[R] "R is not a validated software package.."

[Bert Gunter]

Frank et. al:

I believe this is a bit too facile. 21 CFR Part 11 does necessitate a
software validation **process** -- but this process does not require any
particular software. Rather, it requires that those using whatever software
demonstrate to the FDA's satisfaction that the software does what it's
supposed to do appropriately. This includes a lot more than assuring, say,
the numerical accuracy of computations; I think it also requires
demonstration that the data are "secure," that it is properly transferred
from one source to another, etc. I assume that the statistical validation of
R would be relatively simple, as R already has an extensive test suite, and
it would simply be a matter of providing that test suite info. A bit more
might be required, but I don't think it's such a big deal. 

I think Wensui Liu's characterization of clinical statisticians as having a
mentality "related to job security" is a canard. Although I work in
nonclinical, my observation is that clinical statistics is complex and
difficult, not only because of many challenging statistical issues, but also
because of the labyrinthian complexities of the regulated and extremely
costly environment in which they work. It is certainly a job that I could
not do.

That said, probably the greatest obstacle to change from SAS is neither
obstinacy nor ignorance, but rather inertia: pharmaceutical companies have
over the decades made a huge investment in SAS infrastructure to support the
collection, organization, analysis, and submission of data for clinical
trials. To convert this to anything else would be a herculean task involving
huge expense, risk, and resources. R, S-Plus (and much else -- e.g. numerous
"unvalidated" data mining software packages) are routinely used by clinical
statisticians to better understand their data and for "exploratory" analyses
that are used to supplement official analyses (e.g. for trying to justify
collection of tissue samples or a pivotal study in a patient subpopulation).
But it is difficult for me to see how one could make a business case to
change clinical trial analysis software infrastructure from SAS to S-Plus,
SPSS, or anything else.

**DISCLAINMER** 
My opinions only. They do not in any way represent the view of my company or
its employees.


Bert Gunter
Genentech Nonclinical Statistics
South San Francisco, CA 94404
650-467-7374


-----Original Message-----
From: r-help-bounces at stat.math.ethz.ch
[mailto:r-help-bounces at stat.math.ethz.ch] On Behalf Of Frank E Harrell Jr
Sent: Friday, June 08, 2007 7:45 AM
To: Giovanni Parrinello
Cc: r-help at stat.math.ethz.ch
Subject: Re: [R] "R is not a validated software package.."

Giovanni Parrinello wrote:
> Dear All,
> discussing with a statistician of a pharmaceutical company I received 
> this answer about the statistical package that I have planned to use:
> 
> As R is not a validated software package, we would like to ask if it 
> would rather be possible for you to use SAS, SPSS or another approved 
> statistical software system.
> 
> Could someone suggest me a 'polite' answer?
> TIA
> Giovanni
> 

Search the archives and you'll find a LOT of responses.

Briefly, in my view there are no requirements, just some pharma 
companies that think there are.  FDA is required to accepted all 
submissions, and they get some where only Excel was used, or Minitab, 
and lots more.  There is a session on this at the upcoming R 
International Users Meeting in Iowa in August.  The session will include 
dicussions of federal regulation compliance for R, for those users who 
feel that such compliance is actually needed.

Frank

-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University

______________________________________________
R-help at stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.