NEW YORK (GenomeWeb News) – Combining genomic profiling with standard clinical indicators can refine breast cancer prognosis and potentially guide treatment, new research suggests.
In a retrospective study published online today in the Journal of the American Medical Association, researchers combined conventional predictors of breast cancer outcomes — factors such as patient age, tumor size, and so on — with information about gene expression profiles in nearly a thousand breast cancer tumor samples. Their findings suggest that gene expression patterns can, indeed, define subgroups of women with different prognoses and treatment responses.
“The combination of these two methods, one of which uses the clinical description of a patient’s breast cancer and the other which looks at gene expression at the molecular level in a patient’s tumor, may allow us to [match drugs with patients] with unprecedented accuracy,” senior author Anil Potti, an oncologist affiliated with the Duke Institute for Genome Sciences and the Duke University Medical Center, said in a statement.
In general, breast cancer prognosis is evaluated based on factors such as the patient’s age, tumor size, the level of lymph node involvement, if any, and the degree of metastasis from the primary tumor site. In recent years, a computer system called Adjuvant! has been developed to incorporate these clinicopathological features and make predictions about clinical outcomes.
Adjuvant! also helps doctor’s determine whether adjuvant cancer therapies such as chemotherapy or radiation therapy are warranted or necessary for different patients. But while it can be useful for predicting cancer recurrence, some research indicates that Adjuvant! tends to overestimate cancer recurrence in younger patients.
In an effort to determine whether genomic data can provide additional information to Adjuvant!-based predictions, the researchers did a retrospective study of women with early-stage breast carcinoma who had been followed for 11 years, on average, after their initial assessment.
First, they classified 573 women as low, medium, or high risk for breast cancer recurrence using Standard Version 8.0 Adjuvant! Online. They then added in gene expression data for the tumors from integrating data from Affymetrix GeneChip arrays of different generations and platforms with a Duke-developed program called Chip Comparer. They also used a normalizing algorithm called ComBat to account for batch effects. The results were validated in an additional cohort of 391 women.
Indeed, the researchers did find gene expression clusters that defined subgroups both within and between low-, intermediate-, and high-risk groups. “[T]he molecular traits of patients in the poor prognostic clusters were highly specific and distinct from those of the good prognostic carriers,” the authors wrote.
Identifying these subgroups may not only refine predictions about patient outcomes, the authors suggest, it also provides information about patients’ underlying biology and the tumor microenvironment. That’s because gene expression patterns reveal different genetic pathways that are activated or silenced in different tumors.
For instance, those in the high-risk group with the best outcomes tended to have low expression of cancer risk genes, chromosomal instability, and so on. On the other hand, tumors that have high expression of genes associated with oncogenic pathway activation, wound healing, and so on tend to be associated with poorer outcomes.
The team also found genetic signatures within high-, medium-, and low-risk groups that were associated with different responses to chemotherapy treatments.
Though they noted that the results need to be verified in prospective studies, the authors suggest that their approach to integrating genomic data into existing clinical assessment strategies could individualize and improve breast cancer treatments. The approach is reportedly being tested in ongoing and future clinical trials taking place at Duke.
“This is one of the largest studies in human cancer showing the ability of gene expression profiles to improve risk stratification beyond established risk assessment algorithms that take into account clinicopathological variables,” Northwestern University researchers Chiang-Ching Huang and Markus Bredel wrote in an accompanying editorial in the same issue of JAMA.
Though they also call for prospective confirmation of the results and note that there are other models combining genomic and clinical data that could also be tested, Huang and Bredel said the study “demonstrates the potential value of using microarray-based gene signatures to refine outcome predictions.”