Jennifer Markarian is a contributing editor for 

Jennifer Markarian

	Maintaining data integrity is essential for all biopharmaceutical manufacturing operations. 

 interviewed Dan Nelson, Customer Experience Team member at DataNinja, which provides real-time manufacturing and quality software for the life-science industry, for insight on best practices in maintaining manufacturing data integrity. 

 What do you see as the most important considerations for manufacturing data integrity?

 There are really two main reasons why biopharmaceutical manufacturers care about data integrity: protecting the business and enabling the business. If manufacturing data are missing or inaccurate, then the firm is in violation of GMP regulations; it may mean that the quality of the product cannot be trusted. In serious cases, bad data integrity could lead to an FDA-mandated shutdown of production. I’d say just about every biopharmaceutical manufacturer understands this concern. So, out of fear, they invest heavily in business-protecting activities to ensure data integrity. These investments usually occur in the form of additional manual document reviews, more rigorous quality documentation, hiring more personnel, implementing electronic systems, and so on.

	Many companies thus view data integrity as a law-mandated, paper-intensive, cost center. Stereotypes are changing, however, as innovative companies are leveraging technology to turn data integrity into a business-enabling profit center.

	There are massive amounts of data collected during production, including machine usage; historian data; timestamped, in-process measurements; and materials usage. Forward thinking-manufacturers use these data for business optimization efforts. 

What are some examples of business enabling or optimization activities? 

 Once the business feels protected, it can progress to the next level, where the concern starts to become more about data accessibility. A thought process, for example, may be ‘Now we have production yield and quality data that are reliably collected, and we can trust their integrity. How might we leverage that data to minimize variability?’

	At a large contract manufacturer in Salt Lake City, Utah, for example, one quality manager took it upon her team to try and identify the root causes of batch yield variance. They meticulously went through batch production records, one filing cabinet at a time, to aggregate datasets of the suspected yield variability drivers. When they compared yield data to all the other datasets they discovered that one of their main suppliers was under filling, and that was the most significant correlation to batch yield fluctuation. 

	The practice of dataset correlation analysis is common. It’s how the hard, but often extremely valuable insights are earned. In all industries, the innovators compete with data. 

	Biopharmaceutical manufacturers are leveraging data integrity to answer their toughest operational questions, such as the following:

		Which piece of equipment is causing the most packaging deviations?

		How can I get more timely and accurate financials from manufacturing?

	More life-science manufacturers are stepping away from paper batch records. A paperless system lowers the cost of collecting data and increases the ability to analyze and profit from it.  It enables you to find correlations that otherwise would have been impossible or not cost effective to capture with a paper-based system. 

	Ideally, what you are after is a system that meets all your business-protecting requirements while simultaneously leveraging that data integrity for business-enabling activities. For example, with such a system in place, you could connect an initial entry on an electronic batch record to historian sensors and then process a financial transaction in your ERP [enterprise resource planning system], all in real time. 

 What are the benefits of electronic data systems for improving data integrity? 

 First, electronic systems improve accuracy. With an electronic system, you get real-time verification of operator data entry, instead of catching mistakes/missed entry in the final quality control review. Operators receive immediate feedback if they do anything that could cause a deviation. 

	Second, electronic systems make data searchable. Instead of burying loads of useful manufacturing data in a filing cabinet, an electronic system stores it in a database ready for value-adding insights to be extracted from data trends. No need to type in critical data points. Your analysts can get straight to discovering insights. 

 What are the data integrity risks associated with computerized systems and how can these be mitigated? 

 The data integrity risks associated with computerized systems fall into two main buckets: data loss and data manipulation. Loss of any manufacturing quality data is potentially catastrophic. Because of redundant database architectures and modern cloud applications, computerized systems are superior to paper in terms of total data loss. Computerized systems have their own unique weakness when it comes to data loss; there are more variables that can result in partial data loss. When it comes to paper, all you have to worry about is completion and having a grasp on the physical document. With computerized systems, you might have to deal with Internet disruptions, hardware failures, or human error (e.g., pressing log-out, instead of save). The risks are best mitigated with redundancy, audit trail robustness, and auto-saving. 

	Cloud redundancy can mitigate risk. If lightning stuck your server room and everything there was instantly destroyed, how would it affect your data integrity? Modern computerized applications merge government-level security with cloud agility. So, data are securely stored with multiple redundancies and ensured with multiple failover locations.

	Audit trail robustness is also crucial. With manual, paper-based systems, it’s crucial to know what was done and who did it. We have all reviewed records with an initial missing. It’s a major pain. There is a calculable probability that given the hundreds of initials required on each batch, one out of ‘x’ batch records will be missing an initial. Computerized systems combat this threat by tracking user actions based on user credentials. If a user completes documentation, the system records who did it along with a timestamp automatically.

	Auto-saving is important. A computerized system should not threaten your work based on whether or not you press a save button. This risk should be mitigated with auto-saving functionality. If you don’t have auto-saving, then you’d better focus your efforts on training. 

 How can manipulation of electronic records be detected and prevented?

 Detection of data manipulation has got to be built into the computerized system. Advanced computerized systems leverage MD5 hash encryption to verify that original data integrity remains intact. The MD5 hash is like a digital seal that is uniquely generated based on the contents of the electronic record. If data in the record were manipulated, the MD5 hash would reflect the change and detect the manipulation. Computerized systems for biopharmaceutical manufacturers must comply with FDA’s 

 21 Part 11 regulation for electronic signatures.

	The risk of data manipulation is mitigated through validation and access. It’s hard to manipulate pen on paper; it looks obvious when something appears to be overwritten. How can you trust that computerized systems contain no data manipulation? First, limit access. Make it impossible for people entering the data to manipulate it, whether in person or through a co-worker. Put in place total separation of powers so that there is no way someone could get someone else to manipulate data post-electronic signature. In regulatory jargon, this means you want your computerized system to be ‘open’.  

Articles

Electronic systems allow biopharma manufacturing companies to meet data integrity requirements and optimize practices.

Data Integrity Best Practices

	Both upstream and downstream biopharmaceutical processes have conventionally used stainless-steel equipment, but use of single-use systems (SUS) is becoming more established. An advantage of SUS is its flexibility-modular, disposable components have a lower up-front cost compared to conventional equipment, and the systems can be more easily replicated for installation in different locations to facilitate “in-country” production. One of the most significant differences is that SUS do not require the clean-in-place (CIP) and sterilize-in-place (SIP) systems employed in stainless-steel equipment systems, which saves initial capital costs and the ongoing production time associated with cleaning and sterilization. The process control and automation requirements of SUS also differ from those of stainless-steel equipment. 

 spoke with Michalle Adkins, director of Life Sciences Consulting, and Bill Smizaski, vice-president of Global Life Sciences Operations, both at Emerson, about considerations for process control and automation of biopharmaceutical processes.

What are the primary differences between automating a single-use system compared to a system that uses stainless-steel equipment?

 The complexity of automation is less for SUS because fixed, stainless-steel tanks require CIP/SIP, which increases the complexity of piping and instrumentation. In fact, 60–80% of instrumentation in a typical bioprocessing unit operation is associated with CIP/SIP, not the process itself.

	Although automation of a SUS process is simpler, SUS do require more manual interaction. For example, tubing needs to be connected and operators must make sure that nothing leaks. All parts that come into contact with the product need to be tracked, including disposable parts and reusable parts (e.g., peristatic pumps). Tracking SUS parts is also more complex. These disposable parts may be irradiated by the vendor. For example, tubing sets typically are supplied gamma-irradiated by a third party, and sensors are typically pre-calibrated and gamma irradiated. All this additional information needs to be part of the batch record. Maintaining data integrity and traceability becomes more complex.

	Another difference is in the sensors used to collect data and control the process. SUS employ either disposable or non-product contact instruments for measuring process control variables, such as conductivity, pH, and dissolved oxygen level. With traditional systems, sensors were designed to be highly accurate, long-lasting devices. In SUS, these sensors would be too expensive, because the sensor only needs to last for one batch. Disposable sensors are available today that are less expensive, but in some cases, are not accurate or reliable enough. Although advances have been made, there is room for improvement. To get the reliability and accuracy needed for certain processes, some manufacturers are choosing to use traditional sensors inserted into the process in a plastic, sterile sleeve; these sensors are sterilized for reuse. A new approach is completely noncontact-measurements are taken through the plastic container rather than inserting a probe into the product.

 How can single-use material data be tracked?

 Material tracking and linking material information to a batch record is part of ‘Level 3’ of the ISA 95 model, which includes the definition of manufacturing operations management (MOM) applications. Manufacturing execution systems (MES) that implement MOM can be designed to track both process material consumption and generation as well as single-use (e.g., tubing, instruments) and reusable components consumption for material traceability purposes.

 What are some considerations for automating the process as a whole rather than only at each unit operation?

 Companies can purchase an SUS skid with automation from the original equipment manufacturer and leave the individual skids as ‘islands of automation.’ Others are integrating all the skids throughout the plant using a distributed control system (DCS), while others are using a hybrid of these approaches. There are very few standards yet in bioprocessing, although over time, various options will likely consolidate and evolve into standards.

	The integrated DCS approach has the advantages of providing overall recipe management and putting all the data collected in to production context at the factory level. Not integrating the individual skids leaves the production execution management and data collection isolated and not in the context of a recipe as well as presenting challenges with alarm management, user security access management, and disparate operator interfaces.

 What are some of the challenges with automating a continuous biopharmaceutical process compared to a batch process?

 We are starting to see the move to continuous manufacturing in bioprocessing, although it is in its infancy compared to continuous manufacturing of small-molecule drug products. There are several licensed products that use a continuous perfusion process upstream, but downstream continuous processing is still evolving. The control strategy and automation are important in continuous manufacturing. The keys to successful continuous manufacturing include synchronizing unit operations into one process, defining the process in such a way to quickly achieve steady state, and managing disruptions to the process automatically. In addition, good process analytical technology (PAT) and ‘soft sensors’ (i.e., predicting values based on modeling) are important to ensure product quality. It is important to understand how disturbances affect the process and the product including how residence time would be used to track materials during a change or upset in the process. The industry is still at an early stage, but companies are beginning to do multivariate modelling and gaining the processing knowledge needed for effective continuous manufacturing.

 How will the Internet of Things (IoT) enhance automation?

 There are a couple components to IoT that are interesting for this industry. ‘Smart’ instrumentation has been around for some time, but it has not been pervasive in this industry. However, new data integrity requirements are starting a cultural change to use more of the data available from these devices. For example, a smart instrument measures not only the process variable it is designed to measure, but it tracks and records factors such as drift and vibration that can be used to look at the health of the device in real time. The IoT, as a concept, is a technology enabler to this change by making it easier to target the right information to the right user application. A device with a drift that would have a particular alarm meant for maintenance but not the operator, for example, will be sent to the appropriate person to take action.

	We are also seeing a greater focus in the pharma industry on reliability initiatives. Smart devices connected to equipment and automation, for either stainless-steel systems or SUS, will give an early warning and facilitate an intervention before a deviation or out-of-specification condition occurs.

	The IoT could also help enable continuous process verification (CPV) across the supply chain. Companies have been moving towards CPV rather than annual product reviews. Requirements are emerging, particularly for biopharmaceuticals, to provide CPV across all manufacturing locations for a product both internally and externally (e.g., at a CMO). 

:
	When referring to this article, please cite it as J. Markarian, "Automating Bioprocesses," 

The process control and automation requirements of single-use systems differ from those of stainless-steel equipment.