Premium
Regression Models and Non‐Proportional Hazards in the Analysis of Breast Cancer Survival
Author(s) -
Gore Sheila M.,
Pocock Stuart J.,
Kerr Gillian R.
Publication year - 1984
Publication title -
journal of the royal statistical society: series c (applied statistics)
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.205
H-Index - 72
eISSN - 1467-9876
pISSN - 0035-9254
DOI - 10.2307/2347444
Subject(s) - proportional hazards model , breast cancer , regression analysis , statistics , survival analysis , regression , oncology , mathematics , cancer , medicine
SUMMARY The Western General breast cancer series of 3922 patients sets research methodology for survival data in practical perspective and illustrates that the waning of covariate effects through time is an important phenomenon in medical applications. Non‐monotone convergent hazard functions are associated with most clinical covariates in breast cancer; an unusual hazard pattern according to menopausal state is also reported. These features contraindicate the use of standard regression models for survival such as Weibull and proportional hazards. Inferences about covariate effects are compared under these and a log‐logistic model which implies proportionality of the cumulative odds on death. Regression models are shown to be useful in exploratory analysis. In particular, a step‐function proportional hazards model elucidates the time‐dependent influence of initial covariates and leads to a more appropriate final model, but one whose virtues are balanced by computational difficulty.