Bootstrap failures after upgrading Phoenix (8.5 → 8.8) in 2×2 crossover PopPK model Phoenix NLME

[Bhim_Bahadur]

Hello everyone,
I am facing a bootstrap issue in Phoenix NLME after upgrading from version 8.5 to 8.8 and would appreciate insights from others who may have encountered similar behavior.

Background
I am working on a 2-sequence, 2-period crossover BE study with the following structure:

Each subject receives both Test and Reference
Period 1 → 50% subj first, 50% second formulation
Period 2 → 50% subj second, 50% first formulation after washout
Dataset includes all major columns:
USUBJIDN, TREATN, SequenceN, Period, TAD, AFRLT, Cobs, AMT, EVID, MDV

Data is sorted by USUBJIDN
Model initially used AFRLT (time after first dose):

Period 1: 0–24 h
Period 2: after washout (~168 h onward)

Model behavior

The base model fits without any issue, visual predictive check by TAD or AFRLT all ok. Convergence achieved for all model parameters.
In Phoenix 8.5, bootstrap ran successfully (tested with ~50 replicates)
After upgrading to Phoenix 8.8, running the same dataset , same model file (just simple re-run)and same parameters without any change also results in bootstrap failure.

Here is earlier output smapshot

Now Bootstrap error, without any Stratification, just replicate
It results below error without a single run

Now Bootstrap error, with Stratification as SequenceN

 model runs up to 4 replication then 

I would appreciate your input.

Thank you .

[SimonDavis]

Please can you provide this project to support@certara.com for investigation, thanks, Simon.

[Bhim_Bahadur]

Dear Simon,

I recreated the issue using the standard Phoenix NLME example dataset (Theophylline).

In the original dataset, each subject has an equal number of observed data points (N = 11). To introduce variability in the number of observations per subject, I modified the dataset by removing:

• the last observation for Subject 1, and

• the last two observations for Subject 2.

This resulted in a dataset with unequal numbers of observed (non-missing) values across subjects (summary attached).

I then refitted the same NLME model used in the Phoenix example without any structural changes and proceeded to run the bootstrap analysis. The same error was reproduced during bootstrapping.

Looking forward for your suggestions .

Thank you.

[SimonDavis]

Hi Bhim, we have replicated the problem [SUPPORT-3575] and believe we understand the cause. We are looking to expedite a fix. Will keep you updated as to the likely availability, thanks for bringing this to our attention. Simon.