Data Analysis and Output

The %SURVBIN macro produces three default exploratory plots shown in Figure 12.33. The Kaplan-Meier curves (by endpoint) in Figure 12.33 (top panel) indicates that there is no difference between the treatment arms across the level of the surrogate endpoint. The box plots for the survival times in Figure 12.33 (middle panel) reveal the same pattern. The number of patients per trial and treatment arm is shown in Figure 12.33 (bottom panel). Individual- and trial-level surrogacy, global odds = 4.9108 (4.15794, 5.6638) and R^riai = 0.4417 (0.1564, 0.7269), shown in Figure 12.34 indicate that two- category tumor response is a surrogate of moderate value to overall survival for the colorectal cancer data.

Figure 12.35 shows the parameter estimates for the treatment effects for both the surrogate and true endpoints that were used to estimate trial-level surrogacy.

The SAS Macro °/„SURVCAT

The analysis presented in Section for a categorical (ordinal) and survival endpoint, and in Section, for binary tumor response, was conducted using the SAS macro %SURVCAT. The macro has the general form:


surrog=responder,trt=treat,center=center,trial=center, vars=,patid=patientid)

The macro's arguments are:

  • • true: measurement of the failure-time endpoint.
  • • trueind: censoring indicator (1=event, 0=censoring).
  • • surrog: measurement of the ordinal categorical surrogate (levels 1, 2,...,K).
  • • treat: treatment indicator (0 or 1).
  • • center: unique id (continuous) for units, for which specific treatment effects are estimated.
  • • trial: unique id (continuous) for groups of the units, for which common “baselines” are to be used.
  • • vars: macro variable containing possible covariates for adjustment of the Weibull and proportional odds models. The names of these covariates have to be passed to the program through the macro variable vars.
  • • patid: patient’s identification number.

The macro's output is presented in Chapter 6. Note that if the number of categories is 2, the analysis using the macro %SURVCAT the macro %SURVCAT is identical to the analysis using the macro %SURVBIN.

FIGURE 12.33

Colorectal Cancer Da,ta. Descriptive plots. Panel a: KM curves stratified by the binary surrogate endpoint. Panel b: Survival time distribution by treatment arm across the levels of the binary surrogate. Panel c: Distribution of patients by treatment arm.

FIGURE 12.34

Surrogacy measures with the 95% C.I.

FIGURE 12.35

Colorectal Cancer Data. Evaluation of trial-level surrogacy. Treatment effects upon the true endpoints (log-hazard ratio) versus treatment effects upon the binary surrogate (log-odds ratio). Circle area, is proportional to trial size.

< Prev   CONTENTS   Source   Next >