Rapid development and deployment of high-volume vaccines for pandemic response

Abstract

Overcoming pandemics, such as the current Covid-19 outbreak, requires the manufacture of several billion doses of vaccines within months. This is an extremely challenging task given the constraints in small-scale manufacturing for clinical trials, clinical testing timelines involving multiple phases and large-scale drug substance and drug product manufacturing. To tackle these challenges, regulatory processes are fast-tracked, and rapid-response manufacturing platform technologies are used. Here, we evaluate the current progress, challenges ahead and potential solutions for providing vaccines for pandemic response at an unprecedented scale and rate. Emerging rapid-response vaccine platform technologies, especially RNA platforms, offer a high productivity estimated at over 1 billion doses per year with a small manufacturing footprint and low capital cost facilities. The self-amplifying RNA (saRNA) drug product cost is estimated at below 1 USD/dose. These manufacturing processes and facilities can be decentralized to facilitate production, distribution, but also raw material supply. The RNA platform technology can be complemented by an a priori Quality by Design analysis aided by computational modeling in order to assure product quality and further speed up the regulatory approval processes when these platforms are used for epidemic or pandemic response in the future.

Keywords: Quality by Design; RNA vaccines; bioprocess modeling; distributed manufacturing; pandemic‐response vaccine manufacturing; techno‐economic modeling; vaccine platform technology.

FIGURE 1

Overview of vaccine testing and manufacturing development timelines. A, Pre‐clinical development and clinical testing timelines, costs and success/failure rates for conventional vaccines and for vaccines produced using emerging platform technologies (eg, RNA vaccines).^{[ 13} , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ^{22 ]} Once the platform is developed and used to produce a licensed product, the costs and failure rates for developing further products using the same platform technology would drop substantially. The highest costs and longest development timelines are encountered in phase III clinical trials and the highest failure rates tend to occur in phase II clinical trials. B, Process development and facility construction timelines and cost estimates for conventional and emerging platform technologies with drug substance annual production capacities of tens to hundreds of million doses.^{[ 14} , ¹⁹ , ⁴⁰ , ⁴¹ , ⁴² , ^{59 ]} Process development and facility design is usually initiated during pre‐clinical and clinical testing and investments are usually made as failure risks reduce during clinical development. C, Comparison of overall vaccine production rates for conventional and new platform technologies, considering the development and testing phases presented in parts A and B above. Once fully developed and validated, the new vaccine platform technologies will produce vaccines within weeks to months after antigen identification, which is at least 10‐fold faster than conventional technologies

FIGURE 2

Process flow diagram for saRNA vaccine production based on the in vitro transcription enzymatic reaction. In the upstream process the DNA template is generated, amplified, purified and linearized. In the mid‐stream process the RNA is synthesized following the in vitro transcription reaction using the T7 RNA polymerase enzyme, and 5′ capping of the RNA is achieved co‐transcriptionally using 5′ cap analogues (needed to ensure antigen expression). For downstream purification TFF can be used also in combination with chromatography methods, such as hydroxyapatite chromatography and core bead flow‐through chromatography. In the first TFF step the saRNA and linearized DNA template are retained by the filter and smaller molecular size components, including the T7 RNA polymerase enzyme, flow through the filter. Next, the linearized DNA template is digested using nucleases and then the DNA nucleotides can be separated from the RNA using another TFF step. The obtained drug substance is then formulated predominantly in lipid nanoparticles, however polycationic formulations are also developed and evaluated. Next, the formulated saRNA undergoes quality control and is filled into vials or containers for pandemic‐scale mass vaccination. The vials are then capped, sealed, inspected using automated image processing, labeled and packaged into secondary and tertiary packaging. The entire production process is independent of the RNA sequence, therefore in principle vaccines against virtually any disease can be produced using the same production process^{[ 24} , ⁴⁶ , ⁴⁸ , ⁴⁹ , ^{60 ]}

FIGURE 3

Quality‐by‐design (QbD) framework. The QbD development cycle begins with identifying the patient needs and based on these the Quality Target Product Profile (QTPP) is defined. From the QTPP, the critical quality attributes (CQAs) of the product and their ranges are determined using a risk assessment scoring, based on clinical and non‐clinical data, for both safety and efficacy. Next, based on the CQAs and on understanding the of production process, the critical process parameter (CPP) ranges are defined. Mathematical relations between CPPs and CQAs are established, obtaining this way a mathematical model of the vaccine production process. Using this model, the ranges of CPPs which yield the desired CQAs are determined. Based on these CPP ranges, the design space is determined and therein a sub‐space called the normal operating range (NOR) is defined. The NOR offers the flexibility of modifying operating parameters in the GMP production process, thus allowing optimization to account for inherent biological heterogeneity, instead of “freezing” the GMP process. The QbD bioprocess model can be adapted for advanced process control, using model predictive control and real‐time measurement data from the production process. Such a “digital twin” model can predict CQA values in the following time window (eg, next 5 minutes) and if CQAs are predicted violate the specified ranges, the model can recommend corrective measures, that is, control actions, to prevent CQAs going out of the specified ranges, fixing mistakes before these occur. Thus, computational modeling tools can be integrated with experimental development and QbD follows an iterative development cycle to ensure continuous improvement through the product‐process life cycle^{[ 58 ]}