Rapid i.v. loading with phenytoin with subsequent dose adaptation using non-steady-state serum levels and a Bayesian forecasting computer program to predict maintenance doses

To evaluate the suitability of a phenytoin loading dose regimen; to assess whether dose-individualization was necessary and to investigate the reliability of a Bayesian forecasting method for phenytoin dose adaptation using non-steady-state levels in hospital-admitted patients. An initial loading dose (15 mg phenytoin acid/kg BW) was given i.v. over 4 h, followed by standardized maintenance doses given i.v. in 12-h intervals from days 1 to 5 (175 mg </= 70 kg; 202 mg > 70 kg BW). The evening dose of day 5 was individualized based on three serum trough levels: L1 (after 16 h), L2 (morning day 4) and L3 (morning day 5). Ninety of 136 consecutive patients were evaluable in a prospective study for the standardized phase; 50 of them had additional serum levels in the individualized phase. There was no exclusion of patients with interacting co-medication. Seventy-seven per cent (L1) and 68% (L3) of patients showed therapeutic values (10-20 mg/L). The prediction error of the forecasting was 3.95 mg/L, the root mean squared error 6.27 mg/L (target trough level 11 mg/L). Seventy per cent of the levels (n=50) were within the 68% confidence interval. The effectiveness and safety of the regimen with rapid i.v. loading and the necessity to individualize phenytoin dosing after day 5 were demonstrated.