Salvador García-Muñoz

Handling uncertainty in the establishment of a design space for the manufacture of a pharmaceutical product

Recent trends in the pharmaceutical sector are changing the way processes are designed and executed, moving from: allowing the process to operate in a fixed point, to: allowing a permissible region in the operating space (a.k.a. design-space). This trend is driving product development to design quality into the manufacturing process (Quality by Design) and not to rely solely on testing quality in the product. These changes address the presence of uncertainties entering the process and their negative effect on product quality if the operating conditions are not free to compensate for these disturbances. This work provides a review of methods to address the presence of these uncertainties, ranging from the establishment of multivariate specifications for incoming product, to feed-forward process control. It also presents the development of a feed-forward controller to compensate the process for the observed changes in the properties of the incoming material. The development and testing of this controller is illustrated with a wet-granulation process. The performance of the multiple control strategies is back-propagated to the changes in the acceptance regions for the raw materials. Ultimately, the controller and its model are used to define an integrated design-space that accounts for the network of complex relationships between materials, process conditions and product. This work proposes and demonstrates that the use of mathematics is the only way to reach the maximum potential of the Quality by Design by providing a tool to specify an integral Design Space.

Coating uniformity assessment for colored immediate release tablets using multivariate image analysis

Multivariate image analysis (MIA) was applied to quantitatively assess film-coated tablets providing a cost-effective tool to replace visual inspection. MIA was used to determine the cosmetic end-point of the film-coating step and to calculate the coating level and distribution across tablets. The technique relies on simple digital images of the tablets and multivariate latent variable methods such as Principal Components Analysis (PCA) and Projection to Latent Structures (PLS). This application has precedence in the food industry and is a useful tool in Quality by Design by providing quantitative ranges on coating targets. The technique is illustrated using two sizes of tablets coated at two different scales. As expected, the coating distribution across tablets for the larger scale is broader, and the total amount of coating material is proportional to the surface area of the tablet. The technique is illustrated with off-line images taken with a Single-Lens Reflex digital camera, and with in-line images taken with a webcam installed inside the coater. For the latter case a novel adaptive PCA modeling approach is proposed to handle the real-time images and translate them into indexes that determine the cosmetic end-point of the film-coating step.

An efficient nonlinear programming strategy for PCA models with incomplete data sets

Processing plants can produce large amounts of data that process engineers use for analysis, monitoring, or control. Principal component analysis (PCA) is well suited to analyze large amounts of (possibly) correlated data, and for reducing the dimensionality of the variable space. Failing online sensors, lost historical data, or missing experiments can lead to data sets that have missing values where the current methods for obtaining the PCA model parameters may give questionable results due to the properties of the estimated parameters. This paper proposes a method based on nonlinear programming (NLP) techniques to obtain the parameters of PCA models in the presence of incomplete data sets. We show the relationship that exists between the nonlinear iterative partial least squares (NIPALS) algorithm and the optimality conditions of the squared residuals minimization problem, and how this leads to the modified NIPALS used for the missing value problem. Moreover, we compare the current NIPALS‐based methods with the proposed NLP with a simulation example and an industrial case study, and show how the latter is better suited when there are large amounts of missing values. The solutions obtained with the NLP and the iterative algorithm (IA) are very similar. However when using the NLP‐based method, the loadings and scores are guaranteed to be orthogonal, and the scores will have zero mean. The latter is emphasized in the industrial case study. Also, with the industrial data used here we are able to show that the models obtained with the NLP were easier to interpret. Moreover, when using the NLP many fewer iterations were required to obtain them. Copyright

Multivariate wavelet texture analysis for pharmaceutical solid product characterization

The application of multivariate wavelet texture analysis (MWTA) is presented and discussed as it is applied to three different types of pharmaceutical materials: (a) tablet cores, (b) wet granules and (c) controlled release tablets. The application of MWTA is initially proposed as a quantitative replacement to the human visual judgment of the textural appearance of the different materials. In all cases, the metrics obtained with MWTA agree with visual assessment on the progression of textural features such as erosion and surface roughness. This work further demonstrates that MWTA also represents a useful tool to increase the understanding of the manufacturing process, as it provides diagnostics to relate process parameters with textural features of the material that are difficult or costly to measure otherwise (such as granule size for wet material or surface appearance for a controlled release product). MWTA is also presented as a potential tool for real-time release for those cases where the textural features can be proven to provide accurate enough predictions of the final product performance; as shown here with the obtained prediction of dissolution from the controlled release tablet using the texture of the product as an input.

A framework for in-silico formulation design using multivariate latent variable regression methods.

A comprehensive Quality by Design development paradigm should consider the impact of raw materials and formulation on the final drug product. This work proposes a quantitative approach to simultaneously predict particle, powder, and compact mechanical properties of a pharmaceutical blend, based on that of the raw materials. A new, two-step, multivariate modeling method, referred to as the weighted scores PLS, was developed to address the challenge of predicting the properties of a powder blend while enabling process understanding. The model validation exercise is shown along with selected practical applications. It is shown how the proposed in-silico model exhibits sufficient predictive power to be an important tool in the pharmaceutical development decision making process while requiring minimal experimentation and material usage.

Application of multivariate latent variable modeling to pilot-scale spray drying monitoring and fault detection: Monitoring with fundamental knowledge

Abstract This work describes the use of multivariate latent variable modeling (LVM) to enhance fundamental understanding of the operational space, the scale differences and the common-cause variability present in the operation of a pharmaceutical spray-dryer. LVM provided a real-time process monitoring and fault detection tool for continuous quality assurance. A latent variable model was built and tested using commercially available software in a pilot-scale facility at Bend Research Pharmaceutical Process Development Inc. (BRPPD) in Bend, OR. The key learning from the exercise at the pilot-scale helped identify and understand the normal variability of the commercial scale equipment. A key advantage of the LVM approach is that the variability that drives the process is easily understood in a fundamental way by interpreting the model parameters in light of fundamental engineering knowledge (e.g., transport phenomena, thermodynamics). The understanding of the common-cause variability enables the better understanding of the differences across scales for this unit. In monitoring the process, the faults are not only detected in a statistical way, but also understood in a fundamental way by using the model to track down the driving forces that were involved in detecting such fault (e.g., an abnormal behavior of the gas momentum across the unit).

How do formulation and process parameters impact blend and unit dose uniformity? Further analysis of the product quality research institute blend uniformity working group industry survey

Responses from the second Product Quality Research Institute (PQRI) Blend Uniformity Working Group (BUWG) survey of industry have been reanalyzed to identify potential links between formulation and processing variables and the measured uniformity of blends and unit dosage forms. As expected, the variability of the blend potency and tablet potency data increased with a decrease in the loading of the active pharmaceutical ingredient (API). There was also an inverse relationship between the nominal strength of the unit dose and the blend uniformity data. The data from the PQRI industry survey do not support the commonly held viewpoint that granulation processes are necessary to create and sustain tablet and capsule formulations with a high degree of API uniformity. There was no correlation between the blend or tablet potency variability and the type of process used to manufacture the product. Although it is commonly believed that direct compression processes should be avoided for low API loading formulations because of blend and tablet content uniformity concerns, the data for direct compression processes reported by the respondents to the PQRI survey suggest that such processes are being used routinely to manufacture solid dosage forms of acceptable quality even when the drug loading is quite low.

An optimization-based undeflated PLS (OUPLS) method to handle missing data in the training set

Advances in sensory systems have led to many industrial applications with large amounts of highly correlated data, particularly in chemical and pharmaceutical processes. With these correlated data sets, it becomes important to consider advanced modeling approaches built to deal with correlated inputs in order to understand the underlying sources of variability and how this variability will affect the final quality of the product.

Modelling of Droplet Absorption and Evaporation during Pharmaceutical Tablet Coating

The shelf life of a pharmaceutical tablet is affected by the amount of water that interacts with it during the aqueous film coating process. The purpose of this work is to simulate the spreading, absorption and evaporation of water droplets after impact on a porous tablet core. We divided the spreading, absorption and evaporation phenomena into three separate phases: the kinematic, the capillary and the evaporation phases. For the kinematic phase, we modified 1-D spreading models found in the literature which solve the kinetic energy balance equation. Subsequently, for the capillary phase we solved the Navier-Stokes equation using the lubrication approximation theory. For the evaporation phase, we developed a novel model that treats the tablet as a particle with a wet core surrounded by a dry crust. Our numerical results were in good agreement with recent experimental data found in the literature.

Multivariate Image and Texture Analysis to Investigate the Erosion Mechanism of Film-coated Tablets: An Industrial Case Study

IntroductionThe Quality by Design initiative requires the design space of a process to be based on metrics that are robust and reproducible, and not on qualitative ones that might be easily biased by human perception. Hence, the use of image analysis is attractive for practical industrial applications where the quality assessment is still typically performed by a panel of trained experts.MethodsThe use of multivariate image and texture analysis is proposed in this study to quantitatively characterize the elegance of film-coated tablets. Four unsupervised metrics are developed to quantify both the color uniformity of tablet faces/bands and the surface erosion. To develop robust statistics, more than 7,000 tablets coated in nine different pilot-scale batches are considered. Latent variable modeling is used to regress the measured elegance against coating operating conditions to investigate the driving forces acting on the system and guide pharmaceutical manufacturing, consistently with the Quality by Design framework.ResultsThe model allows one to successfully investigate the causes leading to tablet erosion. Additionally, it is shown that the model space can be effectively used as a monitoring chart of the overall batch elegance in terms of color uniformity and surface erosion, since the batches are found to rank according to the surface roughness of the manufactured tablets.ConclusionsImage analysis has been shown to be an effective process analytical technology for the development of the design space of a film-coating process, where quality assessment on the final product is traditionally based on the judgment of panel of trained experts.