Disruptive technologies in the production of pharmaceuticals

Experts:                Martin Düblin, One One Eleven GmbH and GMP inspector Rainer Gnibl (PhD)

Moderation:         Barbara Peither

Protocol:               Nadja Schaubhut

While modern technologies such as AI, machine learning and the Internet of Things have long been standard in other industries, they are still used very little in the pharmaceutical industry. What are the concerns? Where are solutions already available? What does our future look like? How do the authorities view the new technologies? Would a "reset" in legislation be necessary to move this new world forward?

"Artificial Intelligence" as an umbrella term (Illustration based on diagram 1 in "Perspectives and Good Practices for AI and Continuously Learning Systems in Healthcare", Xavier University, Xavier Health Organization August 2018)

Is it possible to dispense with the four-eyes principle by using robotics? How does the AI "get to the top" or to the authorities and what about the acceptance by the authorities?

The first question does not really need to be asked any more: The dual control principle can already be covered. After all, this is a principle that is intended to reduce the error rate of people. The algorithm controls itself. A system that has "machine learning" constantly requalifies itself. The qualification is automatically confirmed and the error rate decreases. No more qualification phase is necessary. We are heading for this in about five to ten years. However, the regulations must be adapted for implementation. This process is already too slow. A "reset" is needed: coordination processes must be accelerated. To do this, however, we must first create an awareness of the rapid progress of developments in this area and thus of the need for legal adjustments. Interdisciplinary cooperation is urgently needed. Even within the authorities there is no such thing as cooperation. One has to talk about virtualized environments and their qualification!

This brings up the question of so-called "robot laws"...

So far, there is no legal basis for such legislation in Germany or even Europe. Autonomous driving cars were mentioned as an example. Politicians must take urgent action here, because the world is ready for it. Compared to China, Europe is helpless, as both politicians and authorities are too far removed from the issue. In China, the legislative process is faster. There is technological leadership and therefore investment in technology.

Are there approaches for a combination of robotics and PAT (process analysis technology) and are analyses still affordable with full automation?

The current knowledge and procedures must be implemented in the new systems. The foundations have been laid and, according to the representative of the authorities, no authority can refuse to do so. The feasibility is being demonstrated by other industries. The specifications and contents remain the same. However, new processes must be initiated. Sterile drug production was mentioned as an example: The whole chain must work. It can and must be produced around the clock, otherwise the costs would become too high. A high number of batches is therefore a basic requirement. Traceability would also be given, because the documentation is directly included. Diagnostics have been working in this way for a long time. With 24/7 operation with low downtimes, the advantages clearly outweigh the disadvantages in the long term. The legal basis for this is provided not least by Annex 17 of the EU GMP guidelines.

Who is responsible for the drugs and their quality? How do I get the process valid and who confirms it?

Until now, the pharmaceutical company or the persons in the corresponding functions have been responsible for drug quality. This changes with the status of full automation. The supplying companies that provide the automated systems bear a large part of the responsibility. The quality is confirmed by the system. We must define this system and we are clearly required to do so. This makes the training of the QP and the authorities much more IT-intensive and extensive. A rethink must take place: So far, only the results can be verified. Thus, only indirect verification by means of a functional test is possible. In order to find the error in a software programming, the code has to be searched laboriously. Therefore, the programming must become faster and easier to verify. When programming the software, it must be clear from the beginning what the end result should be.

Will the planning also be done by computers when no one is working anymore?

The solutions are already on the market. The planning is done by the system.

What about protecting data from hackers? Can the data be trusted? Who controls the AI?

A residual risk can never be ruled out. Criminal energy can always be present. It is already possible to paralyze the entire production by a hacker attack. Solutions will come gradually.

What happens to the free capacity of people when they stop working? What about individuality, if everything works on the basis of the same algorithms?

In the year 2050/60, all existing knowledge could be available to everyone at any time. This would create equal educational opportunities for all. Life within an acceptable footprint would be possible for everyone. Individuality is not excluded: Those who enjoy doing a job can still do it. In addition, a look at contemporary history shows that trends will come back. Much has been there before and we encounter it again in "retro style".

How can we take away people's fear of new technologies and why do robots always look like humans or animals?

People must be involved. More education is still needed, the subject is often presented in a negative light and fears are stirred up. Each and every individual can contribute to spreading information. Robots look like humans or animals because they are better accepted in tests than, for example, a box with a geometric shape. Moreover, robots primarily replaced human activities.

Bottom line:

• The solutions are largely in place, but regulatory adaptation is lagging behind.
• The risks also affect us already now.
• There is an urgent need to improve interdisciplinary and inter-agency cooperation. The whole process must be reconsidered.
• The public must be more involved, also to remove uncertainties.