OOS Investigation (US FDA): Phase II Full-Scale Investigation

In short: Phase II is the full-scale OOS investigation that follows when Phase I finds no clear laboratory cause. It covers production review, predefined retesting and resampling, correct handling of averages and outliers, and the quality unit’s final batch decision. This post is for QC, QA and production staff and follows FDA’s May 2022 guidance.

See also: Phase I: laboratory investigation (US FDA) and the UK view in OOS investigation Phase II (MHRA).

When Phase II starts

Phase II starts when the Phase I assessment has not shown that laboratory error caused the OOS result and the laboratory data appear accurate. The investigation then follows a predefined procedure. Its goal is to find the root cause and to take corrective and preventive action (CAPA). It must include a review of production and sampling procedures and often needs extra laboratory work. FDA says such investigations should get the highest priority, and they include judging the impact on batches already distributed.

Review of production

The quality unit (QU) leads the investigation and involves every department that could be implicated, such as manufacturing, process development, maintenance and engineering. If manufacturing is at a contract site or at several sites, all sites that may be involved must take part.

The manufacturing records must be fully reviewed, and the written record should contain:

1. A clear statement of why the investigation was opened.
2. A summary of the process steps that may have caused the problem.
3. The result of the documentation review, with the actual or probable cause.
4. The result of a check on whether the problem has happened before.
5. A description of the corrective actions taken.

If this part confirms the OOS result and finds the root cause, the investigation can end and the product can be rejected. A failure investigation must still be completed for other batches or products that may be linked (21 CFR 211.192). If any material was reprocessed after extra testing, the record needs comments and signatures from production and QU staff.

OOS results can also point to a weak product or process design, for example poor formulation robustness, weak raw material control or large variation from a unit operation. Then the product or process must be redesigned to give reproducible quality. FDA also notes that OOS results may come from unauthorised or unvalidated changes to the manufacturing process.

Additional laboratory testing

Phase II may add laboratory work beyond Phase I. The two tools are retesting and resampling.

Retesting

Retesting means testing another portion of the original sample. It must come from the same homogeneous material that gave the OOS result. For a liquid, this may be the original unit or a composite. For a solid, it may be a further weighing from the same sample composite prepared for the original test.

Retesting is appropriate, for example, to investigate an instrument malfunction or a suspected sample handling problem such as a dilution error. The decision should rest on the aim of the testing and sound scientific judgement. The predefined plan should often include a retest by a second analyst who is at least as experienced and qualified in the method as the first.

Avoid testing into compliance. FDA has seen firms repeat tests until a passing result appears and then ignore the OOS result without justification. It calls this unscientific and objectionable. Therefore:

• The maximum number of retests must be fixed in advance in a written SOP, based on sound science and the variability of the method.
• The number must never change depending on the results obtained.
• The SOP must state the point at which testing stops and the batch is evaluated. If results are still unsatisfactory then, the batch is suspect and must be rejected or held for further investigation (21 CFR 211.165(f)).
• Any departure from the SOP should be rare, recorded and justified (21 CFR 211.160(a)). Before extra retesting, the QU should approve a protocol that describes the testing and how the data will be handled.

If a laboratory error is clearly identified, the retest result replaces the original. All original data must still be retained (21 CFR 211.180) and the explanation recorded, initialled and dated, with a discussion of the error and supervisory comments.

If no laboratory or calculation error is found in the first test, there is no scientific basis for setting aside the OOS result in favour of passing retest results. All results, passing and failing, must be reported and considered in the batch release decision.

Resampling

Resampling means analysing other units collected as part of the original sampling, or a new sample taken from the batch if necessary. The original sample should be large enough to allow extra testing if an OOS result occurs. In some cases a new sample may be appropriate, and the examination of extra specimens must follow predetermined procedures and sampling strategies (21 CFR 211.165(c)).

The investigation may conclude that the original sample was prepared improperly and did not represent the batch (21 CFR 211.160(b)(3)). Widely different results from several aliquots of one composite can indicate this, once analytical error has been excluded. Resampling should use the same qualified, validated methods as the first sample. If the investigation shows the original sampling method was inherently inadequate, a new sampling method must be developed, documented and approved by the QU.

Reporting test results

Averaging

Averaging is sometimes right and sometimes wrong.

Appropriate uses

• Averaging several measurements within one test, for example optical rotation, where the average is the reported result. USP prefers averages for microbiological assays because of their natural variability.
• Averaging replicates to give one reportable result, for example consecutive replicate injections of the same HPLC preparation (usually two or three). This counts as one test and one result. The test method approved by the QU should state the number of replicates and the acceptance limits for variability among them. If those limits are not met, the results should not be used.
• Averaging complete, repeated runs of an assay, if the method defines the average as one reportable result. Limits on variability between runs should come from the known variability of the method and be written in the method.

During an OOS investigation, these uses are acceptable only if they were part of the original test that produced the OOS result.

Inappropriate uses

• Averaging hides variability, so individual results should normally be reported separately.
• It is not suitable for powder blend uniformity or content uniformity, where the variability itself is the thing being measured.
• Do not average the original OOS result with retest or resample results. This can mislead, especially when some results fail and some pass. The laboratory must give the QU all individual results.

FDA illustrates this with an assay specification of 90 to 110 percent. An OOS result of 89 followed by 90 and 91 averages 90 and the extra results tend to support the original failure. An OOS result of 80 followed by 85 and 105 also averages 90, but shows high variability and unreliable data. In both cases the individual results, not the average, should be used to judge product quality.

Outlier tests

An outlier is a value markedly different from the rest of a series from a validated method. It may come from a deviation in testing or from natural sample variability, and you must never assume it is a testing error.

• Use of an outlier test must be decided in advance and written in the data interpretation SOP, with the specific test, its parameters and the minimum number of results needed.
• For highly variable biological assays, an outlier test may be suitable, and USP general chapter <111> describes this. USP warns that rejecting values only by their size, without finding a cause, should be done sparingly.
• For validated chemical tests with small variance on a homogeneous sample (such as an assay of a composite), an outlier test cannot identify a cause and must not be used to invalidate the suspect result. Under the 2022 revision it may be used only for information, to see how far a result lies from the mean. FDA’s footnote says outlier testing should not be used to invalidate chemical assays and cites the 1993 Barr Laboratories court opinion.
• Outlier tests do not apply where product variability is being assessed, such as content uniformity, dissolution or release rate. There an apparent outlier may be a true result from a non-uniform product.

The laboratory must give all results to the QU for the final decision. If a contract laboratory finds no assignable cause, it should report all results to the customer on the certificate of analysis, and FDA also recommends giving the customer the OOS investigation report.

Concluding the investigation

The QU evaluates the results, decides the batch quality and makes the release decision, following the relevant SOPs. After a batch is rejected, there is no limit on further testing to find the cause so that corrective action can be taken.

Interpreting the results

An initial OOS result does not automatically mean the batch fails, but it must be investigated and the findings, including retest results, interpreted before a release or rejection decision (21 CFR 211.165).

• Cause found, result invalidated. The result is not used to judge the batch. A result can be invalidated only on the basis of a documented test event that can reasonably be shown to have caused it.
• OOS confirmed. The result is used to judge the batch, which does not meet its standards and should be rejected under 21 CFR 211.165(f). The work becomes a batch failure investigation extended to other batches or products that may be involved.
• Inconclusive. No cause is found and the OOS is not confirmed. The OOS result must still be given full consideration in the disposition decision. The QU may still decide to release the batch.

FDA’s example of a possible release: a composite assay range of 90.0 to 110.0 percent, an initial result of 89.5 percent and seven later retests between 98.8 and 99.1 percent. The laboratory investigation found no error, production review showed no aberration, the process history shows robustness and other batch data agree with the passing retests. The QU might then conclude the OOS result did not reflect true batch quality. Even then, follow-up is needed to stop any laboratory error recurring. Any release despite a non-invalidated OOS result should follow only a full investigation, and the QU should always err on the side of caution.

Cautions

• Averaging preparations from the original sample. If a procedure requires a series of assay results for one reportable result, and some are OOS, some pass, and all are within the method’s variability, the passing results are no more likely to be true than the failing ones. Treat the average as OOS even if it is within specification. This agrees with USP General Notices, under which an article must comply whenever a compendial test is applied.
• Replicates from the same final preparation. If the method sets variability limits and a fixed number of replicate injections, and those limits are met, one replicate alone should not make the reportable result OOS.
• Borderline passing results. A low assay result that is within specification is a concern. Batches must be formulated to contain not less than 100 percent of the labelled amount (21 CFR 211.101(a)). Check whether the result may fall below the limit before expiry.
• Records. Keep all records related to the OOS result, including complete data from all tests (21 CFR 211.194).

Field Alert Reports

For products under an approved NDA or ANDA, FDA says a field alert report (FAR) is required within 3 working days for any failure of a distributed batch to meet an application specification (21 CFR 314.81(b)(1)(ii)). An OOS result is one kind of such information. Unless the result is shown to be invalid within 3 days, submit an initial FAR, and send a follow-up FAR when the investigation ends. This applies to US-marketed application products and is not an Indian CDSCO requirement.

Example SOP

An example format is available for download: SOP for handling of out of specification results. It is an illustrative template and must be adapted to your own quality system and approved before use.

Frequently asked questions

How many retests are allowed?

FDA gives no fixed number. The maximum must be set in advance in a written SOP on scientific grounds and must not change with the results.

Can I average the OOS result with the passing retests?

No. Averaging the original OOS result with retest or resample results hides variability. Report and assess every individual result.

Can an outlier test invalidate an OOS assay result?

No for validated chemical assays. An outlier test cannot find a cause. It may only be used as supporting information.

What if the investigation is inconclusive?

The OOS result still counts in the disposition decision. The QU may release the batch only after a full investigation shows the result does not reflect batch quality, and it should err on the side of caution.

Who decides whether the batch is released?

The quality unit (called the quality control unit in the 2006 guidance), under 21 CFR 211.22.

References

• US FDA, CDER. Guidance for Industry: Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production, Revision 1, May 2022.
• US FDA. 21 CFR Part 211 (including 211.22, 211.101, 211.160, 211.165, 211.180, 211.192, 211.194) and 21 CFR 314.81(b)(1)(ii).
• ICH Q10, Pharmaceutical Quality System (April 2009).
• USP general chapter <111>, Design and Analysis of Biological Assays, as cited in the FDA guidance.

This post is a study summary in original wording and not a substitute for the guidance itself.