by David Hinds
A small study conducted by researchers at the Cleveland Clinic and reported at last week’s American Society of Human Genetics annual meeting found that family health history (FH) often gave different assessments of cancer risk than SNP-based personal genomic (PG) risk assessments obtained from Navigenics. The results have been characterized in the popular press as demonstrating that FH is superior to PG in predicting who is at increased risk of disease. We believe that research in this area is important, but felt that the conclusions of the study were not necessarily supported by the study’s design.
What is family history risk assessment?
An identifying feature of many genetic diseases is that they tend to “run in families”. Families also tend to share similar environmental risk factors. Collecting health information about family members can thus provide important clues about an individual’s risk of disease. Family history risk assessment is typically performed by genetic counselors using structured methods so that the results can be easily compared with guidelines for relating history to disease risk. There are also online tools for recording family histories, and some disease risk calculators can take advantage of this information.
Family history is known to be especially helpful for identifying individuals at substantially elevated risk of disease because their families carry rare, high-risk variants (such as mutations in BRCA1 and BRCA2, for breast and ovarian cancer).
What is SNP-based risk assessment?
SNP-based personal genomic risk assessment is currently based on results of studies demonstrating that specific common genetic variants are correlated with disease risk. These common variants generally have small effects, but in aggregate, may contribute to more cases of disease in the population than rare variants like BRCA1/2. Navigenics and 23andMe use similar methods for combining information from multiple markers to arrive at an estimate of an individual’s risk. As with family history, SNP-based tests only account for a subset of risk factors, and are most informative for the subset of individuals whose estimated risk is substantially different from the population average.
Some personal genomics services, such as 23andMe, include variants other than common SNPs in their risk assessments. In the future, personal genomic risk assessments may routinely apply technologies beyond SNP-based genotyping to form a more comprehensive picture of an individual’s genetic landscape.
What did the study show?
The researchers identified 22 patients in a genomic medicine clinic who either had cancer themselves or a family history of cancer, and their spouses. All 44 received FH assessments, as well as the Navigenics SNP-based PG risk assessment. There was generally poor agreement between the FH results and PG results. In addition, five of the 22 patients were later found to carry high risk mutations for an inherited form of colon cancer, but were not identified as high risk by the PG test.
What does this really mean?
The inconsistency between FH and PG observed in this study is largely due to a fundamental difference between the two types of tests. Family history is most effective at identifying individuals with familial cancer syndromes related to rare mutations with large effects, while SNP-based tests use common variations with smaller effects. The two types of tests are thus measuring very different things.
The two methods also may not be assessing the same underlying conditions. There is no doubt that individuals with hereditary nonpolyposis colorectal cancer or familial adenomatous polyposis have a different spectrum of disease than typical colon cancer patients. Familial syndromes also represent only a small proportion of all cancers (5% of all colon cancers).
Furthermore, the Cleveland Clinic study started from the presumption that FH was the “gold standard”, so any deviations in the PG results were considered errors. In our opinion, this is not an appropriate starting point for performing a head-to-head assessment of test performance. Such an assessment would, ideally, evaluate the accuracy of the two tests in a set of individuals not previously screened for a positive family history, as some of the participants in this study were.
Perhaps the most troubling suggestion from the study was that individuals at high risk based on family history might choose a PG test, and be misled by a result indicating they were at low risk. 23andMe disease risk reports all discuss the limitations of the tests and the importance of family history as well as other environmental and undiscovered genetic risk factors. The suggestion also neglects any benefits of PG testing for the large majority of individuals who do not have a positive family history of disease.
Where do we go from here?
At 23andMe, we are committed to increasing scientific knowledge of the genetic basis of human diseases and traits, and we are conducting research to evaluate and improve the predictive accuracy of our genetic risk models. As we learn more about the genetic factors underlying disease, genetic information should become increasingly valuable in developing risk assessments, and will be complementary to methods such as family history. Ultimately, we believe that the path forward will involve integrated, individualized risk assessment based on family history, personal genetics, and environmental risk factors.