Breast
tumours with intermediate histological grade can be reclassified into
prognostically distinct groups by gene expression profiling.
Sotiriou C,
Wirapati P, Loi S, Viale G, Harris A, Bergh J, Smeds J, Farmer P, Praz
V, Haibe-Kains B, Lallemand F, Desmedt C, Durbecq V, Larsimont D,
Cardoso F, Buyse M, Delorenzi M, Piccart M
Background:
Histological grade (HG) in breast cancer provides important prognostic
information. Low and high grade (HG1 and HG3, respectively) tumours are
known to have good and poor prognosis. However, intermediate grade
(HG2) tumours present a difficulty in clinical decision making, because
their survival profile is not different from that of the total
(non-graded) population and their proportion is large (40%-50%). The
aim of this study was to determine whether the prognostic value of
grade could be refined by using gene expression profiling.
Methods:
Gene expression profiles (GEP) from Affymetrix U133A Genechips were
contrasted between HG 1 (low grade) and 3 (high grade) tumours using a
training set of 64 estrogen receptor (ER)-positive breast cancer
samples. A score called the gene-expression grade index (GGI) based on
the 128 most significant genes was defined and tested on a independent
validation dataset of 129 tumour samples. A multiple independent
external validation was also performed using three publicly available
datasets from van de Vijver, M. J. et al. 2002, Sorlie et al. 2001 and
Sotiriou et al. 2003 using different populations and microarray
platforms (total=474 tumour samples).
Results: HG
1 and 3 tumours were associated with distinct GEP and GGI values. The
GGI clearly splits HG 2 (intermediate grade) tumours into high-risk
(gene-expression grade GG 3) and low-risk (GG 1) groups with
significantly different clinical outcomes (HR: 3.49, 95% CI 2.14-5.69,
p<0.0001), similar to that of HG 3 and 1, respectively (HR:
3.48, 95% CI 2.26-5.36, p<0.0001). Replacing HG with genomic
grade (GG) in the Nottingham Prognostic Index significantly improved
prognostic classification. Similar results were found when we applied
GG to the 3 external independent datasets. We are currently validating
our findings in the TransBig series of 302 tumour samples from 5
different European institutions from which grading was determined based
on central pathology review in Milan.
Conclusions:
Gene-expression-based grading can significantly improve current grading
systems for the prognostic assessment of breast cancer. Reproduction of
these findings across multiple independent datasets and across
different platforms suggests our findings are robust. Refined grading
based on gene expression measurements can have important clinical
implications for breast cancer management in the near future.