December 1, 1995Biometrics
Time-Dependent Effects of Fixed Covariates in Cox Regression
Authors
Discussion
Loading...
Member takes
Implication
Methodological study demonstrates time-varying covariate effects in survival analysis, suggesting enhanced model flexibility over standard proportional hazards assumptions.
Key Points
- To develop a nonparametric extension of the Cox proportional hazards model that allows the effects of fixed covariates to change flexibly over time.
- Assigned individual parameters for covariate effects at each uncensored survival timepoint.
- Estimated parameters by maximizing a penalized partial log-likelihood incorporating a penalty on first-order differences between consecutive coefficients.
- Determined the optimal smoothing parameter using Akaike's Information Criterion and demonstrated the method on ovarian cancer and kidney transplantation datasets.
- The penalized likelihood framework successfully relaxed the proportional hazards assumption by capturing dynamic, time-varying covariate trajectories.
- Demonstration on ovarian cancer and kidney transplantation cohorts confirmed the technique's capability to detect changing hazard dynamics across clinical follow-up periods.
Cite This Study
Verweij et al. (1995) studied this question.