The Center for Global Development aimed to build an online tool to help policy makers see when effective COVID-19 vaccines were likely to be available. Bryden Wood has developed the elements of this tool that model vaccine manufacturing.

Modelling global vaccine manufacturing in an unprecedented pandemic situation is clearly a challenging undertaking so we have adopted an iterative approach. The report we have published (view here) describes the first version of the model. This version aims to demonstrate the method to give initial predictions of global vaccine availability and to start to identify potential pinch points.

A three-part model, using a platform approach

The overall model comprises three parts: one predicts the probability of vaccine candidates in R&D becoming approved for human use; the second predicts when manufacturing could start and the third when enough doses are made to meet the needs of target populations. This report describes the second and third parts. A ‘platform’ approach has been used to differentiate between different vaccine types. In the first version the model elements are linked but not integrated. 

Input data for the manufacturing elements comprises principally of the output of the R&D model; in-house information on the activities required to bring a vaccine into production and CEPI survey data on available vaccine manufacturing capacity. This is supplemented by expert interviews and literature searches.

There are several outputs. Firstly, there are overall predictions of when enough vaccines will have been manufactured to meet WHO global targets for defined populations. Secondly, we predict the proportion of the vaccine doses contributed by different platforms (Protein sub-unit, RNA etc.). By modelling the progression for a single vaccine platform we examine how the platform affects the time to make the required doses.

The model does not consider upstream supply chain restrictions nor the downstream distribution and administration of vaccines. It assumes that all drug products are sterile liquids and that all drug product production capacity is equivalent. Drug substance capacity is divided into categories and different vaccine platforms can only use certain categories. All vaccines within a platform have the same productivity. Each country’s production capacity in a given category is modelled as a single resource and all countries’ capacities contribute to a global total although individual countries can be ‘turned off’.

The scenarios we cover in the report

There is potential to run many scenarios by changing inputs and assumptions. Only three scenarios have been selected for the report; a base case representing the finalised dataset and a set of base assumptions arrived at after model development and testing; a scenario with a more optimistic scheduling assumption that appears to better match with announced manufacturing predictions from lead candidates; and a scenario that uses higher drug product manufacturing capacity and so reduces the time to achieve the larger population targets.

Of the four population targets, target three, where all vulnerable adults are vaccinated, perhaps best represents where social and economic activity would be affected much less by COVID-19. This is predicted to be probable in Q2/Q3 2022 for the global population. Manufacturing activities do not appear to have a big impact on the critical path to the start of production of approved vaccines. The time from the start of production to the smaller targets one and two being achieved does not appear to be constrained by capacity. The time to achieve the larger targets three and four is constrained by drug product capacity. RNA, protein-subunit, non-replicating viral vector and inactivated viral vaccines are the main contributors to the vaccine supply but none of these is dominant.

The base model pools all world manufacturing capacity. We examined the impact of removing major countries/blocs from the model. The results suggest that removal of one of China, the EU, India, Russia or the USA did not have a large impact on the time to vaccinate the world population but if ‘vaccine nationalism’ were severe and all of these were removed years would be added to achieving global vaccination.

Our recommendations

We recommend further development of the model to qualify some of the main assumptions. The lead candidates could be modelled in more detail and the capacity data could be updated and verified if manufacturers could share such information. The way that vaccines are allocated to available capacity could be modified to account for national and commercial constraints and for the impact of technology transfer to multiple sites. Ideally the R&D and manufacturing modelling could be linked with supply and epidemiological models to optimise the entire vaccination system.

Cooperation between manufacturers to pool knowledge and human resources and to standardise on certain techniques or equipment could shorten the time to start manufacturing and to achieving production targets. Explicit measures to make production facilities capable of switching from one vaccine to another, including second generation vaccines, are recommended. Means of expanding drug product capacity, including perhaps rationalisation of container types and formulation, can be considered with a view to shortening the time needed to achieve global vaccination.

Next steps

Bryden Wood are converting the models to an online tool to allow users to use their own input data and to try different scenarios. We will report again.

You will find the current report here.