A new approach to estimate time-to-cure from cancer registries data

Publication date: April 2018
Source:Cancer Epidemiology, Volume 53
Author(s): Olayidé Boussari, Gaëlle Romain, Laurent Remontet, Nadine Bossard, Morgane Mounier, Anne-Marie Bouvier, Christine Binquet, Marc Colonna, Valérie Jooste
BackgroundCure models have been adapted to net survival context to provide important indicators from population-based cancer data, such as the cure fraction and the time-to-cure. However existing methods for computing time-to-cure suffer from some limitations.MethodsCure models in net survival framework were briefly overviewed and a new definition of time-to-cure was introduced as the time TTC at which P(t), the estimated covariate-specific probability of being cured at a given time t after diagnosis, reaches 0.95. We applied flexible parametric cure models to data of four cancer sites provided by the French network of cancer registries (FRANCIM). Then estimates of the time-to-cure by TTC and by two existing methods were derived and compared. Cure fractions and probabilities P(t) were also computed.ResultsDepending on the age group, TTC ranged from to 8 to 10 years for colorectal and pancreatic cancer and was nearly 12 years for breast cancer. In thyroid cancer patients under 55 years at diagnosis, TTC was strikingly 0: the probability of being cured was >0.95 just after diagnosis. This is an interesting result regarding the health insurance premiums of these patients. The estimated values of time-to-cure from the three approaches were close for colorectal cancer only.ConclusionsWe propose a new approach, based on estimated covariate-specific probability of being cured, to estimate time-to-cure. Compared to two existing methods, the new approach seems to be more intuitive and natural and less sensitive to the survival time distribution.

Graphical abstract

from Cancer via ola Kala on Inoreader http://ift.tt/2BWdDqP
via IFTTT