Process Validation - Small-Molecule Drug Substance & Drug Product Continuous Manufacturing Processes

Although concepts from the three-stage PV life cycle can be applied to the CM process, several aspects of PV for the CM process diverge from PV for the traditional batch manufacturing process. Both the article and the Discussion Paper aim to acknowledge differences between CM and traditional batch manufacturing and to review additional considerations for CM processes without creating new definitions. In some cases, the concepts covered in the Discussion Paper may apply universally; however, the intended focus is on small-molecule drug substances and drug products that leverage CM processes, whether for a new product development or the postlaunch conversion from batch manufacturing to CM.

This article summarizes the three sections of the Discussion Paper that are aligned with the three PV life-cycle approach stages: process design, process qualification (PQ), and ongoing process verification (OPV). Additional considerations needed during the development and validation of a CM process are emphasized. Readers are highly encouraged to refer to the entire Discussion Paper for more details as well as a fictional case study that provides additional context for the concepts.

Stage 1—Process Design

In stage 1 of the PV life cycle, the commercial manufacturing process is defined based on knowledge gained through development and scale-up/scale-out activities. The control strategy is defined and refined to ensure that the process is ready to progress to stage 2. One of the primary aims of stage 1 is to develop a control strategy to make sure that the output consistently meets the expectations described in the Quality Target Product Profile (QTPP).

At a high level, the main science- and risk-based steps completed in stage 1 are the same for both traditional batch and CM processes. However, the following are some of the considerations unique to the development of CM processes:

• Process disturbances: CM within a controlled and reproducible operation may have periods of process disturbance (e.g., raw material feed-rate fluctuations during process start-up). The potential for these disturbances should be considered and criteria developed to define and maintain the process in a state of control. When process performance is deemed to be outside of its state of control, a process pause and/or a decision to isolate material can be engaged. Additionally, it is important to recognize the impact that a disturbance in a unit operation could potentially have on downstream unit operations if they are integrated as one module.


• Material diversion: As noted in the previous point, a unique consideration with CM is the concept of material diversion during an identified process upset. A known process upset traveling through a continuous processing equipment set with a known residence time has a known dispersion and, subsequently, a known clearance time on return to normal conditions. Predictive models, including residence time distribution (RTD) models, and process analytical technology (PAT) applications are useful approaches for determining the amount of material that requires diversion.


• Process development using commercial-scale equipment: Stage 1 development entails experimentation to develop an understanding of the process dynamics required for integrated monitoring and control technology. Commercial equipment may be used to perform most stage 1 experimentation at scale, eliminating the need for scale-up activities to assess process risks at the commercial scale required for traditional batch manufacture. Although continuous processes may inherently reduce the need for equipment scale-up activities, it is important to note that because rate (mass per unit time) can become one of the scaling dimensions for a continuous process, there is a burden to characterize critical quality attribute (CQA) performance and processing risks as a function of time/throughput rate (e.g., material buildup on equipment surfaces). It may not always be possible to fully characterize a commercial time scale during stage 1; therefore, this may be carried into stage 2 as a residual risk.


• Spectroscopic tools: Spectroscopic tools, such as PAT, that require advanced chemometric and/or process models may be employed in stage 1, and process automation often becomes an integral part of the control strategy. Real-time release testing (RTRT) strategies integrate these concepts.

To achieve success in CM development, manufacturers will need a cross-functional, collaborative team with a high level of understanding of the concepts listed here.

In general, equipment and utilities qualification and manufacturing facility design involve the same steps for both batch manufacturing and CM. However, the complexity of the qualification activity may be greater for CM because of the integrated nature of the unit operations.

Stage 2—Process Qualification

Before commercially distributing a drug product, the manufacturer is obliged to successfully complete PQ. During the PQ stage of PV, the process design is evaluated to determine whether it is capable of reproducible commercial manufacture. In general, equipment and utilities qualification and manufacturing facility design involve the same steps for both batch and CM. However, the complexity of the qualification activity may be greater for CM because of the integrated nature of the unit operations. The complexity also increases when PAT and automated control and data management systems are utilized. Similarly, although the high-level science- and risk-based steps completed in process performance qualification (PPQ) are the same for both batch and CM processes, there are a few unique PPQ considerations for CM.

CM-specific issues for stage 2 include the following:

• Justification for the quantity of validation batches: When justifying the number of validation batches, the desire for flexibility in batch size and/or mass throughput should be considered. A risk-based approach should be taken when determining whether the intended worst-case run time (defined by time, number of units, mass of product, etc.) and/or selected throughput(s) should be included in the PQ exercise. To do this, consideration would be given to product understanding, process knowledge, and detectability of unexpected performance in relation to the specific risks associated with increases in run time (e.g., microbial growth, material buildup/equipment performance, cleanability). In many cases, a PV with fewer than three batches may be supported, although global acceptance of such an approach remains a topic of discussion. Increases to batch size require thorough risk management; however, with appropriate understanding of the system dynamics, a single batch may be suitable to supplement the initial PPQ.


• Process performance metrics: It is good practice to measure process performance metrics (yield, mass balance, etc.) in stage 2 for process performance monitoring and continuous improvement, rather than treating those metrics as PV acceptance criteria. The robustness measurement techniques explored in stage 1—such as mass balance, yield, and the percentage of time the process remains in a state of control from the planned product collection of the continuous processing batch—would be appropriate process performance metrics.


• Volume of data: Compared with most traditional batch manufacturing operations, CM processes will likely generate more in-process data and, perhaps, more finished-product data. Furthermore, the structure of the data collected (e.g., systematic or nested samples associated with solid dosage uniformity testing) may require more advanced statistical modeling to accurately describe process performance and capability. It is important to clearly dene the scope of data so it is directly tied to the manufacturing control strategy and/or product disposition. The stage 2 statistical analyses and related sampling plans (i.e., defined sampling points and amounts used to support stage 2 statistical analyses) will focus on the CQAs, critical process parameters, in-process controls (IPCs), and other variables relevant to assessing product quality and process control that were identified in the criticality analysis in stage 1.

The data-analysis objectives in stage 2 are the same for CM and batch processes: to evaluate intrabatch variability and capability, and to provide initial assessments of reproducibility and consistency between batches. Based on these assessments, decisions are made regarding 1) the readiness of the process to proceed to stage 3; 2) whether, and to what degree, enhanced sampling (i.e., more frequent sampling and/or greater-than-routine amounts) is needed in stage 3; and 3) if enhanced sampling is required, how many batches will be sampled before another evaluation. Considering the special considerations for PAT and similar tools, and the related implications for statistical methodologies for continuous processes discussed previously, the stage 2 performance evaluation for a CM process is not substantially different than that for a batch process.

Compared with most traditional batch manufacturing operations, CM processes will likely generate more in-process data and, perhaps, more finished-product data.

Stage 3—Ongoing Process Verification

The goals and expectations for OPV are the same for batch and CMf processes: to provide ongoing assurance that the process remains in a state of control by monitoring it through a periodic trending program. This program will help manufacturers understand routine variability, detect unusual variability (i.e., special cause), and enable process improvement to maintain a state of control.

As noted in the previous section on PQ, one of the most distinctive aspects of the CM process relates to the volume of data that may be collected. Given the large amount of data that may be routinely captured for a CM process during OPV, it is important that the trend analysis focus on the critical parameters that are predictive of product quality (e.g., control strategy parameter criticality, IPCs). The process measurements to be statistically trended in the OPV program should be selected using quality risk-management tools and specified in an OPV plan. Careful assessment of statistical assumption violations is needed. The OPV plan should also address the statistical trending tools to be used, the frequency of data review, the duration of data collection, the roles and responsibilities of team members, and which events will trigger actions as well as the actions that must be taken.

An ideal OPV program will capitalize on the additional data generated and allow for a periodic refinement of the control strategy. This type of program provides an opportunity to determine

the root cause of special cause events and further increase the knowledge base of the product, with the optimal goal of continually improving the process.

Other considerations when designing an OPV program for a CM process include model maintenance, which is an additional requirement often associated with CM processes that use technology such as PAT. Process models (such as an RTD), PAT models, and other models supporting RTRT require a model maintenance plan based on their respective risks and roles within the control strategy.

Conclusion

PV in the context of CM is fundamentally the same concept used in the context of batch processes. A life-cycle approach consisting of three stages (process design, PQ, and OPV) forms the basis of the approach. However, multiple unit operations linking in CM processes may lead to an increased volume of data and may require integration of process models and tools such as PAT for feedback/feed-forward controls, as well as the implementation of RTRT strategies. Therefore, opportunities to refine PV for CM processes will require further consideration as the industry continues implementation. The authors of the ISPE Discussion Paper are interested in receiving feedback on the CM topics presented in the full document, including lessons learned through regulatory agency feedback during review and inspection. Please email Wade Jonathan or Robert Levers.

Acknowledgments

This article is a summary from the Discussion Paper with the same title published in January 2019. Contributing authors to the full Discussion Paper posted on the ISPE website include Matthew McMenamin, Manager, Systems and Technology, Glaxo Smith Kline; Carly Evans, Senior Director Regulatory, Akebia Therapeutics; Declan Hurley, Consultant Scientist, Eli Lilly & Co.; and Katherine Giacoletti, Partner, SynoloStats LLC.