We evaluated the degree of tubular differentiation in 172 samples of invasi
ve ductal breast cancer in order to determine numerical thresholds for hist
ological breast cancer grading. The tubular differentiation in each sample
was defined as the fraction of fields showing tubular differentiation (FTD)
. The analysis was based on Kaplan-Meier curves reflecting survival and rec
urrence of disease. univariate and multivariate analyses of Cox's regressio
n, and maximum efficiencies of ROC analysis. The minimum P-value cut-oft fo
r FTD was determined at 59%. The practical interpretation is that tubular d
ifferentiation in the neoplasm observed in at least 60% of microscopical fi
elds in the tumour area indicates favourable prognosis of disease. The rela
tive risks for breast cancer death for patients with FTD below 59% as compa
red with those with FTD above 59% were 6.7- and 6.3-fold (univariate and mu
ltivariate analyses respectively). Another threshold could be determined at
FTD 23%, although this threshold was associated with clearly lower statist
ical significancies. The paper introduces two possible solutions for applic
ation of the thresholds to the morphometric breast cancer grading system. T
he study also emphasizes the clinical relevance of the evaluation of tubula
r differentiation in breast cancer. The consistent morphometric evaluation
method was vital in allowing the full weight of the biological significance
of tubular differentiation to emerge. (C) 2000 Cancer Research Campaign.