Statistical analysis of performance of a liquid chromatographic assay of anticonvulsants involving solvent-demixing extraction and reciprocal internal standardization.

Analysis of variance both factorial and nested was used to validate a HPLC method intended for routine clinical assay of ethosuximide, phenobarbital, phenytoin and carbamazepine. Drugs were salted out, together with the solvent, from 0.5 ml acetonitrile-deproteinized plasma samples with 80-90% recovery. The acetonitrile extraction solution contained a known amount of all four drugs. This added amount of any drug was used when absent from plasma as an internal standard for those present and when present as a calibrator. Results showed that assay precision was acceptable (CV 6%) over and above the therapeutic range when additions did not exceed the lower therapeutic plasma level and if as many replications were made as there were drugs to assay. In return for some loss of sensitivity, reciprocal internal standardization provides increased assay reliability owing to the usual availability of more than one internal standard and to easier identification of interfering chromatographic peaks.