ITT, FAS and PP populations

[Jeewaka Mendis]

Hi

We are running a clinical trial and I wonder what analysis populations to use in the analysis.

According to FDA guidance, I understand ITT (Intention to Treat) is a principle which determines how to define FAS population.

What I think is only a FAS (Full analysis set) and PP (Per Protocol) population is adequate.

but my colleague thinks we need 3 populations, i.e. ITT , FAS and PP defined.

Please clarify.

Thanks in advance

Jeewaka

[John Whittington]

At 20:03 10/09/2017, Jeewaka Mendis wrote:

According to FDA guidance, I understand ITT (Intention to Treat) is a principle which determines how to define FAS population.

What I think is only a FAS (Full analysis set) and PP (Per Protocol) population is adequate.
but my colleague thinks we need 3 populations, i.e. ITT , FAS and PP defined.

My understanding that "FAS" is what many of us would call "modified ITT" (mITT), about which some people are very scornful (because it is not true ITT).

The FDA say ...
"The FAS population consists of all subjects randomly assigned to treatment who received at least one dose of trial medication and have both Baseline and at least one postbaseline Y-MRS assessment."

... whilst ICH-E9 says:
 "In this document the term 'full analysis set' is used to describe the analysis set which is as complete as possible and as close as possible to the intention-to-treat ideal of including all randomised subjects.”
... but then goes on to clarify this by saying (amongst other things) ...
"There are a limited number of circumstances that might lead to excluding randomised subjects from the full analysis set including the failure to satisfy major entry criteria (eligibility violations), the failure to take at least one dose of trial medication and the lack of any data post randomisation."

Hence, FAS is not the same as true ITT - so if their is a requirement for a true ITT analysis, use of a FAS would not satisfy that requirement.

One 'problem' with the definitions of both ITT and FAS is that they refer to 'randomised patients' and therefore cannot, strictly speaking, be applied to non-randomised trials/studies.  It is easy enough to apply essentially the same "ITT" principles, but, in the absence of randomisation, it is necessary to define the point/time at which subjects become part of that population.

Kind Regards,

John

----------------------------------------------------------------

Dr John Whittington,       Voice:    +44 (0) 1296 730225

Mediscience Services       Fax:      +44 (0) 1296 738893

Twyford Manor, Twyford,    E-mail:   Joh...@mediscience.co.uk

Buckingham  MK18 4EL, UK            

----------------------------------------------------------------

[Martin Bland]

The principle of analysis according to intention to treat is that the randomised groups are comparable,  i.e. samples from the same population, and any selection post randomisation makes the groups potentially not comparable, i.e. they may no longer be samples from the same population.

From this viewpoint, both "full analysis set" and "per protocol" samples are not comparable, if any randomised participants are not included.  

The first problem with intention to treat is that we need data on all participants, very difficult  to achieve in medical trials.  We can allow for missing data using methods such as multiple imputation. but these rely on uncheckable assumptions about the data, such as that all missing observations are missing at random.  The second is that departures from the randomised treatment may lead to an underestimate of the treatment effect.  The latter is the argument for FAS and PP approaches.

Personally, I would not use the FAS or PP approaches as being fundamentally flawed, since selection may have made the samples non-comparable.  John and I have disagreed about this in the past.  But you may find it difficult to convince your collaborators of this purist position.

Martin

--
--
To post a new thread to MedStats, send email to MedS...@googlegroups.com .
MedStats' home page is http://groups.google.com/group/MedStats .
Rules: http://groups.google.com/group/MedStats/web/medstats-rules

---
You received this message because you are subscribed to the Google Groups "MedStats" group.
To unsubscribe from this group and stop receiving emails from it, send an email to medstats+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

***************************************************
J. Martin Bland
Prof. of Health Statistics Emeritus
Dept. of Health Sciences
Seebohm Rowntree Building
University of York
Heslington
York YO10 5DD

Email: martin...@york.ac.uk
Phone: 01904 321334     Fax: 01904 321382
Web site: http://martinbland.co.uk/

Statement by the University of York:
This email and its attachments may be confidential and are intended solely for the use of the intended recipient. If you are not the intended recipient of this email and its attachments, you must take no action based upon them, nor must you copy or show them to anyone. Please contact the sender if you believe you have received this email in error. Any views or opinions expressed are solely those of the author and do not necessarily represent those of The University of York.
***************************************************