Any theory based model is suspect. Trapping and detrapping
behaviors are very lot-processing-variable. NBTI has fast,
recoverable and slow, nonrecoverable components to it.
I doubt that you want to take the data needed to fit the
envelope let alone fine-guess circuit response with NBTI in
play. Plus you would need each transistor to comprehend
its bias history uniquely, based on some other DC or transient
simulation.
In my world you don't play footsie with NBTI and hot carrier,
you stay the hell away from it.