- Linkage Home
- Cohort Consortium Flourishing as It Enters Ninth Year
- Norman Boyd Visits as an HREB Distinguished Lecturer
- New Tool Predicts Colorectal Cancer Risk
- NCI Director's Innovation Awards
- DCEG Welcomes Visiting Scholar Norman Breslow
- Changing Guard at The Institutional Review Board
- New Immunity and Inflammation Planning Group
- Robert Hoover Testifies on Cell Phone Use
- Stephen Rappaport Visits DCEG as Distinguished Lecturer
- NIH Research Festival
- Neil Caporaso Honored for Collaboration
- Frontiers in Cancer Prevention Meeting
- The All-Ireland Cancer Conference
- DCEG Fellows Award for Research Excellence
- New Senior Investigators
- Dyskeratosis Congenita: First NIH Clinical Research Workshop
- Scientific Highlights
- DCEG People in the News
- Kelly Bolton Receives Nick Day Prize
- Comings...Goings
- Ihor Masnyk Retires
New Tool Predicts Colorectal Cancer Risk
The home page for the Colorectal Cancer Risk Assessment Tool, at http://wwwqa.cancer.gov/colorectalcancerrisk
In 2008, nearly 149,000 Americans were diagnosed with colorectal cancer and almost 50,000 died from the disease, making it the third most commonly diagnosed cancer and the third leading cause of cancer mortality. Continuing the DCEG tradition of developing cancer risk prediction models, such as the “Gail model” for breast cancer, NCI staff has played a major role in developing the first tool that provides an absolute estimate of colorectal cancer risk in the U.S. population. The details of the statistical model appeared online in the Journal of Clinical Oncology in December. Using a respondent’s answers to a few simple questions, this Web-based interactive tool estimates an individual’s 5-year, 10-year, and lifetime risk of developing colorectal cancer.
Ruth M. Pfeiffer, Ph.D., a senior investigator in the Biostatistics Branch and senior author on the model’s development and validation studies, explained that such a tool offers a variety of applications. “The model might be used by researchers to determine sample sizes and eligibility criteria for screening and prevention trials. The tool can also assist physicians in counseling their patients. Anybody can go to the Web site and plug in their individual data, but interpreting the results is best done in consultation with a health care professional.”
Estimates of relative risks for the model were obtained using data from two large population-based case-control studies. “Because colorectal cancer occurs at three sites—the proximal colon, distal colon, and rectum—each with a different incidence rate, we assumed that some risk factors are probably different,” explained Dr. Andrew Freedman, an epidemiologist in the NCI Division of Cancer Control and Population Sciences and lead author on the model’s development.
“We created three different models for each sex, one for each site, and then combined them,” he said. “We looked at all the known risk factors and picked out those that were most predictive.” Baseline age-specific cancer hazard rates were estimated using NCI’s Surveillance Epidemiology and End Results Program incident rates from 1992 to 2002, so the model is broadly applicable to the U.S. population.
The final step was to develop and refine a short, self-administered risk-assessment questionnaire that captures the information used in the models.
For men, the most pertinent factors affecting risk were a cancer-negative colonoscopy in the previous 10 years, history of colorectal cancer in first-degree relatives, a history of colon polyps, regular use of aspirin or other nonsteroidal anti-inflammatory drugs, cigarette smoking, body mass index, vigorous leisure-time activity, and vegetable consumption. Similar risk factors were seen in women, with the addition of hormone replacement therapy. After a user provides information about each of these factors, “the tool uses an algorithm to calculate the absolute risk, based on age and sex of the individual,” Dr. Freedman explained.
Use of the model is currently limited to non-Hispanic white men and women aged 50 and older. “We had no data on individuals younger than 50 years, so the model starts at age 50, and we also had no data on minorities to provide reliable estimates of risks in these populations,” Dr. Pfeiffer noted. “But efforts are ongoing to expand the model to include African Americans and other racial and ethnic groups.”
The next step was to validate the model using an independent population. Yikyung Park, Sc.D., a staff scientist in the Nutritional Epidemiology Branch and lead author on the validation study, expressed her excitement about using the prospective NIH-AARP Diet and Health Study cohort for this purpose. “This was an excellent use of this cohort, which has so many possible uses.” She explained, “We validated the model in two ways. First, we used it to predict how many cases we would expect to find in the study population and compared that number to the actual number of cases observed in the cohort. We found the fit to be very good. Second, we estimated the model’s ability to predict risk on an individual level and found it to be modest, but similar to that of other absolute risk prediction cancer models.”
Dr. Pfeiffer explained, “It can predict the correct number of events overall but it cannot predict well which specific individuals will get cancer.” She further noted that this is not a screening tool. “It can help a physician decide whether to screen or not, but it cannot replace a colonoscopy.”
Dr. Freedman said that creating the tool was “a tremendous amount of work, but we were lucky to have an interdisciplinary collaborative team that worked very hard on this project. We all realized the importance this new tool will have for public health.”
The Web-based tool can be found at http://wwwqa.cancer.gov/colorectalcancerrisk. Similar tools developed by NCI investigators can be found for melanoma (www.cancer.gov/melanomarisktool), I-131 exposure (http://ntsi131.nci.nih.gov), and breast cancer (www.cancer.gov/bcrisktool).
—Terry Taylor, M.A.
