The broad use of pharmacometrics in patient care: the case of blood transfusion and perioperative management
Nicolás Marco-Ariño (1,2), Sebastian Jaramillo (3), Pedro L Gambús (3), Iñaki F. Trocóniz (1,2)
(1) Pharmacometrics & Systems Pharmacology, Department of Pharmaceutical Technology and Chemistry, School of Pharmacy and Nutrition, University of Navarra, Pamplona, Spain (2) IdiSNA; Navarra Institute for Health Research, Pamplona, Spain (3) Systems Pharmacology Effect Control & Modeling (SPEC-M) Research Group, Anesthesiology Department, Hospital CLINIC de Barcelona, Barcelona, Spain
Introduction:
Despite millions of surgical procedures are performed every year1, multiple areas of perioperative management lack quantitative tools to guide decision-making and rely in indirect signs or clinician´s own criteria2. Modelling and simulation represents a powerful tool to optimise drug administration, particularly for those interventions where careful titration to attain the right therapeutic vs side effect balance is required. In this work we propose a population PK/PD model to characterise the changes in haemoglobin concentrations after surgery, hence supporting decision-making in red blood cells transfusion and perioperative care.
Methods:
Clinical study:
Haemoglobin concentrations were measured before surgery, post-surgery (2 to 6 hours after the initial measure) and every 24h (up to four days) in 156 patients undergoing urological or gynaecological surgery. Of those, 120 subjects were used to build the model, while 36 patients enrolled in a second period of recruitment were used for external model validation. Administered fluid therapy, lost volumes of blood and urine were recorded during the time course of the study.
PK/PD model:
Haemoglobin concentration is the ratio between the mass of haemoglobin and volume of blood. Due to the slow turn-over of haemoglobin compared to the study period, variations in haemoglobin concentration after surgery were considered to be the result of blood losses and an imbalance between fluid intake and elimination. In our model, the initial mass of haemoglobin for a patient was calculated from the individual blood volume and the value of haemoglobin concentration prior to surgery. Haemoglobin losses are subtracted from this amount to account for bleeding occurred during surgery. The resulting mass of haemoglobin is then divided by the blood volume, which was characterised by a two-compartment model with linear elimination, to describe haemoglobin concentrations over time. All data were analysed using the population approach with NONMEM 7.4. Model evaluation was based on goodness of fit plots and visual predictive checks. Surgery and patient-related characteristics were evaluated as potential model covariates using the stepwise covariate model (SCM) method with p-values of 0.05 for the stepwise forward inclusion and 0.01 for the backward elimination. For model validation, the empirical Bayes estimates (EBE) and individual predictions of haemoglobin concentrations were obtained for the subjects in the validation dataset and predicted errors were computed to quantify the precision of the predictions.
Results:
Our model was able to capture the change in haemoglobin concentrations after a surgical procedure. Model evaluation showed adequate model performance. All parameters were estimated with precision [coefficient of variation (CV) <30%] and inter-individual variability ranged from 21 to 73% CV. Fluid elimination was found to be inhibited after surgery in a time dependent manner and was characterised with an IMAX model with a maximum inhibition of 42%, and a time to 50% of maximum inhibition of 9.1 hours. Renal excretion represented only 24% of the total fluid elimination.
The type of surgical procedure was shown to affect the proportion of fluid eliminated via urine with a magnitude ranging from -31% to and 85% increase for certain types of gynaecological surgery. Age was also found to have a moderate effect on total fluid loss (20% change regarding reference value). On the contrary, neither the type of fluid supply (saline or glucose solutions), the volume of haemorrhage nor inflammatory markers were found to correlate with model parameters.
Excellent predictive capacity for the model was demonstrated with 97.7% percent of the predicted errors being smaller than 5%. Particularly, the predicted errors for haemoglobin levels below 10 g/dl (threshold of reference for considering RBC transfusion) were in the same range of magnitude, reinforcing the objective of the model.
Conclusions:
In this work, we have characterised haemoglobin concentrations using variables collected routinely in surgical procedures, providing an opportunity to guide fluid administration and blood transfusion in a simple and cost-effective manner. Altogether, this model provides a quantitative framework in perioperative management, supporting decision-making in this complicated scenario.
References:
[1] Weiser TG, Haynes AB, Molina G, et al. Estimate of the global volume of surgery in 2012: an assessment supporting improved health outcomes. Lancet Elsevier BV; 2015; 385: S11
[2] Hannon T. Do you catch my drift? Transfusion. Blackwell Publishing Inc.; 2014. p. 1448–9
remifentanil. II. Model application. Anesthesiology 1997; 86: 24–33