Life Sciences Insights

Sharing expert knowledge via our latest blog posts

Quality Control challenges in closed systems: implementing AI as an in-process control

Sampling, testing, and preserving your medicinal product in closed systems often present difficult Quality Control challenges. As part of the solution, various AI-based in-process controls can be used. Discover them here!
Quality Control challenges in closed systems - implementing AI as an in-process control - QbD

Table of Contents

A Quality Control (QC) system is essential to ensure the quality and safety of your finished product. Think of In-Process Controls, test methods, and a sampling plan.

In addition to the challenges during QC, further difficulties arise when working in a closed, sterile system during the production of Advanced Therapy Medicinal Products (ATMPs). ATMPs are complex medicines that offer breakthrough opportunities for human use based on (1) tissue engineering, (2) gene therapy, and (3) somatic cell therapy.

To tackle these QC challenges, some traditional analytical approaches will evolve towards machine-learning models (such as random forests, shallow neural networks, Gaussian processes, partial least squares, …). Today, however, most of these promising QC solutions are still in academia.

Curious about the key Quality Control challenges in closed systems and potential solutions to these challenges? Then be sure to read on!

What is a closed system?

A closed system is designed to prevent contamination of the aseptic products by the environment. Three important criteria for checking the readiness of a closed system are

  • the cleanliness level,
  • the bioburden level,
  • and the degree of separation from the environment.

During quality control, closed systems can be opened, which poses several risks. Therefore, a decontamination or sterilization step is necessary before using the process. Safety and sterility must be ensured to keep the products reliable, stable, and trustworthy.

QC challenges in closed systems: sampling, testing, and preserving your product

Testing your product is important to demonstrate that each sample and lot meets the required quality and specification defined in your product’s registration dossier. The Good Manufacturing Practice Guidelines specific to Advanced Therapy Medicinal Products define all required testing for your product prior to marketing.

When you want to test your product, you don’t want to use your entire batch. Therefore, at critical points in your process flow, you take a certain number of samples, which are considered to represent the entire batch. Before you can start collecting samples, there must be a sampling plan. The sampling plan should include the following aspects:

  • How many samples are you going to take?
  • In what part of my process am I going to take samples? What are the critical parts of the process part?
  • A detailed description of the sampling process, including SOPs.
  • The practical limitations that may exist.
  • Possible mitigation measures.
  • How long should all the different samples be kept in accordance with the legislation

The problem with closed systems is that there is no way to just take some samples every now and then. There are specialized machines and tools that can be installed on the process stream, but they are very expensive. So usually tests are done with finished product samples as they just come out of the closed system.

This approach has some disadvantages:

  • You cannot actively monitor your process and correct process deviations as they occur.
  • You cannot test various parameters of your intermediates because they are no longer present at the end of your process.
  • Physical effects on your product and where they occur cannot be precisely traced in your process flow due to the characteristics of your closed system.

There are two types of ATMP therapies.

  • In the first therapy, called autologous and personalized therapies, 1 batch is produced per patient. This means that there is a limited volume of product and only one or a few samples can be taken to leave enough product for your patient.
  • In the second therapy, allogeneic therapy, 1 batch can be used for many patients.

As you can see, sampling, testing and preserving your product in closed systems – especially when it comes to ATMPs – poses tricky Quality Control challenges. Let’s have a look at some ingenious solutions below.

Solutions for Quality Control in closed systems

Soft sensor

‘Soft sensor’ is a combination of  ‘sensor’ and ‘software’. The signal evaluation models of the sensor are implemented in computer programs, which take data streams from hardware sensors as inputs.

‘Soft sensing’ is a data-based analysis method used in quality control, combining these so-called hardware sensors and software-based process models to predict or monitor the state of the process.

Hardware sensors are often used because they can be easily implemented in closed systems (oxygen, pH, spectroscopic sensors such as NIR or Raman, biosensors for nutrients and metabolites, impedance, …).

The machine learning model will then link the data from these sensors to the quality attributes that cannot be monitored or measured. Soft sensing is a very flexible, widely applicable solution.

Soft sensing example: Intelligent cameras

An intelligent camera is a combination of a camera, image processing and machine vision. It is an ideal example of a soft sensor. 

The traditional way is as follows: an expert must decide which features are important to inspect after capturing an image (edges, corners, color spots, curvatures, etc.). 

Then, a rule-based system can classify an object and the resulting system automatically checks whether the product is produced as expected. 

Smart cameras are used, for example, to read barcodes of pharmacological products. Now, people are increasingly moving to the use of machine learning for image analysis. In this case, the smart camera is called a soft sensor.

Want to know more about Soft Sensing? Read our blog on “Pioneering with artificial intelligence to make personalized cell therapy more accessible”.

Inline sensor

Sensors are placed in the process stream or vessel to perform various analyses during the process. These sensors allow for automatic measurements.

Common in-line sensors include pH, oxygen, and biosensors (which use enzymes as part of the sensor) that can measure glucose, lactate, or ammonia.

Some more advanced in-line sensors are optical probes for spectroscopic measurements (NIR, Raman) or probes that quantify the electrical characteristics of the sample (impedance spectroscopy).

At-line sensor

Sensors placed at the line for analysis. For example, a sensor attached to a production line can monitor the products or vials that are passing in a certain timeframe.

Single-use products

Single-use products such as the ‘Takeone Aseptic Sampling System‘ can be used in closed systems such as the ‘CliniMACS Prodigy‘ to reduce cross-contamination and biological contamination. Sterility, accuracy, speed, reliability, and quality can be assured.

More specifically, the Takeone Aspetic Sampling System is used as an effective microbial monitoring program to measure sterility, bioburden, endotoxin levels, etc. Single-use products rely on sampling (and possibly product) and thus are used for destructive analysis.

This solution will only provide information on a few (maybe even just one) moments in the process. No real-time or continuous information is obtained and as such, these single-use products cannot be used for process control.

Conclusion: in-process devices to the rescue

When manufacturing sterile products, such as ATMPs, in a closed system, there are several challenges for testing your product, obtaining and storing samples outside your process flow.

To maintain sterility in the cleanroom or closed system, you need to look for in-process control devices such as soft sensors, smart cameras, ultrasonic and inline sensors.

These tools can perform various checks as early as during your process so that a problem can be detected and corrected on the fly.

Need help with implementing, calibrating, and validating in-process controls? Please don’t hesitate to contact us.

Expert knowledge in ATMPs
QbD is your partner to make ATMPs more stable, robust, and upscalable.

Did you find this article interesting? Thanks for sharing it with your network:

Share on linkedin
Share on twitter
Share on whatsapp
Share on email

More expert content?

Don't miss out!

Table of Contents

Subscribe now for more fresh content.