Abstract Objective We aimed to clarify some previous inconsistencies regarding the role of high density lipoprotein cholesterol (HDL-C) as a CVD protective factor. Methods The SCORE dataset contained data on HDL-C for 104,961 individuals (45% women) without pre-existing coronary heart disease (CHD). These were from 7 pooled European prospective studies. The effect of HDL-C, both in quintiles and as a continuous variable, on risk of CVD and CHD mortality was examined, using Cox proportional hazards model, adjusted for age, total cholesterol, systolic blood pressure, smoking, diabetes and body mass index and stratified by gender, age group, country and category of SCORE CVD risk. Results A strong, graded, independent, inverse relationship between HDL-C and both CVD and CHD mortality was demonstrated. Adjusted hazard ratios per 0.5 mmol/l increase in HDL-C were 0.60 (0.51, 0.69) and 0.76 (0.70, 0.83) in women and men, respectively for the CVD mortality endpoint. The corresponding hazard ratios were 0.53 (0.42, 0.68) and 0.79 (0.64, 0.98) in elderly women and men, respectively. The relationship was significant in all SCORE CVD risk strata and age groups. Conclusions This multivariable analysis, the largest of its kind to date, has confirmed the inverse, independent, strong and graded relationship between HDL-C and both CVD and CHD mortality. We have clarified previous suggestions that the relationship is stronger in women and that it applies in all age groups. This is the first prospective study to demonstrate the independent relationship specifically in healthy elderly women and to show that the relationship holds at all levels of total CVD risk.