Over 1.1 million patients in the nation-wide Swedish Family-Cancer Database were analysed for discordant familial cancers covering 33 sites. Standardised incidence ratios (SIRs) were calculated for patients whose family members had a defined cancer compared to those whose family members did not have that cancer. Three independent tests for each pair of cancer sites were done using different family relationships.
Lung cancer showed 13 significant discordant associations but most of them were with sites for which smoking is a risk factor. An exception was the clustering of lung cancer and endocrine cancers. Four discordant associations reached a minimal significance level of 5 脳 10-6: colorectum-endometrium, breast-ovary, breast-prostate and melanoma-squamous cell carcinoma of the skin. The association of melanoma and nervous system cancer reached a minimal significance of 10鈭?. Discarding lung cancer, all other associations were based on a single test whereby they were liable to be chance associations.
This study showed the extraordinary requirements for statistical power in study of multiple cancer sites. In addition to the smoking related sites, associations between breast and prostate cancers, melanoma and nervous system tumours and lung and endocrine tumours found strong statistical support. Within the present sample size limits, we found no evidence of an overall susceptibility to cancer.