When conducting time-to-event analyses for prospective clinical studies (where data is collected at time intervals), it is appropriate to utilize all information available. Missing data issues can also arise when modeling data of this nature. Conventions such as Last Value Carried Forward (LVCF) may be used to account for the missingness in covariates.
In this paper, we describe an easy method in SAS® for setting up Cox Proportional Hazards models with timevarying covariates utilizing proc PHREG. We also describe the SAS® code used for construction of LVCF. An example of an application of these techniques from a Phase IIB randomized, multi-site HIV prevention trial conducted in the US from 1999 to 2003 (called EXPLORE), are illustrated and discussed.
conference:
Paper Type:
Paper
