Over the last couple of years, I was lucky enough for GreyRigge Associates working together with Paul Nelson to be involved in a project on the use of design of experiments (DOE) for vaccine formulation for a lyophilized live attenuated tetravalent viral vaccine candidate. I was working with extremely talented teams led by David Volkin and Sangeeta Joshi at the Macromolecule and Vaccine Stabilization Centre at the University of Kansas and Jill Livengood, Steve Erb and Linda Strange at Takeda vaccines Inc., Cambridge, MA.
This for me was a great example of how data can be used to drive development of products by bringing together sound formulation science with smart use of statistics, in this case DOE, to achieve the desired outcome. Those working on live attenuated vaccines will be well aware of how difficult virus can be to stabilise. So while live vaccines can bring huge benefits in terms of minimising dosing regimens to one or two doses that still result in effective and safe long term immunity – stabilising them can be a real and painful technical challenge.
On this project, we utilised a semi-empirical screen to look at hundreds of potential excipients to identify those most promising to bring them down to a manageable level. I say semi-empirical as statistics help screen them to identify the excipients that gave a positive effect without the use of a more formal DOE screen. Once we were down to about fifteen candidate excipients, we selected the excipients we wanted to assess and constructed an experiment using a definitive screening design (DSD). These designs are one of my favourite experimental designs in that they allow you to have your cake and eat it; you can screen for critical factors and optimise them simultaneously. And the upshot is, you get to work out what matters, if curvature exists as well as identifying interactions and understand optimal settings all with a minimal number of experiments. Consequently, everyone is happy – the scientist doesn’t have to run too many experiments and the statistician has high quality data. In our case, with the number of excipients we screened in the design we had adequate redundancy with the factors, so this design type was perfect for us. We then looked at potency losses on lyophilisation and storage as responses for our design. These results meant that we could identify potential stabilising candidate formulations for use with our lyophilised live attenuated vaccine.
Using the models and the design space we had generated, we put together a number of candidate formulations. We evaluated them using a cell based potency assay for their ability to stabilise virus against the stresses of freeze-thaw, freeze-drying and accelerated storage (25°C) and real-time (2-8°C) stability studies by measuring the potency of the infectious viral vaccine. Ultimately, we identified an excellent candidate formulation that resulted in only a small loss during lyophilisation as well as demonstrating excellent stability on refrigerated storage. The exercise provided an excellent example of the utility of combining semi-empirical excipient screening and DOE experimental design strategies in the formulation development of lyophilised live attenuated viral vaccine candidate.