The association of breast density (% of breast volume involved by fibro-glandular densities) with the risk of interval cancer (IC) was investigated by reviewing a consecutive series of 346 cancers detected at screening (SDC) during 1996-1999 and of 90 ICs, reported as negative in the same period and diagnosed in the following 2 years, and comparing them to a random sample of 360 healthy controls. The probability of IC was significantly associated with breast density, whatever grouping (0/1-25/26-74/>74%; 0-25/26-60/61-74/>74%; 0-25/26-74/>74%) was considered (chi(2)=30.67-34.08, P<0.<0.01): 27.8% of all ICs were classified in the >74% density class, as compared to 7% of SDC and 5% of healthy controls. No significant association to IC was observed for Wolfe pattern (P2/Dy vs N1/P1: chi(2)=0.30, P=0.960), number of used mammographic views (single oblique vs oblique+craniocaudal: chi(2)=0.02, P=0.90) or screening round (first vs repeat: chi(2)=1.41, P=0.23). Multivariate analysis confirmed the independent association of breast density to IC, the highest risk being observed for >74% density class (OR vs 0% class=13.4, 95% CI 2.7-65.6, OR vs all other density classes=5.1, 95% CI 2.6-10.0). Age showed an independent association too, older women having a lower risk of IC (OR=0.52 95% CI 0.3-09). Breast density (>74%) resulted as being a major determinant of IC. Special screening protocols (shorter rescreening interval, routine use of ultrasonography) might be suggested for these subjects in order to improve screening sensitivity and efficacy.