BACKGROUND:High-density lipoprotein cholesterol (HDL-C) is important in risk assessment for cardiovascular disease or metabolic syndrome; however, different direct HDL-C assays may lead to erroneous risk estimates and potentially misclassify people. METHODS:Data for 30-year HDL-C trends in Finland were obtained from the national FINRISK surveys during 1982-2012 (n=45766) taking into account biases from three external quality assessment programs (EQA). We also compared two different direct HDL-C and turbidimetric apolipoprotein A-I methods using 413 fresh serum samples. RESULTS:HDL-C concentrations in the Finnish population were on average 1.33 (±0.04) mmol/l for men and 1.62 (±0.05) mmol/l for women after bias-correction. Positive HDL-C trends were observed for both sexes with original data, but trends disappeared after bias-correction. Comparison of two direct HDL-C methods demonstrated concentration-dependent difference. When HDL-C concentrations were <1.0 mmol/l, the mean bias was -12.0% (95% CI -13.5 to -10.0) whereas HDL-C concentrations >1.55 mmol/l showed mean bias of 9.0% (95% CI 7.0-10.5). CONCLUSIONS:Accurate reporting of HDL-C concentrations at the population level requires proper and regular attendance to reliable EQA programs. We found evidence for a concentration-dependent difference between some direct HDL-C methods, which may cause misclassification of people in cardiovascular risk assessment.