Amy Bird Knudsen
Assistant Professor in Radiology, Harvard Medical School
Dissertation Title: "Explaining the Secular Trends in Colorectal Cancer Incidence and Mortality with an Empirically Calibrated Microsimulation Model"
Colorectal cancer is the fourth most common cancer and the second leading cause of cancer death in the U.S. The incidence and mortality from colorectal cancer have changed over time. Prior to 1985, incidence was relatively flat at 86 to 90 cases per 100,000. Since 1985, the incidence rate has been on a general decline, falling steadily from 90 cases per 100,000 in 1985 to 69 cases per 100,000 in 1995. Incidence increased slightly during the late 1990s but has been falling steadily since then to a rate of 65 cases per 100,000 in 2001. Colorectal cancer mortality has been falling by approximately 1.5% per year for the past thirty years to a rate of 31 deaths per 100,000 in 2001. Several factors may have contributed to the observed trends in colorectal cancer incidence and mortality, including secular trends in the risk factors for the disease, the dissemination of colorectal cancer screening, and changes in the patterns of care for diagnosed cancer. What is the relative contribution of each of these factors in explaining the incidence and mortality trends? To address this question, we developed a comprehensive population-based microsimulation model of colorectal cancer. The model tracks the development of adenomas and their progression to invasive colorectal cancer and incorporates the effects of risk factors, screening, and treatment on the underlying disease process.
In Chapter 1 we describe the development and calibration of the natural history model of colorectal cancer. Using a likelihood-based approach, we simultaneously fit the model to data on the age-specific prevalence of adenomas by number from autopsy studies, the distribution of findings on colonoscopy by location, size, and histology from screening studies, and the stage-, location-, and age-specific incidence of cancer from the Surveillance, Epidemiology and End Results Program.
In Chapter 2 we use the calibrated model to estimate what the colorectal cancer incidence and mortality over the past thirty years might have been had the secular trends in risk factors, the dissemination of screening, and the changes in the patterns of care not occurred. We found that the secular trends in the risk factors explain only a small fraction of the trend in incidence; the vast majority of the change in incidence over time is attributable to screening. When screening was first disseminated, the incidence rate rose as individuals with asymptomatic cancers were detected. Over time, screening has lead to a reduction in the incidence of the disease by detecting adenomas for removal, thereby preventing cancers from occurring. Changes in the patterns of care for diagnosed cancer explain 59% of the reduction in colorectal cancer mortality; secular trends in risk factors and screening explain 8% and 32% respectively.
Finally, in Chapter 3 we explore whether heterogeneity in the rates of adenoma growth lead to different conclusions about the effectiveness and cost-effectiveness of colorectal cancer screening. We found that heterogeneity mitigates the reductions in incidence and mortality attributable to screening and may lead to more favorable incremental cost-effectiveness ratios but that it does not change the conclusions about which screening strategy is the most effective at reducing the incidence of colorectal cancer and which strategy yields the greatest reduction in colorectal cancer mortality. From a methodologic perspective, together these analyses demonstrate the value of population-based models in informing cancer-control opportunities.