Ascertainment bias and estimating penetrance
Menée auprès de 65 057 patientes atteintes d'un cancer du sein et soumises à un test génétique de prédisposition au cancer (âge moyen : 48,5 ans), cette étude évalue l'association entre la présence de variants dans les gènes testés (21 panels) et le risque de développer un cancer du sein
To the Editor Recently Couch et al1 reported the largest published sample of clinical multigene panel testing. This enormous study of more than 65 000 patients with breast cancer histories demonstrated the “key finding” that a pathologic variant of PALB2 is associated with a high risk for breast cancer development. The authors note in their conclusions that risk estimates derived from their study are “likely to be inflated” compared with population-based studies (in which patients would have lower pretest probabilities of testing positive for a germline alteration) and the authors cite another study involving carriers of PALB2 variants wherein the risk estimates were different. In the accompanying editorial, Obeid et al2 also highlight the same concern regarding “ascertainment bias.” Currently, screening and treatment recommendations for carriers are generally unaffected by pretest probability estimates.3
JAMA Oncology , commentaire, 2016