Computer modelling of adrenal function

The literature pertaining to the modelling of the adrenal-cortical system is reviewed. Two new models describing the behaviour of the adrenal gland under conditions of stimulation from the hormone ACTH are developed. Both are capable of predicting the observed overshoot in cortisol secretion following a step rise in ACTH concentration at the gland. The first relies on depletion of available cholesterol stores in the gland, while the second relies on the proposal of Berger that ACTH stimulates pregnenolone formation and inhibits 17α hydroxyprogesterone formation. Analysis of measurements of ACTH concentrations in sheep show that real differences between the amount of ACTH entering and leaving the gland do occur. This difference is not due to loss of water or ACTH, nor to transport delays. Therefore, ACTH must be taken up by or otherwise destroyed within adrenal tissue. Linear models of the inactivation (within the body) of ACTH are studied. It is found that a model with only two compartments explains the differences between bioassay and immunoassay measurements of ACTH concentration. The first compartment contains biologically active ACTH (time constant of breakdown = 12.9 ± 2.6 min), and the second biologically inactive ACTH which retains immunoactivity (245 ± 105 min). The model may be used to provide estimates of the concentration of biologically active ACTH from immunoassay data. A continuous simulation language for the EAI 640 computer is described. This language allows a simulator to study the equations of a small model interactively using a visual display unit and active teletypewriter. The development of a magnetic tape storage system for the EAI 640 computer is described. The tape system is a backup store for the EAI 260 disc storage system allowing the disc to be kept free for each computer user, and thus allowing larger programs to be executed. The hardware and the software for this system are described. Between animal variation is not often included in simulation studies. When it is included, it is usually estimated by Monte Carlo techniques, which involve computing large numbers of simulations with the values of the model parameters chosen at random. A method - called the "variant function technique!! - is introduced here. The model parameters are replaced by oscillatory functions of time. This permits the statistical behaviour of the model to be estimated from a single extended simulation, for which the computation time needed is strongly dependent upon the number of parameters