The SAS Macro %TWOSTAGECOX

The model discussed in the previous section can be fitted using the SAS macro %TWOSTAGECOX, which has the following generic form:

%TWOSTAGECOX(data=ovarian,true=surv,trueind=survind, surrog=pf s,surrogind=pfsind,trt=treat, trial=center,patid=patient,common=1, robust=1,looa=1);

The macro's specific arguments are;

• • trueind: censoring indicator for the true endpoint (1=event, 0=censoring).
• • surrogind: censoring indicator for the surrogate (1=event, 0=censoring).
• • common: an option that allows the user to choose between the trial-specific baseline hazard function (common=0), or the common baseline hazard function (common=1) in the first stage.
• • robust: an option that allows the user to obtain the robust (or adjusted) Rtrial in the output (1 for robust; 0 for non-robust).

Data Analysis and Output

The macro %TWOSTAGECOX produces two exploratory plots: the patients’ distribution in the trials by treatment arm (shown in Figure 12.20, top panel) and the Kaplan-Meier curves for the true and surrogate endpoints (shown in Figure 12.20, bottom panel).

As shown in Figure 12.21, the estimated trial-level surrogacy is equal to R_rial = 0.9184 (0.8674, 0.9695), which implies that progression-free survival time is a good surrogate for overall survival, which can be clearly seen in Figure 12.22. The result of the leave-one-out analysis is shown in Figure 12.23.