Predictability of phenytoin serum levels by nomograms and clinicians

On the basis of a reliable steady state phenytoin serum concentration on one (lower) dose, the serum concentration on a second (higher) dose was predicted in 28 patients (32 levels) by two previously reported nomograms and 3 physicians who had clinical experience with phenytoin therapy. Both nomograms and all 3 physicians on the average underestimated the serum concentration on the higher dose. Except for 1 physician, this bias was statistically significant (p < 0.05). The mean and SD of the prediction errors (difference between predicted and measured serum concentration) were -2.2 +/- 4.2 and -3.1 +/- 3.6 for the two nomograms and -1.0 +/- 4.0, -2.1 +/- 4.1 and -1.8 +/- 4.2 for the 3 physicians, respectively. Thus, independent of the method used, there was a large uncertainty in the prediction; the approximate 68% confidence interval of the prediction error (mean +/- SD) was almost as large as the therapeutic range of the drug. Because prediction of phenytoin serum levels based on one serum concentration measurement was found unreliable in this study, it is proposed that every patient whose dose had been increased should be under close observation until a new steady state has been confirmed.