O.Ye. Rodionova

Quality control of packed raw materials in pharmaceutical industry.

The possibility of routine testing of pharmaceutical substances directly in warehouses is of great importance for manufactures, especially taking into account the demands of PAT. The application of NIR instruments with remote fiber optic probe makes these measurements simple and rapid. On the other hand carrying out measurements through closed polyethylene bags is a real challenge. To make the whole procedure reliable we propose the special trichotomy classification procedure. The approach is illustrated by a real-world example.

Prediction of the aging of polymer materials

Abstract The article deals with prediction of aging of multiplex polymer systems under conditions where the direct measurements of required properties are either impossible or difficult. To receive the reliable forecast, it is necessary to use physically reasonable models of processes. Solution of such problem and brief mathematical background is described in the article.

Quantitative risk assessment in classification of drugs with identical API content

When combating counterfeits it is equally important to recognize fakes and to avoid misclassification of genuine samples. This study presents a general approach to the problem using a newly-developed method called Data Driven Soft Independent Modeling of Class Analogy. The possibility to collect representative data for both training and validation is of great importance in classification modeling. When fakes are not available, we propose to compose the test set using the legitimate drugs analogs, manufactured by various producers. These analogs should have the identical API and a similar composition of excipients. The approach shows satisfactory results both in revealing counterfeits and in accounting for the future variability of the target class drugs. The presented case studies demonstrate that theoretically predicted misclassification errors can be successfully employed for the science-based risk assessment in drug identification.

Chemometric aided NIR portable instrument for rapid assessment of medicine quality.

The progress in instrumentation technology has led to miniaturization of NIR instruments. Fast systems that contain no moving parts were developed to be used in the field, warehouses, drugstores, etc. At the same time, in general these portable/handheld spectrometers have a lower spectral resolution and a narrower spectral region than stationary ones. Vendors of portable instruments supply their equipment with special software for spectra processing, which aims at simplifying the analysts work to the highest degree possible. Often such software is not fully capable of solving complex problems. In application to a real-world problem of counterfeit drug detection we demonstrate that even impaired spectral data do carry information sufficient for drug authentication. The chemometrics aided approach helps to extract this information and thus to extend the applicability of miniaturized NIR instruments. MicroPhazir-RX NIR spectrometer is used as an example of a portable instrument. The data driven soft independent modeling of class analogy (DD-SIMCA) method is employed for data processing. A representative set of tablets of a calcium channel blocker from 6 different manufacturers is used to illustrate the proposed approach. It is shown that the DD-SIMCA approach yields a better result than the basic method provided by the instrument vendor.

Evolutionary design of experiment for accelerated aging tests

The article is devoted to a special procedure that was developed by the authors and is used to design experiments for accelerated aging of rubber materials. The procedure involves correction of the final design with the help of short term preliminary tests and therefore it is called evolutionary design of experiment. The procedure was tested on seven rubber materials with different recipes and at different temperatures.

Construction of a multivariate calibration by the simple interval calculation method

Simple interval calculation is a method for linear modeling and for constructing interval estimates of predictions in a multivariate calibration. Simple interval calculation gives results in a convenient interval form with regard to all the existing uncertainties: measurement errors of predictors and responses, discrepancies of bilinear modeling, and so on. In addition, the simple interval calculation method opens new opportunities for constructing an informative classification of object significance. The method is based on only the assumption of the error boundedness; it uses linear programming algorithms for data analysis. This approach differs substantially from the conventional regression methods used in chemometrics and, therefore, is poorly understood by analysts. This paper gives an elementary explanation of the simple interval calculation method illustrated by the simplest model and real examples.

Application of the curve resolution method to the preprocessing spectral data in two-layer systems

An approach is proposed for recovering individual spectra of components in two-layer systems using the method of multivariate curve resolution–alternating least squares. The method is applicable to recording spectra in the near infrared region in both transmission and diffuse reflectance modes. The efficiency of the method is tested on a model system polymer film–polymer film and on real samples polymer packaging film–infinite layer of powdered product.

Two approaches to kinetic analysis applied to the prediction of antioxidant activity

Differential scanning calorimetry (DSC) followed by mathematical data processing can be used instead of the conventional method of long term thermal aging in predicting the activity of antioxidants in polyolefins. In this method, a regression relationship is established between the oxidation initial temperatures measured by DSC (X data) and the oxidation induction period values determined by thermal aging (Y data). Two approaches, called hard and soft, are employed in the construction of models. In the first case, nonlinear regression analysis is used in combination with successive Bayesian estimation. The second approach combines partial least squares regression and simple interval calculation. Use of a common data set makes it possible to compare these approaches and to draw inferences as to the cases in which one or the other is preferable.