Gene Expression Profiles Clinical Characteristics And Breast Cancer Outcomes

Differences in breast cancer outcomes have been attributed, in part, to differences in clinical factors such as tumor size and lymph node status that may relate to the timing of cancer detection as well as to cancer biology (30). The importance of cancer biology in the determination of outcome has been recognized by attention to histological grade, ER, PgR, and HER2. However, to a large extent, breast cancer has been viewed as a single disease, with variable features that may affect treatment and outcome. These factors do not sufficiently explain differences in outcomes, particularly among tumors that are ER-positive or intermediate grade.

In addition to the associations with race and age mentioned earlier, several tumor characteristics vary by subtype. In the population-based Carolina Breast Cancer Study, which defined the subtypes using IHC profiles that included ER, PgR, and HER2, there was no significant difference among subtypes in overall stage at presentation; however, there were marginally significant differences (P = 0.04) in the proportions with involved lymph nodes, with highest proportions among the HER2+/ER- (56%), followed by luminal B (47%), basal-like (41%), and luminal A (34%) (20). Infiltrating lobular carcinomas were exclusively seen among the luminal subtypes, although remained the minority comprising only 7-12%. A far more striking difference was seen in the proportions with high-grade tumors. For example, high mitotic index was seen in 87% of basal-like breast cancers, 69% of HER2+ /ER-, but only 31% and 32% of luminal A and luminal B tumors, respectively. Adjusting for the other relevant variables and compared with the referent luminal A subtype, the HER2+/ER- subtype remained 2.2-fold more likely to have involved lymph nodes, 6.8-fold more likely to have marked nuclear pleo-morphism, and 4.3-fold more likely to have high mitotic index. The basal-like subtype was not more likely to have involved lymph nodes, however, remained 9.7-fold more likely to have marked nuclear pleomorphism, 2.5-fold more likely to be poorly differentiated, and 11.0-fold more likely to have a high mitotic index. The luminal B subtype differed little from the luminal A in clinical characteristics, other than a higher (1.7-fold) likelihood of lymph node involvement.

Discovery of biologically distinct subtypes of breast cancer raised the question of whether these subtypes could explain the diversity of cancer behavior seen in clinical practice. To address this question, Sorlie and Perou conducted a uni-variate analysis of correlation between molecular subtype using gene expression array and overall survival based on the five subtypes of breast cancer identified in gene expression studies (7). They examined relapse-free and overall survival among a subset of 49 locally advanced tumors treated with doxorubicin monother-apy. With a median follow-up of 66 months, both relapse-free and overall survival differed significantly among the different breast cancer subtypes. The basal-like and HER2+ /ER- subtypes had the shortest survival times and luminal A tumors the longest, with luminal B having an intermediate prognosis. This association of subtype with outcome has been confirmed in independent datasets using gene expression array and those investigators' intrinsic gene list (8). Using a panel of gene probes derived independently from those of Sorlie, Perou and colleagues, although with significant overlap, Sotiriou et al. (31) studied gene expression among 99 cases of both node-negative and node-positive invasive breast cancer and performed hierarchical clustering analysis. They again identified several subtypes demonstrating luminal or basal characteristics. The breast cancer subtype correlated significantly with survival, with superior survival for the luminal subtypes compared to basal tumors. Once again, ER status was found to correlate strongly with gene expression profile as did tumor grade. There was little correlation between expression profile and lymph node status, tumor size, or menopausal status. Studies using IHC profiling using either a large panel of antibodies (12) or a simpler intrinsic gene list-driven selection of markers (20) have found similar associations with outcome. Interestingly, some studies suggest that in addition to variability in likelihood of metastasis, the site of metastasis may vary with subtype, with basal-like more prone to visceral involvement, particularly of the lung (32,33).

Differences in outcomes that have been linked to demographic factors such as age or ethnicity, may partly be related to the differential representation of breast cancer subtypes among these populations. In other words, it may not be the race or age of a patient that conveys a bad prognosis but the fact that younger and African American patients are more likely to develop the basal-like subtype of breast cancers. As noted earlier, African American ethnicity has been associated with tumor features typical of the basal-like subtype, particularly among premeno-pausal women, which may in part explain poor cancer outcomes compared to those for white women (20-22). The contribution of biology to long recognized risk factors for poor outcomes is further supported by the finding that basal-like tumors were more common among younger patients, of any ethnicity, when analyzed by tissue microarray (12).

Tumor biology as reflected by gene expression appears to be an independent predictor of outcome (12). Genomics technology has also been used to develop pure prognostic panels that have been applied across subtypes. Other groups have sought to use gene expression to identify a limited set of genes correlated with a particular clinical outcome, such as recurrence, which could then be used clinically to guide decision-making. This work will be discussed in detail in a subsequent chapter but reviewed briefly here in light of associations with the breast cancer subtypes.

Researchers in the Netherlands have developed a 70-gene prognostic profile that correctly classifies 90% of tumors destined to recur within five years (19). The expression pattern of these 70 genes was shown to be a strong predictor of both distant disease-free survival and overall survival, and the "poor prognosis" signature, a stronger predictor of metastatic disease than any of the classical clinical criteria (34). Similarly, investigators from Rotterdam have developed a 76-gene prognostic profile that was recently validated in an independent group of node-negative, largely hormone receptor-positive breast cancers (35). Interestingly, the 70-gene and 76-gene prognosticators remain prognostic in multivariate analyses, but have little overlap in the genes they have identified. Paik et al. (36) identified another set of 16 genes out of a group of 250 candidate genes that correlated with prognosis in several clinical trials, including one large homogenous sample of ER-positive, node-negative tumors treated with tamoxifen in a trial from the cooperative group NSABP, trial B-20. This 16-gene model, called the Recurrence Score, was validated in tumor samples from an independent large prospective trial of patients with hormone receptor-positive, node-negative tumors treated with adjuvant tamoxifen on NSABP B-14. These studies illustrate the challenges of determining gene expression profiles of clinical importance either to predict outcomes or to identify drug targets. To some extent, the discrepancies in identified genes may reflect variations in patient population, ascertained tumors, or technical variations in array technology. An even more likely contributor is that common molecular pathways, such as apoptosis or cell cycle regulation, may involve many genes and the identification of any genes whose expression is important in this pathway may provide similar information in terms of prognosis.

Supporting data comes from comparison of several of these well-established prognostic panels with breast cancer subtypes. Using RNA from a 295-patient dataset with known clinical outcomes, investigators at the University of North Carolina compared the "Wound Response" prognostic model (37), the 70-gene prognostic profile (19,34), the 2-gene expression ratio profile (38), and a model replicating the 16 genes of the Recurrence Score (36) for their ability to predict recurrence and with the known breast cancer intrinsic subtypes (39). The Recurrence Score and 2-gene ratio predictors were designed only for ER-positive patients, and therefore, were tested on the 225 ER+ patients from the dataset and on all 295 patients. The investigators found that all models except the 2-gene ratio provided significant accuracy for estimating recurrence risk, and provided prognostic information that was independent of classical factors. Classification of tumors into high-risk and low-risk groups by the 70-gene model and the 16 genes used in the Recurrence Score were identical in 8l% of all cases, even though the models overlap by only one gene (39) As before, the intrinsic subtypes also provided independent prognostic information. Comparison of the prognostic profiles by subtype was intriguing. There were 53 basal-like and 35 HER2+/ER- tumors in the dataset; in these subtypes over 90% had high Recurrence Scores, "poor prognosis" 70-gene signatures, were wound response "activated." Fifty-five luminal B tumors similarly gave fairly uniform poor prognosis signatures, 50 (91%) had high Recurrence Scores, 46 (84%) had poor 70-gene signatures, and 51 (93%) were wound response-activated. The group with the greatest disparity in prognostic profiles was the luminal A subtype. Of 123 tumors with this subtype, 36 (29%) had high Recurrence Scores and poor 70-gene signatures, and 78 (63%) had activated wound response signatures. This study is reassuring in that multiple disparate array-based molecular prognostic assays appear to be capturing similar biological information and are making similar outcome predictions. Among poor-prognosis subtypes such as the HER2+/ER-, basal-like, and luminal B, these assays do not provide additional information, suggesting that new assays that can prognosticate within these groups might be useful. For patients with the luminal A intrinsic groups, the Recurrence Score, 70-gene, and wound-response assays did appear to provide additional information that could be used to guide treatment decisions.

10 Ways To Fight Off Cancer

10 Ways To Fight Off Cancer

Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.

Get My Free Ebook

Post a comment