Understanding the physiology of the hypothalamic-pituitary-adrenal axis to optimise cortisol replacement therapy: A quantitative modelling framework
Davide Bindellini (1, 2), Robin Michelet (1), Johanna Melin (1, 2), Linda B.S. Aulin (1), Wilhelm Huisinga (3), Uta Neumann (4), Oliver Blankenstein (4, 5), Martin J. Whitaker (6), Richard Ross (6), Charlotte Kloft (1)
(1) Dept. of Clinical Pharmacy and Biochemistry, Institute of Pharmacy, Freie Universitaet Berlin, Germany, (2) and Graduate Research Training program PharMetrX, Germany, (3) Institute of Mathematics, University of Potsdam, Germany (4) Charité-Universitätsmedizin Berlin, Berlin, Germany (5) Labor Berlin, Charité Vivantes GmbH, Berlin, Germany (6) University of Sheffield, United Kingdom
Objectives: Congenital adrenal hyperplasia (CAH) is a life-threatening condition characterised by a loss of cortisol secretion, mainly caused by a deficiency in the enzyme 21-hydroxylase. Hydrocortisone (synthetic cortisol), is the recommended drug for cortisol replacement therapy in paediatric CAH patients [1, 2]. The hypothalamic-pituitary-adrenal (HPA) axis controls cortisol secretion through pituitary adrenocorticotropic hormone (ACTH). Furthermore, both ACTH and cortisol have a circadian rhythm characterised by a peak on waking, a rise at midday and nadir at night [3]. Ideally, replacement therapy should mimic physiological cortisol circadian rhythm. Yet, despite many efforts to characterise hydrocortisone PK, a quantitative understanding of the HPA axis regulation in the healthy state, and its alterations in CAH, could provide insights into how to improve cortisol replacement therapy in patients. The aim of this study was to quantify key processes involved in the HPA axis activity regulation in healthy adults.
Methods: Data from 14 healthy adult males [4] were leveraged to develop a physiologically motivated NLME model. ACTH and cortisol concentrations were measured during 24 h following no intervention and during 12 h following dexamethasone (DEX) induced cortisol suppression. The model was parametrised in terms of unbound concentrations using a previously developed cortisol binding model [5]. Cortisol disposition was described using EBEs from a previously developed hydrocortisone PK model in the same individuals [5]. Surge functions were utilised to describe pulsatile secretion of ACTH, together with a baseline secretion. DEX was assumed to fully suppress ACTH pulsatile secretion. The effect of ACTH on cortisol production was tested using linear, log-linear and Emax models. The implementation of a negative feedback loop, for which unbound cortisol suppressed ACTH pulsatile secretion, was tested using an Imax model. IIV was tested on all structural parameters assuming a lognormal distribution. The effect of age and body weight was tested on all parameters for which IIV was included (forward inclusion: p<0.05, backward deletion: p<0.005). The model was evaluated using GOF plots, SIR and VPC.
Results: ACTH concentration-time profiles were best described by a zero-order baseline secretion and first-order elimination, in combination with two surge functions (morning and midday peaks). The morning and midday surges had peak times of 06:16 and 12:02, amplitude of 1210 pmol and 22 pmol, and width of 0.828 h and 2.93 h, respectively. Cortisol production by ACTH was best described using an Emax model with estimated EC50=6.87 pmol/L and Emax=5440 nmol, indicating that these subjects did not reach maximum cortisol production at ACTH peak concentrations. An Imax model best characterised the negative feedback loop of cortisol on ACTH pulsatile secretion with fixed Imax=1 and estimated IC50=3.65 nmol/L (~120 nmol/L total cortisol concentration), showing ACTH secretion is mostly suppressed through the entire daytime. Based on surge amplitude values, a large ACTH overproduction would be observed in the morning when cortisol is not produced (CAH patients). IIV was included on the morning surge amplitude (96.1% CV), Kout (54.6% CV), ACTH baseline concentration (29.6% CV), EC50 (35.9% CV) and IC50 (21.2% CV). Additionally, a correlation between baseline ACTH and EC50 was implemented (correlation coefficient=0.968). No covariates were retained in the model following backward deletion. No model misspecification was observed upon inspection of GOF plots, while parameters precision and predictive performance were considered adequate.
Conclusions: The developed NLME model in healthy volunteers provided physiologically interpretable parameters related to cortisol production under physiological HPA axis activity and suppression by DEX, and subsequent feedback inhibition on ACTH secretion. The characterisation of such processes allows for a deeper quantitative understanding of the whole cortisol pathway and HPA axis regulation. The model can be further extended to healthy children and ultimately to CAH patients, paving the way towards optimisation of cortisol replacement therapy.
References:
[1] Merke D et al. Lancet (2005) 365, 2125-2136
[2] Speiser W et al. JCEM (2018) 103(11), 4043-4088
[3] Krieger D et al. JCEM (1971) 32(2), 266-284
[4] https://clinicaltrials.gov/ct2/show/study/NCT01960530
[5] Melin J et al. ClinPK (2018) 57(4), 515-527