Simulation of Colorectal Cancer: The natural history of disease

Show full item record

Title: Simulation of Colorectal Cancer: The natural history of disease
Author: Cubbage, Daniel Frederick
Advisors: James Wilson, Committee Member
Henry Nuttle, Committee Member
Stephen Roberts, Committee Chair
Abstract: This thesis presents a comprehensive, fully verified, and validated model of the natural history of Colorectal Cancer (CRC). CRC is the fourth most common type of cancer and the second leading cause of cancer death among both men and women. Individuals who develop CRC often fail to detect symptoms until the cancer is in an advanced stage. There are a number of screening methods designed to detect CRC in its early stages or prevent CRC by identifying and removing adenomous polyps, a pre-malignant form. However, because of the long latency of CRC and the time needed for clinical trials, it is not practical to provide clinical trials of all the screening and treatment strategies for CRC. Models offer an alternative means to analyze of screening/surveillance recommendations. Before considering any CRC medical interventions with a model, a model of the natural history of CRC is of fundamental importance. A model of the natural history of CRC requires a compromise of knowledge of CRC and data describing it. These compromises are described by modeling assumptions regarding the actual process of CRC development. To summarize the outcomes, the two primary measurements are the costs associated with the treatments and the years of life, or life-years. These measurements can be modified in several ways, by discounting or adjusting life-years to reflect the quality of life based upon different states of health. Within the medical decision-making community, two primary types of models for CRC have been developed, Markov models and discrete-event simulations. While the Markov models are easy to build and provide a basic analysis of the impact of screening, a more flexible, but more complex, approach is the discrete-event simulation model. One discrete-event simulation is the Vanderbilt Model that is the predecessor to the Vanderbilt-NC State model presented in this thesis. The Vanderbilt-NC State model improves the original Vanderbilt model with enhanced features such as database storage of inputs and Excel outputs. It also models additional important factors such as race, family history, reference year, risk effects, and histology. The object-oriented design allows the discrete-event simulation to follow the adenomas and people through the system, rather than forcing these objects into a process flow. When an individual is created, his natural death and first adenoma development are scheduled. The adenoma object is then responsible for its own progression up to cancer and potential cancer death. It also schedules the next adenoma, which will then follow its own timeline through the simulation. Once the model was constructed, it had to be verified and validated against cancer information from clinical studies. A detailed calibration procedure was implemented to match the model output with the cancer incidence, adenoma prevalence, and people with adenomas. In the process of validation, the Vanderbilt-NC State model outcomes are compared to data from other sources that were not used in the fitting process. The model?s output was compared to the cumulative risk of getting cancer obtained from a national CRC database (SEER). The model was also compared to a previous simulation that sampled from an adjusted lifetime that had the risk of CRC eliminated. This comparison was performed to validate the life-year gain associated with the elimination of CRC. Finally, the Vanderbilt-NC State model was compared to the previous Vanderbilt model. In each of the comparisons, the Vanderbilt-NC State model was consistent with the patterns and magnitudes of outcomes found in the external data. Once the model matched the literature results, analysis could be performed to determine the impact of CRC. According to the model, CRC reduces average lifespan by 0.24 years. This average loss of life comes from approximately 2.5% of the population losing an average of over 10 years of life. The average cost associated with the diagnosis and treatment of the disease is $2,188 per person. The model is the first comprehensive model of CRC and can be extended to consider screening and other medical interventions without direct experimentation using patients.
Date: 2004-06-09
Degree: MS
Discipline: Industrial Engineering
URI: http://www.lib.ncsu.edu/resolver/1840.16/2843


Files in this item

Files Size Format View
etd.pdf 828.0Kb PDF View/Open

This item appears in the following Collection(s)

Show full item record