OmegaStderr and thetacorrelation and covariance output are coming up empty on one of my drugs

Phoenix
1

Hello,

Hope you’re all doing great!

I am running PopPK models for two analytes, with multi dose PK. Drug one is a long half-life drug (20-30 days) and multiple absorption peaks (potential enterohepatic recirculation), and one very short half-life (~8hours). Both are dosed for 3 days (q24), and we have both sparse (n~70) and intensive (n=20) PK sampling.

When I run drug 1, I am not getting any standard error or covariance data on most of my output. Many tables are just empty, and the corresponding graphs just say "not enough data). I thought this was just WNL not providing the output, but I was surprised when I ran the drug 2, which has fewer datapoints, I got most of the tables and data output with standard errors. The data tables have a similar outline, with the same dosing times, schedules, and elapsed times. The main difference are the observations- which tend to be variable in both drugs. Am I doing something wrong? Any input is appreciated!

Thanks,

Sandra

2

Hi Sandra,

This might be hard to troubleshoot without your project. However, in general, this usually happens when the NLME engine did not successfully compute standard errors for your run, so they won’t be returned in the output.

A few things to try:

  • If you are using a method that does not support SEs (like QRPEM or IT2S-EM), SEs will not be available.

  • If you want the engine to try the best available option, use Auto-Detect (PMLAutoSE) so it attempts Sandwich, then Hessian, then Fisher Score. In Phoenix, this setting is in the Run Options Tab, Stderr Method.

Also note: CV% depends on a successful covariance step, so if SEs are not available, CV% will not be reported either.

Hope this helps,

Ana

1 Like