Changes in lipoprotein cholesterol, total plasma cholesterol, and weight prior to the diagnosis of cancer were examined in 103 men who developed cancer in a cohort of 3805 type IIa hyperlipidemic men aged 35-59 enrolled in the Lipid Research Clinics Coronary Primary Prevention Trial. Study measurements were made bimonthly. After adjusting for the effects of the trial intervention and other determinants of lipid levels, the cholesterol levels of the cases diagnosed with nonlocalized cancer dropped below the expected level approaching diagnosis when compared to the entire study population. The decrease averaged 9.3 mg/dl and began about 2 years prior to diagnosis. Weight levels dropped an average of 1.2 kg over the same period. Weight and cholesterol were significantly lower than expected within 8 months of diagnosis (P less than 0.05). No decrease was seen for those diagnosed with localized malignancies. Patterns for low-density lipoprotein cholesterol reflected those of total cholesterol. There was no clear relationship between cancer diagnosis and patterns of change for triglycerides and high-density lipoprotein cholesterol. In the future, investigations of any relationship between a host physiological state and cancer occurrence should account for the metabolic effects of preclinical disease demonstrated here. To protect against spurious conclusions, incident cases occurring within 2 years of measurement should be analyzed separately. In studies of cancer mortality, deaths occurring within 3.5 years of the base-line measurement should be analyzed separately.