Goodness-of-fit and local identifiability of a receptor-binding radio-pharmacokinetic system

Abstract
A four-state nonlinear model describing a radiopharmacokinetic system for a hepatic receptor-binding radiopharmaceutical, [99mTc]-galactosyl-neoglycoalbumin (TcNGA), was tested for goodness-of-fit and local identifiability using scanning data from nine healthy subjects and seven patients with severe liver disease. Based on standard deviations of liver and heart imaging data at equilibria as a measure of observational error, the reduced chi-square ranged from 0.5 to 2.6. Values above 1.2 occurred when the subject moved during the 30 min study. Relative standard errors for each parameter were: TcNGA-receptor forward binding rate constant kb, 13-54%; extra-hepatic plasma volume Ve, 0.8-15.0%; hepatic plasma volume Vh, 0.2-6.5%; hepatic plasma flow F, 54----greater than 1000%; and receptor concentration [R]o, 0.3-13%. The highest standard errors occurred when the amount of TcNGA injected exceeded the total amount of receptor. Therefore, when TcNGA functional imaging was performed without excess patient motion and receptor saturation, the kinetic model provided data fits of low systematic error and yielded high precision estimates of receptor concentration and forward binding rate constant. In summary, optimal performance of the kinetic model occurred when the amount of injected TcNGA resulted in the nonlinear operation of the pharmacokinetic system.