A search of peer reviewed literature through August 2011 was completed, with the following results, for (SNP) testing for breast cancer risk assessment.
Genome-wide association studies (GWAS) examine the entire genome of each of thousands of individuals for single nucleotide polymorphisms (SNPs), single base-pair variations in the DNA sequence at semi-regular intervals, and attempt to associate variant single nucleotide polymorphism (SNP) alleles with particular diseases. Several case-control GWAS have been carried out, primarily in white women, to investigate common risk markers of breast cancer. In recent years, a number of SNPs associated with breast cancer have been reported at a high level of statistical significance and validated in two or more large, independent studies. Recently SNPs associated with breast cancer risk in Asian women have been the subject of more than a half-dozen articles, although these appear exploratory.
Estimates of breast cancer risk, based on SNPs derived from large GWAS and/or from SNPs in other genes known to be associated with breast cancer, are available as laboratory-developed test services from different companies. There is growing literature on these associations although public information on the actual models being offered commercially is sparse. Independent determination of clinical validity in an intended use population to demonstrate clinical validity has not been performed. There are also no studies to suggest that use of SNP-based risk assessment has any impact on health care outcomes.
No peer-reviewed reports have been published in which these commercially available breast cancer risk estimators have been compared to each other to determine if they report similar results on the same individuals specifically for breast cancer. In July 2008, deCODE, 23andme, and Navigenics agreed to work with the Personalized Medicine Coalition (PMC) on a set of standards regarding the scientific validity of their genotyping panels; in the process test individuals were genotyped for three disease associations, but the PMC provides actual information on only one (breast cancer) with very little detail.
Since there are no published studies of commercial SNP-based breast cancer risk predictors, other published studies of the clinical usefulness of other similar combinations of SNPs as risk predictors will be considered here. In 2008, Pharoah et al. considered a combination of seven well-validated SNPs associated with breast cancer, five of which are included in the deCODE BreastCancer test. A model that simply multiplies the individual risks of the seven common SNPs was assumed, and would explain approximately 5% of the total genetic risk of non-familial breast cancer. Applying the model to the population of women in the United Kingdom, the authors concluded that the risk profile provided by the seven SNPs would not provide sufficient discrimination between those who would and would not experience future breast cancer to enable individualized preventive treatment such as tamoxifen. However, the authors did consider the effect on a population screening program that could be personalized with the results of SNP panel testing. They concluded that no women would be included in the high-risk category (currently defined as 20% risk within the next ten years at age 40–49, according to the National Institute for Health and Clinical Excellence), and therefore none would warrant the addition of MRI screening or the consideration of more aggressive intervention on the basis of the SNP panel results.
Wacholder et al. evaluated the performance of a panel of ten SNPs with established associations with breast cancer that had, at the time of the study, been validated in at least three published GWAS. Cases (n=5,590) and controls (n=5,998) from the National Cancer Institute’s Cancer Genetic Markers of Susceptibility GWAS of breast cancer were included in the study (women of primarily European ancestry). The panel contained five SNPs included in the deCODE BreastCancer test. The SNP panel was examined as a risk predictor alone and in addition to readily available components of the Gail model (minus mammographic density and diagnosis of atypical hyperplasia). The authors found that adding the SNP panel to the Gail model resulted in slightly better stratification of a women’s risk than either the SNP panel or the Gail model alone, but that this stratification was not adequate to inform clinical practice. For example, only 34% of the women who actually had breast cancer were actually assigned to the top 20% risk group. The area under the curve (AUC) for the combined SNP and Gail model was 61.8% (50% is random, 100% is perfect).
Reeves et al. evaluated the performance of a panel of seven SNPs with established associations with breast cancer in a study of 10,306 women with breast cancer and 10,383 without cancer in the United Kingdom. The risk panel also contained five SNPs included in the deCODE BreastCancer test and used a similar multiplicative approach. Sensitivity studies were performed using only four SNPs and using ten SNPs, both demonstrating no significant change in performance. While use of the risk score was able to show marked differences in risk between the upper quintile of patients (8.8% cumulative risk to age 70) and the lower quintile of patients (4.4%) these changes were not viewed as clinically useful when compared to patients with an estimated overall background risk of 6.3%. Of note, simple information on patient histories; for example presence of one or two first-degree relatives with breast cancer provided equivalent or superior risk discrimination (9.1% and 15.4% respectively).
Finally, Mealiffe et al. evaluated a seven SNP panel in a nested case-control cohort of 1664 case patients and 1636 controls. Again a multiplicative model was used and, as in the study by Wacholder et al., the genetic risk score was reviewed as a potential replacement for or add-on test to the Gail clinical risk model. These authors employed a relatively new statistic (the net reclassification improvement [NRI]) to evaluate performance. While they concluded that statistically significant improvements could be observed by addition of the genomic risk assessment to the Gail clinical risk assessment, they were unable to posit or demonstrate the observed changes would lead to improved clinical outcomes. They suggested further studies were needed and that benefit might be observed by careful selection of patients (e.g., those who on Gail score analysis exhibited intermediate risk) who might comprise a priori be candidates who would benefit from enhanced or improved risk assessment.
It is assumed that many more risk factors remain to be discovered as the majority of the genetic risk of breast cancer has not been explained by known gene variants and SNPs. One reason more genetic associations have not been found is that even large GWAS are underpowered to detect uncommon genetic variants. For example, known, uncommon variants in the ataxia telangiectasia mutated (ATM) and cell cycle checkpoint kinase2 (CHEK2) genes have not been detected by GWAS, even though they are strongly associated with breast cancer. Additionally, pooled GWAS data have allowed the discovery of significant SNPs that were not significant in the individual studies.
Although there are no guidelines regarding the clinical use of SNP panels for estimating breast cancer risk, the published literature is in general agreement that their use in clinical or screening settings is premature due to a lack of a more complete set of explanatory gene variants and to insufficient discriminatory power at this time. Whether or not additional SNP studies are likely to be informative is under debate, as the study size to detect more and more rare variants becomes prohibitively large. As the cost of whole genome sequencing continues to decrease, some predict that this will become the preferred avenue for researching risk variants. One challenge in sorting through the growing literature on this diagnostic approach is nonstandardization and nontransparency of studies. Janssens et al. have recently published a methods paper providing a road map for optimal reporting and an accompanying detailed article describing good reporting practices.
Recently, Bloss et al. reported on the psychological, behavioral and clinical effects of risk scanning in 3,639 subjects followed for a short-term period (mean of 5.6 months; SD [standard deviation] of 2.4 months). These investigators evaluated the patients’ anxiety, intake of dietary fat, and exercise based on information from genomic testing. They concluded there were no significant changes before and after testing. They also noted no increase in the number of screening tests obtained in enrolled patients. While more than half of patients participating in the study indicated the intent to have screening tests performed in the future, during the course of the study itself, there was no actual increase observed.
Common, SNPs have been shown to be significantly associated with breast cancer and to individually convey slightly elevated risk of breast cancer compared to the general population risk. Panels of well-documented and validated SNPs are commercially available, with results synthesized into breast cancer risk estimate that have not been clinically validated and for which clinical utility (in spite of potential statistical predictive ability) is questionable. The majority of these tests are commercially available as direct-to-consumer (DTC) tests. The application of such risk panels to individual patient management or to population screening programs is premature due to the uncertain performance of these profiles in the intended use populations and the expectation that the majority of the genetic risk of breast cancer has yet to be explained by undiscovered gene variants and SNPs. The discrimination offered by the limited genetic factors currently known is insufficient to inform clinical practice. Therefore, the use of this testing is considered experimental, investigational and unproven.
Disclaimer for coding information on Medical Policies
Procedure and diagnosis codes on Medical Policy documents are included only as a general reference tool for each policy. They may not be all-inclusive.
The presence or absence of procedure, service, supply, device or diagnosis codes in a Medical Policy document has no relevance for determination of benefit coverage for members or reimbursement for providers. Only the written coverage position in a medical policy should be used for such determinations.
Benefit coverage determinations based on written Medical Policy coverage positions must include review of the member’s benefit contract or Summary Plan Description (SPD) for defined coverage vs. non-coverage, benefit exclusions, and benefit limitations such as dollar or duration caps.