• R&D explores boundaries of processes for "understanding". R&D focus is different. Use PATs for "process understanding", versus PATs for "process control" in manufacturing. Provides opportunity for "process control".
• Not all PATs will make it to manufacturing. Identify the PATs that should be used in manufacturing.
• Demonstrate suitability of PAT measurement for intended use. e.g. For predicting end product quality attributes.
• Demonstrate it is validatable, e.g. Sensor suitability, location of sensors, number of sensors, as well as traditional measurement attributes.
• PAT performance requirements are more rigorous if the intended use of PATs, either individually or combined, is to replace end product testing.

PAT R&D efforts in manufacturing:

• Points stated above in R&D section will apply with the following additions:
• Transferability of PAT should be investigated. Equipment design, scale up issues, interface change, ongoing calibration, maintenance, equipment qualification, safety (e.g. operator, potential contamination of product).
• Refinement of models will be needed, e.g. "process signature" developed in R&D, based on limited information.
• Concept of PAT can be submitted as a "protocol" in an original NDA, or as a prior approval supplement. Implementation of PAT can be done through a less burdensome filing mechanism, e.g. CBE supplement, or Annual Report.

Routine manufacturing using PATs:

Level of built-in redundancy?

• Body of PAT information should have equivalent or better informing power than the corresponding conventional approved end product tests.
• Recommend setting-up a table, showing relationship between PAT testing and current testing methodology.
• Parallel PAT testing and conventional testing (in-process and/or release tests) should be performed for a sufficient number of batches (a minimum of three batches in the absence of historical manufacturing data).
• Level of redundancy is often a business decision.

Identify steps for resolving OOS observations:

• PATs may allow selective rejection, or partial batch release.
• Within batch trend information available with PATs facilitates resolution of OOS observations.
• Until PAT test results are approved for regulatory purposes, the approved conventional test results supersede the PAT results.
• If the OOS result is traced to instrument failure, then traditional approved analytical methods can be utilized for batch release in lieu of PAT based result.

Using on-line NIR (for blend, dry, content, dissolution) and HPLC as examples of PAT, please outline the essential experiments (hypothesis or questions) that should be conducted by a company to successfully develop and validate these tools for use in manufacturing operations.

Criteria should be based on

• Product performance
• Adequate Process Control
• Assurance of Product Quality

PATs either individually or in aggregate are predictive of final product quality.

• Process and Product signatures are a sum of multiple measurements.
• Demonstrate link between PAT parameter and product characteristics.
• Surrogate should be scientifically based.
• Acceptable variability in population should be established.

When and to what extent would FDA involvement facilitate PAT R&D and application projects?

• Issue guidance. Define terms, provide a glossary.
• Develop training programs; both internal and external.
• Develop workshops; and include case studies, mock submissions.
• Provide meetings between Agency and Applicants.
• Global harmonization, ICH guidance.