The objective of the study was to determine the sensitivity and specificity of the National Cholesterol Education Program (NCEP) guidelines in identifying at-risk individuals for hypercholesterolemia, using population-based data from the Manitoba Heart Health Survey (MHHS). The MHHS surveyed a representative sample of adult residents of the province of Manitoba, Canada, aged 18-74 (n = 2792), 2212 of whom underwent complete lipoprotein analyses: total cholesterol (TC), high-density (HDL) and low-density (LDL) lipoprotein cholesterol, and triglycerides (TG) after overnight fasting. Following NCEP criteria for risk categories, 618 individuals could be considered as "at risk" based on their HDL and/or LDL levels. However, if the NCEP algorithms were followed, i.e. initial screen for TC only, and determination of HDL and LDL for selected individuals with borderline and high TC, only 438 of the total "at risk" individuals would have been identified (sensitivity = 71%). 1203 of 1594 individuals with acceptable HDL and LDL levels were not considered "at risk" on screening (specificity = 76%). If "at risk" status was determined solely on LDL criteria, the NCEP guidelines result in a higher sensitivity (94%) and comparable specificity (76%). However, when only HDL criteria were used, the sensitivity and specificity declined to 38% and 63% respectively. Compared to a simulated screening study based on the Lipid Research Clinic data reported by Bush and Riedel (1991) the estimates of sensitivity from the MHHS were high. Current NCEP guidelines appear to balance optimally the sensitivity and specificity for selecting individuals based on TC levels for further complete lipoprotein analysis.(ABSTRACT TRUNCATED AT 250 WORDS)