Precision Medicine in Critical Care Requires an Understanding of Pharmacokinetic Variability*

    loading  Checking for direct PDF access through Ovid


Pharmacotherapy in the PICU remains largely trial and error—the application of standardized weight- and renal function-based dosing algorithms and reactive calibration by intensivists and pharmacists. Better prediction and individualized medication dosing is a core principle of precision medicine but remains a challenge in the ICU and in pediatrics where pharmacology data are sparse. Pharmacokinetic changes are well-known to be drug-specific and time-sensitive following acute disease progression or injury, resuscitation, and recovery. It is therefore critical to understand these relationships, especially for antibiotics like vancomycin where concentration-effect relationships and the importance of early individualized therapy are established (1, 2).
Treatments that independently alter drug pharmacokinetics add an additional layer of complexity. Targeted temperature management is recommended in neonates with hypoxic-ischemic encephalopathy (HIE) and for infants and children with out-of-hospital cardiac arrest (3, 4). Many studies have documented that hypothermia targeting a core temperature of 32–34°C decreases phase I cytochrome p450 metabolism resulting in increases in concentrations of hepatically eliminated medications like midazolam (5, 6), fentanyl (5), and phenobarbital (7). For drugs with long elimination half-lives such as phenytoin, consequences may also persist for extended time periods beyond rewarming (8). Increasingly, there is also an appreciation of pharmacokinetic interactions between hypothermia and the underlying injury (e.g., asphyxia or cardiac arrest) that exceed either alone (7, 9). Less consistent information is available regarding the impact of cooling on passive drug elimination processes such as glomerular filtration. Recent data suggest that the clearance of gentamicin which is primarily eliminated unchanged in the urine is reduced in neonates receiving hypothermia for HIE (10).
In this issue of Pediatric Critical Care Medicine, Zane et al (11) evaluated vancomycin disposition in 52 children who were treated with either hypothermia (32–34°C) or normothermia (36.3–37.6°C) after cardiac arrest. They report reductions in clearance of up to 80% with reduced or poor renal function, up to 25% with hypothermia treatment, and up to 84% when these covariates are combined. Limitations of the retrospective research and its small sample size are offset by the robust pharmacometric analysis performed and the observation that vancomycin is surprisingly understudied during therapeutic hypothermia, despite it being a workhorse in the ICU for the treatment of methicillin-resistant gram-positive infections. The observation that vancomycin clearance was significantly reduced with decreasing glomerular filtration rate (GFR) was expected as vancomycin is primarily renally eliminated (80–90%). The reduction in clearance attributed to therapeutic hypothermia was smaller in comparison and contradicts a study of therapeutic hypothermia in acute brain injury that reported no differences vancomycin pharmacokinetics versus what was calculated using population data (12).
This work shows the utility of nonlinear mixed effects modeling to understand pharmacokinetics in critically ill pediatric populations. This approach 1) allowed the use of levels already obtained as part of routine clinical care in a patient population where intensive sampling would otherwise not have been feasible; 2) provided for an evaluation of temperature, renal function, and weight as continuous covariates; and 3) resulted in a quantification of the relative impact of individual covariates on pharmacokinetic parameters (13). The latter is notable because it enabled simulations of expected trough concentrations and is useful for dosing algorithm development.
Specific weaknesses should also be noted. Most importantly, although the modeling was successful, the parameter estimates reveal significant unexplained interindividual variability in clearance of nearly 50%. The small number of patients and samples likely led to the decision to use a prespecified covariate model. In this situation, low powered covariates may falsely appear clinical important and an evaluation of additional covariates that may have further explained variability was not conducted.

Related Topics

    loading  Loading Related Articles