Manufacturing a drug is one of the most controlled and regulated processes, where the quality of the end product is the main concern. In addition to the strict control of the manufacturing process, it is also important to control every material that enters the manufacturing process. The quality of every in-coming material is controlled with laboratory analyses and MID IR or NIR spectra. The latter are currently used only for the purpose of identification, even though these spectra carry a lot of information about the chemical and physical properties of the analyte. Therefore we wanted to check if by analyzing these spectra with multivariate analysis tools, variability between batches of produced raw material and even within a single selected batch could be detected. In addition to that, we have made models by using partial least squares, which predict chemical or physical properties of selected incoming materials from ID spectra.
By using principal component analysis, variability within a single batch of anhydrous lactose and hydroxipropil metilcellulose (HPMC) have been detected. Such variability cannot be detected with other analytical tools, since they are performed on an average sample that contains the analyte from 3 packaging units, whereas spectra are measured on every packaging unit. Additionally, we have detected variability between batches produced in the previous three years of anhydrous lactose, selected active ingredient and HPMC. Variability between batches of the active ingredient was not as significant as the variability between spectra of lactose and HPMC. However, smaller variability is expected, since the production of active ingredients is much more controlled.
By using the partial least square method we have made models, that predict physical or chemical properties from MID IR or NIR spectra. These models could, in the future, replace laboratory analyses, that are used to determine raw materials properties. This way human error while performing analyses could be avoided and in-coming materials could be released faster.
|