This work suggested that both flexible modelling of non-proportional hazards and the relative survival approach are important. In particular, there are considerable differences between the results obtained from modelling non-proportionality of hazards in relative survival and in the conventional Cox model. The application to colon-cancer data also demonstrated the advantages of using restricted cubic spline functions for modelling non-proportional hazards in relative survival analysis. First, the restricted cubic spline model provided a better fit than the model with a linear time-by-covariate interaction. Moreover, the model based on restricted cubic spline functions is more parsimonious than a piecewise PH relative survival model and allows to represent both simple and complex patterns of changes with a single parsimonious model without requiring a priori definition of the functional form of the variation of the hazard ratio. Of course using a spline function requires the determination of the the number and the location of the knots .

Go to Comment Form