DOE Gains Ground in Biopharmaceutical Development

Shanley,Agnes;

DOE Gains Ground in Biopharmaceutical Development

Published on: December 2, 2019

Agnes Shanley

Pharmaceutical Technology, Pharmaceutical Technology-12-02-2019, Volume 43, Issue 12

Pages: 26-27

As equipment evolves and use of PAT increases, DOE is becoming an integral part of upstream bioprocess development.

Editor’s Note: This article was published in Pharmaceutical Technology Europe’s December 2019 print issue.

loped in the 1920s, design of experiments (DoE) has become an indispensable tool for drug developers who seek to gain deeper insights into their products and processes. Instead of taking a univariate approach, changing one factor at a time and studying the impact on results, DoE uses statistical methods to perform multivariate analysis. This approach allows users to assess the impact of various factors on each other and on an outcome or product, allowing them all to be studied all at once.

Applied by Toyota engineers to understand the impact of variability on product quality, the method was first used for chemical development in the 1950s. Over the past few decades, its use in pharmaceutical development has grown dramatically, spurred by guidelines such as International Council for Harmonization (ICH) Q8 (R2), and FDA-driven initiatives such as process analytical technology (PAT) and pharmaceutical quality by design (QbD).

Although the mathematical underpinnings of DoE have been around for a hundred years, advances in computing and graphics, and software improvements made by vendors such as Umetrics, JMP, Stat-Ease, SaaS, and Camo have increased the power and user-friendliness of DoE software, according to spectroscopist and PAT expert Emil Ciurczak, head of DoraMaxx Consulting. Combined with the miniaturization of analytical instrumentation and the increasing sophistication of PAT, users can now run complex studies within labs or on the plant floor, and get results that would once have taken weeks, within days.

DOE at work

Design of experiments (DoE) can be used in applications ranging from pre-formulation and formulation to finished drug manufacturing, says Ciurczak. To give an example of how DoE can enrich understanding in preformulation, consider tests that might be designed to study the interaction between an API and a single excipient, and how they affect stability.

Using a univariate method, without DoE, one could take a mixture of active ingredient and excipient and put it through different stability studies (i.e., exposing it to heat, cold, light, and other variables)and hold it for 4–20 weeks and look for the breakdown point.This would provide information on interactions between the API and excipient, Ciurczak explains, but only one combination could be tested at a time. The time required for testing would be significant, after which high-performance liquid chromatography (HPLC) and thin layer chromatography (TLC) testing would still be required, and, overall tests would take several months to complete.

Using DoE, one could choose to consider the API, plus seven of the excipients that are most often used with that active, and test for eight different parameters, says Ciurczak. This would mean 8 times 7 mixtures, which would be put in vials in test tube racks. Near Infrared (NIR) or Raman spectroscopy could then be used to test these samples every day and to plot any changes. After that point, the vials would be tested with chromatography. Not only would this approach be much faster, it would be much more exact, and show clearly when each different combination of API and excipient reacted.

In drug synthesis, where chemists look for the highest yield, the best chemical for increasing yield may also be the least stable, DoE could be used to select the most stable alternative with the highest yield. In formulation, DoE would help determine the best formulation based on stability and release pattern, Ciurczak explains.

In pharmaceutical quality by design (QbD), DoE methods are used to define the design space for a product, correlating critical quality attributes and process parameters, allowing product quality to be optimized by staying within specific boundaries. For these tests, one would need to know the process parameters and to test a wide range of combinations of API with various excipients, disintegrants, lubricants, and other additives, and to use a full-sized batch for testing.

As Ciurczak explains, the Plackett Burman approach (1) would often be used for the first phase of testing, to develop a matrix. Optimum levels would be set for each parameter, with a deviation of plus or minus 10%. The Partial Least Squares regression method would then be used to show whether deviation is positive or negative.

A second stage of testing would then determine upper and lower limits to the design space, allowing formulators to tweak formulations and to ensure that variations in ingredients or raw materials (e.g., between different lots of the same excipient or excipients from different suppliers) did not have an impact on product quality. Not only could this approach allow testing to be accomplished in a few days, says Ciurczak, but results could be applied to optimization in other areas. For example, using DoE and stability data, product shelf life could be maximized to extend expiry dates and reduce costs, Ciurczak says.

Equipment advances drive DoE

In the small-molecule world, a major advance was made in the 1990s when near-Infrared (NIR) spectrometers were made small and versatile enough to be used in the lab or on the production floor. Pfizer and Carl Zeiss pioneered the first such PAT devices in the 1990s using a benchtop NIR device, a V-blender, and a rotator, and employing DoE to study solid dosage blends. Similar setups were later used by other companies for small-molecule solid dosage form testing.

Not only did this approach allow the first new drug applications incorporating PAT methods to be submitted to FDA, it represented a breakthrough by allowing development and quality testing to be done continuously for batch products using a multivariate approach to analyze data. Now that more companies are evaluating or using continuous processes to make solid dosage forms, applying DoE with PAT and QbD in continuous testing can eliminate the need for large and expensive pilot plants, offering considerable savings, says Ciurczak.

An analogous trend has been happening in upstream biopharmaceutical process development, as equipment has been developed to facilitate the use of PAT and DoE much earlier in the process. Within the past five years, high-throughput miniature bioreactors, which were previously used in pre-clinical screening, have been replacing bench-scale bioreactors in scale-down exercises. Scale-down is an important step in scaling up processes for current good manufacturing practices (cGMP) manufacturing in late-stage clinical development.

The devices eliminate the problems found with microtiter plates, spin tubes, and shake flasks, which can be difficult to automate (2), and allow more than 24 bioreactors to be run in parallel. They are also fitted with analytics that can take continuous readings on such crucial variables as pH, dissolved oxygen, and pressure for each of many samples.

Although developers have not yet filed new drug applications (NDAs) with FDA mentioning this technology, more companies including Genentech, Merck, Biogen, Lonza, and FujiFilm are working on this approach to scale-down (3). Developers are using devices such as Sartorius Stedim’s ambr 15 and 250 bioreactors (4). Starting in 2015, Sartorius Stedim Stedim Biotech has integrated Umetrics’ DOE software with its ambr 15 and 250 high throughput minibioreactors, later incorporated Nova Biomedical’s BioProfile FLEX2 automated cell culture analyzer, to simplify PAT and QbD work, and make scaleup and process optimization more efficient (5). The approach is also being evaluated for use in gene and cell therapy development.

If anything, these developments show the durability and importance of DoE as a way to make drug development (whether for small molecules or biopharmaceuticals) more systematic. Equipment and software evolution promises to make these techniques, as well as QbD approaches, easier to apply in the future.