Mitigating bias and dealing with multiple time scales in cohort studies: studying medications and complications of diabetes

Cohort studies are an important and powerful tool of epidemiologic research. When based on a representative cohort, observational cohort studies provide results of a high external validity given that the internal validity is not impaired by bias. Diabetes and cancer represent two prevalent, complex, diverse and potentially fatal chronic diseases, occurs more often than could be expected by chance only. Cancer and diabetes share common risk factors, such as obesity and smoking, but also antidiabetic medications may play a role. I studied the relationship between the use of antidiabetic medications and cancer risk in two retrospectively conducted observational cohort studies, when mitigating bias through the study design and analytical methods. In Study I, I studied the risk of cancer in 23 394 individuals from the National FINRISK cohorts. I assessed the risk along time since the initiation of anti-diabetic medication and adjusted for several risk factors, including smoking and body mass index. I found no association between cancer risk and the use of antidiabetic medication. In Study II on the CARING (CAncer Risk and INsulin Analogues) five-country (Denmark, Finland, Norway, Sweden, UK) cohort of 327 112 new insulin users, the risks of ten site-specific cancers and any cancer were scrutinized by contrasting the cumulative exposures to human insulin and insulin analogues glargine and detemir, when mitigating biases involved in previous observational studies. We found no consistent differences in the studied risks as assessed for insulin glargine or insulin detemir use relative to use of human insulin. Due to their longitudinal nature, cohort studies involve at least one time scale at which the time-dependent dynamics of a phenomenon can be studied. There are often several relevant time scales, but the traditional statistical methods of survival analysis, such as Cox’s proportional hazards model, rely on a single time scale. In the methodological part of this thesis, I addressed the issue of multiple time scales. In Study III, I addressed the issue of multiple time scales in cohort studies by introducing and evaluating a nonparametric Bayesian model for estimation of intensity on two time scales jointly. Because of the built-in smoothing and borrowing of strength in two dimensions, the model outperformed two other methods when applied to simulated data. In Study IV, I used the Bayesian intensity model to explore the time-dependent dynamics of end-stage-renal-disease and death without end-stage-renal disease in 11 810 individuals with type 1 diabetes from the nation-wide FinDM study. The time-dependent dynamics of these outcomes were modelled on two and three time scales jointly, including age, diabetes duration and calendar time. The study demonstrated that the estimation of multidimensional hazard allows for addressing both empirical and methodological questions.