed BE. Go and have a party!
In planning the next study consider using more conservative assumptions (especially about the drop-out rate).

❝ CVintrasubject of AUC and Cmax were below 10%. […] Can we use the posthoc CV to show that the power is more than 80% even with 15 subjects ?


If you look at power-curves you will see that the CV is more important than the sample size. It might well be that the CV (lower than the one assumed in study planning) increased power more than what you “lost” due to the high drop-out rate. But:
Since <80% “post hoc power” does not invalidate a study which demonstrated BE – the opposite is true as well: 80% “post hoc power” does not further support a study which already showed BE.
BE is BE in any case. Full stop.*

If the questions (like many similar ones related to power in the past) came from a Malaysian regulator, tell the group that I’ll be happy to perform some basic statistical training for them. No offense, but these same old stories suck.


  • A reminder for whoever asked these questions: BE is defined as the inclusion of a confidence interval within an acceptance range. Power does not show up in this definition and is not mentioned in any guideline worldwide.
    Homework: 5,000–10,000 BE-studies are performed worldwide each year. How many of them pass BE with power ~60% and are still accepted by regulators?

Edit: See also this post.

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
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,636 registered users;
83 visitors (0 registered, 83 guests [including 2 identified bots]).
Forum time: 22:49 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