Propensity score analysis with a time-dependent intervention is an acceptable although not an optimal analytical approach when treatment selection bias and survivor bias coexist