Mechanism-Based Modeling of Perioperative Variations in Hemoglobin Concentration in Patients Undergoing Laparoscopic Surgery

Abstract

BACKGROUND: Hemoglobin concentration ([Hb]) in the perioperative setting should be interpreted in the context of the variables and processes that may affect it to differentiate the dilution effects caused by changes in intravascular volume. However, it is unclear what variables and processes affect [Hb]. Here, we modeled the perioperative variations in [Hb] to identify the variables and processes that govern [Hb] and to describe their effects. METHODS: We first constructed a mechanistic framework based on the main variables and processes related to the perioperative [Hb] variations. We then prospectively studied patients undergoing laparoscopic surgery, divided into 2 consecutive cohorts for the development and validation of the model. The study protocol consisted of serial measurements of [Hb] along with recordings of hemoglobin mass loss, blood volume loss, fluid infusion, urine volume, and inflammatory biomarkers measurements, up to 96 hours postoperatively. Mathematical fitting was performed using nonlinear mixed-effects. Additionally, we performed simulations to explore the effects of blood loss and fluid therapy protocols on [Hb]. RESULTS: We studied 154 patients: 118 enrolled in the development group and 36 in the validation group. We characterized the perioperative course of [Hb] using a mass balance model that accounted for hemoglobin losses during surgery, and a 2-compartment model that estimated fluid kinetics and intravascular volume changes. During model development, we found that urinary fluid elimination represented only 24% of the total fluid elimination, and that total fluid elimination was inhibited after surgery in a time-dependent manner and influenced by age. Also, covariate evaluation showed a significant association between the type of surgery and proportion of fluid eliminated via urine. In contrast, neither the type of infused solution, blood volume loss nor inflammatory biomarkers were found to correlate with model parameters. In the validation analysis, the model demonstrated a considerable predictive capacity, with 95% of the predicted [Hb] within −4.4 and +5.5 g/L. Simulations demonstrated that hemoglobin mass loss determined most of the postoperative changes in [Hb], while intravascular volume changes due to fluid infusion, distribution, and elimination induced smaller but clinically relevant variations. Simulated patients receiving standard fluid therapy protocols exhibited a hemodilution effect that resulted in a [Hb] decrease between 7 and 15 g/L at the end of surgery, and which was responsible for the lowest [Hb] value during the perioperative period. CONCLUSIONS: Our model provides a mechanistic and quantitative understanding of the causes underlying the perioperative [Hb] variations.