Four genes by RT-PCR predicts distant relapse for women given adjuvant tamoxifen.

Desmedt C, Loi SM, Haibe-Kains B, Soree A, Lallemand F, Durbecq V, Larsimont D, Tutt A, Ellis P, Gillett C, Ryder K, Harris A, Cardoso F, Martiat PH, Piccart MJ, Sotiriou C

Background: There have been many gene classifiers developed recently that claim to predict clinical outcome better than currently used clinical factors. Whilst microarray technology offers huge advantages in furthering our knowledge of breast cancer biology, for present-day use, clinically applicable and useful tools may need only the quantitative information that is provided by genes that are already well known to convey prognostic/predictive information in breast cancer.

Material and methods: Our aim was to assess if the results of a reverse-transcriptase-polymerase-chain-reaction (RT-PCR) assay for the genes ER, PgR, HER2/Neu and Ki67 in frozen tissue could correlate with distant disease recurrence in 175 breast cancer patients from 2 institutions treated with adjuvant tamoxifen only and followed for a median of 9.4 years. All RT-PCR experiments were performed in duplicate.

Results: Adequate RT-PCR profiles were obtained for 154 of 175 tumors. We constructed a score from a linear combination of the coefficients from a multivariate Cox analysis including the 4 genes (likelihood ratio test, overall p-value<0.001) and the expression of those genes. When considering the score as a binary variable with the median as cutoff, a hazard ratio (HR) of 3.7 (95% CI: 1.8-7.6, p<0.001) was reached with 97%, 95% and 91% of the patients being distant metastasis free at 3, 5 and 10 years in the low risk group and 88%, 74% and 65% in the high risk group respectively. In a Cox model that included the binary score and traditional prognostic variables (tumor size, grade, age, nodal status), the score was the only significant predictor of distant metastasis at .05 level with a HR of 2.7 (95%CI: 1.2-6.2, p=0.019) and tumor size was borderline significant (p=0.052). Similar HR were found when considering the score as a continuous variable both in univariate and multivariate analysis.

Discussion: We are currently validating this score on a further 100 samples and comparing with RT-PCR results on corresponding paraffin-embedded tumour tissue. A comparison between RT-PCR and immunohistochemistry is also ongoing. This simple test using the RT-PCR value of 4 genes could provide an immediately applicable and cheap clinical tool that can assist clinicians in discriminating high and low risk prognostic groups for ER+ women treated with tamoxifen.