NONMEM Users Network Archive

Hosted by Cognigen

RE: NONMEM 7 Update

From: Bauer, Robert <Robert.Bauer>
Date: Wed, 8 Jul 2009 17:40:11 -0400

Allow me to respond to questions 2 and 3 regarding source code legacy.

The new methods were originally written by me for S-ADAPT, so as author,
I was free to use them for other purposes. The reworking of the code to
fit NONMEM was then performed by me as a paid employee of ICON. This
code for NONMEM is therefore fully owned by ICON, and while it functions
mathematically similar to that in S-ADAPT, is now quite different in
appearance from that in S-ADAPT.

Nonetheless, S-ADAPT itself (which is not owned by ICON, and therefore
has no claim on S-ADAPT) and its source code that I wrote for it, along
with the legacy ADAPT II code that originally came from Dave D'Argenio
and Alan Schumitzky, continues to be open-source, and continues to be
forbidden from being sold for profit as is. This is in accordance with
the agreement that I arranged with Dave D'Argenio.

The FORTRAN algorithms I wrote for the SAEM method were also originally
written for S-ADAPT. They are completely based on my reading the
MONOLIX manual on Marc Lavielle's description of the kernel density
setups, and the general literature on Bayesian analysis. I never read
the MATLAB source code of Monolix to implement these algorithms.

Robert J. Bauer, Ph.D.
Vice President, Pharmacometrics
ICON Development Solutions
Tel: (215) 616-6428
Mob: (925) 286-0769
Email: Robert.Bauer

-----Original Message-----
From: owner-nmusers
On Behalf Of Nick Holford
Sent: Saturday, July 04, 2009 2:01 PM
To: nmusers
Subject: Re: [NMusers] NONMEM 7 Update


I think some interesting issues have been raised about Icon's plan to
distribute encrypted source code. I would like to ask what does this
really hide
and how can it benefit Icon and the user community?

First -- what is being hidden?

1. The 'old' estimation methods (METHOD=ZERO, METHOD=CONDITIONAL) =
their various options such as LAPLACE were revealed to previous
of NONMEM. As far as I can tell these 'old' methods, in NONMEM7,
very similarly to earlier versions but have been just been tweaked with
changes in convergence options (NSIG, SIGL).

2. The 'new' ITS and MCPEM methods presumably came from the S-ADAPT
(developed by Bob Bauer) whose licensing agreement says "No part of the
may be repackaged into a non open-source version, and no part of the
may be sold for profit.". It remains to be seen how these new methods
will help
users but I am hopeful that incorporating these methods will give us new

opportunities to improve the science.

3. I cannot be sure about the origins of the NONMEM SAEM method in
Monolix but the SAEM code in Monolix is distributed under a Free
license agreement and it seems likely that this has been helpful for the
implentation. I am hopeful that the pioneering efforts of the Monolix
team will
benefit NONMEM users.

Thus the core methods of NONMEM, both old and new, are already available
to others who may want to compete with Icon so I wonder what the
encrypted source will really hide?

It is clear that the IMSL source code cannot be distributed to users
because that is the contract that Icon have with IMSL. Would it be
possible to distribute the licensed NONMEM source code with object code
libraries for the IMSL routines?

Serge asked " How many people took advantage from the fact Nonmem was
distributed as open source?". Under the earlier NONMEM license I was
able to
make modifications to the code to fix minor problems with compilers,
the output formatting and compute interesting things like the
determinant of the
information matrix. These changes could be made without understanding
the whole
NONMEM system and no doubt others like me without formal training in
fortran or
statistical software development were also able to personalise things.

After the major recoding effort of NONMEM into a modern fortran dialect
(thanks to Icon) it is likely that the code will no longer be
"functionally encrypted" and indeed might be usefully extended by others
developing new methods in this area. So even if only a few people have
been able to take advantage of the source code for NONMEM in the past it
is more likely that in the future more people will be able to contribute
to its development. Furthermore, inability to access the source code
will also
make it harder for 3rd party developers to help users (e.g. NMQUAL) to
fix known
bugs and to create qualified installations using different SIZES.

Second -- How does it aid Icon?
Hiding the source code might reduce the potential user base and thus
fees for Icon because statistical algorithm development has frequently
had its
roots outside of proprietary implementations. As Thierry pointed out
S-Plus with R) isn't there a risk that hiding NONMEM will encourage
others to
develop an competing alternative?

Once again I want to congratulate Bob, Tom and Alison for their efforts
to modernize NONMEM and prepare us for a new and more powerful tool. But
hiding the source code seems against the community spirit of nmusers who
freely of their experiences and expertise and the advantages, if any, to
remain obscure.

Best wishes,


Nick Holford, Professor Clinical Pharmacology
Dept Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New
mobile: +33 64 271-6369 (Apr 6-Jul 20 2009)

ICON plc made the following annotations.
This e-mail transmission may contain confidential or legally privileged i=
that is intended only for the individual or entity named in the e-mail ad=
dress. If you
are not the intended recipient, you are hereby notified that any disclosu=
re, copying,
distribution, or reliance upon the contents of this e-mail is strictly pr=
ohibited. If
you have received this e-mail transmission in error, please reply to the =
sender, so that
ICON plc can arrange for proper delivery, and then please delete the mess=
Thank You,
ICON plc
South County Business Park
Dublin 18
Registered number: 145835

Received on Wed Jul 08 2009 - 17:40:11 EDT

The NONMEM Users Network is maintained by ICON plc. Requests to subscribe to the network should be sent to:

Once subscribed, you may contribute to the discussion by emailing: