Stochastic Models for Improving Screening and Surveillance Decisions for Prostate Cancer Care

Markov model;simulation;partially observable Markov decision process;prostate cancer;biomarkers;MRI;Industrial and Operations Engineering;Engineering;Industrial & Operations Engineering

Barnett, ChristineSeiford, Lawrence Martin ;

University of Michigan

关键词: Markov model;    simulation;    partially observable Markov decision process;    prostate cancer;    biomarkers;    MRI;    Industrial and Operations Engineering;    Engineering;    Industrial & Operations Engineering;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/136969/clbarnet_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Recent advances in the development of new technologies for the early detection and treatment of cancer have the potential to improve patient survival and lower the cost of treatment by catching cancer at an early stage. However, there is little research investigating the health and economic implications of these new technologies. For example, magnetic resonance imaging (MRI) and new biomarker tests have been proposed as potential minimally invasive ways to achieve early detection of prostate cancer. These new technologies vary in their sensitivity and specificity leading to both false-positive and false-negative results that can have serious health implications for patients. Moreover, due to the high cost and imperfect nature of these new tests, whether and when to use these tests is unclear. We present stochastic models for prostate cancer disease onset and progression that incorporates partial observability of a patient;;s prostate cancer health status. We used statistical learning algorithms and clinical datasets combined with expert clinical knowledge of urologists at the University of Michigan to estimate and validate the models. The models can simulate progression through prostate cancer states to mortality from prostate cancer or other causes for a population of patients. New technologies, such as MRI and biomarker tests, are incorporated into the model using a probabilistic representation of test outcomes to represent the information these tests provide about the true health status of the patient. Since these technologies can be used in varying ways, the choice of tests and optimal times to initiate tests are treated as decision variables in the model. We calibrated and validated our models using several data sources and subsequently used our models to design optimal testing strategies that trade-off the harms and benefits of using these new technologies. Our results show that these new technologies can lead to significantly improved health outcomes and they are cost-effective relative to established norms for societal willingness-to-pay. We have also used these models to provide important insights about the optimal timing of prostate biopsies for men with low-risk prostate cancer undergoing active surveillance. By using new technologies to better select men for biopsy and by improving active surveillance strategies, physicians can reduce the harms of prostate cancer screening (e.g., unnecessary biopsies and overtreatment of low-risk disease) while continuing to reduce prostate cancer deaths through screening and early detection. The methodological approaches we present in this thesis could be applied to many other chronic diseases, including bladder, breast, and colorectal cancer.