Understanding the Impact of Chemical Variability and Calibration Algorithms on Prediction of Solid Fraction of Roller...

Area of Science:

• Pharmaceutical Technology
• Analytical Chemistry
• Process Analytical Technology (PAT)

Background:

• Near-infrared (NIR) spectroscopy is a valuable tool for pharmaceutical analysis.
• Predicting solid fraction (SF) in dry granulated ribbons is crucial for manufacturing consistency.
• Formulation variability can significantly impact analytical method performance.

Purpose of the Study:

• To develop and evaluate a near-infrared (NIR) method for predicting solid fraction (SF) in dry granulated ribbons.
• To investigate the impact of unmodeled chemical variability on NIR method performance.
• To compare the robustness of Partial Least Squares (PLS) and spectral slope algorithms in predicting SF.

Main Methods:

• NIR spectra were collected from calibration and test compacts with varying excipient and active pharmaceutical ingredient (API) concentrations.
• Partial Least Squares (PLS) and spectral slope algorithms were employed to model SF.
• Method performance was assessed under conditions of chemical and processing variability.

Main Results:

• Unmodeled chemical variation introduced spectral artifacts like new peaks and baseline shifts.
• The spectral slope algorithm demonstrated superior robustness compared to PLS when predicting SF with increasing API load.
• PLS model robustness was compromised by chemical variability affecting both spectral baseline and peak absorbance.

Conclusions:

• Spectral slope analysis is a more resilient approach for NIR-based SF prediction in the presence of unmodeled chemical variability.
• Understanding variability risks enables the development of NIR methods as API-sparing techniques for low-dose drug products.
• This work supports the application of PAT for efficient and robust pharmaceutical manufacturing.