Extensions of flexible survival models to marginal structural and Markov multi-state modeling

点击翻译按钮获取中文摘要

Prognostic studies of the progression of various diseases are essential to identify patients at high risk of mortality and to assess the effectiveness of alternative treatments in preventing deaths and increasing survival. The clinical outcomes for patients with different clinical and socio-demographic characteristics and/or receiving different treatments are compared, to derive conclusions regarding the impact of these characteristics and treatments on disease progression and mortality. The validity of such conclusions depends crucially on the accuracy of statistical methods employed to analyze the data. Sophisticated statistical methods are necessary to account for the following challenges: (a) several patient characteristics and/or treatments may influence the outcome, (b) some characteristics are correlated with each other, (c) treatment often depends on patient's characteristics that may also affect the clinical outcome. The proposed research project relies on recent progress in both statistical theory and computational techniques to develop new, flexible statistical methods to address further challenges related to (d) the multitude of possible longitudinal pathways of disease evolution and of clinical outcomes (for example, deaths due to different causes), and changes over time (e) in some important clinical risk factors and the type and/or dose of the prescribed treatment. By allowing clinical researchers to draw correct conclusions even from such complex data, the proposed project will advance understanding of the progression of complex diseases. This may ultimately enhance the quality of clinical prognosis and the effectiveness of the treatments.