Equivalence Criteria of Impurities for API Process Validation
Introduction
This guidance provides recommendations for demonstrating equivalence of impurities to historic batches during validation of API processes for small molecules. Demonstrating equivalence is needed in all API validations, but this guidance is applicable to APIs prepared by chemical synthesis. Additional analysis (not described here) may be needed to evaluate physical attributes of the API, where requirements are defined for the corresponding Drug Product. A change in an API process should be evaluated to determine if revalidation of the drug product process may be necessary. Process validation should confirm that the impurity profile (e.g. for process-related impurities and volatile impurities, i.e. residual solvents) for each API is within the limits specified. Therefore, for all equivalency studies, it is expected that the results of the validation batch testing be within registered specifications. It is also expected by ICH Q7a that the impurity profile for validation batches be comparable to or better than historical data. This is consistent with expectations for post-approval changes to API processes. For older processes, the registered specifications may or may not accurately reflect the most recent historical data. In cases where the specifications may not be reflective of more recent historical data, it is recommended that additional criteria, such as meeting the upper statistical limit of historical data, be considered for validation equivalence criteria.
It is recommended that equivalence criteria include:
There are no new unqualified impurities exceeding 0.1% (or similar threshold per ICH Q3a; and
One of the three numbered conditions that follow is met.
* Existing impurities meet registered specifications when:
– There are fewer than ten historical reference batches, or;
– Reference data may not be representative of current process capability, or;
– The quantitative reference data do not exhibit a symmetrical distribution.
* Existing impurities meet specifications and are within historical ranges (minimum and maximum values) when:
– There are ten or more reference batches with a non-Gaussian distribution of quantitative data, but there is confidence that the data are representative of current process capability.
* Existing impurities meet specifications and are within the statistical limit (mean +/- three standard deviations) when compared to reference data when:
– There are ten or more reference batches with a normal (Gaussian) distribution of data, and there is confidence that the data are representative of current process capability.
A more detailed discussion of these guidelines follows, along with recommendations to assist with statistical evaluations. Other specific recommendations include selection of reference batches and analytical tests for the equivalence comparison. A more detailed summary of recommendations is provided in Appendix I, and a statistical evaluation of sample data to illustrate one approach for performing the equivalence comparison is provided in Appendix II.
Recommendations and Rationales
For validation of a process to prepare a new API, the impurity profile should be comparable to or better than the profile determined during process development, or for batches used for clinical or toxicological studies. For evaluation of a newly developed or modified process to prepare an API that is already commercially distributed, the comparison provides assurance that the process produces material that is equivalent to (or better than) acceptable material prepared in the past by an existing process, with respect to impurities.
The need to evaluate equivalence for isolated process intermediates should be considered on a case-by-case basis.
For some validations, insufficient reference batches are available for a meaningful comparison.
Meeting established limits is considered adequate for the equivalence comparison in these situations.
For other validations, the availability of adequate reference batch data makes the use of statisticalacceptance criteria more desirable because it enables comparison of the validation batches to established process capability data.
Selection of appropriate reference batches:
New Products (new API or intermediate process at first manufacturing site) Batches prepared during process development should be selected. These may include batches prepared in Production equipment, and those prepared at laboratory and pilot scale. They should be batches made by the same process and may include pivotal clinical batches, and those used for toxicological and/or stability studies. In many cases, the number of acceptable development batches may be relatively small.
Existing APIs
a) Major changes undergoing revalidation, or first-time validation – Reference batches should be selected from plant batches prepared prior to the validation to be performed.
b) Site transfers – Reference batches should be selected from those prepared at the sending site.
c) New processes – Reference batches should be selected from those prepared by the previous process used to prepare commercially released API. Other considerations for selecting historical batches for the reference data set:
Batches should be selected from those prepared at commercial scale, and from consecutively prepared acceptable batches that represent the process as it has most recently been run.
The impact of past changes to the process or to analytical methods should be considered when selecting historical batches for the data set.
Rejected, reprocessed, and reworked batches generally should be excluded from the historical set because they are generally not representative of the results expected from normal processing. Removing such batches for cause is permissible (thus breaking the consecutive nature of the batches), but batches should never be arbitrarily added to or excluded for the purpose of influencing the historical ranges of analytical results.
The historical reference batches should be included or referenced in the validation protocol as part of the pre-determined acceptance criteria.
Selection of analytical tests for equivalence comparison: Results from routine analytical release testing should be examined when performing the equivalence comparison. Results from testing of the validation batches will be compared to historical results obtained using the same analytical methods. A change in an assay method thus requires careful consideration, unless it has already been shown to give equivalent results to the earlier method.
Other considerations for selecting analytical tests for equivalence comparisons:
Evaluating the analytical results for a chemical quality attribute (usually process impurities) and an attribute related to drying of the product (usually residual moisture and/or solvents), depending on the relevance to the API’s critical quality attributes.
Other tests provide different measures of the critical quality attributes of the material being evaluated. Tests other than those noted above are recommended for inclusion in the equivalence evaluation only when that test method examines a quality attribute of particular importance to the API or drug product (6).
Select tests that provide quantitative results. Tests that provide qualitative results (“Meets Test”, “Positive”) are generally of less value to equivalence evaluations.
Number of Reference Batches and Acceptance Criteria
From a statistical perspective, it is desirable to have at least ten or more batches in the reference data set to be confident in determining if a normal distribution is defined by the data. Statistical evaluations may be performed for the equivalence comparison using as few as ten reference batches in the historical data set, but from a statistical perspective, there is more confidence in the statistical distribution with a larger number of reference batches (e.g. 30 or more).
When fewer than ten batches are available for the reference data set, it is preferable to compare results for the validation batches to historical ranges for a given assay (maximum and minimum values). This is also recommended when there is doubt that the reference batches accurately represent process capability.
Reference data should be examined to determine if a normal (Gaussian) pattern is evident. If there are sufficient reference batches and examination of the data reveals a normal distribution, a “mean +/-3 standard deviations” (+3 SD) statistical acceptance criterion may be used to evaluate the test results from the validation batches. When this criterion is selected for representative data that reveal a normal distribution, there is less than a 1% chance that a test result will fall outside the +3 SD range. The probability of failure may increase when this criterion is selected despite fewer reference batches, and/or when the data do not show the pattern of a normal distribution.
Appendix I of this guidance provides recommendations for selecting acceptance criteria, based on the number of reference batches and the distribution of data for the reference batches.
Other Considerations for Selecting Acceptance Criteria
For impurities, the additional criterion that no new unqualified impurity exceeding 0.1% is present when compared to previously produced material is required (1). Any unqualified impurity not meeting this criterion would prompt investigation, and explanation in the validation report. Note that the threshold of NMT 0.1% is a generalization; a different limit may be required, depending on the product dose.
For residual solvents, a statistical acceptance criterion may not be appropriate in the validation of process changes that impact product drying. An acceptance criterion of “meets specifications” may be preferable in this situation. A process change that involves introduction of one or more new solvents to the process requires different considerations, and will usually require that registration specifications and/or ICH (Q3C) guidelines be met for the new residual solvent(s).
Other acceptance criteria may be used if justified in the validation protocol, based on a risk assessment. An example of a reference data table for evaluating equivalence data is provided in Appendix II.