The diagnosis of familial hypercholesterolemia (FH) in unselected children is difficult due to the frequent overlap of cholesterol values in affected and non-affected and the paucity of physical signs. Nevertheless, detection and treatment of FH in childhood has been advocated to prevent atherosclerosis in these patients. Here, we report the results of a screening program in a cohort of 157 unrelated, hypercholesterolemic (HC) children (age range 2-15 years; mean 8.3+/-3.4 years) carried out by a combination of family study and molecular analysis of the LDLR gene. On the basis of the familial phenotype, 27 (17.2%) were classified as probable FH and 49 (31.2%) as affected by FCHL. Among probable FH children, 14 (51.8%) carried mutant LDLR alleles, giving an overall 8.9% prevalence of FH. Most of LDLR variants were already reported, but three new mutations G266C, T368M, and D451Y were identified. Beside increased TC and LDL-C (p<0.001), FH children showed decreased HDL-C (p<0.05) and higher prevalence of family history of CAD when compared to non-FH children. None presented tendon xanthomas. We estimated that LDL-C >3.9 mmol/L was the best cut off value for diagnosing FH in these children, showing 79% sensitivity and 71.0% specificity. We propose the use of a LDL-C cut off level associated with a family study to identify FH among HC children.