Validation of a computational platform for the analysis of the physiologic mechanisms of a human experimental model of hemorrhage

Abstract

Computational models of integrative physiology may serve as a framework for understanding the complex adaptive responses essential for homeostasis in critical illness and resuscitation and may provide insights for design of diagnostics and therapeutics. In this study a computer model of human physiology was compared to results obtained from experiments using Lower Body Negative Pressure (LBNP) analog model of human hemorrhage. LBNP has been demonstrated to produce physiologic changes in humans consistent with hemorrhage. The computer model contains over 4000 parameters that describe the detailed integration of physiology based upon basic physical principles and established biologic interactions. The LBNP protocol consisted of a 5 min rest period (0 mmHg) followed by 5 min of chamber decompression of the lower body to −15, −30, −45, and −60 mmHg and additional increments of −10 mmHg every 5 min until the onset of hemodynamic decompensation ( n = 20). Physiologic parameters recorded include mean arterial pressure (MAP), cardiac output (CO), and venous oxygen saturation (SVO 2; from peripheral venous blood), during the last 30 s at each LBNP level. The computer model analytic procedure recreates the investigational protocol for a virtual individual in an In Silico environment. After baseline normalization, the model predicted measurements for MAP, CO, and SVO 2 were compared to those observed through the entire range of LBNP. Differences were evaluated using standard statistical performance error measurements (median performance error (PE) <5%). The simulation results closely tracked the average changes observed during LBNP. The predicted MAP fell outside the standard error measurement for the experimental data at only LBNP −30 mmHg while CO was more variable. The predicted SVO 2 fell outside the standard error measurement for the experimental data only during the post-LBNP recovery point. However, the statistical median PE measurement was found to be within the 5% objective error measure (1.3% for MAP, −3.5% for CO, and 3.95% for SVO 2). The computer model was found to accurately predict the experimental results observed using LBNP. The model should be explored as a platform for studying concepts and physiologic mechanisms of hemorrhage including its diagnosis and treatment.