By Allison Given Chunyk and Franklin Spriggs
Design of experiments (DoE) is an experimental approach that allows for efficient, simultaneous evaluation of multiple factors. The subsequent statistical analysis of the resulting data provides information on the relationship between the tested factors and various output parameters. The concept of DoE was first described by Sir Ronald Fisher in 1935 while working at an agricultural research facility in England. Since then, many industries, including the pharmaceutical industry, have utilized DoE to improve processes and reduce experimental error. While DoE has been utilized in certain pharmaceutical fields for some time, it has recently been adopted in process development and ligand binding assay (LBA) laboratories.
A main advantage of DoE is the ability to understand potential interactions between factors that would not be apparent in a classic one-factor-at-a-time (OFAT) experiment. Unfortunately, extensive adoption of DoE in LBA bioanalytical laboratories has been limited, primarily due to concerns of the complexities of DoE design, experimentation, and data analysis.
An increase in the utilization of DoE in the laboratory setting is timely, considering the fiscal realities that continue to encumber today’s research scientists. Whether it is ongoing LOEs (loss of exclusivity), reduced budget or head count, etc., we continue to work in a “more with less” environment. Innovation, however, is at an all-time high, and DoE can play a large role in this by advancing the quality of our work and diminishing the amount of time it takes to obtain that quality. Since extensive analysis using DoE can be more complicated than a traditional approach to assay/experimental design, usually, the largest hurdle for implementation is convincing scientists to introduce this methodology in their workflow.
But by combining elegant statistical programs with automated platforms, much of that complexity is removed from the process. With a small amount of training, DoE is easily understandable and relatively simple to apply. There are many opportunities to learn DoE including personalized, on-site training as well as webinars by the companies that offer statistical program packages. AAPS is also sponsoring the workshop Design of Experiments for Bioanlysis and Manufacturing, which will be taught by members of the pharma/biotech industry (including AIT Bioscience, Bristol Myers Squibb, EMD Millipore, Exponent, Pfizer, and SAS). The workshop will be taught using a problem-based learning (PBL) strategy to include DoE set-up with previously obtained data in both the LBA and manufacturing/process development settings as well as with hands-on experimentation. The workshop will be held June 7 in San Francisco, in conjunction with the AAPS National Biotechnology Conference.
This graph illustrates the improvement in LBA (ligand binding assay) performance when comparing multiple plate-based assays developed using OFAT vs. Gyros assays developed using DoE. By combining new technologies and approaches, we can improve robustness, lessen assay development and sample analysis timelines as well as reduce sample volume (Gyros technology)—all contributing to the more with less ideal. Robustness was measured as a function of reduction in TE (Total Error). TE is the sum of the %CV and the absolute value of the %Bias for high and low concentration quality controls (QC).