Indirect estimation of the unbound fraction of cyclosporine in plasma.

ABSTRACT

The unbound fraction (fU) of cyclosporine in plasma is approximately 0.02. The measurement of cyclosporine fU requires a laborious equilibrium dialysis procedure, which is not practical in a clinical setting. A mathematical model was developed to estimate cyclosporine fU from concentrations of serum lipoproteins, the major binding proteins for cyclosporine. Values of fU were determined ex vivo in 126 plasma samples obtained from 58 recipients of heart and lung transplants, using equilibrium dialysis. Concentrations of serum lipids, measured using standard enzymatic techniques, were used as concentration markers for serum lipoproteins. Patients were randomly assigned to either of two equal-sized groups. One group (subgroup 1) was used to evaluate the parameters of the model, and the other group (subgroup 2) was used to examine its predictive performance. The parameters were estimated using least squares non-linear regression. A model incorporating concentrations of serum HDL- and LDL-cholesterol, serum albumin, and time after transplantation gave the best fit. For subgroup 2, mean prediction error (ME), a measure of bias, and root mean squared error (RMSE) and median absolute error (MAE), measures of precision, and their 95% confidence intervals were estimated. For the best fit model, ME was 0.07 x 10(-2) (-0.065 x 10(-2) - 0.1 x 10(-2)), indicating that the model provided an unbiased estimate of the value of cyclosporine fU. Root mean squared error and MAE were 0.536 x 10(-2) (0.398 x 10(-2) - 0.645 x 10(-2)) and 0.27 x 10(-2) (0.226 x 10(-2) - 0.409 x 10(-2)), respectively. Prediction error was normally distributed; approximately 30% of the prediction errors were <10% and <5% of prediction errors were >50%. This model has shown a reasonable predictive performance in the patients with cardiac transplants studied; however, its predictive performance will need to be validated in a larger number of recipients of transplants of various types.