QUANTITATIVE ANALYSIS OF POLYMORPHIC FORMS OF MEBENDAZOLE IN PHARMACEUTICAL RAW MATERIALS USING NEAR-INFRARED SPECTROSCOPY

Mebendazole (MBZ) is a broad spectrum anthelmintic drug and has three different polymorphic forms (A, B and C). Form A has no anthelmintic effect when it exceeds 30% w/w. Form B show higher toxicity and solubility. Form C does not have the drawbacks of forms A and B and is stable at room temperature, consequently it is pharmaceutically preferred. All polymorphic forms of MBZ, however, have been reported in pharmaceutical raw materials. NIR spectroscopy can be performed efficiently in determining polymorphs due to inherent technical features. NIR does not access information directly from the crystal lattice, but may detect changes in the vibrational modes of substance in the solid state. This technique allows fast, precise and accurate analysis, required for process control. Furthermore, when combined with chemometric tools, it can identify, classify and quantify different pharmaceutical samples.
This work evaluates the feasibility of using NIR spectroscopy for quantification of three polymorphic forms of MBZ in pharmaceutical raw materials.
Thirty ternary mixtures of polymorphic forms of MBZ were prepared, varying forms A and C from 0 to 100% (w/w), and form B from 0 to 30% (w/w). MBZ samples of forms A and C were donated by Formil (Brazil) and Form B was obtained by recrystallization of form C.
Reflectance NIR spectra were acquired using a FTNIR Frontier spectrometer (Perkin Elmer), at the spectral range from 12000 to 4000 cm-1 with an average of 32 scans at a spectral resolution of 4 cm-1. For comparison, Mid-Infrared (MIR) spectra acquired using a sampling accessory (UATR – universal attenuated total reflectance) was also tested.
Partial Least Square (PLS) regression models were built using preprocessed spectral data (7000-4000 cm-1) by Standard Normal Variation (SNV), Multiplicative Scatter Correction (MSC) and first derivative (Savitzky-Golay filter, 2nd order polynomial and 11 window point). PLS models were also developed employing the variables selected by the Jack-Knife (JK) algorithm. The predictive ability of the models was evaluated by the Root Mean Squared Error of Prediction (RMSEP) obtained from the external validation set. An F-test (95% confidence level) was used to access statistical differences among the models.
NIR spectra from the three polymorphic forms of MBZ showed different spectral behavior, mainly for the spectral band assigned to the first overtone of N-H stretching modes at around 6850 cm-1. For the polymorphic forms A and C, the lowest RMSEP values, 2.3% w/w and 1.4% w/w, were obtained using preprocessed spectra with MSC and SNV, respectively. On the other hand, for the Form B, the lowest RMSEP value (1.2% w/w) was obtained using derivative spectra and significant variables selected by JK. All these models showed R² values above of 0.97. In addition, there was no evidence of the presence of systematic errors in the models. For the MIR region, the better RMSEP values obtained were 3.3%, 1.3% and 3.0% for forms A, B and C, respectively. These results are worse than those obtained with NIR spectroscopy, indicating that NIR is more suitable for quantifying polymorphic forms of MBZ in pharmaceutical raw materials.