How to reduce time to market by >50%

It usually takes 5-10 years and 1-3 billion USD to reach market approval for drug products. This creates a huge bottleneck when it comes to providing affordable drugs and preventing drug shortages. Besides conducting clinical trials to understand the effectiveness and safety of a new product, a process needs to be developed and approved by regulatory agencies that consistently delivers high product quality. This process development including process characterization (PCS) takes 2-4 years and needs several hundred experiments adding up to a multi-million-dollar undertaking!

In this article we would like to introduce a self-learning digital twin that enables the reduction of required experiments in PCS by more than 50% and hence also time to market.

How does it work?

Think about workers on a conveyor belt: if the second worker fixes most of the errors of the first worker, it is not necessary to look at the first worker in too much detail, however, it is more useful to understand to what extent the second worker can be challenged to compensate for possible errors. 

This analogy can be applied to biopharmaceutical processes. We replace the workers here on the conveyor belt by unit operations. The same principle still applies: if we understand which unit operation(s) produces poor product quality we can save runs on all the previous unit operations.

It is as simple as that!

To do this conceptually, we need to employ two types of experiments:

• A. Experiments that have been conducted at a specific unit operation to understand the impact of its PPs (Process Parameters) on the product quality delivered by that unit operation

• B. Experiments that help to understand how much the input material quality impacts on the product quality measured after the unit operation

In the past we focused only on the first type of experiments (type A). The reasons for this may be due to specific company business practices as different unit operations are handled by different people and departments. This leads to silos of data and knowledge.

Isolated clearance and spiking studies (type B) have also been conducted. However, this data has not been generated and analysed systematically, in addition the relevant information has not been evaluated together with type A experiments of process development at later stages, i.e. process characterization. This may also be due to another aspect of company business practice that leads to this information gap between the stages of the product life cycle.  

The self-learning digital twin

We have recently established a recommender system¹ based on well-established digital twin technology² that tells us which experiment to do at which unit of operation. Moreover, as new data arrives the end-to-end digital twin of the entire manufacturing process is updated and if required additional runs can be suggested (Figure 1). This routine is self-learning and stops as soon as sufficient process understanding has been reached to achieve filing of the process.  

In a simulation study that is based upon an industry relevant biopharmaceutical manufacturing process we could show that we can reduce the number of overall required experiments by more than 50% to achieve the same process knowledge regarding the final product quality (Figure 2). If you are interested in more technical details please give this paper a read¹! 

How can your company benefit from it?

In the past, process development was conducted by studying individual unit operations separately. These silos stopped us in unlocking the full efficiency and led to increased costs and prolonged time-to-market. Over the last 7 years we have had the pleasure to support over 20 products from more than 12 different companies on their complete process development journey until market approval³ ⁴ ⁵. In nearly all of these projects we have employed an end-to-end digital twin to calculate acceptance criteria and to define a control strategy.  

As described in this article, the same technology can now be used in a self-learning manner as a design tool to:

• Plan for the right experiments and save more than 50% of unnecessary runs¹
• At the same time increase the process knowledge to establish more robust processes¹
• 50% faster market approval of the manufacturing process

Many CMOs and pharma companies adopt this technology as the time to market becomes increasingly important to stay competitive. If you want to reduce development costs and shortcut time-to-market, we invite you to fasten your seatbelts on the journey with us!

Get in touch with us for more information or a free demo!

Get in touch!