Colorectal Cancer

The CISNET Colorectal Working Group (CWG) consists of three modeling groups (MISCAN, SIM-CRC and CRC-SPIN) and a Coordinating Center. The investigators are focused on the long-term objective of reducing the impact of colorectal cancer (CRC) by providing the information needed to address key policy questions and prioritize future research. The models describe the natural history of colorectal cancer using different underlying structures, which enables model comparisons that explore the impact of structural uncertainty. Despite large increases in screening in the past two decades, colorectal cancer remains the second leading cause of cancer death in the United States. The CWG’s research has shown that more than half of these deaths could be prevented by better use of available screening interventions. In this cycle, the CWG will continue to use microsimulation modeling to evaluate and help prioritize interventions to further reduce the burden of colorectal cancer.

Highlights

The MISCAN model led a manuscript titled “Effectiveness and Cost-Effectiveness of Colorectal Cancer Screening With a Blood Test That Meets the Centers for Medicare & Medicaid Services Coverage Decision” was accepted for publication in Gastroenterology.

The CRC-SPIN team led an analysis which was published in eLife in 2023 entitled, “Projected Long-Term Effects Of Colorectal Cancer Screening Disruptions Following the COVID-19 Pandemic.”

The SIM-CRC team led an analysis which calibrated the natural history model of CRC to simulated epidemiological data with different degrees of uncertainty and obtained the joint posterior distribution of the parameters using a Bayesian approach.

Investigators

Comparative Modeling of Colorectal Cancer: Informing Health Policies and Prioritizing Future Research


Principal Investigator:

Memorial Sloan Kettering Cancer Center

University of Utah Health
Co-Principal Investigators:

Erasmus MC

Massachusetts General Hospital

University of Minnesota

Stanford University

Fred Hutchinson Cancer Research Center
Co-Investigators:

University of Chicago, Argonne National Laboratory

University of Chicago, Argonne National Laboratory

RAND

Grant Number: U01CA253913

Abstract & Aims

Abstract: Colorectal cancer (CRC) is the second leading cause of cancer death in the United States. The long-term goal of our project is to reduce the population burden of CRC by providing the information needed to address key policy questions across the CRC control continuum in an accessible and transparent manner.

To accomplish this goal we will use our population-based microsimulation models to:

  1. Evaluate the impact of screening as practiced in the US;
  2. Inform the debate about the increase in CRC incidence before age 50;
  3. Consider the effectiveness of precision of screening and surveillance;
  4. Address other emerging issues and opportunities in CRC control such as genomic biomarkers; and
  5. Use novel methods to improve model accessibility and transparency.

Our team will fill critical gaps in knowledge, enabling decision makers to act.
New evidence that we will incorporate in our models to better inform CRC control opportunities will be:

  1. Updated information on screening patterns in the US (in collaboration with the Population-based Research Optimizing Screening through Personalized Regimen, or PROSPR),
  2. Data on the increased risk of CRC in persons under age 50 (in collaboration with Rebecca Siegal of the American Cancer Society, who did the seminal work in this area), and
  3. State-of-the art colonoscopy screening data to incorporate alternative carcinogenesis pathways in the natural history models (in collaboration with the New Hampshire Colonoscopy Registry).

We will synthesize and incorporate the growing body of evidence in the literature to assess the clinical utility of personalized screening and treatment, as well as the potential role for novel computer-aided detection and diagnosis modalities. We will expand our models to project clinical and resource-based outcomes for middle-income countries that are considering the implementation of a screening program. Lastly, there is a critical need to make our models assessable and transparent. To this end we will use high performance computing approaches to develop and apply deep learning methods for model calibration and model emulation, which will aid in model sharing. The three participating modeling groups are well positioned to carry out this work, bringing a wealth of experience, expertise, and insight to issues related to microsimulation modeling of CRC, and have a proven track record of collaboration and disseminating our work to health policy decision makers.

Major Analyses & Contributions

Comparative Analyses

Cost Effectiveness of Stool DNA Screening for Colorectal Cancer

Stool DNA testing for the screening of colorectal cancer represents one of the new screening approaches which has been considered by the Centers for Medicare & Medicaid Services, often the first adopter of innovative screening tests.  CISNET modelers utilizing two different models analyzed the current conditions for which stool DNA could be cost-effective compared to current, reimbursable screening tests.  At the current estimated cost and a screening interval of every 3 5 years, modelers found stool DNA tests yielded fewer life-years and higher costs than established, recommended screening methods in both a cohort of CMS persons as well as those of average screening age, 50 years and older.  (Lansdorp-Vogelaar 2010).  Until overall cost of stool DNA screening decreases, this screening strategy does not represent a cost-effective alternative to the majority of currently reimbursable screening tests.

Using Microsimulation modeling to Inform Health Policy: Estimation of benefits, burden, and harms of CRC screening strategies for the US Preventive Services Task Force

The US Preventive Services Task Force requested a decision analysis from the CISNET colorectal working group to inform their updated recommendation process for CRC screening. This is the second time CISNET work has been used in the USPSTF recommendation process, and it was evaluated alongside a systematic review of clinical evidence. The CISNET team modeled 204 unique screening strategies, including ages to begin and end screening, screening intervals, and screening modality. CISNET models found that for screening between ages 50 and 75, there were four model-recommendable strategies: colonoscopy every 10 years, annual FIT, sigmoidoscopy every 5 years combined with annual FIT, and CT-colonography every 5 years. The supporting modeling paper points out the need for further clinical evidence on an earlier start age for CRC screening and test-specific adherence over multiple rounds of screening. Overall, however, while the models differed slightly in terms of absolute outcomes, they yielded consistent relative predictions across screening modalities and similar rankings within the different classes of modalities. This lends strength to the argument that model results are a valid source of data for informing health policy (Knudsen 2016). An interactive web resource for this paper can be found at: https://resources.cisnet.cancer.gov/projects/#crcr/uspstf/abstract.

Model Validation: accuracy of model predictions based on randomized trial data and implications for use

Models require assumptions about unobservable processes; in the case of CRC, this includes inputs of the natural history of disease. Because of these assumptions, it is important to assess a model's predictive accuracy based on clinical trial data. The CISNET group decided to validate all three CRC models against outcomes from the United Kingdom Flexible Sigmoidoscopy Screening (UKFSS) Trial, which was a randomized trial examining one-time flexible sigmoidoscopy. In this validation exercise, all three models accurately predicted the relative reduction in CRC mortality 10 years after screening. 10-year CRC incidence was accurately predicted in the two models with longer average preclinical duration. The MISCAN model underestimated the incidence reduction due to shorter dwell time estimates, and has since been recalibrated to yield predictions that are consistent with both mortality and incidence reductions. This exercise proved that model accuracy can be established only through external validation, and that such exercises are essential for the accuracy of any decision model (Rutter 2016).

Modeling Benefits and Harms of Screening Cessation Ages Based on Comorbid Conditions

The US Preventive Services Task Force has recommended against screening past the age of 74 for breast and colorectal cancers. However, as life expectancy increases and people are living longer, this may no longer be an effective screening stop age for everyone. In order to determine whether we could risk stratify screening cessation age, we modeled the benefits and harms of screening based on comorbid conditions. This comparative analysis was unique in that we modeled for three CISNET cancer sites: CRC, breast, and prostate. We found that comorbid conditions are an important determinant of both benefits and harms of screening. For persons with no, mild, moderate, and severe comorbid conditions, screening until ages 76, 74, 72, and 66 years, respectively, resulted in harms and benefits similar to average-health persons. These findings may help tailor personalized screening regimens for persons with differential comorbid conditions who want to continue screening past the recommended stop age of 74. (Lansdorp-Vogelaar 2014).

Other Achievements

Colorectal Cancer Other Achievements: Highlights

When is it appropriate to screen someone older than 75 for colorectal cancer?

The US Preventive Services Task Force currently recommends screening for average-risk individuals from ages 50-75. Benefit after the age of 75 was given a C rating, suggesting that screening at older ages should be based on professional judgment and patient preference. To determine whether there might be a way to personalize colonoscopy screening for elderly individuals, van Hees et al. used the MISCAN microsimulation model to test factors that might contribute to differential risk. They found that less intensive screening history, higher background risk for colorectal cancer, and fewer comorbidities were associated with cost-effective screening at older ages. It would therefore be more effective and cost-effective if we were able to risk stratify patients by these factors rather than basing screening decisions on age alone, which can be inefficient in older ages (van Hees 2015). In addition, in unscreened elderly persons with no comorbid conditions, colorectal cancer screening is always cost-effective up to age 86; with moderate comorbid conditions to age 83; and with severe comorbid conditions to age 80. Therefore, colorectal cancer screening should always be considered beyond age 75 for previously unscreened individuals (van Hees 2014).

What are some of the issues with colorectal cancer screening in Medicare beneficiaries?

The amount and intensity of colorectal cancer screening for Medicare beneficiaries depends largely on a person's pre-Medicare screening history. Some individuals never get screened before they become Medicare-eligible, while others screen more intensively than necessary. We addressed both of these issues using our CISNET models. First, Goede et al. demonstrated that increased colorectal cancer screening in the pre-Medicare population could reduce both colorectal cancer incidence and mortality when the recipient enters Medicare age, and that the additional screening costs would be largely offset by long-term Medicare treatment savings (Goede 2015). In addition, van Hees led a study on the appropriateness of more intensive colonoscopy screening than recommended in Medicare beneficiaries. They found that this is an unfavorable screening strategy both from a societal and personal level. This practice should therefore be highly discouraged (van Hees 2014).

How would increased adherence rates for screening affect colorectal cancer incidence and mortality?

The National Colorectal Cancer Roundtable (NCCRT) is a national coalition of organizations aimed at reducing colorectal cancer incidence and mortality in the US. The NCCRT announced an initiative aiming to increase colorectal cancer screening rates in the US to 80% by 2018. Current estimates of adherence to colorectal cancer screening are between 65-69%. Meester et al. used the MISCAN model to estimate the public health benefits of achieving the 80% by 2018 goal. They found that this would have a considerable impact, averting approximately 280,000 new cancer cases and 200,000 cancer deaths in less than 20 years (Meester 2015).

How should we rescreen individuals with an initial negative colonoscopy?

Colonoscopy is a recommended test for colorectal cancer screening, as well as being used for follow-up of individuals with positive results on other screening tests. However, colonoscopy can cause complications and is also more resource-intensive than many of the other colorectal cancer screening methods. Understanding this, Knudsen et al. used the SimCRC model to evaluate alternative management strategies for 50-year-olds with neither adenomas nor colorectal cancer at their initial screening colonoscopy. They found that rescreening with any sort of test was more effective than not rescreening at all. Furthermore, they found that rescreening at age 60 (the recommended 10-year interval after a negative colonoscopy result at age 50) with annual high-sensitivity gFOBT or FIT, or CTC every 5 years provides approximately the same benefit in terms of life-years gained, with fewer complications and at a lower cost than rescreening with colonoscopy. Therefore, they concluded that it might be reasonable to use other methods to rescreen individuals with negative initial colonoscopy results (Knudsen 2012).

How can we address state disparities in screening for colorectal cancer?

There are many state-based disparities in colorectal cancer incidence and mortality across the US. For example, Northeastern states have shown more progress in reducing colorectal cancer incidence and mortality rates than Southern states. In addition, several states across the US are implementing initiatives to provide access to colorectal cancer screening for their low-income, uninsured populations. However, states differ in risk factors, budgets, and screening rates. The CISNET team has decided to tackle some of these issues using our microsimulation models of colorectal cancer. In one paper, Lansdorp-Vogelaar et al. used the cases of New Jersey and Louisiana as examples to evaluate some possible causes in state-based differences in colorectal cancer. They found that these disparities could be eliminated if Louisiana could attain New Jersey's level of risk factors, screening, and survival (Lansdorp-Vogelaar 2015). Using South Carolina as an example, van der Steen et al. assessed which screening test would be best for a state-based colorectal cancer screening initiative with a limited budget. They found that a FIT-based screening program would prevent more colorectal cancer deaths than a colonoscopy-based program. This is especially true when a state's budget supports screening of only a fraction of the target population. For example, in South Carolina, using a FIT-based program resulted in nearly eight times more individuals being screening and approximately four times as many colorectal cancer deaths prevented and life-years gained than the colonoscopy program (van der Steen 2015).

Model Profiles & Registry

Model profiles are standardized documents that facilitate the comparison of models and their results. The Joint Profile provided includes profiles for all colorectal cancer models. Individual profiles for each model are also provided and may be more current than the joint profile document.

The following model profiles have been developed by CISNET members for colorectal cancer:

For additional high-level information about the colorectal models visit the colorectal overview page on the CISNET Model Registry.