We investigated whether decline over time in serum cholesterol was associated with the risk of death from cancer in French men. We studied 6,230 working men, age 43-52 years in 1967-1972, who had at least three annual measurements of serum cholesterol. We estimated individual change over time in serum total cholesterol using within-person linear regression. During an average of 17 years of follow-up after the last examination, 747 subjects died from cancer. The multivariate-adjusted relative risks for subjects in the fourth (highest increase in serum total cholesterol), third, and second quartiles, compared with men in the first quartile (who had a decrease in serum total cholesterol), were 0.70 [95% confidence interval (CI) = 0.56-0.87], 0.71 (95% CI = 0.57-0.88), and 0.74 (95% CI = 0.61-0.91), respectively. The group with the highest decline in cholesterol displayed an excess risk for most cancer sites. These associations were more pronounced in subjects whose weight remained stable or decreased over time than in those who gained weight.