How to model temporal changes in epidemiological data : Treatment trajectories in men with prostate cancer

Abstract: Owing to the improvements in detection, diagnostics, and treatment, many men currently diagnosed with prostate cancer (PCa) have a low risk of PCa death. For many men PCa has become a chronic disease with a small risk of progression that remains even decades after date of diag-nosis. In this thesis PCa was used as an example of a chronic disease, since it holds all the main characteristics required by the WHO definition of a chronic disease. Against this background, the aim of this PhD thesis was to create models that can be used to quantify the probability of dif-ferent treatment trajectories and to assess the duration of certain disease states, while accounting for patient characteristics, disease severity, and primary treatment.In Paper I a state transition model was developed for prediction of dis-ease trajectories. The developed state transition model showed good consistency with a follow-up spanning up to 30 years.In Paper II a state transition model was developed and validated using age, Charlson comorbidity index, and a drug comorbidity index (DCI) based on filled drug prescriptions collected at a population-based level to estimate life expectancy.In Paper III the state transition model proposed in Paper I was used to assess 30-year PCa trajectories in men managed with active surveillance, in order to identify the ideal candidates for this management strategy.In Paper IV the state transition model from Paper I was updated includ-ing the castration resistant PCa (CRPC) state as an additional state and estimating the duration of the CRPC state as well as its outcomes.In Paper V, the updated state transition model from Paper IV was used to model long term outcomes for men with PCa managed with watchful waiting (WW). Since WW is currently recommended for men with PCa and life expectancy less than 10 years, the state transition model from Paper II was used to estimate life expectancy.