OBJECTIVE To validate to what extent the isolated determination of total Cholesterol (TC) is effective when seeking to predict coronary risk. METHODS An observational crossover study of the analytic determinations of the clinics which systematically request TC and HDL-Cholesterol (HDL)--case-finding method. METHODS Health Centre. METHODS 631 analytic determinations, with samples from people who attended a Health Centre between May and November 1992, were studied. RESULTS As proof of certainty the Atherogenic Index (AI) was used for the relative risks (RR) of suffering a coronary event in line with the Framingham study. The confidence limits (CL) were calculated to 95% in order to quantify random error and permit comparison. On varying the cut-off points of TC the indicators changed, being more sensitive (S) and less specific (E) with the lower figures: 180 mg/dl, RR > 1, S = 97.5% (CL: 100-94.7) and E = 30.5% (36.8-24.2); RR > 2, S = 100%, E = 22.1% (26.9-17.3) and RR > 3, S = 100%, E = 20.8% (25.3-16.3). As values of TC increase, S diminishes and E increases: 250 mg/dl, RR > 1, S = 48.3% (57.2-39.4), E = 87.2% (91.8-82.6); RR > 2, S = 58.6% (76.5-40.7), E = 77.2% (82-72.4) and RR > 3, S = 63.6% (92-35.2), E = 75.3% (80.1-70.5). CONCLUSIONS HDL must be determined if TC is -200 mg/dl. If everyone with RR > 2 is to be detected, HDL-cholesterol from TC > or = 180 mg/dl must be measured.