Prediction of early distant relapses on tamoxifen in early-stage breast cancer (BC): A potential tool for adjuvant aromatase inhibitor (AI) tailoring.


Loi S, Piccart M, Haibe-Kains B, Desmedt C, Harris A, Bergh J, Ellis P, Miller L, Liu E, Sotiriou C

Background: The majority of early-stage BC express estrogen receptors (ER) and receive tamoxifen in the adjuvant setting. Yet up to 40% of these patients will relapse on tamoxifen and develop incurable metastatic disease. Recent evidence from three large randomised controlled trials exploring the role of AI in the adjuvant setting shows a benefit from the novel strategy, however the optimal sequence and duration of AI/tamoxifen treatment is unknown. Therefore, it is vital that we learn to identify those women at higher risk of tamoxifen resistance. Our aim was to identify genes that could predict for this subset of women.

Methods: We determined the gene expression profiles from 105 tamoxifen-only treated ER positive early stage BC (training set) using Affymetrix U133 A and B chips. Within this group, 30 (29%) patients developed distant recurrence at a median time to relapse of 3.8 years (yrs) and 75 (71%) remained disease free at a median of 5.7 yrs of follow-up. The independent validation set consisted of 64 ER+ tamoxifen only treated BC patients (17 recurrences) from a different institution (Karolinska, Sweden).

Results: Cox proportional hazards regression analysis identified 50 genes that were significantly associated with distant relapse on tamoxifen. The probability of selecting these genes by chance was estimated to be p=0.001 after 1000 random permutations. These genes were powerful predictors of clinical outcome in the independent validation set, dividing the cohort into 2 statistically different prognostic groups (p=0.02) in the poor prognostic group (tamoxifen resistant) the median time to distant relapse was 2.4yrs compared with 7yrs in the good prognostic one (tamoxifen sensitive). We are currently validating this signature in another independent dataset (Guys hospital, UK).

Conclusions: These results suggest that a group of genes can identify BC patients at risk of early distant relapse on tamoxifen. These patients could be the ideal candidates for upfront AIs, while the others would be considered for sequential tamoxifen/AI. This hypothesis will be tested in an upcoming prospective clinical trial “MINDACT”.