Multi-state AFT Regression

Hi everyone!

Could anyone point me towards an implementation of buckley james that can handle left-truncation? Or perhaps rms can be finagled into handling this?

I’d like to estimate a clock-reset multi-state semi-parametric accelerated failure time model. If I could assume proportional hazards, I’d use mstate+coxph, but 1) it’s not the right model and 2) I’d like proper residuals.

Thanks in advance,