Category: Manufacturing and Bioprocessing
Purpose: The development of pharmaceutical nanoformulations has accelerated over past decade. However, nano-sized drug carriers continue to meet substantial regulatory and clinical translation challenges. In order to address some of these key challenges in early development, we adopted a quality by design approach to microemulsions as a model nanoformulation. Microemulsions are optically transparent, thermodynamically stable emulsions that typically range from 20 – 100 nm in diameter. Microemulsions are typcially produced using low-energy methods such as water titration, and several reports have found that the process parameters of water addition rate and stir rate during water titration impact final microemulsion diameter. Sensitivity to these processing parameters may hinder microemulsion scale-up. We hypothesized that a quality by design approach would enable us to identify a robust microemulsion platform resistant to changes in processing parameters, thus facilitating microemulsion scale-up.
Methods: Failure modes, effects, and criticality analysis (FMECA) was utilized to identify the process parameters that were most likely to impact (1) microemulsion diameter, (2) polydispersity index (PDI), (3) stability under thermal cycling, and (4) shelf life. These process parameters were used to develop a 15-run, D-optimal screening design of experiments (DoE). Based upon statistical analysis of the screening DoE, 15 additional runs were added to the DoE. This augmentation was performed to refine the design space and facilitate investigation of microemulsion composition main effects and second order interactions. Results from the augmented DoE were used to develop multiple linear regression (MLR) models that predicted microemulsion diameter, PDI, and shelf life as a function of microemulsion composition. Additionally, logistic regression (LR) models were developed to predict the probability that a microemulsion would meet thermal cycling and/or shelf life specifications. Based upon these analyses, a select microemulsion was scaled up ten-fold from 100mL to 1,000mL in triplicate. This select formulation also incorporated the poorly water soluble compound resveratrol at a concentration of 17.5mM. JMP Pro 13 statistical software was used to develop all DoEs, MLR and LR models. All microemulsion diameter and PDI measurements were performed using a Zetasizer Nano, ZS (Malvern, UK).
Results: In the screening design of experiments, 7 of the 15 runs met all critical quality attribute (CQA) specifications for diameter, PDI, thermal cycling stability, and shelf life. MLR models were developed to predict microemulsion diameter, PDI, and percent diameter increase over 30 days storage as a function of microemulsion composition, stir rate, and water addition rate. Using these models, we determined that if microemulsion maximum internal phase was decreased from 13.5% to 12.0%, microemulsion colloidal properties were dependent solely upon microemulsion composition. The DoE was augmented to include additional runs that enabled us to focus on microemulsion composition main effects and second order interactions. Accurate MLR models were developed for diameter (R2=0.9419), PDI (R2=0.8949), and 30-day percent diameter increase (R2=0.9637) (Figure 1). We found that microemulsion oil content contributed the most significantly to all three studied CQA specifications. Further, the diameter and PDI MLR models contained the same terms, suggesting that there is a correlation between microemulsion diameter and PDI. LR models were developed to predict the probability that a microemulsion formulation would meet the CQA specifications for thermal cycling, shelf life, or both thermal cycling and shelf life as a function of either microemulsion diameter or microemulsion PDI 24h after production. We found that including both thermal cycling and shelf life as the response in the LR model improved model accuracy such that there were zero misclassifications in both training and validation sets (Figure 2). A select microemulsion was successfully scaled up ten-fold to 1L in triplicate (Figure 3) without impacting microemulsion diameter, PDI, stability under thermal cycling, or shelf life. Finally, this select microemulsion formulation incorporated the natural anti-inflammatory compound resveratrol at a concentration of 17.5mM.
Conclusion: In the presented work, we utilized a QbD approach to identify a design space in which microemulsion colloidal properties were dependent solely upon microemulsion composition, which facilitated microemulsion scale-up. The unique combination of MLR and LR was powerful in this specific application, as it could be used to predict not just the basic colloidal properties (diameter and PDI), but also the probability that a formulation will pass quality control testing. Extensive quality control analyses improved the predictive accuracy of the LR models, suggesting that rigorous quality control testing early in formulation development can aid in the timely identification of unsuitable formulations. The presented work is an example meant to demonstrate the usefulness of adapting QbD approaches to nanoformulation development. Specifically, nano-formulations such as those presented here have the potential to improve the solubility and bioavailability of BCS class II drugs such as resveratrol.
Eric Lambert– Graduate Research Assistant, Duquesne University, Pittsburgh, Pennsylvania
Allison Kachel– Pittsburgh, Pennsylvania
James Drennen– Assoc. Dean Research and Graduate Programs, Duquesne University, Pittsburgh, Pennsylvania
Jelena Janjic– Pittsburgh, Pennsylvania