Using the Multiple Attribute Method for Process Development and Quality Control

Published on: 
Pharmaceutical Technology, Pharmaceutical Technology-05-02-2019, Volume 43, Issue 5
Pages: 58, 60–61

More manufacturers are embracing MAM, which simplifies biopharmaceutical product quality testing, and facilitates the measurement and monitoring of critical quality attributes.

Complex biopharmaceuticals such as monoclonal antibodies can be extremely sensitive, and subject to alterations due to post-translational modifications (PTM) and degradation of the product during storage. Conventionally, a large battery of tests is required to detect impurities, but each method can only detect a limited set of problems. Several years ago, a team at Amgen began to work on a method that would simplify quality control (QC) by reducing the number of tests required, and that would allow developers to assess critical quality attributes (CQA) throughout product development.

Since initial work with Thermo Fisher, Amgen is now using this method, called multiple attribute method (MAM), in QC and process development. Biopharmaceutical manufacturers of all sizes, and a growing number of contract development and manufacturing organizations (CDMOs), are also evaluating or working with the method. Analytical equipment, software, and reagent vendors are currently offering products tailored for use with MAM. 

Richard Rogers, who worked on the Amgen team that pioneered MAM and is now principal scientist at Just Biotherapeutics, discussed MAM’s history with Pharmaceutical Technology.

Taking MS into the QC lab

PharmTech: Why was the MAM approach developed?

Rogers: Scientists at Amgen had been working on ways to improve QC testing before I joined the company in 2012. Originally, the project was called “Mass Spectroscopy (MS) in QC,” because its goal was to replace some of the current QC release methods with a mass spec-based method.

After I joined the group, we developed a proof of principle, since we could demonstrate a correlation between MS data and conventional product quality assays. Within a short time, we were able to expand the method and our goals for it. It was no longer just about using MS in QC, but seeing how many attributes we could target with the method.

PharmTech: How did you get high-level support for the project?

Rogers: Senior management at Amgen had been very supportive of this approach to quality control. Support from the top level of the organization helped win over QC staffers who, initially, were very hesitant about using the method.

PharmTech: How does peptide mapping figure into MAM?

Rogers: Peptide mapping has been used for decades. The approach involves digesting a test sample of a biopharmaceutical, and then using an algorithm to search the data to identify peptides and post-translational modifications that are present on the product. After a molecule has been characterized, attributes are then monitored by relative quantitation, which compares the modified peptides with parent peptides.

What differentiates MAM from conventional peptide mapping is the purity component. Conventional release assays such as capillary electrophoresis and cation exchange chromatography are considered purity assays. In order to replace these traditional QC release methods, MAM had to have a comparable purity component associated with it.

To address this issue, a new peak detection feature was incorporated into the method, enabling a way to perform differential analysis between a reference standard injection and each test sample. Instead of leveraging a person to interpret an electropherogram (a chart generated when electrophoresis is used) or an ultra-violet (UV) chromatogram, we use software to go through and identify differences between the test sample and reference standard. The mass spectrometer acquires full scan data. MAM does not need every peak to be completely resolved. We are able to leverage software to see everything underneath a single peak. A single peak in the total ion chromatogram could consist of three to five species.

PharmTech: Did you develop the required software inhouse?

Rogers: Initially, we worked exclusively with Thermo Fisher Scientific. We acquired data using an Exactive Plus MS and then used Pinpoint for PTM quantification and Sieve for new peak detection. Those two pieces of software still exist but aren’t being supported by Thermo anymore. Other analytical software vendors, including Protein Metrics, Genedata, Sciex, Agilent, Bruker, and Waters, are also incorporating MAM analysis into their software platforms.

Currently, Thermo has integrated PTM quantification and new peak-detection capabilities into its Biopharma Finder software. The software is not yet 21 Code of Federal Regulations Part 11-compliant, but its attribute monitoring method can be incorporated into Thermo’s compliant software, Chromeleon. So, if you are working on Thermo’s platform  and transferring an MAM release method, you would develop the method in Biopharma Finder and then transfer it to Chromeleon for use in QC.

Gaining buyin for the method

PharmTech: Many new technology advocates at pharma companies find management resistant to change. This used to be true for process analytical technology (PAT). What was different about Amgen’s managers?

Rogers: Amgen has always had a large technology development effort. The company’s senior managers wanted us to test new methods for QC, and it all came down to us being able to demonstrate proof of principle. What helped was the fact that many senior managers at the company had strong science backgrounds, for example, vice-president Jim Thomas, could critically evaluate data and see the benefit of doing these types of assays in QC. 


PharmTech: How did you get the QC staff to become more comfortable using this approach? 

Rogers: The Exactive series of MS tools is user friendly, with straightforward calibration and overall robustness. As a result, equipment uptime using these tools is as good or better than that for a conventional ultra high-pressure liquid chromatography (UHPLC) assay.

Fortunately, the QC department had already run UV-based peptide maps before, and had also quantified attributes based on the UV trace. As a result, they were confident doing trypsin digestions and transferring in a reverse-phase gradient to separate peptides. All we had to do was to make analysis and reporting easier for them.

In Chromeleon, we created a report template and transferred it to them directly. They would prepare a sample, then run system suitability tests on it and make sure they passed all assay acceptance criteria. They would then run the test sample along with the reference standard, and could perform attribute analytics and new peak detection on the fly without having to set up these methods themselves.


Cost and time savings

PharmTech: What savings can be achieved with MAM?  

Rogers: To get some idea of the potential savings, consider a molecule in which glycans are important, and what is involved for MAM as compared with traditional methods. Imagine that there are four relatively standard QC assays required to release that molecule: a release glycan map, identity test, charge variant test, and reduced capillary electrophoresis. Those are four separate assays that will have to be qualified and validated in order to take the molecule into QC. If a new peak shows up in any one of the purity assays, that sample would need to be fractionated to isolate the new species. 

With MAM, you are running a digested peptide map anyway. The method’s real value stems from the fact that it uses one digested sample to cover all those assays and in a very analogous way, to get site-specific information for where the protein is being modified or where it is degrading. MAM data saves time and effort because it directly provides the new peak’s identity.

PharmTech: What types of advances in model and software development were required to make MAM possible initially, and what improvements have been made since then?

Rogers: In the beginning of our development efforts, it was surprising that we could prove proof-of-concept using Pinpoint and Sieve software. These two pieces of software were originally developed for applications that are quite different from what we were trying to do. Pinpoint, for example, was developed for multiple-reaction-monitoring (MRM)–based applications, while Sieve was designed for proteomics-based differential analysis. We were able to leverage both of those and show that this method could be successful.

Thermo’s incorporation of both attribute analytics and new peak detection into Biopharma Finder [and connecting it to Chromelon] has been a great step forward. Incorporating these capabilities into a 21 CFR-compliant framework has allowed us to put this into a QC environment. 

Users’ consortium plans

PharmTech: What are your plans for the MAM users consortium? 

Rogers: We want vendors to keep innovating and improving, and we would like to work with them on developing MAM solutions, because that will make the method better, in the end. The consortium’s main goal, however, is to enable the biopharma community to leverage MAM for product characterization as well as product release from QC. We are trying to improve how we do mass spec as a biopharma community, to teach people how to acquire better data and how to analyze those data more effectively using this approach.

We can turn around data much quicker by leveraging this method, and can also discover potential critical quality attributes that we didn’t even know were there by leveraging new peak detection. 

The Consortium is focusing on the free exchange of information. We currently have more than 220 individual members with more than 70 company members, ranging from FDA and the National Institute for Standards and Technology (NIST) from the government side, to biopharma companies, software vendors, instrument vendors, reagent vendors, and CDMOs.

Standardizing methods

PharmTech: What innovative approaches have you used recently?

Rogers: In 2018, we did a Round Robin to see how well members were able to apply standardized MAM methods. Participants filled out a spreadsheet template and evaluated a reference standard, then a second version of that standard that had been spiked with peptides, then a pH stress sample, and then an unknown sample. The idea was to go through all the samples and look for new peaks and compare them to the reference.

About half of the groups were able to pass the criteria. We see this test as the first step in teaching the industry how to apply this method in the process development lab.

PharmTech: Are any companies officially using this method for QC and development, and have you published best practices on method transfer for CDMOs?

Rogers: We haven’t published anything yet on method transfer, although members are working in that area. So far, a couple of members have put MAM into their QC operations. Amgen is the farthest along. FDA representatives have already visited Amgen’s QC lab to see its MAM setup and to learn more about how it works.

Article Details

Pharmaceutical Technology
Volume 43, Number 5
May 2019
Pages: 58, 60–61


When referring to this article, please cite it as A. Shanley, "Using the Multiple Attribute Method for Process Development and Quality Control," Pharmaceutical Technology 43 (5) 2019.