Methods of quantifying cancer risk
The overarching goal of cancer risk assessment is to individualize cancer risk management recommendations based on personalized risk. Methods to calculate risk utilize health history information and risk factor and family history data often in combination with emerging biologic and genetic/genomic evidence to establish predictions. Multiple methodologies are used to calculate risk, including statistical models, prevalence data from specific populations, penetrance data when a documented deleterious mutation has been identified in a family, Mendelian inheritance, and Bayesian analysis. All models have distinct capabilities, weaknesses, and limitations based on the methodology, sample size, and/or population used to create the model. Methods to individually quantify risk encompass two primary areas: the probability of harboring a deleterious mutation in a cancer susceptibility gene and the risk of developing a specific form of cancer.
Risk of harboring a deleterious mutation in a cancer susceptibility gene
The decision to offer genetic testing for cancer susceptibility is complex and can be aided in part by objectively assessing an individual's and/or family's probability of harboring a mutation. Predicting the probability of harboring a mutation in a cancer susceptibility gene can be done using several strategies, including empiric data, statistical models, population prevalence data, Mendel's laws, Bayesian analysis, and specific health information, such as tumor-specific features.[60,61] All of these methods are gene specific or cancer-syndrome specific and are employed only after a thorough assessment has been completed and genetic differential diagnoses have been established.
If a gene or hereditary cancer syndrome is suspected, models specific to that disorder can be used to determine whether genetic testing may be informative. (Refer to the PDQ summaries on the Genetics of Breast and Ovarian Cancer; Genetics of Colorectal Cancer; or the Genetics of Skin Cancer for more information about cancer syndrome-specific probability models.) The key to using specific models or prevalence data is to apply the model or statistics only in the population best suited for its use. For instance, a model or prevalence data derived from a population study of individuals older than 35 years may not accurately be applied in a population aged 35 years and younger. Care must be taken when interpreting the data obtained from various risk models because they differ with regard to what is actually being estimated. Some models estimate the risk of a mutation being present in the family; others estimate the risk of a mutation being present in the individual being counseled. Some models estimate the risk of specific cancers developing in an individual, while others estimate more than one of the data above. (Refer to NCI's Risk Prediction Models Web site or the disease-specific PDQ cancer genetics summaries for more information about specific cancer risk prediction and mutation probability models.) Other important considerations include critical family constructs, which can significantly impact model reliability, such as small family size or male-dominated families when the cancer risks are predominately female in origin, adoption, and early deaths from other causes.[61,62] In addition, most models provide gene and/or syndrome-specific probabilities but do not account for the possibility that the personal and/or family history of cancer may be conferred by an as-yet-unidentified cancer susceptibility gene. In the absence of a documented mutation in the family, critical assessment of the personal and family history is essential in determining the usefulness and limitations of probability estimates used to aid in the decisions regarding indications for genetic testing.[60,61,63]