Batch and continuous blending of particulate material studied by near-infrared spectroscopy

Background: Pharmaceutical manufacturing is moving towards real-time release of the \nproducts. This objective can only be achieved by clearly understanding the process and by \nimplementing suitable technologies for manufacturing and for process control. Near-infrared \n(NIR) spectroscopy is one technology that has attracted lot of attention from the pharmaceutical \nindustry since it can analyze bulk solids without any pretreatment, therefore reducing or \neliminating wet chemistry analysis. NIR spectroscopy is a powerful tool for the monitoring unit \noperations were bulk material is involved i.e. blending of powders. \nBlending of powders is a complex and poorly understood unit operation. In the pharmaceutical \nindustry blending has been performed batchwise and controlled by thief sampling. Thief \nsampling is an invasive process which is tedious and tends to introduce bias; therefore an \nalternative sampling method was highly needed. Here is where NIR found a perfect match with \nblend uniformity monitoring, thus NIR implementation offers several advantages: thief sampling \nis avoided, the process is continuously monitored, detection of blend-end point, and fast \nidentification of process deviations. \nNIR spectral data need to be correlated with the parameter of interest (physical or chemical). \nThese computations are done by multivariate data analysis (MVDA). MVDA and NIR are a \npowerful combination for in-process control and their use has been promoted by the health \nauthorities through the Process Analytical technology (PAT) initiative by the FDA. \n \nPurpose: This thesis is focused on the study of powder blending, which is an essential unit \noperation for the manufacture of solid dosage forms. The aim was to develop two quantitative \nmethods for the monitoring of the active ingredient concentration. One method was developed \nfor blend uniformity monitoring of a batch mixing process, and a second method for a \ncontinuous mixing process. \nThis study also tackles the relevance of the physical presentation of the powder on the final \nblend quality, by studying the influence of the particle size and the effect of the previous \nmanufacturing steps on the NIR spectral data. \n \nMethods: Particle size was studied by NIR in diffuse reflectance mode, using Kubelka-Munk \nfunction and the transformation of reflectance of absorbance values, in order to focus the \nanalysis on the physical properties. Furthermore, an off-line NIR model was developed for the \nquantification of the mean particle size. Segregation tendencies due to particle size \nincompatibilities were studied. \nBlend uniformity monitoring of a batch pharmaceutical mixing was achieved through a NIR off- \nline calibration method, which was used for the in-line drug quantification of a production scale \nmixing process. \nNIR in diffuse reflectance mode was used in the study of a continuous blending system. The \neffect of the process parameters, i.e. flow rate and stirring rate, was analyzed. Moreover, a NIR \nmethod for the in-line drug quantification was developed. \nNIR was implemented in a powder stream, in which the mass of powder measured by NIR was \nestimated. \n \nResults and discussion: Regarding particle size, incompatibilities due to different particle size \nranges between the formulation ingredients lead to severe segregation. Particle size and \ncohesion determined the quality of the powder blend; slight cohesion and broader particle size \ndistribution improved the robustness of the final blend. NIR showed high sensitivity to particle \nsize variations, thus it was possible to develop a quantitative model for the mean particle size \ndetermination with a prediction error of 16 micrometers. \nConcerning batch mixing, an off-line calibration was generated for the quantification of two \nactive ingredients contained in the formulation. The prediction errors varied from 0.4 to 2.3% \nm/m for each of the drugs respectively. Special emphasis was given on the proper wavelength \nselection for the quantitative analysis in order to focus the analysis on the active ingredients \nquantification. \nIn relation to continuous blending of particulate material, a quantitative NIR model was \ndeveloped for the in-line prediction of the active ingredient concentration. The NIR model was \ntested under different process conditions of feeding rate and stirring rate. High stirring rates \nproduce higher scattering of the NIR predictions. This was directly associated with the \nacceleration of the particles at the outlet of the blender affecting the dwell time of the particles \nwith the NIR probe. The NIR model showed to be robust to moderate feed rate increments; \nhowever the NIR model under-predicted the drug concentration under moderate feed rate \nreductions of 30 kg/h. Furthermore, the continuous blending phases were clearly identified by \nprincipal component analysis, moving block of standard deviation, and relative standard \ndeviation, all of them giving consistent results. \nNIR measurements in a powder stream involved the scanning of powder flowing in a chute. The \nflow of bulk solids is a complex phenomenon in which powder moves at a certain velocity. The \nmotion of particles produces changes in the density and distribution of the voids. In this study, \nthe velocity of the powder sliding down an inclined chute was measured and used for the \nestimation of the NIR measured mass. The mass observed during one NIR measurement was \nestimated to be less than one tablet. \n \nConclusions: This study proved the feasibility of applying NIR spectroscopy for the blend \nuniformity monitoring of batch and continuous powder mixing. Understanding the critical \nparameters of powder mixing lead to a robust process and reliable analytical methods. NIR \nproved to be a valuable and versatile analytical tool in the measurement of bulk solids.