What Today’s Biopharma Leaders Need to Know about Data Analytics

Data analytics is more than just a tool to help pharma and biopharma companies set and reach KPIs. It’s also a fundamental part of a digital transformation strategy that leads to more effective products, faster time to market, and larger profit margins. 

^{This article is posted on our Science Snippets Blog} 

Achieving success in the competitive and highly regulated pharmaceutical industry demands that today’s company leaders embrace the latest approaches to data analytics. But getting there may require addressing and overcoming various internal challenges, not the least of which may be data silos or a lack of data analytics insights across departments and processes.

In many pharma and biopharma organizations, a major stumbling block to digital transformation is the lack of connection or insight into the data created by different departments or even different processes within the same departments – in other words – data silos.  Breaking down data silos and setting up infrastructure to gain insights from data at every level – from R&D to production to clinical trials – requires a willingness to adopt the right sort of technology. 

Data analytics for manufacturing alone has the potential to be game-changing.  Consider the supply chain with multiple production lines, devices, and processes in which any failure or mistake can be costly, or even lead to a product being recalled. 

Data analytics can help prevent that scenario, allowing your teams to react quickly if any deviations in the production process happen – before it affects the downstream process.  But even more, data analytics can help you create optimal processes, develop robust formulas and select the right molecules from the start. 

Advanced data analytics software for multivariate data analysis (MVDA), Design of Experiments (DOE), and real-time analysis support:

• Robust formula creation
• Equipment health estimation
• Predictive/preventive maintenance
• Predictive analytics 
• Quality by Design (QbD)
• Real-time process monitoring
• Process analytical technologies (PAT)

Data Analytics for Manufacturing Applications

Digitalizing and embracing a robust approach to data analytics requires an understanding of how the data can and should be used, and what sort of applications or methodologies provide the most value.  

How is it done? You start with multivariate data analysis software that combines reliable statistical methods, processes, and tools to help you create dependable processes, look for deviations and understand the relationship between different critical process parameters, raw material attributes, and critical quality attributes.

You can then take it a step further and use predictive analytics to get information about an impending problem before it reaches a critical stage.  With real-time and predictive analytics, your production teams can make adjustments to processes with minimal downtime or know in advance what the lifecycle of equipment or instruments might be in order to prevent interruptions.

Combine that with Design of Experiments (DOE) software that helps optimize process development and employ a Quality by Design (QbD) approach favored by regulators, and your production processes have jumped one of the biggest hurdles manufacturers face in moving toward digital transformation and pharma 4.0: aligning process information with supply chain and using metrics to gain an end-to-end picture of your product lifecycle and enable company-wide forecasting.

The right data analytics tools and instruments are also essential for the adoption of Process Analytical Technology (PAT), which is playing an increasingly important role in the pharmaceutical and biopharmaceutical industries due to regulatory pressure and the push to reduce costs while managing quality. 

What is Multivariate Data Analysis (MVDA)?

Multivariate data analysis (MVDA) is a statistical technique used to analyze data that is generated from more than one variable (or source). MVDA allows you to create summary indexes that represent the entire set, in order to better understand the whole.  One good example of this is a stock index. For example, the Dow Jones index is a summary of the stocks of the individual companies weighted together using a specific algorithmic function (shown below).

The Dow Jones index is a summary of the stocks of the individual companies weighted together using a specific algorithmic function.

Data analytics used for process modeling in manufacturing is conceptually very similar. As with the stock index, you want to calculate the summary indexes for how the process is performing. This summary can apply to a continuous process, a batch process, or a hybrid of the two, or any other kind of data table.

What a process summary index has in common with the Dow Jones index is that it presents a trend over time. You monitor it to see how it goes up or down. If you are a process engineer or manager, you may take various actions or refrain from actions based on how this summary index looks over time. 

Why Use Multivariate Data Analysis

The reasons to use multivariate analysis include:

• Multivariate data analysis helps simplify data without losing any important details.
• MVDA makes it possible to group and sort multiple variables based on their unique features.
• Multivariate data analysis helps you understand which variables are dependent or independent (which have an impact on others).
• MVDA helps you understand the relationship between all the variables and predict the behavior of the variables based on observations.
• MVDA provides a valid way to create and test statistical hypotheses to determine whether or not assumptions are true.

Multivariate vs. Univariate Data Analysis

Historically, many pharmaceutical manufacturers have taken a univariate approach to evaluate and even develop their R&D and production processes. That means taking measurements or monitoring one single process variable at a time to ensure it meets the correct specifications or stays within the acceptable limits. But that’s no longer the most effective or valid approach.

Taking process development to the next level requires monitoring all of the individual process variables at once, both to create a big picture of the process status, but also to know when one step or parameter might be about to impact another.   

In its simplest form, the difference between univariate, bivariate, and multivariate data analysis is:

• Univariate- one variable at a time is analyzed. The objective is to describe the variable. For example How many cells in the bioreactor are viable?
• Bivariate- Two variables are analyzed together for any possible association or empirical relationship. For example: What is the correlation between “viable” cells and “temperature”
• Multivariate- More than two variables are analyzed together for any possible association or interactions. For example: What is the correlation between “viable” cells, temperature, pH, oxygen saturation, growth time, and other variables? 

In pharmaceutical manufacturing processes, or indeed all manufacturing processes, it is vital to understand the relationship between parameters, which means using more advanced types of data analysis, including regression analysis or PCA:

• Principal component analysis (PCA) for data summary/overview 
• Partial least squares (PLS) and orthogonal PLS (OPLS) for regression analysis

Multivariate models summarize multiple sets of variables into one easy-to-understand chart.

What is the Design of Experiments (DOE)?

For pharmaceutical and biopharma companies, building quality into your products from an early stage is a key factor in regulatory approval and market success. Design of Experiments (DOE) is an essential tool for achieving both regulatory compliance and faster time to market.

In a highly competitive market, being able to shave months or even years off the research and development process while still delivering quality and stable products can be the key to achieving commercial success.

DOE helps scientists and process specialists confidently create robust and reproducible processes that shorten the time to market for (bio)pharma products.

DOE helps support R&D and production in many critical ways including:

• Understand a process.  Investigate many process parameters and their cause-and-effect relationships. For example, evaluate media composition against total cell density (TCD) and viable cell density (VCD).
• Perform process optimization. Select the right factors and ranges of process parameters to find process optimums, such as pH range and temperature ranges that optimize a certain key performance indicator (KPI), such as titer.
• Build a design space. Define regions where product quality is assured.
• Validate and characterize process stability and robustness. Identify critical process parameters (CPPs) that may be sensitive to small factor changes or which need to be controlled to achieve robustness.
• Get to market faster. Reduce the number of experiments you need to perform by increasing your knowledge about your process.
• Save costs. Identify factors that may reduce material consumption and waiting time.
• Fulfill regulatory compliance. Create reproducible and documented processes.
• Follow the Quality by Design (QbD) approach. Build quality into your process as outlined by ICH Q8, Q9, Q10.

DOE defines the optimal design space, which can be viewed as a set of points that sit within a 3-D space (or cube) on a plot. The larger the optimal design space for a product is, the easier it is for production to stay compliant.

Read more: Why DOE is essential in the Biopharma industry

Quality by Design (QbD) and Process Optimization

Adopting a Quality by Design (QbD) approach to process development makes good business sense. In addition to creating more robust formulas and well-documented processes, following a QbD approach can help ease regulatory compliance and ensure a stable process for long-term production optimization. Since its adoption over 10 years ago, QbD is increasingly viewed as the best-practice approach for process development among regulators in the biotech industry. [1]

QbD differs from past approaches to product development by building quality into every step of the process, rather than relying on testing to achieve quality. In essence, QbD is a statistical approach to development that focuses on process understanding and control by assessing variables that may impact quality. That means a QbD approach hinges on being able to analyze a broad set of data effectively. 

A QbD-based approach provides a high degree of assurance that a pharmaceutical manufacturing process is adjustable within a design space and therefore robust, and that it is managed with a control strategy that uses modern statistical process control methods. It enables a lifecycle approach to validation and continuous process verification. 

DOE supports a quality by design (QbD) approach to product development that is favored by regulatory agencies. By building experiment certainty and reproducibility into your process, you can be confident in the robustness of your formulation and the quality of your final product.

Real-Time Process Monitoring and Control

If your process models are statistically accurate enough, you will be able to determine when a production process is deviating from any normal operating condition, and perhaps even predict when a current process might start deviating from accepted conditions.

Using advanced data analytics models in real-time opens up a whole new world of possibilities for improving your production processes. Not only does real-time process monitoring provide a level of confidence in your process performance, but it can also help improve the overall quality of your production output.

That’s why a multivariate data analysis tool for real-time process monitoring is so important. A real-time analysis tool allows your production floor team to know when a process is performing optimally or to see immediately when a deviation occurs. This early warning allows them to take the necessary steps right away to correct any issues that might cause a batch to be rendered unusable or to stop contamination of a downstream process.

A real-time data analysis solution utilizes regression models to summarize all of the individual data points from various operations into multivariate models that can be monitored in real-time. This becomes very efficient in the control room because instead of looking at a large number of individual parameters or signals, you have a small set of summary parameters that let you monitor all the variables at the same time.
 

Real-time process monitoring software (like SIMCA^®-online) provides warnings and alerts when processes deviate from optimal settings.

Continuous Process Manufacturing Introduces Control

Continuous manufacturing is one of the key trends within the pharmaceutical industry, both for the production of ‘classical’ drugs as well as large molecules. Using a statically sound and reproducible method of process development allows companies to shift from traditional batch processing to continuous process manufacturing. 

Regulatory bodies such as the FDA see the promise of these processes and actively encourage companies to implement this manufacturing strategy. This is illustrated by the 2019 published draft guidance from the FDA for the production of small molecule drugs using continuous processes.

The FDA considers multivariate models to be surrogates for traditional release tests, for example, a stand-in for dissolution. [3] Real-time release testing provides for increased assurance of quality as well as increased manufacturing flexibility and efficiency. This means:

• Shorter cycle time
• Reduced inventory
• Reduction in end-product testing
• Reduction in manufacturing cost

It also allows manufacturers to leverage enhanced process understanding to make corrective actions in real-time.

The FDA encourages manufacturers to adopt continuous manufacturing using advanced multivariate data models, because it improves process control, reduces variability and improves product quality and consistency.

PAT Adoption Is Increasing

PAT and other types of advanced models for process control are pivotal to support the industry trend towards continuous manufacturing. PAT also reduces regulatory compliance concerns,  as the sort of documentation, data management and audit trails that PAT provides is what regulatory bodies are increasingly looking for to show compliance.

The use of Process Analytical Technologies (PAT) has become more and more important in the pharma industry. Some of the top reasons include:

• PAT and other quality programs are expected by the FDA and other regulatory agencies
• The increasing push towards continuous process manufacturing
• PAT is accepted as providing for more robust bioprocessing and overall being a cost-effective product, and minimizing quality defects 
• Continued improvement in sensors, probes, software, and analytical equipment

Strong business cases can be made for wide adoption to maximize yields, obtain consistent high-quality, including its usefulness in reducing errors, eliminating batch failures, and diminishing loss from processes gone astray. 

The continued improvement and reduced costs of sensor probes over the last five years or so is another reason for increasing levels of PAT adoption. The availability of analytical software associated with these technologies has also become more widespread, both for upstream and downstream bioprocessing, and covers a number of different kinds of therapeutics, as well.

Keeping Your Processes in Control

Whether your process deviation is caused by human error, contamination, equipment or sensor malfunction, or data that is out of range, knowing what to do to correct the problem, and whether your batch can be saved or process corrected in time to save the final product, depends on how quickly you detect it and whether you have the statistical data to back up your decisions. Having the right statistical tools for creating the optimal design space, establishing effective control parameters, and continuously validating your process is important.

Data analytics software embedded into the R&D and manufacturing process makes digital transformation achievable and yields products that are profitable and able to endure regulatory scrutiny. 

View a Case Example

Want to know more about how DOE software can help scientists set statically reliable robustness studies that meet International Conference on Harmonization Q8 (ICHQ8) Quality by Design standards?

Download the Growth Story featuring a Hoffmann-Roche case study.

Get the Case Story
 

References

1. Horst, J.P., Turimella, S.L., Metsers, F. et al. Implementation of Quality by Design (QbD) Principles in Regulatory Dossiers of Medicinal Products in the European Union (EU) Between 2014 and 2019. Ther Innov Regul Sci 55, 583–590 (2021). https://doi.org/10.1007/s43441-020-00254-9
2.  ICH Q8 (R2). Pharmaceutical development, ICH Harmonized Tripartite Guidelines. International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use. 2009.
3. Quality Considerations for Continuous  Manufacturing , Guidance for Industry , DRAFT GUIDANCE , U.S. Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research (CDER), February 2019 https://www.fda.gov/media/121314/download