![](img015.GIF) |
This Table compares the model using
parametric time-by-covariate interactions and the piecewise PH model in terms of the
results of LRT tests of the PH hypothesis. These results confirm the need to incorporate
modelling of non-proportional hazards in the relative survival analyses , as the PH
hypothesis is rejected in at least two of the three analyses for all prognostic factors
except gender. Moreover, this table shows the failure of linear and logarithmic
time-by-covariate interaction models to indicate significant time-dependent effects for
stage 3 probably due to the inability of these models to represent non-monotonic changes.
The test based on the piecewise PH model will be less efficient than 1-df tests based on
more parsimonious linear time-by-covariate interaction models. |