A protocol using stable isotopes, developed to measure apolipoprotein B turnover in humans, was tested by intravenous infusion of [2H3]-leucine into 5 normolipidemic volunteers during a 14-h fast. Tracer-to-tracee ratio curves were analyzed by four different approaches: linear regression, monoexponential regression, a minimal compartmental model (3 compartments) and a complex model (4 compartments and two shunt pathways). The three-compartment model was validated by qualitative analysis of data obtained after injection of a [2H3]-leucine bolus. This simple model gave an FCR of 0.48 +/- 0.05 h-1 for VLDL, 0.62 +/- 0.08 h-1 for IDL and 0.022 +/- 0.002 h-1 for LDL. The total production rate of apolipoprotein B in plasma was 24.8 +/- 6.5 mg.kg-1.day-1. Kinetic parameters were similar for the complex model which showed no improvement in fit. Lower estimates were observed with the non-compartmental approaches.