U.S. Flag An official website of the United States government

Colorectal Cancer Mortality Projections

Simulation Models for Colorectal Cancer

Sections on this page:

Summary: learn more about the models we used to make these projections

Simulation models are useful to both estimate up-to-date mortality trends, because of the normal reporting delays for mortality data, and to estimate future mortality trends.

For colorectal cancer (CRC), two mathematical models are used to simulate the impact on CRC mortality of changing cancer prevention, early detection, and use of chemotherapy treatments among populations. The two models use somewhat different assumptions about the development of CRC, and how risk factors, screening, and treatment influence the course of the disease and ultimately mortality.

For the purposes of policy and program planning, the Web site takes an estimated “average” of the different assumptions from the two models to provide a simple and clear set of findings about what the impact of different cancer control options will have on CRC mortality.

To see specific results from each model, go to the Compare Mortality Projection Models section of Interactive Graphs

Model Overview

These mortality projections are based on two computer models that simulate colorectal cancer (CRC) disease progression. The two models are:

  • SimCRC - based at the University of Minnesota, Minneapolis, MN and Massachusetts General Hospital, Boston, Massachusetts
  • MISCAN - a collaboration between Memorial Sloan-Kettering Cancer Center, New York, NY and Erasmus MC, the Netherlands

The models simulate colorectal disease progression in a large population of individuals from birth to death. As each individual ages, there is a chance that an adenomatous polyp – a benign precursor lesion that leads to colorectal cancer – develops. One or more adenomas can occur in any individual and each can independently develop into colorectal cancer (CRC).

The simulated population of individuals reflects observed distributions of characteristics in the real population. For example, the proportion of smokers among simulated individuals in a certain age group (cohort) reflects the proportion of smokers in that age group in the general population. The proportion of overweight individuals is similarly represented, and the models account for the fact that it is less likely for smokers to be overweight than non-smokers. As the individual “ages”, one variable to be calculated is whether that individual develops an adenoma. The calculation, while taking into account the positive or negative affects associated with characteristics like smoking, assures that the rate of adenoma development in the simulated population reflects our best knowledge about the rate of development in the real population (which is impossible to observe directly).

Various statistics can be computed from this simulated population (e.g., standardized mortality rates) in the same way such rates are generated from empirical data.

Diagram of the population simulation model. How changes in upstream objectives such as risk factors, screening and chemotherapy go into the CISNET models and project effects on the downstream objective, colorectal cancer mortality.We modeled interventions and their effects as follows:

  • Risk factor behavior influences the risk of onset of adenomas and/or the progression of large adenomas to preclinical cancer. The probabilities of adenoma onset or adenoma progression are modified up or down depending on a simulated individual’s risk factor profile.
  • Screening and surveillance affect all preclinical disease stages, possibly leading to the removal of an adenoma (potentially preventing CRC) or to the early detection of a carcinoma with a more favorable prognosis.
  • Chemotherapy can postpone or even prevent death from CRC. Improvements in chemotherapy are modeled by a reduction in cancer-specific mortality rates based on published hazard ratios.

The models incorporate specific objectives for these interventions, termed upstream objectives. The simulations project the effects of changes to upstream factors on CRC mortality, the downstream objective.

The SimCRC and MISCAN models share many characteristics. Both models simulate the US population from the 1970s to 2020 according to basic demographics, derived from census data and life table inputs. These data inputs determine the size of cohorts of simulated individuals. There are cohorts by age, race, and sex, enabling the model outcomes to be broken down by male/female and black/white.

Both models assume that all colorectal cancers arise from adenomas. Risk factor trends, screening dissemination, and chemotherapy utilization are the same for both models. Both models are calibrated to the same data concerning adenoma prevalence, CRC incidence, and population characteristics.

The primary differences between the two models are:

  • The MISCAN model assumes heterogeneity in adenoma growth and cancer progression, distinguishing progressive adenomas (adenomas that eventually can become cancer) and non-progressive adenomas (adenomas that cannot become cancer). The SimCRC model assumes that all adenomas have the ability to progress to cancer (although most will not during the life span of the individual).
  • Because of the differences in adenoma growth assumptions, the two models differ in terms of dwell time, defined as the average time that a cancer and its precursor adenoma lesion were present before a cancer was clinically diagnosed. The SimCRC model has a longer dwell time than the MISCAN model. One effect of this difference is that it takes longer for changes in risk factors to appear in mortality projections.
  • In the SimCRC model, a risk factor can influence adenoma onset and progression. In the MISCAN model, risk factors influence adenoma onset only.

Return to top

Model Details

Natural History

The simulations model development of CRC according to the disease free –> adenoma –> carcinoma sequence as illustrated below [Morson, 1974; Muto et al., 1975]. Development of an adenoma depends on age, sex, race, genetic and other propensity and risk factors. Location in the colon is an attribute of a given adenoma, which becomes a factor in the disease progression and the chance that it will be found by screening.

Natural history diagram of CRC cancer showing that risk factors and screening influence progression at the adenoma preclinical phase, screening influences the preclinical cancer phase progression, and chemotherapy affects the clinical cancer phase.

Adenomas progress from small (1-5 mm) to medium (6-9 mm) to large (10+ mm) size. Some adenomas eventually become malignant, transforming to stage I preclinical cancer. The cancer then has a chance of progressing through the stages (from stage I to stage IV). Cancer is considered preclinical if it has not yet been detected. Once detected, it becomes clinical (and treatment begins).

Little is known about how fast this progression occurs. It is estimated that 30% to 50% of the population have one or more adenomas, but it is difficult to measure dwell time in a real population because, by definition, it is the period during which the condition is undiagnosed.

Current long-term endoscopy studies will provide us more information about dwell time, but results from these studies (Atkin, PLCO, SCORE) are not yet available. We are continuing to monitor the latest research to determine whether we need to adjust either or both models.

Return to top

Risk Factors

Both models incorporate the effects of eight modifiable risk factors associated with CRC:

  • Smoking
  • Obesity (BMI)
  • Physical activity
  • Fruit and vegetable intake
  • Red meat intake
  • Multivitamin use (proxy for folate intake)
  • Aspirin/non-steroidal anti-inflammatory (NSAID) use
  • Postmenopausal hormone replacement therapy (HRT) (women only)

The prevalence of risk factors in the population is based on the National Health and Nutrition Examination Survey (NHANES) and varies according to the scenario being run. A risk profile is assigned to each simulated individual in the population. This assignment is random but reflects the overall prevalence of risk factors for the particular scenario and the individual’s age and birth year.

Risk factors affect an individual’s chance of developing a pre-cancerous adenoma. Each risk factor is assigned a number above or below 1, reflecting the magnitude of the risk or benefit of that behavior. These factors are applied when determining whether the individual will develop an adenoma. In the SimCRC model, risk factors also affect the chance that an adenoma will transform into stage I cancer.

Return to top


Both models incorporate the commonly used screening tests for CRC:

  • fecal occult blood test (FOBT)
  • flexible sigmoidoscopy
  • colonoscopy

Whether or not a simulated individual gets a screening test, and which test he/she gets, is determined by dissemination characteristics taken from National Health Interview Study (NHIS) data and the scenario being run.

Screening can detect both adenomas and preclinical cancer. Size and location, as well as the sensitivity of the test, play a part in determining whether a given screening test will detect a given cancer or lesion. Pre-cancerous adenomas may be removed when detected, and therefore do not develop into cancer.

Return to top


The probability that a simulated person with a new diagnosis of colorectal cancer receives chemotherapy is modeled as a function of stage at diagnosis, age, sex, race and calendar year and is dependent on the scenario being run. An individual with CRC may die from the illness or from other causes. Chemotherapy decreases the chance that the individual will die of CRC, by a percentage specific to the stage of the cancer, the chemotherapy regimen and the time since diagnosis.

Chemotherapy patterns are based on analyses of theSEER-Medicare linked dataset for patients diagnosed 1991-2002, and are extrapolated for patients aged less than 65 years at diagnosis, based on patterns of care studies.

Return to top


Model Profiles (https://cisnet.cancer.gov/resources/profiles.html) - standardized descriptions of simulation models developed by each CISNET consortium member to aid comparison of models and their results.

Return to top

NIH… Turning Discovery Into Health®