Predictive performance of serum digoxin concentration in patients with congestive heart failure by a hyperbolic model based on creatinine clearance

Objectives: To formulate a simple equation for determining the daily dose requirements of digoxin by inclusion of creatinine clearance (Ccr) values as an explanatory variable. Methods: We included 235 routine monitoring and clinical laboratory test data (steady-state serum digoxin concentration and Ccr values), obtained from hospitalized patients receiving digoxin for treatment of congestive heart failure. The 107 data sets were fitted to a hyperbolic model to account for the relation between the ratio of serum digoxin level to the daily dose (L/D) and the Ccr values determined by six methods. Their correlation coefficients (r) were computed by non-linear regression analysis. To evaluate the validity of the best-fitting model, the predictive performance of the L/D ratios was compared with those given by seven reference models previously published, using another 128 data sets. Results: The hyperbolic model involving the Ccr values estimated by Cockcroft and Gault's equation showed the closest correlation (r = 0.81) between the actual and estimated Ccr values. Mean prediction error (ME), a measure of bias, of the L/D ratio (0.018 ng/mL) was almost negligible when other data were fitted to the proposed model, and this ME value proved to be much smaller than those calculated from the previously published prediction models. Mean absolute prediction error, a measure of precision, by the proposed model was also satisfactory for prediction. Conclusions: The newly developed model provided good predictive performance of serum digoxin level. Taking simplicity in practical use into account, the clinical application of the proposed model will allow for accurate and rapid determination of the initial maintenance dosing regimen of digoxin based on the individual Ccr value, without actual measurement of its serum concentration.