In a previous qualitative study it has been shown that the incorporati
on of pre-existing collagen bundles from the reticular dermis into the
bulk of melanoma lesions metastatic to the skin indicates rapid syste
mic spread. In the present study the amount of pre-existing dermal col
lagen in the bulk of the melanoma lesions in 267 cases of primary mela
noma of the skin with a Clark level of at least III was quantitatively
assessed using automated image analysis based on RGB (red, green and
blue) colour images of sections stained with haematoxylin and eosin. T
here was a weak correlation between the amount of pre-existing collage
n and the Clark level and the Breslow index. With regard to prognosis,
a large amount of pre-existing collagen (>0.13 mm(2) per index slide)
was significantly associated with a particularly poor outcome (24-mon
th survival rate: 71 +/- 17% compared with 96 +/- 2%; log rank test: P
< 0.001). It is clear that a large amount of pre-existing collagen bu
ndles occurring as a particular feature of tumour-stroma interaction i
ndicates high metastatic capacity in primary malignant melanoma. (C) 1
998 Lippincott-Raven Publishers.