[R] Use of R in clinical trials

Douglas Bates bates at stat.wisc.edu
Thu Feb 18 20:14:12 CET 2010

On Thu, Feb 18, 2010 at 12:36 PM, Bert Gunter <gunter.berton at gene.com> wrote:
> The key dates are 1938 and 1962. The FDC act of 1938 essentially mandated
> (demonstration of) safety. The tox testing infrastructure grew from that.At
> that time, there were no computers, little data, little statistics
> methodology. Statistics played little role -- as is still mainly the case
> today for safety. Any safety findings whatever in safety testing raise a
> flag; statistical significance in the multiple testing framework is
> irrelevant.

> 1962 saw the Kefauver-Harris Amendments that mandated demonstration of
> efficacy. That was the key. The whole clinical trial framework and the
> relevant statistical design and analysis infrastructure flowed from that
> regulatory requirement. SAS's development soon after was therefore the first
> direct response to the statistical software needs that resulted. Note also,
> that statistical software was in its infancy at this time: before SAS there
> was Fortran and COBOL; there was no statistical software.

> So, as you can see, there essentially was **no** "before SAS".

> (Corrections/additional information welcome!)

My recollection is that the BMD programs (which, in a later version,
became BMDP) predated SAS and were specifically for BioMeDical
analysis.  Early statistical software was oriented to applications
areas: SPSS (Statistical Package for the Social Sciences) was the
predominant system used in the social sciences, BMD(P) in biomedical
areas and SAS in agricultural/life sciences settings.  Eventually the
more coherent framework and comparative ease-of-use of SAS (yes, I am
saying that with a straight face - in the days of batch jobs submitted
on punched cards with data residing on magnetic tape, there were
different standards of ease-of-use) won over more users in medical

> Bert Gunter
> Genentech Nonclinical Biostatistics
> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On
> Behalf Of Christopher W. Ryan
> Sent: Thursday, February 18, 2010 10:09 AM
> To: r-help at r-project.org
> Cc: p.dalgaard at biostat.ku.dk
> Subject: Re: [R] Use of R in clinical trials
> Pure Food and Drug Act: 1906
> FDA: 1930s
> founding of SAS: early 1970s
> (from the history websites of SAS and FDA)
> What did pharmaceutical companies use for data analysis before there was
> SAS? And was there much angst over the change to SAS from whatever was
> in use before?
> Or was there not such emphasis on and need for thorough data analysis
> back then?
> --Chris
> Christopher W. Ryan, MD
> SUNY Upstate Medical University Clinical Campus at Binghamton
> 425 Robinson Street, Binghamton, NY  13904
> cryanatbinghamtondotedu
> "If you want to build a ship, don't drum up the men to gather wood,
> divide the work and give orders. Instead, teach them to yearn for the
> vast and endless sea."  [Antoine de St. Exupery]
> Bert Gunter wrote:
>> DISCLAIMER: This represents my personal view and in no way reflects that
> of
>> my company.
>> Warning: This is a long harangue that contains no useful information on R.
>> May be wise to delete without reading.
>> ----------
>> Sorry folks, I still don't understand your comments. As Cody's original
> post
>> pointed out, there are a host of factors other than ease of
> programmability
>> or even quality of results that weigh against any change. To reiterate,
> all
>> companies have a huge infrastructure of **validated SAS code** that would
>> have to be replaced. This, in itself, would take years and cost tens of
>> millions of dollars at least. Also to reiterate, it's not only
>> statistical/reporting functionality but even more the integration into the
>> existing clinical database systems that would have to be rewritten **and
>> validated**. All this would have to be done while continuing full steam on
>> existing submissions. It is therefore not surprising to me that no pharma
>> company in its right mind even contemplates undertaking such an effort.
>> To put these things into perspective. Let's say Pfizer has 200 SAS
>> programmers (it's probably more, as they are a large Pharma, but I dunno).
>> If each programmer costs, conservatively, $200K U.S. per year fully
> loaded,
>> that's $40 million U.S. for SAS Programmers. And this is probably a severe
>> underestimate. So the $14M quoted below is chicken feed -- it doesn't even
>> make the radar.
>> To add further perspective, a single (large) pivotal clinical trial can
>> easily cost $250M . A delay in approval due to fooling around trying to
>> shift to a whole new software system could easily cause hundreds of
> million
>> to billions if it means a competitor gets to the marketplace first. So, to
>> repeat, SAS costs are chicken feed.
>> Yes, I suppose that the present system institutionalizes mediocrity. How
>> could it be otherwise in any such large scale enterprise? Continuity,
>> reliability, and robustness are all orders of magnitude more important for
>> both the FDA and Pharma to get safe and efficacious drugs to the public.
>> Constantly hopping onto the latest and greatest "craze" (yes, I exaggerate
>> here!) would be dangerous, unacceptable, and would probably delay drug
>> approvals. I consider this another example of the Kuhnsian paradigm
> (Thomas
>> Kuhn: "The Structure of Scientific Revolutions")in action.
>> This is **not** to say that there is not a useful role for R (or STATA or
>> ...) to play in clinical trial submissions or, more generally, in drug
>> research and development. There certainly is. For the record, I use R
>> exclusively in my (nonclinical statistics) work. Nor is to say that all
>> change must be avoided. That would be downright dangerous. But let's
> please
>> keep these issues in perspective. One's enthusiasm for R's manifold
> virtues
>> should not replace common sense and logic. That, too, would be
> unfortunate.
>> Since I've freely blustered, I am now a fair target. So I welcome forceful
>> rebuttals and criticisms and, as I've said what I wanted to, I will not
>> respond. You have the last word.
>> Bert Gunter
>> Genentech Nonclinical Biostatistics
> ______________________________________________
> R-help at r-project.org 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.
> ______________________________________________
> R-help at r-project.org 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.

More information about the R-help mailing list