The inverse relationship between plasma levels of high density lipoprotein (HDL) and coronary heart disease rates has suggested that HDL might influence body stores of cholesterol. Therefore, we have investigated potential relationships between the parameters of body cholesterol metabolism and the plasma levels of HDL cholesterol and the major HDL apoproteins. The study involved 55 human subjects who underwent long-term cholesterol turnover studies, as well as plasma lipoprotein and apolipoprotein assays. In order to maximize the likelihood of detecting existing relationships, the subjects were selected to span a wide range of plasma levels of lipids, lipoproteins, and apolipoproteins. Single univariate correlation analyses suggested weak but statistically significant inverse relationships of HDL cholesterol and apoA-I levels with the following model parameters: production rate (PR), the mass of rapidly exchanging body cholesterol (M1), the minimum estimate of the mass of slowly exchanging body cholesterol (M3min), and of the mass of total exchangeable body cholesterol (Mtotmin). These correlations, however, were quantitatively quite small (/r/ = 0.28-0.42) in comparison to the strength of the univariate relationships between body weight and PR (r = 0.76), M1 (r = 0.61), M3min (r = 0.58), and Mtotmin (r = 0.78). Correlations for apoA-II and apoE levels were even smaller than those for apoA-I and HDL cholesterol. In additional analyses using multivariate approaches, HDL cholesterol, apoA-I, apoA-II, and apoE levels were all found not to be independent determinants of the parameters of body cholesterol metabolism (/partial r/ less than 0.17, P greater than 0.3 in all cases). Thus the weak univariate correlations reflect relationships of HDL cholesterol and apoA-I levels with physiological variables, such as body size, which are primarily related to the model parameters. We conclude that plasma levels of HDL cholesterol and apoproteins A-I, A-II, and E are not quantitatively important independent determinants of the mass of slowly exchanging body cholesterol or of other parameters of long-term cholesterol turnover in humans. These studies give no support to the hypothesis that the inverse relationship between HDL cholesterol levels and coronary heart disease rates is mediated via an influence of HDL on body stores of cholesterol.

During the past few years, kinetic studies have provided a considerable amount of information about cholesterol turnover and metabolism in intact humans. These studies have generally involved analysis of the turnover of plasma cholesterol following injection of radioactively labeled cholesterol, in experiments of about 10-12 weeks duration (Goodman and Noble 1968; Nestel et al. 1969; Goodman et al. 1973a), or of much longer duration (Samuel and Perl 1970; Goodman et al. 1973b; Samuel and Lieberman 1973; Smith et al. 1976). In 1976, we reported the results of long-term studies (32-49 weeks) of the turnover of plasma cholesterol in 8 normal and 16 hyperlipidemic subjects (Smith et al. 1976). We now report the results of similar long-term studies carried out in a total population of 54 subjects.

Total body turnover of cholesterol was studied in 54 subjects by fitting a three-pool mathematical model to plasma decay curves of 32--49 weeks duration following [14C]cholesterol injection. Fifteen subjects were normal, 10 hypercholesterolemic, 21 hypertriglyceridemic, and 8 had both hypercholesterolemia and hypertriglyceridemia; 21 had a familial form of hyperlipidemia. In every subject in this heterogeneous population, the three-pool model gave the best fit for the data. An extensive search was conducted for relationships between model parameters and physiological variables (body size, serum lipid levels, age, and sex). Both linear and nonlinear relationships, and those involving interactions between pairs of variables, were explored. Fifty different forms of the model parameters and 53 forms of the physiological variables were examined. To guard against declaring statistical significance when none was present, subjects were first randomly divided into two matched groups. In the first (hypothesis-generating) group of 36 subjects, more than 100,000 regression equations were considered for each form of the model parameters. Twenty-one highly significant equations were found that were then tested in the second group (hypothesis-testing, 18 subjects). Eighteen of the 21 equations were found to be significant; of these, 6 were selected that accounted for a large part of the observed variation in the four model parameters for which equations were found (production rate (PR), and the sizes of pool 1, pool 3, and total exchangeable body cholesterol). The major determinant of cholesterol PR was body weight alone (r = 0.80). No function of serum lipid levels significantly influenced PR. Both body weight and serum cholesterol level entered into the equations for cholesterol mass. Age influenced the size of pool 3. Serum triglyceride level only had an effect on the size of pool 1. Since these equations were generated in one group of subjects and tested in another, they can be considered a confirmed set of predictive equations.