Question about sparse sampling

[js3110]

I have a quick question regarding sparse NCA calculations with PKNCA:

Lets say I run an NCA using the following grouping for conc and dose respectively:

PKNCAconc: conc ~time | matrix + drug + usubjid / analyte
PKNCAdose: dose ~time| drug + usubjid
and I select sparse = TRUE.

How will the sparse sampling be grouped? Will it combine all samples by usubjid, or group them by analyte, or drug treatment, or matrix, analyte, and drug treatment? The vignette does not describe how sparse sampling works in more complex data scenarios.