THE NEW TREND IN THE GMP ENVIRONMENT: TRENDING
While a "trend" shows the general direction in which a situation is changing / developing or in the way that people behave, "trending" is more collecting information and attempting to spot a pattern (trend) in this information. And this was exactly what this year's European GMP Conference in June was about - as initially presented by Dr Franz Schönfeld from the government of Upper Franconia, Germany.
According to Dr. Schönfeld trending is also part of a continuous improvement process as required in ICH Q10 and part of the quality risk management process of ICH Q9.
In the regulatory environment trending is used in the context of a risk-based approach during inspections. Concerning this he mentioned the following documents/initiatives:
PIC/S document PI 037-1
Compilation of Community Procedures on Inspection and Exchange of Information "A model for Risk Based Plannings of Inspections of Pharmaceutical Manufacturers"
Interim and pre-inspection compliance reports (MHRA)
FDA: Quality metrics programme and
In Germany compliance factors (CF1-CF3) are assigned based on the outcome of the last inspection.
In the part "Regulatory requirements" he listed quotations from the EU GMP Guideline Parts I and II in which trending is addressed. These are the chapters 1.10, 6.7, 6.9, 6.16, 6.32, 6.35, chapter 8.19 (all in part I) and chapter 2.60 (part II). But due to the variety of trends, the complexity of the various data and statistical tools there are no detailed recommendations or instructions from the authorities on how to perform a trend analysis. Even in the FDA Guidance on Investigating Out-of-Specification (OOS) results an "out of trend" is only indicated in a footnote, said Dr. Schönfeld. As a consequence there are only few regulatory requirements - which he also highlighted.
The general GMP requirements from chapter 4 (Documentation) of the EU GMP Guideline Part I generally apply to trending. This means there should be written policies, procedures, protocols and reports and the associated records of measures or results resulting from trending and OOT investigations. Furthermore, a trend analysis is a statistical tool which follows the same principles of documentation as any other analytical test: Thus, the requirements are: a specification, a test protocol and a test report. But what should be written in the specification, test protocol and test report?
These questions were also answered by Dr. Schönfeld. The specification should describe inter alia which data (such as quality metrics, KPI) shall be trended, which trend (such as drift to the specification limits, increased degradation in the course of ongoing stability studies, process instability) shall be identified and how to define acceptance criteria.
He showed different control charts (Regression Control Chart, Slope Control Chart, SPC taking Western Electric Rules/Westinghouse Rules into consideration) as examples for the implementation of test protocols.
Dr. Schönfeld stressed the importance of establishing proper specifications and test protocols in order to guarantee that no false trends or no trends are detected. Therefore statistical tests should also be applied in the context of trend analyses in order to exclude that the supposed trend really is a pattern owed to the natural variability of measurements. As examples he mentioned Wald- Wolfowitz runs test, Grubbs' test Shapiro-Wilk test, Kolmogorow-Smirnov test and Chi-square tests each of which have different sensibilities and which can be used for the identification of normally distributed data.
In the test report Dr. Schönfeld expects a statement whether the analytical results or the trend (e.g. in the case of stability programme) are within the expectation. If this is not the case an OOT investigation should be carried out during which at least the following questions have to be clarified:
- Are written procedures in place for confirming or invalidating OOT results?
- Is there an assignable root cause considered for the confirmation or invalidation of the OOT result?
- Are any CAPA measures required due to the invalidation or confirmation of the OOT result?
Further important questions that should be clarified in the report:
- Is there a likelihood of future OOS results due to the trend?
- Is there any data that require more detailed attention?
Dr. Schönfeld considers that a good trending reflects the manufacturer's profound knowledge of the processes and hence builds confidence in the manufacturer's quality risk management. Moreover, trending may mitigate the risk of:
- failure to comply with marketing authorisation
- limitation or failure to comply with GMP
- regulatory action (product recall, withdrawal of the GMP-certificate)
- and related legal costs with an uncertain outcome
Other benefits of trending may be to avoid shortages while focussing on less robust processes and products.
The findings of inspections are always fascinating. According to Dr. Schönfeld:
- not all of the required data are trended (e.g. no critical IPC testing)
- trending is limited to a graphic presentation of data
- trending is performed with disregard of the underlying amount of data and the resulting uncertainty of the result (e.g. PQRs are prepared with a minimum of 3 batches)
- there is no specification/definition of a trend
- the specification and/or test protocol do not take into account statistical methods
- trending tools are not properly applied (e.g. only 3 Sigma rule of the "Westinghouse Rules" is applied)
- significant negative trends are not reported to the competent authorities but only OOS-results (especially in the case of ongoing stability studies).
In return Dr. Ingolf Stückrath, Head of Production, Sanofi- Aventis GmbH presented trending in the context of stage 3 of the process validation lifecycle (continued process verification).
At the beginning of his lecture Dr. Stückrath showed how Six Sigma was implemented at Sanofi Aventis' (2001-2005). The implementation was even honoured with the Six Sigma Excellence Award by IQ in 2005. Dr. Stückrath said that the highly structured approach to a six sigma project by the DMAIC model is really useful:
- Define: What is important?
- Measure: How are we doing?
- Analyze: What is wrong?
- Improve: What needs to be done?
- Control: How do we guarantee performance?
Dr. Stückrath generally subdivides processes in the context of a process analysis into four categories. If processes run within (Cpk > 1) or outside specification limits (Cpk < 1) and are they in statistical control or not. He then explained the process validation lifecycle by means of a case study putting the focus on stage 3 continued process verification. He cited the relevant passages from the FDA Process Validation Guidance which explicitly requires an ongoing programme as well as a statistical evaluation regarding a trend of the data. Afterwards he presented the observations of an inspection carried out by the Canadian authorities which criticise that a statistical treatment of the process data including confidence intervals is missing. For the future, the answer assured a defined procedure concerning the evaluations of the process and IPC data in order to recognize trends early enough to react before an OOS material is obtained. Accordingly, a project was started. Dr. Stückrath talked about this project.
The following questions were discussed in the course of the project:
- Which data should be evaluated? and
- What should be the objective of the evaluation?
Should the procedure only satisfy the authorities' expectations or rather be a tool which really helps to control the process? The following data could be included in a trending:
- All data defined as critical
- Not all CMC data
- Data which are critical for the process
Furthermore, the following question was discussed in the course of the project: Which analysis should be chosen? This could range from Excel based analytics to a full statistical analysis using statistical software.
Result of all these considerations
Next he presented a decision tree of IPC controls and other critical and specifically selected process parameters which retrospectively analyses 30 batches in total every 10 batches alternating. Evaluation criteria are whether trends could be identified in the course of the statistical process control and if the process has a Cpk > 1. If one of the criteria is answered in the affirmative a form will be filled out on which the further steps (such as the handling of singular events) are documented. In the case of deviations the evaluation of the data is performed by a process engineer. The completed form is approved by the Head of Production and the Head of Quality. This very pragmatic procedure was discussed critically by the participants.
Conclusion: Trending is/becomes a trend in the GMP environment. Especially in the validation environment statistical trend evaluation is required already since 2011 because of the FDA Process Validation Guidance for stage 3 of the process validation lifecycle. The revised Annex 15 requires something comparable too, as of 1 October 2015. The findings of inspections demonstrate that the sources of errors can be manifold when trending. This covers the whole spectrum from the non-inclusion of all required data and the non-consideration of statistical methods to the wrong application of these methods.
At least Dr. Schönfeld will concentrate more on the topic trending during the inspections he will carry out in the future - and he won't remain the only inspector doing this.
Author:
Sven Pommeranz
CONCEPT HEIDELBERG
