U.S. scientists at the National Cancer Institute in Bethesda have created a statistical formula that allows them to approximate a woman’s risk of developing cancer under a 10 year and 20 year model. What does this mean and how successful are these models?
Writing in the journal PLOS Medicine, the researchers describe how they were able to use previous research that indicates key risk factors for cancer to create a prediction of a woman’s future cancer risk.
Those variables, among others, include:
- body mass index
- number of children
- how long the woman in question took birth control pills
- use of hormone therapy to alleviate menopause symptoms
- family history of cancer
- cigarette and alcohol consumption
The researchers used data on white, non-Hispanic women all of 50 years of age or older as collected during two large U.S. cohort studies on cancer screening and diet and health, and U.S. cancer incidence and mortality rates. These were provided by the Surveillance, Epidemiology and End Results Program.
The researchers applied their formula to almost 122,000 women and began calculating their cancer risks.
Among a group of 56,638 women, the researchers forecast that 2,930 breast cancer cases would develop. The actual figure was 2,934.
That number was slightly more accurate than other assessment tools but the model performed less well in some specific areas, for instance by “significantly” underestimating breast cancer cases in premenopausal women.
Why that occurred isn’t fully understood, though, if the variable or variables that led to this discrepancy could be included in the model in the future, there’s no reason just yet to think this problem could not be overcome.
As a result, this kind of model could have significant implications for clinicians.
Lead researcher Dr Ruth Pfeiffer explains why, saying, “These models might assist in clinical decision making related to the risks of these cancers. These include designing cancer prevention trials, assessing the absolute burden of disease in the population and in sub groups and gauging the potential absolute reductions in risk from preventive strategies.”
Essentially, the statistical models could allow doctors to better ascertain what demographics are at high risk and more precisely select courses of treatment for individual patients.
Though the breast cancer model performed best, the scientists also attempted to calculate a model for ovarian cancer, which predicted a total of 406 cases against the real life figure of 377 cases, and endometrial cancer, or cancer that forms in the tissue lining the uterus — the researchers anticipated 640 cases against 532 actual diagnosed cases.
It should be noted, however, that this was the first time researchers anywhere have created a model to predict an absolute risk for endometrial cancer. The predictor models could further be used to inform wider health decisions for women and in fact empower women to be more secure about their cancer prevention and treatment choices. For instance, the drug tamoxifen is used to prevent breast cancer but has been shown to actually increase the risk of endometrial cancer.
A woman with no above average statistical likelihood of developing endometrial cancer, as calculated by these models, might therefore feel confident they are making the right decision in taking tamoxifen if their risk of breast cancer is significantly high. At the same time, those patients who also have a statistical likelihood of developing endometrial cancer may choose other measures so as to mitigate their endometrial cancer risk.
The statistical models produced here have been welcomed by wider scientists who have praised the idea of developing risk assessment models for cancer, though they caution that assessing the accuracy of those models comes with its own set of challenges.
Still, the models represent a chance at yet another tool in the fight against cancer, and for that reason are an exciting prospect.