Sampling weights reflecting the survey design were used to produce national estimates.10 Adjustments for nonresponse in NHANES III did not reveal evidence of nonresponse bias (T. M. Ezzati, MS, and M. Khare, MS, unpublished data, 1992). The weights were adjusted to reduce bias from nonresponse at the interview and examination stages. The weighted total was 154,987,478 adults aged 20 years or older as of October 1991, the midpoint of NHANES III. To assess potential nonresponse bias for LDL-C level, the weighted distributions of individuals in the selected sample, individuals in the interviewed sample, and individuals with and without LDL-C values in the fasting morning subsample for various demographic and health-related factors were obtained.11 To estimate the total number of participants with an elevated cholesterol level who might require lipid-lowering drug therapy, we considered 4 scenarios in which dietary intervention would uniformly reduce LDL-C levels by 0%, 5%, 10%, or 15%, which was consistent with previous analyses conducted by Sempos et al.3 In addition, we examined the percentage reduction in LDL-C level required by individuals qualifying for drug therapy according to the ATP II guidelines, assuming a 10% reduction in LDL-C level with dietary therapy. A sensitivity analysis was performed on the range of 20% to 50% LDL-C reduction needed to achieve ATP II goals by aggregating people across CHD risk strata. Computer software (Microsoft Access 97; Microsoft Corp, Redmond, Wash) was used to perform descriptive analyses and summation of weights based on various criteria. For quality control, the analyses were reconducted using additional software (Statistical Export and Tabulation System; US Department of Health and Human Services, National Center for Health Statistics, Hyattsville, Md).