Predicting survival in patients with colorectal cancer
Menée en Grande-Bretagne à partir de données portant sur 44 145 patients atteints d'un cancer colorectal (âge : 15 à 99 ans), puis validée à partir de données (issues de la base QResearch et du registre national des cancers) portant sur 15 214 et 437 821 patients complémentaires, cette étude évalue la performance d'un modèle mathématique intégrant 16 variables cliniques pour estimer la survie
New models will help, but shouldn’t be used in isolationBy estimating the probability of given outcomes for individuals based on a combination of clinical and sociodemographic characteristics, the growing number of risk prediction models have the potential to support decision making by patients and clinicians.In a linked paper (doi:10.1136/bmj.j2497), Hippisley-Cox and Coupland use data from a large UK primary care database (QResearch) to develop models to estimate survival in men and women after a diagnosis of colorectal cancer.1 They then validate them in a separate set of patients within the same database and in the Public Health England cancer registry. Using established statistical measures of performance, they show that the models are reasonably good at ranking people according to their survival, and the predicted survival estimates closely match those observed in the study populations and other studies.Compared with existing models, these new ones have several advantages.23 Firstly, they are applicable to all patients with colorectal cancer, whereas existing models apply to …
BMJ , éditorial, 2016