Crossroads in Breast Cancer: The Intersection of Clinical Uncertainty and Molecular Profiling
Published Online: February 26, 2013
Kim Farina, PhD
Molecular oncology has brought a new understanding of carcinogenesis and has revolutionized the way we think about treating cancer.
Oncologists believe they are at a turning point in terms of how to use molecular assays now and in the not-toodistant future. “We are beginning to realize that we have only just scratched the surface,” said Peter Ravdin, MD, PhD, UT Health Science Center San Antonio, who chaired an educational session titled The Practical Use of Molecular Profiling at the 2012 CTCR-AACR San Antonio Breast Cancer Symposium (SABCS). “It is pretty exciting, actually.” The field is moving away from making decisions based on qualitative, descriptive prognostic information and moving toward precision medicine that is based on quantitative predictions.
Lawrence Solin, MD, Albert Einstein Medical Center, talked about molecular profiling of in situ carcinoma. Recent studies suggest that in situ disease is relatively advanced in its molecular progression to invasion, despite its distinct histologic appearance. National Comprehensive Cancer Network guidelines do not recommend molecular profiling for ductal carcinoma in situ (DCIS) or lobular carcinoma in situ, although estrogen receptor (ER) status is recognized as a tumor marker for DCIS.1
Whole breast irradiation (XRT) is known to reduce local recurrence among women who undergo lumpectomy for DCIS. Randomized trial data have demonstrated that radiation plus tamoxifen reduces risk for patients with estrogen receptor (ER)-positive tumors. In practice, it is difficult to distinguish low-risk patients using clinical and pathological factors. “These are aggressive treatments for low-risk patients who rarely die of this disease,” said Dr Solin.
Researchers who have developed a 12-gene signature assay for DCIS hope their system will address the problem of who to treat with adjuvant XRT. Initial results of their findings were presented at the 2011 SABCS and are under review for publication.2 Investigators anticipate that this type of assay, once incorporated into widespread use, will support individualized treatment decision making and identification of molecular biology underlying DCIS.
Dr Solin participated in an analysis that compared the costeffectiveness of basing decisions of whether to use adjuvant XRT on the 12-gene score versus routine clinical practice. Results of the analysis were presented as a poster at the 2012 SABCS.3 The average cost of the 12-gene assay was lower than that of standard clinical assessment by approximately $1000 per patient. Changing strategy from using the assay to clinical assessment was associated with an incremental cost-effectiveness ratio of approximately $95,000 per quality-adjusted life-year. Estimates assumed that XRT benefits were independent of biology and that ≥75% of patients would not receive XRT.
For early-stage breast cancer, decisions regarding the use of adjuvant chemotherapy have considered hormone status of the tumor. With respect to tumor markers for breast cancer, the American Society of Clinical Oncology recommends ER and progesterone receptor (PgR) testing for decisions on endocrine therapy, and human epidermal growth factor (HER2) testing for decisions regarding anti-HER2 and anthracycline therapy.4 For prognosis, the 21- gene recurrence score (Oncotype DX breast cancer assay, Genomic Health, Inc), urokinase plasminogen activator, and plasminogen activator inhibitor for prognosis are recommended. The 21-gene recurrence score may also be used to help make treatment decisions; patients identified with a highrisk recurrence score are more likely to benefit from chemotherapy. This is not surprising, explained Antonio Wolff, MD, Johns Hopkins Kimmel Cancer Center, because nearly one-third of the non-reference genes in the assay are markers of proliferation, a trait that makes tumors particularly vulnerable to chemotherapy.
Gene expression array analyses have further refined tumor classifications. As additional molecular information is emerging, it is becoming clear that breast cancer should be viewed as a biologic continuum, not as discrete categories. “We know that [these categories] are not enough,” said Dr Wolff. The validated predictive biomarkers have a strong negative predictive value; if the markers are not expressed, targeted therapy is not indicated. With a modest positive predictive value, however, they don’t specifically ensure that patients who have receptor-positive tumor markers will benefit. “They also don’t tell us in absolute terms whether to give it. For prognostic utility, you have to incorporate clinicopathologic measures of prognosis such as tumor size and node status for baseline assessment of risk.”
At this point, clinicians and researchers are realizing that standard clinical measures and molecular assays do not provide the same information. “One assay, one method, is not necessarily better than the other,” said Dr Wolff, emphasizing that clinical context matters. “The question is whether they can be complementary. What we need now are exercises where we are able to combine both molecular and clinical standard information and try to improve on the prediction of what to do.”
To answer some questions about how to use these assays, the field is awaiting the results of 2 completed prospective randomized trials: the Trial Assigning IndividuaLized Options for Treatment (Rx) (TailoRX) and the Microarray for Node-Negative Disease may Avoid Chemotherapy (MINDACT).5 TailoRX will attempt to answer the question of whether adding adjuvant chemotherapy to endocrine therapy will improve outcomes of patients with ERpositive, node-negative breast cancer and an intermediate-risk recurrence score (by the 21-gene assay). MINDACT will examine outcomes when chemotherapy is added to endocrine therapy in patients with discordant tumor risk assessments as determined by Adjuvant! Online and the 70-gene signature (MammaPrint, Agendia).
Tumor classification will improve with continued research. Building on established molecular information, the Cancer Genome Atlas recently reported the molecular portrait of over 800 breast tumor tissue samples, incorporating a host of biological factors, including phenotype, genotype, proteomics, epigenetics, copy number variation, and mRNA expression.6 The analyses confirmed the existence of 4 major molecular subtypes of breast cancer: HER-2 enriched, luminal A, luminal B, and basal-like. “Much of the prognostic variability seems to be happening within these groups,” said Dr Wolff. Yet, the usefulness of the data set is limited without a correlation to clinical outcomes. “As we move into the era of precision medicine, I think we need to begin to correlate this kind of information with clinical outcomes so that we can begin to make decisions,” he said. Along with other studies, this study is also beginning to tease out different molecular subtypes underlying triple-negative breast cancer.6,7
Multiple studies have been examining how and to what extent available assays are being put into practice. One of the largest of these studies used the NCCN registry database to assess the adoption of gene expression profiling (GEP) and use of chemotherapy for women with hormone-receptorpositive breast cancer at 17 centers.8 They found an absolute increase of 12% in GEP use (P <.01) from 2006 to 2008 and an absolute decrease of about 5% in chemotherapy use (P <.01) over the same time period (N = 7375). Testing was associated with lower odds of chemotherapy use (odds ratio [OR], 0.70; 95% confidence interval [CI], 0.62- 0.80), overall. However, Dr Wolff noted an interesting observation: physicians might have ordered the 21-gene assay to help confirm their underlying intent. Patients with node-positive and large node-negative cancers were less likely to receive chemotherapy (OR, 0.11; 95% CI, 0.07-0.17) while those with small, node-negative cancers were more likely to receive chemotherapy after GEP testing (OR, 11.1; 95% CI, 5.39-22.99). This suggested an underlying assumption by the clinician that one test might be more informative than the other. Patient factors also may have influenced GEP test ordering; the odds of being tested were lower for blacks versus whites (OR, 0.70; 95% CI, 0.54-0.92) and for those with high school education or lower versus those with a higher education (OR, 0.63; 95% CI, 0.52-0.76).
Despite a great deal of exciting investigational activity on molecular profiling of metastatic disease, relatively little has carried over to actual routine clinical use. “Sadly enough, there is no genomic test that is used routinely in the management of metastatic breast cancer today,” lamented Lajos Pusztai, MD, Yale Cancer Center. More often, for metastatic breast cancer, molecular markers are being used as patient selection criteria or to enrich a specific population in clinical trials.
In some select cases, repeating ER, PgR, or HER2 tests can be justified; for example, when the diagnosis of a solitary nodule is in doubt, on technical grounds, in clinical trial, or when the clinical course of disease suggests different biology than the original receptor test result indicates. Considering test results in the clinical context is critically important. “For example, if a patient originally diagnosed with ERnegative breast cancer has recurrence 5 or more years after original diagnosis, this would be atypical,” explained Dr Pusztai. “In this setting, repeating a biopsy could be very important. If the patient turns out to be hormone receptor– positive on repeat testing, you could extend the life of that patient by years by offering hormone therapy.”
It is important to understand that a single repeat of the same test does not improve accuracy of the result. A 90% accurate test repeated twice on the same sample yields concordant results about 81% of the time (0.9 x 0.9) because 10% of the 90% of correct classifications from the first round will be misclassified in the second round and vice versa.
Because there is a cognizance bias to act on the most recent results, discordant results can be problematic and potentially dangerous, said Dr Pusztai. Yet, outside of a known technical error in the first test, there is no reason why the most recent test would be more accurate than the first. He recommends that if a patient has at least 1 positive result, at least 1 targeted agent should be tried.
PDF is available on the last page.