A multitude of clinical, pathological and biological parameters have b
een reliably associated with prognosis in breast cancer patients. Rece
ntly much interest has been engendered by multifactorial computing met
hods attempting to produce higher prognostic accuracy as well as being
of clinical utility for treatment selection. Clinical, pathological a
nd a number of more recent biological parameters were evaluated retros
pectively in a population of 196 breast cancer patients who had been t
reated with first line surgery and followed for a median of 7.3 years.
Clinical tumour size, menopausal status, type of treatment as well as
tumour grade, pathological tumour size, node invasion, the presence o
f vascular emboli and steroid hormone receptor status were evaluated t
ogether with gene amplification (by Southern blot) of c-erbB2/neu, c-m
yc and int-2/FGF3 as well as overexpression (by immunohistochemistry,
IHC) of c-erbB2/neu, EGF receptor, CSF-1 and CSF-1 receptor. The prese
nce and abundance of inflammatory cell infiltrates (T, B cells and mon
ocytes) were also evaluated by IHC. The risk of cancer related death a
s tested in univariate and multivariate analyses was consistently high
er in patients with positive axillary nodes and in those whose tumours
showed evidence of int-2/FGF3 gene amplification, of overexpression o
f c-erbB2/neu at the cellular membrane, of vascular invasion by tumour
cells and of abundant CD45RO+ T cells infiltrates. A prognostic score
was calculated for each patient by computing the prognostic index ass
ociated with each of these five parameters and risk profiles were esta
blished by order of increasing risk. Survival curves drawn for three g
roups of high, intermediate or lowest risk showed a highly significant
poorer survival for the highest risk group. The model we present, alt
hough far from the complexity of a 'neural network' analysis, does dis
criminate effectively between low, moderate and high risk groups. Futu
re prospective studies should test these independent prognostic marker
s together with more recently established markers in an attempt to det
ermine not only the most predictive but also the most cost-effective '
prognostikit' for breast cancer patients. Such a molecular prognostic
index will allow the evaluation of current therapies in the light of s
pecific molecular alterations and may aid the design of new therapeuti
c approaches geared to the genetic profile of a given tumour.