Lifecycle of NIR Analytical Methods: Regulatory Perspective

Transcription

1 Lifecycle of NIR Analytical Methods: Regulatory Perspective Bogdan Kurtyka, Ph.D. FDA/OPQ/OPF IFPAC Annual Meeting January 27, 2015

2 Outline Method lifecycle Impact of selected elements of lifecycle on method performance Considerations for submission Conclusions 2

3 Method Lifecycle Planning Verification Update Data Collection Calibration Routine Use Revalidation Validation Continuous verification Implementation Retirement Reporting 3

4 Planning Very important for development of accurate and robust methods Early planning allows for: Exact definition of method requirements, scope of validation, and validation acceptance limits Selection of proper test points, analyzer, and interface Design of equipment (probes, location of windows) Development of procedures for collection of samples used as reference Aligning next steps (such as data collection) with development (e.g., DOE runs) Discussion of chosen strategies with FDA 4

5 Considerations for Data Collection Follow planning activity for calibration and validation samples Include sources of variance Span the entire design space Scale and equipment variations Collect spectra from multiple analyzers Obtain uniform distribution of spectra over the analysis range 5

6 Considerations for Calibration Proper calibration assures method specificity, accuracy, and robustness Balance of accuracy and robustness: Overfitting can improve apparent accuracy However, robustness is likely to be compromised Some procedural solutions can help to improve accuracy: Model developed at target instead of design space Verification procedure triggered when deviating form target conditions 6

7 Considerations for Validation External validation Comparing predictions of NIR and reference method can simulate future analysis and confirms method error Independent validation set helps establish model robustness Independence of the validation set Time span between calibration and validation data collection Accuracy at the prediction limit(s) Desirable to understand limitations of validation and identify appropriate mitigation steps 7

8 Does FDA Require Parallel Testing? No, but FDA expects that applicant provides sufficient data for external validation Applicant needs to demonstrate that the NIR method is valid at commercial scale In general, validation involves comparing NIR prediction with the reference method for a sufficient number of commercial scale batches Adequate maintenance and update procedures are developed and implemented 8

9 Submission Considerations Information reviewed during evaluation of NIR methods used in the Control Strategy includes NIR analytical procedure Method development Method validation For an NIR method submitted as alternative to a regulatory method Specification should indicate which method will be used during routine commercial operation 9

10 Factors affecting method performance over lifecycle NIR predictions may change with time as new sources of variability are introduced New raw material suppliers Changes in process parameters, not previously included in the model calibration New data added to increase robustness or prevent false negatives Results of periodical evaluation 10

11 Other Reasons for Update Some NIR methods may require update even shortly after approval Reasons for update may include Instrument repair Scale up Method transfer to a new manufacturing site (even on same instrument type or unit) 11

12 Checking Continued Method Suitability Periodic verification against reference method Change control to detect changes in the process or material that may effect method results 12

13 Maintenance and Update Procedures Maintenance and update procedures should be developed as a part of a site quality system The procedures should consider Capability of early detection of problems Timely corrective action Recommended to provide a summary of model maintenance plan in the application for high impact methods For example, methods supporting RTRT 13

14 Potential Elements of Maintenance and Update Procedures (1) Frequency and methods of periodical model evaluation Actions resulting from OOS, new suppliers, process changes, new data, failed evaluation (level of change) Depth of validation done on updated model, depending on level of change 14

15 Potential Elements of Maintenance and Update Procedures (2) If the method includes end-point criteria, they should be re-evaluated to reflect performance of the new model Model update and recalibration Level of validation of updated methods Reporting mechanism 15

16 Reporting Changes Selection of reporting category for post approval change depends on potential impact of the change on product quality (21 CFR ) Consideration factors can include: Impact of the change on performance of the analytical procedure (e.g., accuracy, specificity) Role of the analytical procedure in the control strategy (e.g., only for monitoring or to support RTRT) Can leverage comparability protocols for lowered reporting categories (21 CFR (e)) 16

17 Conclusions In order to improve likelihood of initial implementation and continued success of NIR method it is important to consider: Careful planning for model maintenance Including full scale data in the model or appropriate surrogates Including possible sources of variance in the model Comprehensive validation on full scale data Post-deployment model verification in real process environment 17

18 Thank you Questions?