(510e) A First Step Towards the Development of a Nano-Production Line Using Multidisciplinary Quality By Design Approaches

The increasing molecular understanding of diseases, scientific advances in the development of large molecules, such as proteins, monoclonal antibodies and nucleic acids, and the global Corona pandemic have brought nanomedicine into the focus of research. Currently, fabrication of nano-delivery systems is a multi-step process using either non-continuous bulk processes or (parallelized) microfluidic technologies [1]. The latter include mixing, ultrafiltration to remove organic solvents, buffer exchange, and dilution to initial and final concentration, followed by bioburden filtration, sterile filtration and filling [2]. The smallest quality changes during such a multistage manufacturing process can lead to massive changes that impair efficacy and application. Despite enormous progress in research, industrial production processes are lagging behind, yet must not become a bottleneck in the future in order to keep pace with medical progress and ensure the best medical treatment for patients. Accordingly, there is a huge strategic need for the development of robust, scalable and versatile nano-manufacturing platforms that enable continuous, fast, high-yield, controllable and flexibly configurable production of nanomedicines at the large but also at the small scale.

This study presents a first step towards the development of a continuous and flexible nano-production line for reproducible production of nanoparticles with defined properties using multidisciplinary Quality by Design (QbD). The objective was to systematically identify and predict the ideal input parameters (i.e., process conditions and formulation composition) to optimize the size, size distribution, zeta potential, morphology, shape, internal structure, etc. of lipid nanoparticles according to regulatory guidelines (i.e., ICH Q8 (R2) and FDA-2017-D-0759). Design of Experiments (DoE) studies were conducted for this purpose.

To this end, drug-free solid lipid nanoparticles (SLN) consisting of Tween® 80 as stabilizer and one solid lipid respectively, i.e., either Palmitic acid, Compritol® 888 ATO or Geleol^TM Mono- and Diglycerides NF were studied. In addition, more complex nanostructured lipid carriers (NLC) comprising one of the solid lipids and one of the liquid lipids, i.e., oleic acid, linoleic acid or medium-chain triglycerides stabilized with Tween® 80 were investigated. Both lipid formulations were dispersed by high-shear mixing (HSM) and further processed with high-pressure homogenization (HPH). Central Composite Face-centered (CCF) quadratic experimental design with multiple linear regression (MLR) were chosen to investigate the main, interaction, and quadratic effects of process conditions and formulation characteristics on SLN and NLC properties. After the initial risk assessment (RA), the total lipid concentration (i.e., 10 % w/w), batch size (i.e., 75 g) and process temperatures (i.e., 10°C above the melting point of the materials used) were set as fixed parameters. In contrast, the stabilizer concentration (i.e., 1-4 % w/w), the HSM speed (i.e., 8000-16000 rpm) and time (i.e., 15-60 s) as well as the HPH pressure (i.e., 250-750 bar) and the number of cycles (i.e., 1-10 (SLN) and 3-10 (NLC)) were considered as input in the DoE. In addition, for NLC preparation, different ratios of solid to liquid lipids (i.e., 9:1, 8:2, and 7:3) were tested. As suggested by MODDE®, a total of 29 experiments (26 runs with 3 center points to access the model reproducibility) were conducted for each formulation (i.e., 12). The particle size (i.e., d10, d50 and d90) and size distribution (i.e., span-value) was selected as response using Laser diffraction for characterization. For the statistical analysis the MODDE® software was used and coefficient plots, a summary of fit considering R2 (i.e., coefficient of determination), Q2 (i.e., goodness of prediction), model validity and reproducibility were created.

After the completion of the experimental data, the models were simplified by excluding small and insignificant interaction terms. Obtained coefficients plots for SLN formulations identified the applied number of cycles and the stabilizer concentration as well as their quadratic effects as the most significant terms of the model. In addition, interaction effects of pressure and number of cycles and number of cycles and stabilizer concentration show a minor effect on particle size and size distribution. The same was observed for the NLC formulations. In addition, the matrix composition (i.e., lipid ratio) and its quadratic effect also proved to have a significant influence. The summary of fit for SLN and NLC showed high R2-values (i.e., 0.803 and 0.913), which indicate high fit between data and model and high Q2-values (i.e., 0.595 and 0.850), which represent reliability of the prediction. Since R2 and Q2 were > 0.5 and differences between R2 and Q2 were < 0.3, the obtained models are considered highly valid. Moreover, via center point experiments in triplicate, reproducibility of the models was proven to be high (i.e., 1).

In summary, gaining proper process knowledge of the individual processes involved and a sound understanding of the individual formulation components in terms of processability and product quality can help bridge the gap between academic and batch-based research and industrially relevant large-scale production. Furthermore, this knowledge can be used to flexibly design continuous processes and monitor and control the critical process parameters and formulation properties during production using process analytical technologies (PAT) and control software.

References:

1. Sheperd, S.J. Microfluidic formulation of nanoparticles for biomedical applications, Biomaterials 274, 120826 (2021).
2. European medicines agency, Assesment report Onpattro, EMA/554262/2018 (2018).