SAT-738 Sources of Error in Estimation of Cortisol Half-Life Using Conventional, Single-Compartment Model: Bias Due to Variation in CBG Concentration

Abstract

Most reports of cortisol half-life in the literature report a range of 90-130 min, which results are based on descriptive model that assumes mono-exponential decay of a single, total cortisol compartment. Free cortisol half-life has been similarly assessed using a descriptive single compartment model (1). However, the descriptive model is not physiologic in view of the rapid exchange between protein-bound and free cortisol compartments and evidence that metabolic elimination is restricted to the free cortisol compartment. In the present study, we sought to explore potential limitations of the descriptive, single-compartment model for cortisol elimination by assessing the influence of CBG concentration ([CBG]) on cortisol half-life estimates obtained using the descriptive model. We studied the influence of [CBG] and other variables on descriptive cortisol half-life using a Monte Carlo simulation of cortisol concentration decay curves developed using data from healthy controls (1). Total cortisol concentration ([TF]) curves were generated on the basis of 4 predictor variables: (i) [CBG], (ii) albumin concentration, (iii) [TF] at time zero following iv bolus (total cortisol at time 0, y-intercept), and (iv) free cortisol half-life central to a mechanistic (dynamic, 3-compartment) model (2). Simulations used a multivariable normal distribution and selected means, SDs, and correlation structure among these 4 variables in healthy controls. After generation of a series of cortisol decay curves (n=1000), half-lives for total and free cortisol were solved using the conventional (descriptive, single-compartment) model. The influence of predictor variables on conventional half-life estimates were assessed using standardized beta (STB) coefficients, which represent change in the SD of the outcome (numerator, i.e. total or free cortisol half-life obtained by descriptive model) for each SD change in a predictor (denominator) in a multivariable context (3). For total cortisol half-life (descriptive model) STBs were 0.91 ([CBG]), 0.73 (free cortisol half-life), -0.37 (y-intercept), and 0.04 ([albumin]) (all P<0.001). For free cortisol half-life (descriptive model), STBs were 0.98 ([CBG]), 0.73 (free cortisol half-life), -0.78 (y-intercept), and 0.11 [albumin]) (all P<0.001). We conclude that the conventional descriptive model for estimation of cortisol has significant limitations, including inaccuracy and systematic bias related to the influence of CBG concentration on half-life estimates. By inference, a similar bias confounds interpretation of the half-life obtained using conventional single-compartment model of other hormones associated with high-affinity serum binding proteins. References: (1) Perogamvros et al. Clin Endo 2011;74:30-36, (2) Keenan et al. Am J Physiol Endocrinol Metab 2004;287:E652-E661 (3) Dorin et al., J Endocrinol Soc 2017 July;1(7):945-56.

Four-compartment diffusion model of cortisol disposition: Comparison to three alternative models in current clinical use

“…The mathematical comparison of t 3 and t 4 is to be determined. However, using Monte Carlo simulations to generate cortisol concentration elimination data (19), the ordered sequence of solutions t3>t4 was consistently observed (data not shown).…”