Impact Of Early Detection In Prostate Cancer Incidence: A Joint Modeling Approach Of Stage And Age At Diagnosis

• Activity Number: 589
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Chen Hu*+ and Alexander Tsodikov
• Companies: University of Michigan and University of Michigan
• Keywords: Prostate cancer incidence ; Cancer epidemiology ; Screening ; Semiparametric ; SEER ; Disease natural history

Abstract

Early detection of prostate cancer, such as prostate-specific antigen (PSA) screening, led to a sharp spike in incidence and accompanied by equally pronounced improvement in patient prognoses at diagnosis. Such observation questions the casual connection between early detection and mortality reduction due to the so-called "over-diagnosed" cancers. The impacts of screening on the disease natural history can be justified through statistical models of prostate cancer incidence based on lead time, over-diagnosis and disease progression. In this paper, we propose a semiparametric population model for cancer stage progression and diagnosis to evaluate impacts of screening programs. The disease natural history model is constructed through a series of semiparametric regression with time-dependent covariates such that we can jointly model the correlated response of stage and age at diagnosis as stage-specific cancer incidences. The proposed model provides estimates of lead time, fractions of over-diagnosis and disease progression changes due to early detection. The proposed model is applied on SEER prostate cancer incidence data from 1977 to 2000.