Objective: Healthy subjects and patients with rheumatoid arthritis (RA) exhibit circadian rhythms of the neuroendocrine-immune system. Understanding circadian dynamics is complex due to the nonlinear behavior of the neuroendocrine-immune network. This study was undertaken to seek and test a mathematical model for studying this network.
Methods: We established a quantitative computational model to simulate nonlinear interactions between key factors in the neuroendocrine-immune system, such as plasma tumor necrosis factor (TNF), plasma cortisol (and adrenal cholesterol store), and plasma noradrenaline (NA) (and presynaptic NA store).
Results: The model was nicely fitted with measured reference data on healthy subjects and RA patients. Although the individual circadian pacemakers of cortisol, NA, and TNF were installed without a phase shift, the relative phase shift between these factors evolved as a consequence of the modeled network interactions. Combined long-term and short-term TNF increase (the "RA model") increased cortisol plasma levels for only a few days, and cholesterol stores started to become markedly depleted. This nicely demonstrated the phenomenon of inadequate cortisol secretion relative to plasma TNF levels, as a consequence of adrenal deficiency. Using the RA model, treatment with glucocorticoids between midnight and 2:00 AM was found to have the strongest inhibitory effect on TNF secretion, which supports recent studies on RA therapy. Long-term reduction of TNF levels by simulation of anti-TNF therapy normalized cholesterol stores under "RA" conditions.
Conclusion: These first in silico studies of the neuroendocrine-immune system in rheumatology demonstrate that computational biology in medicine, making use of large collections of experimental data, supports understanding of the pathophysiology of complex nonlinear systems.