Using Mathematical Models for Cancer Control: An Example of Colorectal Cancer

by Karen M. Kuntz, Sc.D., Professor,
Division of Health Policy and Management,
University of Minnesota School of Public Health

Colorectal cancer is the second most common cause of cancer-related deaths in the United States and is one of the few preventable cancers. It is generally accepted that most, if not all, colorectal cancers arise from a precancerous condition: the adenomatous polyp. Polyps can be identified by a screening test and removed, thereby preventing cancer from developing. In addition, colorectal cancer can be prevented through risk factor modifications, such as changes in smoking, diet, physical activity, or multivitamin use. For individuals who are diagnosed with colorectal cancer, advancements in chemotherapy regimens over the past few of decades significantly reduce the chance that the patient will actually die of his or her cancer.

Since 1978, colorectal cancer mortality has been declining approximately 1.5 percent per year. While this trend is promising and is a step in the right direction in fighting the war on cancer, it is not clear what advancements have contributed most to this decline. To best guide cancer control opportunities in the future, it is important to know the degree to which improvements in the past are responsible for the observed decline in colorectal cancer mortality. With a goal to address such questions, the National Cancer Institute (NCI) formed the Cancer Intervention and Surveillance Modeling Network (CISNET), which is a consortium of NCI-sponsored investigators whose focus is to use mathematical modeling to improve our understanding of the impact of cancer control interventions on population trends in mortality. By using a mathematical representation of the disease process and its interactions with risk factors, screening and treatment, CISNET modelers thereby create a "public health laboratory" with which to examine questions such as: What are the predicted cancer mortality trends in the absence of any screening?

Population-based Simulation Model

For a CISNET project that focuses on colorectal cancer, we developed a population-based simulation model -- SimCRC -- to analyze the potential contributors to observed colorectal cancer mortality trend. SimCRC incorporates the U.S. population trends for colorectal cancer risk factors (e.g., decreased smoking rates), screening practice patterns, and chemotherapy dissemination rates, as well as their effects on colorectal cancer incidence and mortality. Using SimCRC to simulate the U.S. population from 1978 to 2000, our predictions for the number of colorectal cancer deaths per 100,000 are similar to the U.S. vital statistics. One at a time we restrict the risk factor levels, screening practice patterns, and chemotherapy regimens from changing over time. We are then able to project what the colorectal cancer mortality would have been in the absence of these changes.

Between 1978 and 2000 colorectal cancer incidence declined from 28 to 21 deaths per 100,000 -- a decline of 25 percent. We estimate that 8 percent of this decline was due to net improvements in risk factors over this time period, while 80 percent was due to screening and 17 percent was due to advancements and increased utilization of chemotherapy for cancer patients.* While most risk factor trends have been improving over time (e.g., decline in smoking rates, increase in multivitamin use), others have not (e.g., increase in obesity prevelance). The net effect of the long-term trends in the risk factors for colorectal cancer, however, has been a reduction in risk.

Our results suggest that cancer control policies should focus their efforts on increasing screening dissemination rates. In spite of relative low participation rates in the U.S. (e.g., only 39 percent of the population 50 years of age and older in 2000 reported ever having a colorectal endoscopy), the dissemination of screening has played a significant role in decreasing colorectal cancer mortality. Widespread adoption of screening could make significant inroads at reducing the burden of colorectal cancer.

SimCRC and Healthy People 2010

SimCRC has also been used to project future outcomes to evaluate "Healthy People 2010" goals -- a set of health promotion and disease prevention objectives for the U.S. population put forth by the federal government. In partnership with the NCI and the Centers for Disease Control and Prevention's Mid-Course Review of the cancer goals, we evaluated whether the nation will meet the Healthy People 2010 goal of reducing the colorectal cancer mortality rate to 13.9 deaths per 100,000 under alternative projections for CRC risk factors, screening, and treatment.

We used SimCRC to project colorectal cancer mortality rates in the year 2010 under three scenarios: (1) risk factor and screening trends continue at their current pace, a"continuing scenario:" (2) Healthy People 2010 goals for risk factors and screening are met, a "Healthy People scenario;" and (3) difficult yet realistic goals for colorectal cancer risk factors, screening, and treatment are achieved, an "optimistic scenario."

We found that if risk factor, screening, and treatment trends continue at their current pace (i.e., continuing scenario), the colorectal cancer mortality rate among white men is projected to fall from 21.5 deaths per 100,000 in 2004 to 17.5 in 2010. If the Healthy People 2010 goals for risk factors and screening are met, the Healthy People 2010 colorectal cancer mortality goal is not achievable in 2010 but would be reached by the year 2015. However, many of the Healthy People 2010 risk factor goals are unrealistic (e.g., increase the proportion at a healthy weight from 22 percent of white men in 2004 to 60 percent in 2010); a similar reduction could be achieved in the year 2015 if more realistic goals are set for risk factors, more aggressive goals are set for colorectal cancer screening, and if all who are eligible for chemotherapy receive the best-available treatment (i.e., optimistic scenario).

The existence of several CISNET models provide an important framework for examining questions related to cancer control interventions, priority setting, and quality of care initiatives.

For more information about CISNET go to: www.cisnet.cancer.gov.

* Percentages don't sum to 100 because each set of improvements was implemented in the absence of the other improvements (e.g., without improvements in screening there is more room for improvement with changes in risk factors.)