Defining the Clinical Utility of Gene Expression Assays in Breast Cancer: The Intersection of Science and Art in Clinical Decision Making

The “art of medicine” is a commonly used phrase, acknowledging that although modern medical practice involves the application of science, the application of scientific principles may vary widely. For example, although there is unequivocal scientific evidence that adjuvant chemotherapy reduces recurrence and mortality in early-stage breast cancer, clinicians often weight multiple factors when considering an adjuvant chemotherapy recommendation. The conundrum often faced by clinicians and patients about the role of chemotherapy in early-stage breast cancer is reflected by the musing of Yogi Berra, well known for his skills both as a New York Yankee and unwitting philosopher—although it is clear that a path must be chosen, it is unclear which path that should be. Enhanced knowledge of the biology of breast cancer, and technological advances which facilitate application of that knowledge in clinical specimens, offers potential to empower clinicians and patients toward making more informed decisions. Toward that end, it has been nearly a decade since distinct breast cancer genotypes were first characterized, and 5 years since the first multiparameter gene expression assays became commercially available for early-stage breast cancer, including a 70-gene assay and a 21-gene assay. Numerous other multiparameter assays have been developed, and several others are also commercially available. Some evidence suggests that these assays provide comparable prognostic information. Although there is compelling evidence regarding the clinical validity of these assays in providing prognostic information about distant disease recurrence, there is little information about their clinical utility—that is, how does the test result influence clinical decision making, and do patients benefit from that change? In the current issue of Journal of Clinical Oncology, Lo et al provide important new information regarding the clinical utility of one of these assays, the 21-gene recurrence score (RS), and Mamounas et al provide new information regarding the clinical validity of the same assay in predicting locoregional rather than distant disease recurrence. Details regarding the 21-gene RS assay and its development were recently reviewed in JCO. Most of the validation studies focused on the relationship between RS and distant recurrence. However, isolated locoregional recurrence (LRR) is a common and important cause of morbidity and patient distress, and is known to be associated with distant recurrence and mortality. LRR is also one of the elements contributing to the disease-free survival end point commonly used in clinical trials (which also includes distant recurrence, contralateral breast cancer, and second primary cancers). In low-risk patients with estrogen receptor (ER) –positive, lymph node–negative disease treated with breast conservation and tamoxifen, isolated LRR occurs in approximately 8% of all patients at 10 years, and accounts for up to 40% of all recurrence events. Up to one half of LRRs will never be associated with distant recurrence, and would therefore not be reflected in the current recurrence estimates predicted by the RS. In this issue of JCO, Mamounas et al from the National Surgical Adjuvant Breast and Bowel Project provide new information regarding the relationship between RS and LRR after breast conservation or mastectomy in a low risk population with ER-positive, lymph node– negative disease. RS was an independent significant predictor of LRR, as were age and type of initial locoregional treatment (mastectomy v breast conservation treatment). The relationship between RS and LRR was driven largely by a strong association among women treated with tamoxifen younger than 50 years irrespective of the type of local therapy, and in older patients treated by mastectomy (but not breast conservation). The authors cautiously and appropriately conclude that these findings should be regarded as hypothesis generating, which require validation in other trials. Nevertheless, because of known association between distant and local recurrence, this information confirms our suspicion that the biologic features driving distant recurrence reflected by specific genomic signatures are also predictive of LRR. Indeed, another report has likewise provided evidence that another genomic signature predictive of distant recurrence is also predictive of LRR. Multiparameter assays have thus far been used in clinical practice primarily to allow physicians and patients to make more informed decisions about adjuvant chemotherapy, which may be explained by several factors. First, unlike ER and progesterone receptor testing for endocrine therapy and HER2 testing for anti-HER2 directed therapy, there are no clinicopathologic features predictive of chemotherapy benefit. Second, current practice guidelines recommend adjuvant chemotherapy for the majority of patients with operable breast cancer, including those at low risk of recurrence. Third, many patients are willing to accept chemotherapy even if there is a low likelihood of benefit. Given the therapeutic landscape, how can the RS assay or other multiparamater gene expression assays influence clinical decision making? Treatment sparing. In this scenario, the clinical features indicate that treatment is indicated, but the test result indicates that treatment is not likely to be beneficial and may be spared. Treatment selection. The clinical features indicate that treatment is not indicated, but the test result indicates that treatment is very likely Editorials