ISoP ISoP on Twitter | ACoP ACoP on Twitter | PAGE | WCoP | PAGANZ | PAGJa

What are the main differences between PK NCA data and NONMEM PopPK data?


I need to summarize the differences between PK NCA data and NONMEM PopPK data. Any input will be appreciated. Thanks.
The differences that I’m aware of:

  1.        Typically NCA data only requires time since last dose and time since first dose is nice to have for data checking. NONMEM PK data requires time since first dose for modeling, and sometime may use time since last dose based on assumptions regarding steady state.
  2.        Phase II/III data: usually NCA is not needed unless there are intensive PK samples. All PK samples including sparse PK can be included in NONMEM PK data. 
  3.        NCA analysis itself doesn’t need covariates. Covariates are needed only for summary tables & figures. In  NONMEM PK data, it may include more covariates than the ones needed for PK summary tables & figures. 
  4.        Usually urine data is included for NCA, but not for PopPK modeling. 
  5.        NONMEM PopPK data has to meet the NONMEM data requirements including dosing records and reserved columns.


Jing - can you please provide some context for asking this question. The points that you made look like this information will be used by data management teams or for educational purposes, is that correct?


Hi Jing:

Welcome to the discussion forum!

To your questions above

  1. A normal single dose NCA dataset would have the time characterized from the last dose (time 0), or time since the start of dosing for multiple dose analyses. NONMEM (or perhaps pop pk is a better way to say it as other population pk tools exist outside of NONMEM, such as Monolix ) may require more complete dosing information, depending on the type of data used. In both cases (NCA and pop pk) assumptions about time to steady state are the same (usually).
  2. Normally we handle Phase II/III data with a population approach as the samples are often sparse and or taken at different visits.
  3. NCA does use covariates, (especially if one considers dose as a covariate). In some cases, folks even do metanalysis across NCA parameter estimates from various studies in a program to look at impact of age, gender, much like they would for a population pk approach.
  4. While urine can certainly be included in an NCA dataset, I would not say “normally” in my experience. Later stage Clin Pharm trials may not collect urine at all. Also, urine data can be included with population datasets and even used simultaneously in the model estimations. In both cases, it depends on the questions being asked by the analysis.
  5. Population datasets, like NCA, require careful data preparation, and the field names normally are specified within the code, so they need to match in the dataset for analysis to be successful.


Vijay - a better understanding of the differences will help standardize NONMEM PopPK data and may help streamline the workflow as well. Thanks.


Hi Brian,
Thanks very much for your input!
Regarding 3), by NCA data, I meant the input data for NCA analysis. In your response, ‘NCA parameter estimates’ means PK parameters estimated from NCA like AUC, Cmax, etc., right?
Thanks again.


I think Brian did a very good job of summarizing some of the differences between the two approaches. You can think of NCA (most often used in early phase) as an approach used when we know little about the drug and want to generate some basic quick descriptors of the nature of the kinetics of the drug.

In a very basic sense all I need to perform NCA is dose amount, time from dose, and concentration of a serial set of samples. So an NCA dataset can be pretty basic. As Brian pointed out we often use patient characteristics to at the very least summarize the basic parameters we calculate, so the dataset can grow in complexity as the needs do. The results from NCA are dose event based (i.e., the resultant parameters from a single dose event or a steady state dose event are single point representations) so linking that AUC/exposure estimate to some sort of continuous time based covariate is difficult and usually not a requirement.

That does not mean that we could not want a dataset with time based covariate type linkages, because an underlying overall project goal might be to combine several single study datasets to generate a meta study early population based analysis.

Brian also mentions the use of urine data in NCA, we may use total amount excreted (derived from the urine data) in a population based dataset, but it would be a rare compound and a rare need where we would feel it necessary to try and model urine data with population methodology.

I hear the underlying message behind your post, and I infer that to mean, “do I have the ability to push back on what I perceive as requests for an overly complex dataset that is supposedly only going to be used for NCA analysis” The answer would depend on what the scientist wanted to do with the data overall. Maybe NCA is the primary goal, but off the main developmental chain, the request for the more complex information is intended to support early phase model development which then would necessitate a more complex dataset. I think it is completely fair to have the conversation with the scientist to determine if the level of complexity is warranted.

Finally, Brian is correct in his mention that the dataset constraints are much more rigid for the population based analysis programs than they are for the NCA programs. So while standards can come into play for both, there will be more extensive sets of “rules” for a population based dataset than an NCA one.

Is that where you were wanting to go with the discussion or did I miss the mark?



Hi Jing,
I agree with Brian and DBRadtke.
Some extra tips: To help standardize NONMEM PopPK data, normally for NONMEM popPK dataset, it combines the dm, ex, pc, lb and their supplementary domains, etc. NCA dataset will come from same domains, but if you would like to analyze PK in WinNonlin, then dosing information could be a single dataset that doesn’t need to be append to the concentration dataset.


Thanks very much Dave for your response! I appreciate the nice summary of the NCA data from very basic to complex scenario. Also it is nice to know urine data may be needed in a population based dataset even though it is uncommon. Thanks again.


Thanks Hongxia for the extra tips!