Dynamic Chemotherapy Scheduling for Metastatic Colorectal Cancer Patients: Assessments and Improvements
MetadataShow full item record
Colorectal cancer (CRC) is the third most common and second deadliest type of cancer affecting both sexes. Patients with late-stage CRC undergo sequential chemotherapy treatments, which often yield undesirable toxic outcomes affecting patients' quality of life. Despite recent advancements in screening and treatment methods, only about 13% of metastatic CRC (mCRC) patients survive longer than 5 years. Existing guidelines for mCRC treatment offer flexibility in the decision making but may lead to sub-optimal outcomes. Tumors' drug resistance and high cost of chemotherapy treatments bring extra challenges to physicians, yet continue to be either ignored or overlooked from existing guidelines. In effort to address these challenges, this dissertation leverages the unexploited amount of existing data describing the effectiveness of chemotherapy on CRC patients, recording prior clinical trials in a patient-, disease-, and treatment-specific database. Through this database, analytical models were developed to offer optimal solutions to decision makers for mCRC chemotherapy scheduling. Advancements in chemotherapy treatment have improved long term survival for mCRC while revoking financial concerns. In the first part of this dissertation, we analyze the cost-effectiveness of clinically accepted combinations of up to three lines of therapies from 8 chemotherapy regimens to evaluate the progress made in CRC treatment and examine how treatment sequencing affects the effectiveness of treatment plans. My analyses show that improvements in health outcomes may come at a high incremental cost for mCRC patients and are highly dependent on sequencing of treatments. While single and two lines of therapies can be cost-effective, majority of the cost-effective sequences consist of three-lines of therapies. The important role of treatment sequencing in the effectiveness of a treatment plan galvanized the development of analytical models addressing the chemotherapy scheduling for mCRC patients under uncertain health progression. We developed two finite-horizon, discrete-time Markov decision process models under different assumptions that incorporate treatments' induced toxicity, risk of disease progression and the accumulating knowledge from frequent routine tests in the decision making process, providing a novel chemotherapy treatment decision making framework. We also propose a procedure that reduces the computational effort needed to solve the model without compromising accuracy of our results. The goals of these models are to maximize the total expected lifetime of the patient. Using clinical data, we derive optimal policies along with insights to assist decision makers in making more informative decisions. The final part of this research studies the survival benefits from introducing chemotherapy free periods in the treatment schedules of mCRC patients. We developed a Markov decision process model that adheres to the principles of current clinical practice. The objective of the model is to maximize the total expected lifetime of a mCRC patient. Findings from numerical results are consistent with past clinical trials with regard to the value of chemotherapy breaks. This dissertation presents a novel direction in mCRC chemotherapy scheduling research. Important factors such as toxicity and tumors' drug resistance influence the effectiveness of chemotherapy treatment. The developed research acknowledges their significance and addresses the chemotherapy scheduling problem embracing the challenges arising from these factors. Albeit this dissertation's focus is the chemotherapy scheduling for mCRC patients, the proposed methodologies can be used to address other types of cancer and/or treatment options. The developed models and their solution methods are generalizable to other vexing problems arising in cancer decision making.