Biomarkers are widely used in epidemiology, yet there are few reliability studies to assess the appropriateness of using these biomarkers for the assessment of exposure-disease relationships. The aim of the study was to assess the reliability of 20 biomarkers in serum collected from two Dutch centres (Utrecht and Bilthoven) participating in the European Prospective Investigation into Cancer and Nutrition (EPIC) at two points several years apart. Blood samples were collected from 30 men from Bilthoven and 35 women from Utrecht. Ferritin, total iron, total iron-binding capacity, transferrin saturation, transferrin, C-reactive protein, bilirubin, cholesterol, triglycerides, apo lipoprotein-A, apo lipoprotein-B, high-density lipoproteins, low-density lipoproteins, uric acid, creatinine, reactive oxygen metabolites, the ferric-reducing ability of plasma, protein thiol oxidation, fructosamine, and vitamin D biomarkers in serum were analysed from the blood samples at the two points of time. For all biomarkers, except C-reactive protein, there were no substantial changes in the mean levels over time. Uric acid, ferritin, creatinine, HDL, and apo lipoprotein-B levels consistently showed the highest reliability for men and women (intra-class correlation = 0.69-0.86). Among women, the ferric-reducing ability of plasma, and protein thiol oxidation had poor reliability; and among men iron-related biomarkers (except serum ferritin) had poor reliability. With the exception of a few gender-specific differences, most of the 20 biomarkers performed well and can be considered to have sufficient reliability to be used in future cohort studies.