Abstract
Five modified multivariate calibration models based on classical least squares (CLS) in addition to traditional CLS model are developed and validated for assaying cyclobenzaprine HCl (CZ) with its major degradants; dibenzocycloheptatrienone (DZ) and anthraquinone (AQ), whether in its pure form or in pharmaceutical dosage form. The five models are net analyte processing CLS (NAP-CLS), orthogonal signal correction CLS (OSC-CLS), direct orthogonal signal correction CLS (DOSC-CLS) and hybrid linear analysis following the strategy of Xu and Schechter (HLA-XS) or Goicoechea et al. (HLA-GO). The five modified CLS models in addition to traditional CLS were subjected to a comparative study through manipulation of ultra-violet absorption data in the region of 220–350 nm. Three factor four level experimental design was adopted which results in 16 mixtures calibration set covering various concentrations of CZ, DZ and AQ. An extra validation set, composed of nine mixtures, was prepared for validation of the prediction power of the presented models. Experimental results showed high capability of the proposed modified CLS models for assaying CZ successfully without any interference from the co-existing degradation products (DZ and AQ). A statistical comparison between the results of CZ analysis in its dosage form by the six CLS based models and the reported HPLC method was carried out presenting no significant difference in regards to precision and accuracy. Significance of CLS based models is a consequent of their high quantitative and qualitative power for assaying multi-components mixtures.
RMSEP plot for the prediction of validation test set samples for cyclobenzaprine HCl using the proposed six models: 1-CLS, 2-NAP-CLS, OSC-CLS, 4-DOSC-CLS, 5-HLA-XS and 6-HLA-GO. [Display omitted]
•Improved predictions via modified Classical Least Squares based models.•Classical Least Squares based models have both quantitative and qualitative power.•Better quality control analyses of pharmaceuticals through CLS based models.