Prostate cancer risk assessment model: a scoring model based on the Swedish Family-Cancer Database.

Roudgari H, Hemminki K, Brandt A, Sundquist J, Fallah M.

Source

German Cancer Research Centre, Division of Molecular Genetic Epidemiology, Im Neuenheimer Feld 580, 69120 Heidelberg, Germany; m.fallah@dkfz.de.

Abstract

Background Many prostate cancer (PC) risk assessment models have been developed, however almost none include familial history. Aim To produce a risk assessment model for PC based on familial background of related cancers. Method 976 859 independent index men aged ≥30 in year 1998 and their family members in the Swedish Family-Cancer Database (FCD2010) were randomly divided into development (60%) and validation (40%) datasets (follow-up=10 years). The HR from Cox model was used to extrapolate risk scores. Results Specified scores were: for PC in situ at age <60 years in index man, 5; for PC at age <60 years in each first-degree relative (FDR), 15; for PC at age ≥60 years in each FDR, 10; for PC at age <60 years in each second-degree relative, 5; for breast cancer in each FDR, 2; for oesophageal carcinoma in situ in index man, 2; and for oesophagus cancer in each FDR, 2. Based on the findings, if the milestone age for a PC screening programme was 60 years or more, the recommended starting age for the men with the score-group 6-10 would be 54 years; score-group 11-15, 52 years; score-group 16-20, 50 years; score-group 21-25, 44 years; and for the score-group 26+ it should start before age 40. The concordance index in development and validation sets was 0.885 (95% CI 0.883 to 0.888). No significant difference was found between curves from development and validation datasets (internally validated using twofold validation and bootstrapping). Conclusion Familial history of relevant malignancies can be used as risk factors to estimate a man's prior risk of developing PC. The prostate cancer risk assessment model could satisfactorily assess risk of developing prostate cancer.