INFORMATIVE CENSORING IN A MULTIPLICATIVE RELATIVE SURVIVAL MODEL: APPLICATION FOR KIDNEY TRANSPLANT RECIPIENTS

Vanessa Rousseau (France), Yohann Foucher (France), Magali Giral (France), Jean-Paul Soulillou (France) and Jean-Pierre Daures (France)

Received April 20, 2009

Abstract

Relative survival makes it possible to adjust the survival of a specific population relative to the survival of a reference population. In general, relative survival models are either additive or multiplicative: in the additive case the excess mortality rate is the difference between the specific mortality rate of a pathology and the mortality rate of the reference population, whereas in the multiplicative case it is the ratio (and it is named the relative mortality rate). In both cases, independence is usually assumed between the delay of occurrence of the event and the censoring. In some situations, such as unplanned or non-administrative censoring, this assumption is not valid. With a view on multiplicative models in particular (due to our interest in etiologic epidemiology and our search for explicative factors of particular events), results are developed for the case when censoring can be considered informative. These results are applied to the study of the survival of kidney transplant recipients, where patients who return to dialysis have been censored, which is in itself informative.

Keywords and phrases: relative survival, informative censoring, Cox model, semi-Markov model, kidney transplant recipients.