Forum de statistique : conférence « Estimating overdiagnosis in cancer screening studies: a statistical conversation about bias, identifiability and biologic mechanism »

- Forum de statistiques
Prof. Marc Ryser

Department of Population Health Sciences
Duke University, Durham, North Carolina, USA

Abstract: It is generally accepted that some screen-detected cancers are overdiagnosed and would not progress to symptomatic cancer if left untreated. However, because overdiagnosis is not directly observable, obtaining precise estimates constitutes a methodological challenge. In the first part of the talk we discuss the most common methods used to estimate overdiagnosis, namely excess-incidence and model-based approaches. We explain why most excess-incidence studies are prone to bias and consider model-based approaches that build on and extend methods for interval censored incidence as observed in cancer screening trials. Our extension permits preclinical tumors that are non-progressive but raises parameter identifiability issues. We discuss identifiability and model misspecification and propose concrete solutions to address these challenges. In the second part of the talk, we revisit the problem of cancer overdiagnosis through the lens of cancer biology. In particular, we seek to reconcile the current paradigm of cancer initiation and progression with the statistical overdiagnosis estimates from cancer screening studies and highlight the need for multi-scale approaches that integrate evidence from multiple sources.

Date et heure
Jeudi, 21 Mars, 2019 - 11:00
Salle Delachaux (étage 01), Unisanté, Bâtiment Biopôle 2, route de la Corniche 10, 1010 Lausanne

Retrouver sur notre site

Personne de contact: 

Institut universitaire de médecine sociale et préventive
Route de la Corniche 10, 1010 Lausanne - Switzerland
+41 21 314 72 72 |

Go to top