Data is prevalent through all functions of the Life Science industry. Its value impacts many disciplines throughout an organization, and while the use of data by discipline varies, it always involves:

• Gathering and aggregating multiple types of data from multiple sources
• Making that data available across an organization
• Subjecting it to statistical analysis
• Extracting knowledge from it
• Using those insights to correct problems and capitalize on opportunities

An area of life science that has recently received significant attention around data, its integrity and management processes, is – manufacturing. One key contributor to this focus is the 2011 publication of the FDA’s ‘Process Validation: General Principles and Practices’.

Within this guideline, the third and final stage of process validation – ‘Continued Process Verification (CPV)’ – states that manufacturers must provide continual assurance that their manufacturing process remains in control over time. Each manufacturer must put plans and processes in place to collect and analyze end-to-end production and process data to ensure product outputs are within predetermined quality limits. The plan must also ensure the process remains in statistical process control (SPC) – noting that while there may not be “failures” of an individual batch, a process may start to drift out of control, leading to the likelihood of future batch failures. The goal of SPC, and its incorporation into CPV, is to prevent those failures from occurring.

Although the FDA does not specify how an organization should execute CPV, Stage 3 of the Guidance suggests, “A system for detecting departures from the process as designed is essential to accomplish this goal.” While it is possible to meet the requirements of CPV through manual efforts, or use of common Office applications, these approaches are likely sources of errors, can lead to long lead times for report availability, and are labor intensive.

Life Science manufacturers and CMOs who secure and streamline their data management and process control strategy for CPV by leveraging the latest data management systems are on the forefront of ensuring compliance, timeliness and operational efficiency throughout their manufacturing operations.

The current environment of manufacturing data management systems can be complex. Data is collected by different groups, sites and partner organizations. Additionally, data is managed in multiple places – spreadsheets, paper-based systems, and disparate data silos.

This paper will help you streamline your search for the right system. It provides key insights on how purpose-built data management systems can be incorporated into your CPV program to meet FDA requirements and protect and improve your business. The following content refers to many of the required elements discussed in BPOG’s ‘Continued Process Verification Paper with Example Plan’.

Starting with the basics, a successful CPV program begins with a formal CPV plan. The components of a plan include:

• Process Definition
• Data Collection
• Limits Management
• Process Trending & Analysis

Process Definition

A key component of all CPV plans is the monitoring plan, defining the data elements – “parameters” – that will be measured and reported for that product. These parameters – Critical Process Parameters (CPPs), Critical Quality Attributes (CQAs), Critical Material Attributes (CMAs), etc. – are typically determined during Stages 1 and 2 of the overall process validation efforts.

With the transition to commercial manufacturing covered by Stage 3 of the process validation guideline, a CPV plan must be developed that includes an overall process description and flow chart of the highest-level process steps, or unit operations. The parameters noted above are gathered and documented for each step within the process definition.

CPV software solutions must allow process experts to easily create this process definition, and to represent the unit operations along with nested phases and sub-phases. Once this process definition tree has been created, the solution must allow the intuitive insertion of all the critical and key process parameters and material/quality attributes (“parameters”) in their appropriate place. The process definition should also allow for the capture of additional information for each process step such as flow charts, information supporting process changes, links to relevant internal or external documents that would be useful context for tech transfer or on-boarding new employees.

For each parameter, the following metadata should be easily captured:

• Expected number of results per batch (one, multiple time independent values or time profile data)
• Data type, e.g., text or numeric
• Units of measure
• Parameter class, e.g., KPP, CMA
• Additional contextual information, e.g., risk assessment details

Key Solution Components

Each CPV plan will be different, based on the product being manufactured, the manufacturing process, prior variability, etc. However, at a minimum, all CPV initiatives will have the following requirements:

• Batch Data Collection – gaining access to all of the results for the defined parameters (CPPs, CQAs, etc.) on a batch-by-batch basis
• Limit Management – providing data storage and management for specification and target control limits with valid date ranges, metadata, etc. for each defined parameter in the monitoring/CPV plan
• Process Trending and Analysis – CPV guidelines describe a minimum of control charts and process capability for compliance with the guideline’s expectation of statistical process control reporting

Batch Data Collection

As with any solution that reports on manufacturing, data integrity and consistency with 21 CFR Part 11 / Annex 11 are mandatory. This requires that the data be entered into a secure, validatable system that includes electronic signatures and has a complete audit trail. There must also be capabilities built into the solution that ensure other aspects of data integrity. At a minimum, these capabilities should include independent data approval workflows and techniques to reduce data entry errors such as numeric precision requirements, and data entry ranges that would alert the data enterer and data approver that a data entry error may exist.

The data entry interface should be automatically configured based on the information provided during the process definition phase. In addition to the parameter characteristics mentioned earlier, it should be possible to provide additional details, e.g., expected decimal places and expected value ranges for numeric parameters, and dropdown lists for string parameters. With this small amount of additional information, the ideal solution would require no development or configuration tasks between process definition and beginning data entry.

Limit Management

Some specification limits are known very early on in the life cycle of a product, e.g., for drug substance CQAs or in-process controls. Others, such as for CPPs, may be derived from small-scale process characterization studies but later adjusted during scale-up, PPQ and/or follow-up experiments.

Statistical target control limits require a sufficient number of batches in order to accurately represent the expected long-term variability of the process. The exact strategy for establishing initial and long-term control limits varies from company to company and is commonly documented in an SOP. These limits are reassessed, and adjusted as necessary, on a periodic basis or in response to known process changes.

For effective CPV analysis and reporting, the multiple versions of both specification and target control limits must be easily available for use in process trending and analysis. Management of date ranges – effectivity dates, obsolete dates – is required so the limits can be matched to each manufacturing batch, based on manufacturing date. A full solution to limit management must also consider that specification limits can vary by sub-product, e.g., different formulations / strengths, or by geographic market that a batch will be delivered to. The target control limits might vary by equipment ID and/or some other category; a Limit Management system must be flexible enough to manage the possible variations.

Finally, for every limit version, there should be a way to capture the rationale behind the limit values, whether it is through simple text comments, internal or external reference URLs, attachments or notes, so that the institutional, cumulative knowledge behind the changes is not lost and can be easily referenced during periodic reviews of process performance, investigations, or audits.

Process Trending & Analysis

The ultimate goal of any CPV program is to use statistical process control to monitor processes in a timely manner. These techniques verify that the process remains in its validated state by identifying any process shifts/trends early enough to take corrective actions.

As noted above, CPV requires control charts and process capability. While there are many software products – both at the desktop level and the enterprise level – that can create these types of charts, most of these still have challenges for the support of CPV. The best solution for CPV trending and analysis has many of the following characteristics:

• is easily validated
• has ready access to both batch data and limit data
• supports segmented control charts to accommodate changes in limits over the date range being charted

Without a purpose-built, off-the-shelf solution, achieving the desired goal typically involves developing custom software or significant manual efforts, each of which significantly increases expense and makes validation much more difficult.

Summary

The first step for CPV is proper planning – a key expected deliverable from each organization’s CPV efforts is the plan itself. Beyond that plan, each organization will need a specific plan on the optimal approach to create, review and act on the actual outputs of the CPV plan, i.e., the analytics represented in the control charts and process capability.

Adopting a solution that allows you to achieve this goal and maintain product quality and compliance across the product life cycle will significantly simplify what is historically a time-consuming and costly process.

The right product data management system collects and transforms the specific data required for CPV into the requisite outputs with the least amount of resources – people, time, money.

To get the most value from a system, it is important to first understand the components of CPV planning and key features and functionality a system needs to deliver. These include:

• 21 CFR Part 11 compliant
• Easily validatable
• Full audit trail
• Multi-user, role-based access
• Approval flows and e-signatures
• Scalable