LAUSR

Purpose: This study evaluated the predictive performances of four equations (Sheiner‐Tozer [ST], Kane‐modified ST, Anderson‐modified ST, and Cheng‐modified ST) used to estimate free phenytoin concentrations in a medical intensive care unit (MICU) and assessed the impact of exogenously administered albumin. Methods: Thirty MICU subjects receiving phenytoin were retrospectively evaluated. Predictive performances were assessed by mean absolute error (MAE), mean prediction error (MPE), and root mean squared error (RMSE); linear regression of predicted vs. actual unbound concentrations; and Bland‐Altman plots. Parameters were further delineated by recent exogenous albumin administration. Results: The measured unbound phenytoin concentration was 2.14 ± 0.84 &mgr;g/mL for all 90 levels, 1.89 ± 0.92 &mgr;g/mL for the 58 levels without albumin, and 2.58 ± 0.83 &mgr;g/mL (p < 0.0001 vs. without albumin) for the 32 levels after exogenous albumin. R2 values were below 0.4 for all equations. The ST equation over‐predicted unbound concentrations whereas all other equations under‐estimated unbound concentrations. All equations possessed bias and lacked precision. Bland‐Altman plots demonstrated greatest bias with the ST equation. Albumin administration introduced additional bias, limited precision, reduced R2 values, and completely negated the performances of the ST and Kane‐modified ST equations. Conclusion: Estimating unbound concentrations with equations in the MICU population is discouraged. HighlightsSeveral equations are frequently used in the critically ill to estimate unbound phenytoin levels despite not being validatedThis evaluation demonstrated poor predictive performances of four equations in medical ICU patientsThe administration of exogenous albumin completely negated the performances of two equations