Facebook
Twitter
LinkedIn
Email
 |
Checking for direct PDF access through Ovid | |
Excerpt
The introduction of the micro-emulsion formulation of cyclosporin, Neoral®, has renewed interest in alternative metrics for monitoring therapy. Since the first demonstration that AUC could be estimated accurately from a limited number of blood samples1 there have been more than 40 publications using this technique to calculate AUC. However, the majority of the authors have used simple linear regression to develop their models and most have used r2, incorrectly, as a measure of prediction accuracy. Since we know that AUC cannot be measured without error it is more correct to allow for this uncertainty when developing limited strategies for cyclosporin AUC.
The training set was 20, 12 hour, profiles from 20 heart transplant patients. Each profile had 9 concentration-time points giving 36 possible two time-point combinations. Limited sampling algorithms were determined for each time-point combination using weighted and unweighted linear regression. Seven power weighting functions were compared (power of 0 ≡ unweighted) using a testing set of 89 profiles from an additional 17 patients.
When the algorithms were tested on the second data set the weighted linear regression showed improved accuracy of the prediction. However, the improvement in accuracy was small and is unlikely to be clinically significant.
