Jennifer Markarian is manufacturing reporter for Pharmaceutical Technology.

Jennifer Markarian

	Multivariate data analysis (MVDA) is being used to effectively handle complex datasets generated by process analytical technology (PAT) in biopharmaceutical process development and manufacturing. From these large datasets, MVDA can be used to identify the parameters that are causing most of the variability in a process. These parameters can then be controlled to improve consistency in the process and product quality, noted Rathore and Singh in an article reviewing use of MVDA in bioprocessing (1). The authors gave an example of MVDA used in scale-up of a cell-culture process from 2 L to  2000 L. Data analysis showed that osmolality and ammonia levels were changed upon scale-up, indicating a change in cell-culture performance. The change in osmolality resulted from the buildup of carbon dioxide due to less-efficient gas transfer at the larger scale.

	MVDA can also be used to establish comparability of processes and products, which is crucial for biosimilar development, noted Rathore and Singh (1). Chemometric algorithms can be used to compare different phases of manufacturing. Partial-least squares analysis of data in a laboratory-scale and a production-scale fermenter, for example, identified which variables were responsible for differences between the two scales.

	Sartorius Stedim Biotech (SSB) recently integrated chemometric software into its microbioreactors used for bioprocess development. "Providing a consistent, scalable platform for design-of-experiment (DOE) studies and data analysis will help scientists develop robust and flexible manufacturing processes based on single-use bioreactor technology," said Mario Becker, director of marketing, PAT and Automation at SSB, in a press release (2). He said that the MVDA toolkit would help reduce risk in bioprocess development and achieve more rapid, cost-effective production. The combination of microbioreactors and chemometric software can be used for media optimization, process parameter screening applications, and Monte Carlo simulations as part of a quality-by-design program to identify the desired operating region for manufacturing-scale processes, commented Barney Zoro, ambr15 product manager at SSB, in the press release.

      
	1. A.S. Rathore and S.K. Singh, "Use of Multivariate Data Analysis in Bioprocessing," 

www.biopharminternational.com/use-multivariate-data-analysis-bioprocessing

, Apr. 1, 2015.
	2. Sartorius Stedim Biotech, "

New ambr Bioreactor Systems Enhanced with Software for Design of Experiments

," Press Release, (Goettingen, Germany and Cambridge, UK, Feb. 25, 2015). 

Articles

Multivariate data analysis (MVDA) is being used to effectively handle complex datasets generated by process analytical technology (PAT) in biopharmaceutical process development and manufacturing.

Multivariate Data Analysis Finds Use In Biopharmaceutical Process Development And Scale-Up

	Process modeling is becoming established as a method to design and optimize pharmaceutical processes using data from process analytical technology (PAT). Modeling software can be implemented to monitor the process and increase process understanding and, if desired, to control the process.

	A process control model takes process data inputs, such as temperature and flow rate, from traditional sensors, as well as spectral and other data from PAT, and calculates a predicted output. In an open loop, process operators evaluate this predicted data and decide whether a change to the process is necessary. In closed-loop, advanced process control (APC), a process controller uses the model to make changes automatically, to keep the process in specification.

	One use of process modeling and control is to evaluate the quality characteristics of incoming raw materials and determine how these characteristics will affect the final product. "A raw material could be within specification, yet result in an out-of-specification end product," notes Jakob Mohr Christensen, general manager at MKS Umetrics. "The modeling software can be used to determine how process parameters could be changed to account for variability in a raw material or, in some cases, to identify whether a given material should not be used at all. The software can be used to build quality control charts for the materials that go into the process."

	Modeling software can be used to meet various goals, including improving the quality or economics of a product, or determining a design space. To build a process model, users identify the relevant variables and extract representative data, either from the existing process or, if the process is new, from designed experiments, explains Chris McCready, principal engineer at MKS Umetrics

 With Umetrics’ offline SIMCA modeling software, users refine the model by analyzing data and reviewing it with process experts to compare the model to existing process knowledge. Next, they verify the model on the system using SIMCAonline software, while the process is running, and the controller is commissioned to monitor and/or control the process.

	Currently, most of the pharmaceutical clients who use Umetrics' SIMCA-online process modeling software use an open-loop mode. "This 'advised future' mode is typically the first step of applying process control models to a pharmaceutical process, but the software can also be programmed to allow closed-loop control," explains McCready. "Although APC is a mature technology, a challenge is in applying these methods to pharmaceutical processes where the regulatory environment was written for release of batches. A few companies are building continuous processes with advanced control, while others are waiting to see."


	Umetrics SIMCA modeling software uses finite impulse response (FIR) modeling to handle "data lagging," which is how the state of the process at a given time point is influenced by what was happening in the process at an earlier time. "Some variables have longer lag times than others," explains Lennart Eriksson, senior principal scientist and lecturer at MKS Umetrics. "The most appropriate lag sequence must be identified by testing each case as part of the overall process of refining the process model." FIR modeling in a continuous process can be compared to the method of "data unfolding" in a batch process, explained McCready in a presentation at IFPAC. The Umetrics software can be configured to either batch or continuous or even hybrids of the two. For example, in a biopharmaceutical perfusion process, the software is used to monitor the initial cell growth, which is a batch process, and then to monitor and control the steady-state, continuous perfusion process. 


	The topic of control systems for continuous processing will be discussed on the 

 (New York City) on Tuesday, April 21. Doug Hausner, Associate Director for Industrial Relations and Business Development at the Engineering Research Center for Structured Organic Particulate Systems, Rutgers University, will present "Meeting Regulatory Expectations for Continuous Processing Sensing and Control," and will review technology and regulatory expectations for implementing next-generation continuous processes in solid-dosage pharmaceutical manufacturing.

	To hear Chris McCready explain the steps of implementing process control software, tune in to 

Use of modeling software can help improve process understanding, and can be used in open- or closed-loop control.

How Modeling Software Can Help Process Development

	
	Continued advances in inline process analytical technology (PAT) are making real-time data available for process engineers to better understand and optimize pharmaceutical processes. "In the past, there was a vacuum of real-time information, and process engineers had to wait for and rely on laboratory data that didn't necessarily show what had really happened in the process," says Luke Kiernan, technical services director at 

. "Now, inline PAT provides information that can be used to find the true root causes of batch failure, which allows processors to implement the most effective corrective actions." In addition to troubleshooting, inline PAT data can be used to increase process understanding and as an input to process modeling and control software.

	Accuracy and robustness of analytical instrumentation are important, and-particularly in pharmaceutical processing-non-product contact sensors are easier to integrate. "Nondisruptive technologies, which don't require cutting into validated equipment for installation, reduce the change-control burden," explains Ian Jones, founder and CEO of 


	Innopharma Labs' Eyecon particle characterizer is a real-time, non-product contact, high-speed machine vision system that takes images of particles in processes such as granulation, coating, spheronization, and milling. The images are analyzed to measure particle size distribution, morphology, and surface roughness. Jones and his colleagues presented case studies demonstrating Eyecon in fluidized-bed granulation and continuous wet granulation at the IFPAC 2015 conference held in Jan. 2015 in Arlington, VA, and an article, "

Monitoring Fluid-Bed Granulation and Milling Processes In-Line with Real-Time Imaging,

" has been published in the March issue of 

	Inline imaging is more accurate than at-line sieve analysis, which is the simple method long used to measure particle size distribution, notes Jones. Imaging is also a direct measurement that requires little calibration compared to online laser diffraction particle characterization techniques, which extrapolate from measurements to calculate particle size and require product contact. "The beauty of imaging is that it is an eye into a process," explains Jones. "It is helpful to actually see the product, because you can compare what you see with what instrument data is saying about the process. Imaging is powerfully improving our understanding of the process because we can see in real time how particles are changing." 


	Near-infrared spectroscopy (NIR) is another PAT commonly used in solid-dosage manufacturing processes to measure properties such as moisture content and blend uniformity. Multiple presentations at IFPAC discussed use of NIR for applications including API crystallization and tablet blend uniformity.

	NIR has typically been a single-point measurement, but last year Innopharma Labs introduced the Multieye, a multipoint NIR spectrometer that can measure up to four points at the same time, which results in a larger sampling area that gives a better representation of the product. From a regulatory perspective, it is easier to demonstrate that multipoint analysis is a representative measurement, says Kiernan.

	NIR is complementary to imaging and the two can be used to cross-validate each other, adds Jones. "If you see an unusual spike in NIR, for example, you can look at the imaging data for a possible reason."


	Integrating PAT into a process requires customization for each process and equipment setup. For example, the type of probe needs to be selected carefully and integrated into a system to obtain consistent, representative sampling, explains Jones.

	Making the most use of PAT data is another challenge. Process monitoring and increasing understanding are typically the first step. Using data for feedback to control the process is the next step that, so far, a few companies have taken using process modeling and control software systems. 

Real-time, noncontact imaging and spectroscopy techniques provide insight into pharmaceutical processes.

Advances in Process Analytical Technology Improve Understanding

	
	A transformation in thinking about how to manage data for controlling manufacturing processes is needed to obtain improvements in efficiency and quality in the pharmaceutical industry, said thought leaders at the International Society of Automation (ISA) Food and Pharmaceutical Industry Division (FPID) symposium on Automation & Serialization in the Pharmaceutical & Food Industry, held Feb. 10-11, 2015 in Philadelphia, PA. 

	Quality must be controlled throughout the supply chain by managing production information from the raw material through to the patient in a holistic, end-to-end process, said Paul McKenzie, currently vice-president of Ethicon global R&D and recently head of Janssen's Manufacturing and Technical Operations. One problem is that information management is often still paper and people based, rather than automated and electronic. Product lifecycle management (PLM) seeks to organize and automate product information handling end to end.  However, today’s systems are typically facility based, and various stages of a drug product may be manufactured at many different plants, including both internal facilities and external suppliers or contract facilities. In addition, different facilities may use different information management systems or applications. "We spend too much time recreating the same information in different systems, rather than having one, product and process data-based system," said McKenzie. The solution to the need to manage information from multiple plants, he proposed, is to base production information and recipes on the ISA S88 standard for batch process control, which provides an organized way of defining and describing processes and equipment.  Over the past year, Janssen implemented this standard for one of its products. Now, with fully integrated data that feed back in real time, process engineers can optimize the process with cycle-time analysis and troubleshoot out-of-specification deviations. "The key is expression of the product end-to-end, drilling down to the plant and back up with data," said McKenzie. The open S88 standard allows greater transparency with outside partners, because information can be communicated easily. With a standard "language" for data, all parties are seeing the "same truth," whether they are in different companies, at different facilities, or simply in different departments (e.g., quality control or process), he explained.

	The S88 recipe structure can also be applied to R&D. An R&D "data warehouse" that can hold data in a standardized recipe context can be used for knowledge management, and data can be drawn from it for data analysis or to compile regulatory filings, said Shaun McWeeney, principal engineer in the API large molecule group at Janssen, in a separate presentation. "Data warehousing makes it easier to analyze data. We don't need to waste time copying and pasting into spreadsheets to make graphs," he explained. Data warehousing is broadly used in other industries, such as retail, and it can help pharmaceutical companies be much more efficient, noted McWeeney. 

	Electronic production records (EPR) provide another tool for improving efficiency of data management in pharmaceutical manufacturing. Implementing EPR for a site-wide equipment monitoring system enabled electronic reports with review-by-exception and electronic signature approvals; saving several thousand hours per year of employee time previously spent manually logging parameter data and reviewing monthly reports fully paid for the infrastructure needed, noted Ken Kovacs, quality business systems manager at Fujirebio Diagnostics, in a presentation. The system was initially implemented using the ISA95 functional hierarchy model. That first project was completed three years ago, Kovacs said, and provided an infrastructure to add equipment both last year and this year. Having more accessible data allows the company to better analyze equipment performance. "We can address issues that weren't necessarily excursions, but that can be improved to make equipment more efficient. Looking at 24-hour, 7-day-a-week data also gives us a better understanding of the impact of equipment cycling," added Kovacs. The company is now applying the ISA88 process model to expand this system’s infrastructure for use in manufacturing operations management.

Advances in process modeling and use of open-architecture control models will lead to more efficient process control methods.

Forward Thinking in Data Management

	Improving understanding and practice of visual inspection of parenteral drugs is a high priority for pharmaceutical companies and industry associations. The Parenteral Drug Association (PDA) and the United States Pharmacopeial Convention (USP) have been working to educate and provide guidance to the industry in the hope of establishing consistent, practical guidelines and best practices that will lead to improved quality and reduced recalls due to visual defects. 

 spoke with John Shabushnig, PhD, principal consultant at Insight Pharma Consulting, group leader for PDA’s Visual Inspection of Parenterals Interest Group, and a member of the USP Visual Inspection Expert Panel, about guidance from USP and industry best practices.
	
	

General Chapter <790> ‘Visible Particulates in Injections’ became official in August 2014 (1). Can you comment on the sampling plan and acceptance criteria that the chapter provides to help define that a product is ‘essentially free’ from visible particulates?
	
	

: I think the ANSI/ASQ Z1.4 (2) or ISO 2859 (3) sampling plans are widely used and generally accepted by industry and regulators. These are the same general plans used to sample components before use as well as in support of batch release. The actual sampling plan chosen from these standards to achieve the level of protection specified is based on the size of the batch or lot produced. There are varying levels of understanding of how these plans should be applied and the level of protection the various sampling plans provide. The USP Expert Panel believes an acceptable quality level of 0.65% is an appropriate general value for application to particles in injections based on the current understanding of the risk of potential harm to patients. I think it is still early to know how widely accepted this value is or will be. Based on the regulatory emphasis and large number of recalls associated with particles in recent years, many companies have implemented tighter internal limits to define ‘essentially free’ [from visible particulates]. I also see some confusion about the application of the special sampling plans referenced in the standards and recommended for use when supplemental testing is required. Such testing is performed when the container or the formulation makes thorough inspection difficult. This testing is in addition to the non-destructive inspection of all units in the batch (100% inspection). Since these tests are destructive, such as reconstituting a powder or removing the contents from a dark amber vial or translucent plastic container, the number of units tested must be smaller than what is used for non-destructive testing.

	The new chapter was discussed extensively at the PDA Visual Inspection Interest Group meetings this year as well as at the PDA Visual Inspection Forum. USP offered a live webinar on this chapter along with another on the subvisible particle chapters <787> (4) and <788> (5) in December 2014. I expect education activities and further discussion to continue in 2015 through both of these organizations.
	
	

: Is there consensus on ‘essentially free’ when applied to particulates? What effect have you seen in industry practice as a result of this guidance?

: I believe there is a developing consensus on what is meant by ‘essentially free’ as now defined in 

 <790>. Certainly there may be special circumstances where a tighter limit may be required for certain high-risk patient groups or routes of administration. These are discussed in an industry position paper drafted by a team of inspection experts and medical officers brought together by PDA earlier this year. This paper, an ‘Industry Perspective on the Medical Risk of Visible Particles in Injectable Drug Products,’ is scheduled for publication in the PDA 

 early in 2015. It supports the general concepts in 

 <790> and provides guidance on assessing patient risk for the inadvertent administration of small numbers of particles.

: What are the key points of the draft information chapter on visual inspection?

 <1790> ‘Visual Inspection of Injections’ was recently published (6) in draft form with the opportunity for all stakeholders to comment. 

 <790> benefited from this commenting and revision process. General chapters by design are typically short and specific to the test method and acceptance criteria. Information chapters permit further discussion on the points to consider and the concepts behind the methods described in the general chapters. 

 <1790> contains discussion of the key elements of an effective inspection process. It addresses visual inspection in general, including container and closure defects, as well as particles, which are the central focus of 

 <790>. It contains discussion on patient risk and risk factors. It further develops the concepts of a lifecycle approach to inspection with the goal of defect prevention through continuous process improvement. It emphasizes gaining process knowledge through inspection, and not just sorting and removing nonconforming units. It provides more information on defect types and classifications as well as discussing the three common modes of visual inspection: human manual visual inspection, semi-automated inspection, and fully automated visual inspection. There is also discussion on the qualification of human inspectors and the validation of automated systems.

: What are the most important actions companies need to take to be sure that their inspection systems are following good practices?

: I believe it is important to understand the specific inspection process in use for the products that a company manufactures. What can be detected and removed during 100% inspection will vary with product formulation and with different primary packaging types. This also is important to determine if supplemental destructive testing, as previously described in 

 <790>, is needed. Such testing may involve dissolution of a powder or freeze-dried cake or removal of the contents of a darkly colored or non-transparent container to assess if previously unseen particles are present. It is important to have a good list of defects so it is clear what the inspectors need to find. This is based on the nature of the product as well as an understanding of observed defects and potential production process and materials failure modes. This list, often supported by photographs and physical samples, provides a good foundation for inspector training and qualification as well as equipment validation. Finally, I also believe it is important to track and trend inspection results to support continuous process improvement.

: What are some of the key results from the PDA VI benchmarking survey?

: Inspection practices are generally in alignment with the current conditions specified in 

 <790>. Typical rejection rates are between 1-2%. The most common reason for rejection continues to be particles, with fibers being the type of particle most often observed.

: Inspector training seems to be key. What best practices do you recommend for inspector training and qualification?

: As stated earlier, start with a good list of expected defects supported by photographs and physical examples. Provide sufficient time to practice the inspection technique to gain proficiency. Then document inspection ability with a good test set. This set should be different than the training set and not have too high a defect rate so as to provide an experience comparable to routine inspection. When assessing performance, be sure to set a time limit on completing the test that aligns with normal production inspection rates and also have a limit on the number of good units that can be falsely rejected, so that test results are not biased.

: What are some unique inspection challenges for biopharmaceutical products?

: Biopharmaceutical products are often sensitive to shear. Care must be taken when swirling the product so as not to damage the therapeutic proteins. 

 <787> ‘Subvisible Particulate Matter in Therapeutic Protein Injections’ (4) was recently published to address some of these issues for subvisible particle measurement. Concerns of damaging the drug molecule are especially true when considering a high-speed spin available in most semi-automated and automated systems when inspecting for particles in the visible size range. The other issue that arises is that of ‘inherent’ particles. These are particles formed from protein agglomerates. Through clinical testing and the regulatory approval process, such particles may be shown to be safe and insignificant to efficacy. The difficulty is then ignoring these insignificant particles but still remaining sensitive to extrinsic and intrinsic particles.

: I would like to acknowledge my fellow members of the USP Visual Inspection Expert Panel, who wrote Chapters <790> and <1790>: D. Scott Aldrich, Ultramikro; John Ayres, Eli Lilly; Roy Cherris, Bridge Associates International; Desmond Hunt, USP; Steve Langille, US FDA; Russell Madsen, The Williamsburg Group; and Deborah Shnek, Amgen.


	1. USP General Chapter <790> “Visible Particulates in Injections”(US Pharmacopeial Convention, Rockville, MD, 2014).
	2. American Society for Quality, ANSI/ASQ Z1.4 

Sampling Procedures and Tables for Inspection by Attributes 

Sampling procedures for inspection by attributes--Part 1: Sampling schemes indexed by acceptance quality limit (AQL) for lot-by-lot inspectio

 General Chapter <787> “Subvisible Particulate Matter in Therapeutic Protein Injections” (US Pharmacopeial Convention, Rockville, MD, 2014).
	5. 

 General Chapter <788> “Particulate Matter in Injections” (US Pharmacopeial Convention, Rockville, MD, 2014).
	6. 

 Draft Chapter <1790>, “Visual Inspection of Injections,” Pharmacopeial Forum 41 (1) Jan-Feb. 2015. 

When referring to this article, please cite as J. Markarian, “Improving Visual Inspection Practices,” 

New guidelines and best practices may lead to improved quality and reduced recalls due to visual defects.

Improving Visual Inspection Practices

	
	Pre-sterilized, nested syringes are an alternative to bulk syringes that must be sterilized at the filling line. As the injectables market grows globally, there is growing use of pre-sterilized syringes because they are "off-the-shelf" systems, comments Kevin Wrigley, product and service manager at Vetter. Bulk syringes, which offer more flexibility, remain in high demand for products with specific primary packaging requirements. "Pre-sterilized syringes offer standardized systems and are often easier to handle because of the tubs they are nested in," says Wrigley. "Bulk syringes on the other hand, offer a wider range of choices for product-specific processes or primary packaging material like closure components. Ultimately, the decision regarding the drug-delivery system will depend on the specific filling requirements of the product, market volume, and the business strategy of the pharma/biotech company when launching a product." Choosing the most suitable packaging should begin in the earliest clinical stages, adds Wrigley. "Once a drug manufacturer has decided on a pre-sterilized system, process adjustments or format changes might be limited because of the pre-set packaging and tub characteristics."

	Advantages of pre-sterilized syringes include lower investment costs and smaller-footprint equipment for filling and closing. Fewer processing steps and less handling help reduce damage. "Since syringes are even more prone to damage and much more difficult to transport than other glass containers, they were the first packaging type to be offered as pre-sterilized solution in nests," notes Klaus Ullherr, product manager at Bosch Packaging Technology.


	Pre-sterilized vials or cartridges are now also available, and, to date, are typically used in smaller batches, such as in formulation and development laboratories, according to Markus Hörsch, sales and marketing director at Bausch+Ströbel. The company's SFM 5110 filling system is designed for these applications and processes all pre-sterilized containers (vials, cartridges, or syringes) in nests inside tubs.

	Schott's adaptiQ system for ready-to-use pharmaceutical vials consists of a patented nest that securely holds up to 100 clean and sterile vials in an industry-standard tub. It is suitable for both small and large batches, and is advantageous for smaller batches because it minimizes setup time for packaging changes. Besides simplifying the process, the nested packaging protects the glass containers from scratches caused by vial-to-vial and vial-to-machine contact, which reduces the reject rate and also preserves the cosmetic quality of the vials, noted the company in a press release. Schott is advancing to full-scale production at its US facility, the company announced in a Jan. 12, 2015 

	Bosch recently launched a filling and closing machine for pre-sterilized nested syringes, vials, and cartridges. "The FXS Combi is the first machine of its kind with an integrated capping station for vials and cartridges," says Ullherr. "The capping of vials and cartridges takes place directly after the stopper insertion. This ensures the highest pharmaceutical safety, as the container is immediately and safely closed after filling."

	A challenge in setting up a line for pre-sterilized syringes is how to transfer the sterile syringes to the point of filling without compromising sterility. Syringes, nested in tubs, are sealed with a plastic layer and also protected by a bag. Combining the filling and closing machine with an upstream tub and bag opener eliminates contamination risk, notes Ullher.

	Although most syringes are glass, plastic syringes are also available pre-sterilized. Some small adjustments might need to be made to filling and closing machines to handle plastic syringes. "Depending on the plastic material, plastic can be more delicate and requires a gentler handling," notes Ullherr. "Plastic syringes might require a centering plate made of plastic instead of metal, and they can also have a slightly different diameter than glass syringes."


	For changing from one product to another in filling equipment, manufacturers can choose from clean-in-place/sterilize-in-place (CIP/SIP) systems and single-use, disposable systems. "The use of CIP/SIP is recommended for the production of large batches and toxic products, while contract manufacturers and smaller batches will work with single-use systems, as they can eliminate the loss of capacity resulting from lengthy cleaning validation as well as the risk of contamination between batch runs," comments Ullherr. In 2014, Bosch expanded its PreVAS prevalidated, pre-assembled, and pre-sterilized single-use filling systems, which the company developed in cooperation with Sartorius Stedim Biotech. For example, a new mobile filling station can be combined with existing filling lines. In this version, the product bag is situated outside of the isolator, and the filling tubes are guided through the isolator wall using a newly developed multi-tubing port, before being connected to the filling station, Bosch said in a 

Pre-sterilized, nested syringes and vials are seeing increased use in sterile filling.