Modelling short and long term consequences of changes in diagnostic activity and treatment

Since the late 90’s the diagnostic activity for prostate cancer has increased in Sweden, primarily due to increased use of PSA testing, and this has led to a large increase in diagnoses. Simultaneously, there have been changes in treatment strategies, and more effective treatments have been introduced. This thesis aims to increase the understanding of short and long term consequences of these changes by use of high quality data on virtually all men diagnosed with prostate cancer in Sweden. In paper I, the survival of men with metastatic prostate cancer at diagnosis was investigatedby use of survival models, including Kaplan-Meier analyses and Cox proportional hazards regression.The median survival from diagnosis increased with 6 months when comparing mendiagnosed 1998-2001 with men diagnosed 2010-2015, and the risk of death decreased with 13%, while median levels of prostate specific antigen at diagnosis dropped with up to 50%. In paper II, the interplay between diagnostic activity, incidence and risk of death by prostate cancer was modelled using a discrete time model. Data on diagnostic activity, e.g. in termsof testing frequencies, was not available and therefore a proxy for the diagnostic activity wasused. The hazards were estimated within the framework of generalized additive models. Two simulations were performed, assuming low and high diagnostic activity respectively, to compare incidence and mortality from 2017-2060. Higher diagnostic activity, compared to lower, led to more men being diagnosed, primarily with lower risk prostate cancer, but in the long run it led to fewer men diagnosed with metastatic disease and fewer prostate cancer deaths