Category: Formulation and Quality
Purpose: Lyophilization is a process in which the solvent is removed by sublimation after the solution is frozen. The amount of residual moisture content (MC) is crucially related to the stability of formulation over shelf life. Moreover, the moisture level during lyophilization is an important parameter to determine the end of critical steps for process development and optimization. Karl Fischer titration (KF) is the most commonly used method to measure MC. However, the reagents for KF titration are toxic. In addition, the sample preparation is destructive and time-consuming. Thus, a toxic reagent free, non-invasive and time-saving method is desired to facilitate the MC determination. Moisture is one of the most evaluated Critical Quality Attributes quantified by NIRS, due to its strong signal in the NIR spectral range. Herein, we developed a NIR method to predict MC in a lyophilized drug product using a model compound A.
Methods: To create a series of lyophilized vials with different moisture contents, lyophilized vials were opened to a humidified environment to increase the moisture content. Vials were also placed in a vacuum oven to decrease the moisture content. At predetermined time points, vails were taken out, stoppered, capped and crimped. NIR spectra were recorded using a Bruker FT-NIR spectrometer. Sample vials were measured in reflection mode through the bottom of the vials. Three spectra were recorded for each vial using 32 cm-1 resolution/64 scans setting. Subsequently, the moisture content in each vial was measured by coulometric Karl Fischer titration method. A combination of Savitzky-Glay secondary derivative and standard normal variate method was used to pretreat the spectra. A principle component analysis and a partial least squares regression were used to analyze the data qualitatively and quantitatively, respectively.
Results: A moisture content range between 0.052% and 1.953% was achieved by sample preparation. In raw spectra, the absorption peaks of moisture were observed around 5200 cm-1 and 7150 cm-1. The 5200 cm-1 band was caused by a combination of OH stretching and HOH bending. The absorption at 7150 cm-1 was caused by first overtone of OH stretching. The spectral range of 5083.87 to 5407.88 cm-1 was used for subsequent analysis. For the PCA analysis, the first factor explained 100% of the variance of the spectral change. The shape of the factor highly correlated with the water band with peaks around 5190 and 5300 cm-1. For PLS regression, 49 samples were randomly selected as calibration set. A calibration model with two PLS factors was developed from 0.052% to 1.953%. Excellent predictive performances were observed. An r of 0.9986 indicated a good correlation between the NIR and reference KF methods. The y-intercept was 0.001% indicating the absence of constant systematic errors. The slope was 0.9972, which was close to 1. The RMSEC and RMSECV were 0.026% and 0.027%, respectively, indicating a good model accuracy. Twenty-one samples were used as validation set. The moisture content range of validation samples was from 0.055% to 1.899%, which was included in the calibration range. Excellent predictive performance was also observed. An r of 0.9992, a y-intercept of 0.003%, and a slope of 0.9895 were observed. The RMSEP was 0.021%.
Conclusion: NIRS has been demonstrated as an efficient and accurate PAT tool to determine the moisture content in lyophilized product. Compared to commonly used Karl-Fischer method, NIRS is a toxic reagent free, non-invasive and time saving method.