Background It has been proposed that cancer is more common in some families than in others, but the hypothesis lacks population level support. We use a novel approach by studying any cancers in large three-generation families and thus are able to find risks even though penetrance is low.

Methods Individuals in the nation-wide Swedish Family-Cancer Database were organised in three generations and the relative risk (RR) of cancer was calculated to the persons in the third generation by the numbers of patients with cancer in generations 1, 2 and 3.

Results The RRs for any cancer in generation 3 increased by the numbers of affected relatives, reaching 1.61 when at least seven relatives were diagnosed. The median patient had two affected relatives, and 7.0% had five or more affected relatives with an RR of 1.46, which translated to an absolute risk of 21.5% compared with 14.7% in population by age 65 years. For prostate cancer, the RR was 2.85 with four or more affected family members with any cancer, and it increased to 14.42 with four or more concordant cancers in family members. RRs for prostate cancer were approximately equal (2.70 vs 2.85) if a man had one relative with prostate cancer or four or more relatives diagnosed with any cancer.

Conclusions A strong family history of cancer, regardless of tumour type, increases cancer risk of family members and calls for mechanistic explanations. Our data provide tools for counselling of patients with cancer with both low and high familiar risks.