Best Practices for Creating Population PK Data Sets

I have posted the best practices for creating submission ready pharmacometric data sets at the PKPD programming forum. I invite everyone to share your thoughts and experience on this topic.

  1. Best Practice User Guide for Esubmission of NONMEM data sets.
    Best Practice User Guide for NONMEM Data Set Esubmission.docx (23.1 KB)

  2. Standard Imputation Rules and Best Practices for Creating Population PK Data Sets.
    Best Practices for Creating Pharmacometric Analysis Data Sets

For more information on PKPD Programming Forum refer to the link

Neelima Thanneer, Bristol-Myers Squibb


Hi Neelima,

Thank you very much for sharing.

I also have some suggestions for generating NONMEM adpm (analysis data set for pharmacometrics analysis) data set from SDTM dm, ex and pc and their supplimental domains as below. I call it adpm instead of pm for the reason of item 3.

  1. The USUBJID in the SDTM is usually character type. So we need to create a numeric ID by coverting the USUBJID.
  2. The FLAG variable nomenclature can follow the ADaM ANLzzFL nomenclature. “zz” is 01, 02, 03, etc. Thus we can use ANL01FL, ANL02FL, and ANL03FL colume to apply different flags.
  3. Because pm data is not raw data, it is analysis-ready data. So I would not include it in the SDTM data sets, and I would prefer to include it in the ADaM data sets. Thus I call it adpm instead of pm. For the same reason, I don’t think it need to be finalized before data-lock or apply the data-lock for it.

If we use WinNonlin NLME to do the pharmacometric analysis, instead of NONMEM, the dosing date/time data set can be separated from the concentration data set.

Hope it helps.



Hi Michelle,
Thanks for sharing.
Agree with #1.
Regarding #2 given the issues with population PK data sets it would be easy to understand and summarize if all the flags for exclusion are captured in one variable. If there are 20 different flags then going by ANLzzFL convention there should be 20 variables.

Are you creating pharmacometric analysis data set in 100% ADaM format? If so are you using BDS structure? I will be interested to know if you have implemented BDS structure on population PK data sets.

I am working with “Pharmacometric Data Standards” workstream with a goal to come up with standards for POPPK data sets. The intent is to make them complaint with CDISC standards. So, if you are already submitting POPPK data sets in ADaM format it would be valuable to know the challenges and how you resolved them.


Hi Neelima,

Thank you for your feedback. That’s is very useful. I’ve not done any POPPK data sets submission so far. Will wait for other experts to respond this.


Hi all,
Rebooting this thread in case any updates have occurred:

I’ve currently been tasked with ensuring that my employer’s PK vendor appropriately creates PopPK datasets for submission to FDA/EMEA. I’ve been scouring the web for any recent opinions from regulatory authorities on the matter, and came across your great group and library (thanks!).
Has anyone seen/heard any information recently regarding compliant format of PopPK datasets?
To be specific, has anything come from CDISC after the meeting mentioned in this post:


@justinwilkins and @andrijana may be able to provide you with more information regarding the current state of pop pk dat astandards. They are working with a group to develop the type of standards you describe.

1 Like