Investigation of using robust statistics: Weighted, parametric & non parametric linear regression of derivative/discrete Fourier transform HPLC convoluted peak responses for handling standard addition method data in bioanalysis

Errors that may be encountered during the use of standard addition method (SAM), especially in the estimation of low drug concentrations in biological fluids, are overcome by the combination of chemometric treatment of data along with the application of weighted regression (WR) as a robust statistic method. The study compares the effect of combining different chemometric and statistical methods of regression in handling data of the SAM, for HPLC analysis of a model drug, Ceftazidime (CTZ), in biological fluids. The regression methods, least square parametric (LSPR), nonparametric (NPR) and WR were applied to HPLC response data after chemometric treatment. The chemometric treatment involves the use of D1& D2 methods followed by convolution of the resulting derivative curves using 8-points sin xi polynomials (discrete Fourier function). Combination of chemometric treatment of data along with the application of WR method was found to be promising in analyzing low drug concentrations in urine. A significant improvement in the precision and accuracy was achieved on applying WR after chemometric treatment relative to LSPR or NPR methods. In-vivo study was performed and the method was applied for the estimation of CTZ elimination in urine with time.