The Health Outcome

The National Cancer Institute estimates there were more than 40,000 deaths due to breast cancer in the U.S. in 2004 (http://www.cancer.org/downloads/MED/Page6.pdf). Breast cancer is the most frequently diagnosed cancer in women and the second most frequent cause of cancer-related deaths in women (lung cancer is first). Treatments include surgery, radiation, and chemotherapy. Tamoxifen is a drug that is used to treat and prevent breast cancer. It is thought to work by attaching itself to estrogen receptors, thus decreasing the amount of estrogen available in the breast tissue. With decreased availability of estrogen, tumors stop growing. Tamoxifen has some undesirable side effects, such as inducing menopause and increasing the risk of endometrial cancer in women over 50.

Higher mortality has been associated with recurrence of breast cancer at distant sites. Being able to predict the likelihood of recurrence of breast cancer at distant sites has implications for more individualized treatment, preventive measures and follow-up. Using gene expression analysis of breast tumors from women who were treated with chemotherapy and tamoxifen and who had no lymph node involvement, this study set out to design a valid and reliable tool for predicting high, medium and low likelihood of recurrence.

The Finding

The authors of this study showed that a recurrence score based on expression of 16 genes in tumor tissue can be an accurate tool to quantify the likelihood of distant site recurrence in tamoxifen-treated female patients whose breast cancer was estrogen-receptor positive and had not spread to the lymph nodes. The study was a retrospective cohort design using 668 tumor blocks from patients enrolled in the National Surgical Adjuvant Breast and Bowel Project (NSABP) trials B14 and B20. The estrogen-receptor-positive tumors studied were from patients with no lymph node involvement who were treated with tamoxifen during the trials.

First, using a reverse transcriptase polymerase chain reaction (RT-PCR) method, the expression of hundreds of genes in RNA from fixed paraffin embedded tumor tissue was quantified. Based on these measurements, on the published literature, pathway analysis and genomic databases, 250 candidate genes were selected.

Next, the authors performed three independent clinical studies on a total of 447 breast cancer patients to see if there was any relationship among the 250 candidate genes and recurrence. Twenty-three genes showed a statistically significant relationship to tumor recurrence across the three studies. The authors further narrowed the list when they developed a Recurrence Score model based on the 16 genes that (1) showed consistent statistical significance across multiple studies and (2) showed consistent primer/probe performance in the assay. The 16 genes were: Ki-67, Cyclin B1, MYBL2, STK15, Survivin, ER, PR, Bcl2, SCUBE2, HER2, GRB7, Stromolysin 3, Cathepsin L2, GSTM1, CD68, and BAG1. Five reference genes were included to normalize the gene expression measurements: ACTB, GAPDH, RPLPO, GUS and TFRC.

Last, the tumors in this study were graded for likelihood of recurrence independently by three pathologists. The overall agreement in tumor grade among the three pathologists was 43 percent.

Recurrence-free was defined as not having had a tumor at a distant site for ten years after surgery. Recurrence scores of less than 18 were considered low risk; intermediate scores were 18 or higher but less than 31 and scores of 31 or higher were considered high risk.

The authors showed, using Kaplan-Meier Estimates, that the patients in the low risk-of-recurrence category (51 percent of total patients in the study) had a rate of distant site recurrence at 10 years out from surgery of 6.8 percent (95 percent confidence interval 4.0 – 9.6). Patients with a high risk-of-recurrence score (27 percent of all patients in the study) had a 30.5 percent rate of distant recurrence (23.6-37.4) at 10 years out. The recurrence score was also predictive of the relapse-free interval and overall survival rate (P<0.001 for both).

In a multivariate Cox proportional hazards model that did not include recurrence score, only age at surgery, not tumor size, was significant for recurrence (P=0.004). In a multivariate Cox model including patient age, size of tumor and recurrence score, recurrence score was predictive of distant recurrence independent of age and tumor size (P<0.001). The recurrence score can also be used as a continuous predictor because the likelihood of distant recurrence at 10 years increased continuously as the recurrence score increased.

Public Health Implications

Limitations of this study are that it is unknown if the genes used in the calculation of the recurrence score correlate with recurrence in the population studied because they show a relation with the natural history of breast cancer, because they predict responsiveness to tamoxifen, or both. Also, tumor grading is subjective to some extent and reproducibility of results is only moderate at best. Additionally, the presence of the genes studied cannot be used to determine which women should take tamoxifen.

References

1. Paik, S, et al. A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. New England Journal of Medicine. 2004: 351:2817-2826.