Abstract P3-04-03: A seven-gene profile predicting benefit of postmastectomy radiotherapy independently of nodal status in high risk breast cancer

Background: Recommendations for postmastectomy radiotherapy (PMRT) are well established in patients estimated to have a high risk of loco-regional recurrence (LR) (e.g. tumor size > 5 cm or ≥ 4 positive lymph nodes). International recommendations for patients with 1–3 positive lymph nodes suggest that PMRT should be restricted to younger patients and patients with other poor prognostic features. However, large randomized trials, including the DBCG82 trials (Danish Breast Cancer Cooperative Group), have previously shown a substantial overall survival benefit after PMRT in patients with low risk of LR, and shown that the largest translation of LR reduction into breast cancer mortality reduction occurs within the most favorable prognosis group. Our hypothesis is that a more refined partitioning of patients likely to benefit from PMRT can be established through identification of genes whose transcription interacts with PMRT to modify the hazard of LR.

Material and methods: The DBCG82bc cohort constitutes high risk patients (tumor size > 5 cm and/or positive nodes and/or invasion in skin or pectoral fascia) diagnosed between 1983–89, treated with mastectomy and partial axillary lymph nodes dissection and randomized to +/− PMRT. From 267 DBCG82bc patients, frozen tumor samples were available. Whole genome arrays (Applied Biosystem Human Genome Survey Microarray v2.0®, Applied Biosystem, Foster City, USA) were successful in 195 samples. Genes, whose expression levels interacted with PMRT on the association with LR, were identified through a two step Cox Proportional Hazard model with lasso penalty. 11 node negative patients were excluded from the subsequent analysis of 184 node positive patients (1–3 pos. nodes: 102 pts, ≥ 4 pos. nodes: 82 pts).

Results: Seven genes were identified whose expression interact with the effect of PMRT, and a specialized index was generated based on the expression levels of these genes. Patients were ranked according to the size of the index, and divided into quartiles with 25% of the patients designated as having a “high index” and 75% a “low index”. Among patients not receiving PMRT, a low index was in both nodal groups (1–3 vs. ≥ 4 positive nodes) associated with a significantly higher risk of LR compared to patients with a high index. In both nodal groups, PMRT significantly reduced the risk of LR in patients with a low index; equalizing the risk to patients with a high index, who showed no additional LR reduction by PMRT. In the group of 1–3 positive nodes, PMRT raised the local control rate (LCR) after 10 years from 47.5 % to 91.8% in the low index group (p = 0.0001), whereas the change in the high index group was non-significant (92.3% vs. 100.0%). In the group of patients with ≥ 4 positive nodes, PMRT raised LCR from 16.6% to 80.0% in the low index group (p = 0.0001), and no effect on LCR was seen in the high index group (83.3% vs. 87.5%, n.s.).

Conclusion: A seven gene-profile attaining prognostic and predictive impact, irrespective of number of positive nodes, was identified. The profile allowed the identification of 25% of the patients not showing any additional benefit from PMRT in terms of LR. The gene-profile may provide a method to identify patients expected to benefit from PMRT.

Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P3-04-03.