A prospective study has been conducted on 4954 female volunteers from the Island of Guernsey between 1977 and 1985 to examine risk factors for breast cancer and their relationship to mammographic parenchymal patterns as assessed by Wolfe's method of grading. Up to September 1988, 69 women had developed breast cancer, 11 of whom were prevalent cases being diagnosed within six months of mammography. The remaining incident cases were diagnosed six to 126 months (median 65 months) after entry to the study. Univariate analysis showed that the distribution of Wolfe grades in the population was significantly associated with menopausal status, age, parity, adiposity, age at menarche, age at first childbirth and use of oral contraception, but not with a family history of breast cancer. Multivariate analysis of the data for these variables from either pre- and/or post-menopausal women indicated that age, parity and adiposity were significantly related to Wolfe grade pattern. Age had an opposite effect in pre- compared with postmenopausal women thus the probability of either a P2 or DY pattern increased with increasing age in premenopausal but decreased in postmenopausal women so that incidence peaked around the menopause. Other variables did not achieve significance in the multivariate analysis. Odds ratios (ORs) were calculated for women with P2 or DY patterns using those with N1 or P1 grades as the reference group. The ORs were determined at two censoring times; one at five years and the other to include the most recent follow-up of this cohort. The ORs were adjusted for years of follow-up, age and adiposity and in postmenopausal women adjustment was also made for age at menarche.(ABSTRACT TRUNCATED AT 250 WORDS)