Spectral resolution of quaternary components in a sinus and congestion mixture; Multivariate algorithms to approach extremes of concentration levels

Sinus and congestion mixture of three drugs and an impurity was studied for their spectral resolution using four multivariate algorithms. The studied drugs present in extremes of low and high concentrations. Low concentration levels of phenylephrine HCl and doxylamine succinate with high concentration of paracetamol along with its official impurity was a challenging mixture for pharmaceutical analysis. The developed algorithms are principal component regression, partial least squares, concentration residuals augmented classical least squares and artificial neural networks. Models were compared for their calibration errors in general and individual calibration as well as their prediction of external validation samples. Concentration levels in the designed mixture were carefully considered to recede the challenging dosage form ratio. The value of root mean square error of prediction and percentage recoveries were used to describe the analytical performance of the proposed models. Artificial neural networks provided the lower most error with good recoveries for all samples without outlier. Correlation coefficients between the predicted spectra by concentration residuals augmented classical least squares and the pure ones revealed the ability for qualitative analysis. The proposed chemometrics have been successfully used for pharmaceutical application and compared favorably with the official methods.