Wide concentration range investigation of recovery, precision and error structure in liquid chromatography

Using a typical HPLC assay, the characteristics of recovery, system precision and repeatability were investigated over a wide concentration range. In the presence of a constant amount of typical tablet excipients, the antidiabetic drug glibenclamide was analyzed in the range from 0.24 to 0.005 mg/mL (18 concentration levels, 6 independent sample preparations each). On the basis of a typical concentration for an HPLC glibenclamide assay of 0.2 mg/mL, this corresponds to a relative amount of 120-0.025% label claim. In the range from 120 to 0.075%, the recovery was found to be quite constant and systematically heightened mainly due to the evaporation from vials during centrifuging and the displacement of solvent volume by the added matrix. Both system precision and repeatability remain almost constant in the interval from 120 to 10% at a R.S.D.% of 0.31 and 0.70%, respectively, indicating that the sample preparation is the major error source in this range (0.63%). Between 10 and 0.25%, a linear relationship between the logarithmized concentration and the repeatability was noted. However, for lower amounts close to the limit of quantitation, the R.S.D.% of measurements increases much more distinctly. This increase is caused by a strong rise of the system precision. At this concentration range, system precision and repeatability are not significantly different any longer. This leads to the conclusion that with the injection error being constant the peak integration error becomes the dominating error source at low concentrations, e.g. at concentrations below the five-fold of the LOQ. The results obtained here agree well with earlier published data. As the quantitation limit of 0.05% can be regarded as typical for a pharmaceutical impurity control test, generalizations of these findings from this extensive data set should be possible. In this context, peak integration and improvements of the signal-to-noise ratio are the most promising measures to improve an unsatisfactory precision in LC. (C) 2004 Elsevier B.V. All rights reserved.