Estimation, resolution and solvability of linear compartmental models: cholesterol metabolism in man

The paper presents a method to test metabolic linear multicompartment models for their mathematical solvability and their power of resolution with respect to individual parameters. The method can be seen as a means of predicting not only the model theoretical identifiability but also the expected quality of estimation before undertaking any actual serial experiments. The compartmental model of human cholesterol metabolism is taken as a demonstrative problem. The analysis shows that it is possible with tracer techniques to measure the cholesterol turnover, the cholesterol fluxes between exchangeable cholesterol pools, the cholesterol masses in individual pools, and the endogenous cholesterol synthesis in individual pools if certain conditions are met. In any complex kinetic situation, performance of a systematic model analysis is recommended before embarking on large-scale experimentation.