The use of regression models for making covariance adjustments in the comparison of survival curves is illustrated by application to a clinical trial of maintenance therapy for childhood leukemia. Three models are considered; the log linear exponential (Glasser [1967]); Cox's [1972] nonparametric generalization of this; and the linear exponential (Feigl and Zelen [1965]). Age and white blood count at diagnosis are both shown to be important for prognosis; adjustment for the latter variable has marked effects on the treatment comparisons. both advantages and disadvantages with the regression approach are noted.
Abstract