The Sheffield table is a simple method for targeting a specified level of CHD risk, and indicates the total cholesterol (TC) level conferring that risk in an individual. We have examined whether inclusion of the TC:HDL ratio rather than TC alone improves prediction. The sensitivity and specificity of the TC and the TC:HDL tables, targeted at a CHD event rate of 3% per year, were compared to the full Framingham equation in 216 men and women aged 35-70 years with TC ≥5.5mmol/l who were studied prospectively. The 'treat' and 'no treat' decisions from the table for men and women yielded groups with a mean CHD risk by the Framingham equation of 4.1% per year and 1.8% per year respectively for the TC table, and 3.6% per year and 1.5% per year respectively for the TC:HDL table. The TC table had sensitivity of 45% and specificity 98%, and the TC:HDL table had sensitivity of 100% and specificity 94%. The TC:HDL table thus improved sensitivity with no significant loss in specificity. In men only (n=126), both tables were then compared to Joint European Task Force Guidelines and to the targeting of patients with TC ≥ 6.5mmol/l using the PROCAM risk function as an external standard. The sensitivity and specificity of the various methods is shown in the table. Method CHD risk targeted (% per year) sensitivity (%) specificity (%) CHD risk treated CHD risk not treated Sheffield TC 3 52 96 4.5 1.9 Sheffield TC/HDL 3 97 82 3.8 1.3 Joint Euro Task Force 3 100 26 2.6 0.6 Joint Euro Task Force 2 98 37 2.6 0.6 TC ≥6.5mmol/l 3 63 51 2.4 2.1 The TC:HDL table is highly sensitive and specific when compared to the complete Framingham equation and to the independently derived PROCAM equation, and identifies accurately individuals at high and low CHD risk. It is as sensitive but significantly more specific than Joint European Task Force recommendations. Targeting patients by cholesterol threshold alone is unacceptably inaccurate. It remains to be shown whether use of the TC:HDL ratio is accepted readily by ordinary doctors.