Senior Modeler Rosa & Co. LLC Estepona, Andalucia, Spain
Disclosure(s):
Alvaro Ruiz-Martinez, PhD: No financial relationships to disclose
Objectives: Cushing’s disease is a rare condition that results from having too much cortisol in the blood. The most common cause is a pituitary noncancerous tumor (adenoma) secreting adrenocorticotropic hormone (ACTH), which causes increased secretion of cortisol [1]. The biological pathway regulating cortisol production in the hypothalamic-pituitary-adrenal (HPA) axis has been modeled previously [2-4], but hourly ACTH and cortisol oscillations are not captured easily. Also, tissue growth occurs over the course of weeks to months and stops if hormone concentrations decrease to normal levels. For these reasons, a proper Cushing’s disease model should be stable and accurate from hours to months. This work recalibrates the model presented in [5] to represent the HPA axis dynamics accurately and expands it with the aim of studying Cushing’s disease.
Methods: The starting point model [5] represents corticotropin-releasing hormone (CRH), ACTH, and cortisol with appropriate feedback regulation of hormone production. It also includes an artificial construct for setting a circadian rhythm. To this structure, we have added tissue mass, outcome measures, and drug or hormone treatments. Since neither the original calibration nor the recalibration from a posterior study [6] function for more than a 24-hour simulation, we have developed an alternative calibration process that gives a better fitting and a stable concentration pattern every 24-hour cycle. In our approach, the fitting of a healthy virtual patient (VP) is performed using a 72-hour dataset by repeating the original 24-hour dataset three times in a row and assigning more weight to the high concentration datapoints. Besides a healthy VP, a Cushing’s disease VP (CD VP) has also been generated and calibrated for longer-term simulations. Both can be used to simulate clinical trials lasting for weeks to months. Dexamethasone and CRH tests, which are used in the diagnosis of Cushing’s disease, have been implemented to test that both VPs match clinical data.
Results: For the untreated healthy VP case, model results from [5] and [6] show most of the hormone concentration peaks and ultradian oscillations rapidly decreasing after the first 24 hours of simulation, which is not consistent with clinical data. Our improved calibration, however, solves that issue and renders the model into an efficient tool to perform long-term scenarios. Similarly, the CD VP, that represents an average Cushing’s disease patient, has time-averaged hormone concentrations that are in agreement with the data [7]. The model reproduces and is consistent with dexamethasone (DEX) and CRH test results for both VPs. For the former, the model captures a greater than 50% inhibition of cortisol concentration for an 8 mg DEX dose reported in literature [8]. For the latter, it captures the fold change of ACTH and cortisol concentrations for a standard 8 mg ovine CRH test [9] and predicts the fold change in human.
Conclusions: The HPA axis model simulations match published single-day clinical data and remain stable and accurate in long term scenarios. The model also replicates a healthy and a Cushing’s disease patient and both VP responses to CRH and dexamethasone tests are in agreement with results reported in literature.
Citations: [1] S. Takayasu et al. Cancers (Basel) (2023) 15 (2) [2] A.N. Churilov and J.G. Milton. Sci Rep (2022) 12 (1):8480 [3] P. Caruso. Virginia Polytechnic Institute and State University (2022) [4] Y. Li et al. J Pharm Sci. (2023) Aug 18;113(1):33–46 [5] E.O. Bangsgaard and J.T. Ottesen. Math Biosci (2017) 287:24-35 [6] C. Parker et al. Entropy (Basel) (2022) 24 (12) [7] Liu et al. J Clin Endocrinol Metab (1987) 64 (5):1027-35 [8] M. Detomas et al. Front Endocrinol (Lausanne) (2022) 13:955945 [9] J.H. Schluger et al. Neuropsychopharmacology (2003) 28 (5):985-94
