Development of Near-Infrared Methods to Analyze Moisture Levels in Lyophilized Pharmaceutical Products

In the pharmaceutical industry, measuring residual moisture in lyophilized parenteral products is crucial for process control and ensuring that samples from production lots fulfil specified release parameters. For developmental purposes, such quantifications are essential during stability studies and for optimizing the freeze/pump cycles employed in the freeze-drying process.

Conventional techniques for moisture measurement include gas chromatography, Karl Fischer titration, and thermo-gravimetric analysis, also referred to as loss-on-drying (LOD). Although these techniques provide benefits in terms of system cost, solvent specificity, and sensitivity, they are time intensive and destructive to the sample under analysis.

Near-infrared (NIR) spectroscopy is a suitable technique to rapidly measure moisture in freeze-dried materials. This article discusses the development of NIR techniques for residual moisture measurement in typical lyophilized pharmaceutical products.

Advantages of NIR Spectroscopy

Analyses with NIR spectroscopy typically take below 30s. Calibration for measuring trace amounts of moisture is possible, thanks to the strong overtone absorption bands for water at 1450nm and 1940nm.

The measurement may be carried out through the bottom of the vial in a diffuse reflectance configuration, considering the fact that the borosilicate glass commonly utilized in serum vials and ampoules is ‘invisible’ in the NIR region. Hence, little or no sample preparation is required, while eliminating the requirement for expensive or toxic reagents.

NIR spectroscopy is a non-invasive and non-destructive technique, making it especially advantageous to analyze samples comprising costly active ingredients, or for sample analysis in stability studies to enable the use of the same vials for subsequent potency assays.

Experimental Procedure

A FOSS NIRSystems Model 6500 spectrometer featuring a Rapid Content™ Analyzer (RCA) was used to acquire the NIR spectra (Figure 1). The RCA sampling accessory employs a mechanical iris to place vials over the monochromatic incident beam and an array of lead-sulfide detectors that can analyze diffusely reflected NIR radiation between 1100nm and 2500nm. Spectral data were signal averaged utilizing 32 scans. Since this device is no longer available, it is recommended to use the NIRS XDS Rapid Content™ Analyzer.

Figure 1. RCA sampling module used to analyze lyophilized material.

The Vision™ software was used to perform data acquisition and analysis. Using 46 sample vials that were divided into ten groups (A to J), a calibration set was prepared. The moisture content is incrementally increased by spiking the samples in each group with specified amounts of water (Figure 2). The samples were then stored for two days to reach equilibrium with the sample in the vial by allowing the moisture to completely evaporate.

Figure 2. Schematic of the technique used to spike sample vials with water.

During this period, the water was not allowed to come into direct contact with the freeze-dried cake by maintaining the tilted orientation of the vials. This approach readily allows for preparing the calibration sets, which have a ‘box-car’ distribution, with approximately the same count of samples in each ‘group’ of moisture levels (A to J).

Experimental Results

Karl Fischer Titration of Calibration Samples

The Karl Fischer analysis results are delineated in Figure 3. The measured values are in line with the desired increase for each group, showing the ability to easily control the experimental method used for spiking the samples. Nevertheless, the measured moisture values were greater than anticipated in three samples, D5, E2, and G3.

This may be due to the inclusion of excess water or high initial moisture content in each vial before spiking. Excluding these samples, the average moisture content before spiking was estimated to be roughly 0.7% by deducting the desired moisture increase from the actual measured value for each vial.

Figure 3. Karl Fischer analysis results of calibration set samples.

NIR Spectral Data

Figure 4 illustrates the NIR spectra of 10 calibration samples consisting of different amounts of moisture as depicted in the inset bar graph, clearly showing the correlation between an increase in moisture content and the corresponding changes in the spectral features around the band at 1940nm.

Figure 4. Diffuse reflectance near-infrared spectra of lyophilized samples containing varying amounts of moisture as indicated in the inset bar graph.

Second derivative spectra are often employed in the development of quantitative regression equations. This math pretreatment is applied to correct for sloping backgrounds and varying offsets associated with variations in the reflectivity of each group sample.

These differences are due to physical parameters, such as differences in cake appearance and particle size distributions. The second derivative spectra of the same 10 samples depicted in Figure 4 are delineated in Figure 5.

Figure 5. Second derivative spectra corresponding to the absorption spectra of Figure 4.

Regression Equation Development

Spectra of the calibration samples are generally split into two unique groups for the development of quantitative models. Samples in the ‘training-set’ are employed for developing the regression equations, while samples in the ‘test-set’ are employed to assess the equations. Spectra of a specific sample must not be used in both sets.

To obtain a calibration for analyzing samples containing moisture values in a range close to the specified acceptance limit (~ 2% H₂O), multiple linear regression (MLR) algorithms were used to second derivative spectra of calibration samples from groups A to F.

This provided calibration Equation A, which is optimized for samples containing low moisture content (0-3.5% H₂O). The divisor term reduces the effects caused by multiplicative scatter due to pathlength difference in the samples.

Equation A - for samples having low moisture content is,

NIR Predicted moisture in % = 0.806 - 19. 411 (A” 1842nm/A” 2124nm)

The application of MLR algorithms to second derivative spectra of all of the samples in the training set yielded Equation B, which is optimized for samples consisting of high moisture content (up to 15% H₂O).

Equation B - for samples having high residual moisture is,

NIR Predicted moisture in % = 1.427+6.472 (A” 1842nm/A” 2162nm)

Both regression equations yielded outstanding coefficients of determination, with R2 > 0.99. Figures 6 and 7 show NIR predicted moisture values estimated using Equation A vs. Karl Fischer results and NIR predicted moisture values estimated using Equation B vs. Karl Fischer results, respectively.

Figure 6. NIR predicted moisture values estimated using Equation A vs. Karl Fischer results.

Figure 7. NIR predicted moisture values estimated using Equation A vs. Karl Fischer results.

NIR Analysis of Defect Samples Containing Melts

‘Meltback’ or a collapse of the freeze-dried cake due to partial dissolution and the presence of high residual moisture levels in the vials is a common defect observed in lyophilized products.

Melts can also occur during the freeze-drying process when the temperature and/or pressure are not kept below the transition point at which liquefaction takes place. A sample consisting of a melt is regarded as a ‘critical defect’ due to the possibility of drug degradation, which ultimately affects the efficacy of the final product.

The NIR technique was applied for analyzing a group of 15 production samples containing melts. Each sample had high levels of moisture content, ranging from 5% to 20% H₂O. The The NIR technique was able to determine these samples despite the fact that there was a significant compromise in the sample matrix. Furthermore, the diffusely reflected NIR radiation from a sample having a melt may be very different from a sample that does not have a melt.

Although the NIR technique was not able to determine the actual moisture levels of these samples, it provided meaningful results. This demonstrated the potential of the NIR technique to be utilized as a qualitative inspection tool. Since the technique is non-destructive, it is possible to automate the analyses, thereby enabling the analysis of every production sample.

Conclusion

The results clearly demonstrate the ability of the NIR spectroscopy technique to analyze residual moisture in lyophilized products. The speed and simplicity of the method facilities the analysis of very large sample sets.

The automation of the analysis allows for including residual moisture analysis as an integral process of the quality assurance analysis of all lyophilized vials.

This information has been sourced, reviewed and adapted from materials provided by Metrohm AG.

For more information on this source, please visit Metrohm AG.