Summary of Judge Wolin's Interpretation of GMP Issues Contained In The Court's Ruling In USA vs Barr Laboratories

2-4-93

1. USP STANDARDS

The court ruled that USP's established standards are absolute and that firms cannot stretch the USP standards.  These standards provide established criteria upon which firms release their product.

2. FAILURE (OUT-OF-SPECIFICATION) LABORATORY RESULTS

Judge Wolin preferred to use the term "out-of-specification" (OOS) laboratory results rather than the term "product failure" which is more common to FDA's investigators.  He ruled that an OOS result identified as a laboratory error by a failure investigation or an outlier test ¹, or overcome by retesting ² is not a product failure.  OOS results fall into three categories:

A. LABORATORY ERRORS

Laboratory errors occur when analysts make mistakes in following the method of analysis, uses incorrect standards, and/or simply miscalculates the data.

Judge Wolin provided specific guidance about the matter of determining when an error can be designated a laboratory error.  Laboratory errors must be determined through a failure investigation to identify the cause of the OOS.  Once the nature of the OOS result has been identified, it can be classified into one of the three categories above.  He states that the inquiry may vary with the object under investigation.

B. LABORATORY INVESTIGATIONS

The court said that the exact cause of analyst error or mistake can be difficult to pin down and that it is unrealistic to expect that analyst error will always be determined and documented, and he ruled that the "laboratory investigation consists of more than a retest".  The inability to identify an error's cause with confidence affects retesting procedures, not the investigation inquiry required for the initial OOS result.

The analyst should follow a written procedure, checking off each step as it is completed during the analytical procedure.  Laboratory test data must be recorded in notebooks; use of scrap paper and loose paper is to be avoided.  These measures enhance the investigation process.

The court specifically identified procedures that must be followed when single and multiple OOS results are investigated.

For the single OOS result, the investigation must include the following steps and these inquiries must be conducted before there is a retest of the sample:

An alternative means to invalidate an initial OOS result provided  the failure investigation proves inconclusive is the "outlier" test.  The Court placed specific restrictions on the use of this test.

1. Firms cannot frequently reject results on this basis
2. The USP standards govern its use in specific cases
3. The test cannot be used for chemical testing results ³

A full scale inquiry is required for multiple OOS results.  This inquiry involves quality control and quality assurance personnel in addition to laboratory workers to identify exact process or non-process related errors.

The court ruled that when the laboratory investigation is inconclusive (reason for the error is not identified):

1. Cannot conduct 2 retests and base release on average of three tests
2. Cannot use outlier test in chemical tests
3. Cannot use a resample to assume a sampling or preparation error
4. Will allow a retest of different tablets from the same sample when a retest is considered appropriate (see criteria elsewhere)

C. FORMAL INVESTIGATIONS

Judge Wolin ruled that formal investigations extending beyond the laboratory must follow the government's outline with particular attention to corrective action.  He said the company must:

D. INVESTIGATION DOCUMENTATION

Analyst's mistakes, such as calculation errors, should be specified with particularity and supported by evidence.  Investigations along with conclusions reached must be preserved with written documentation that enumerates each step of the review in the form of a "computer generated flow sheet".  This writing should be preserved in an investigation or failure report and placed into a central file.

E. INVESTIGATION TIMEFRAMES

All failure investigations must be performed within 30 business days of the problem's occurrence and recorded and written into a "failure or investigation report".

F. PRODUCT FAILURES

An OOS laboratory result can be overcome (disregarded) when laboratory error has been documented.  However, non-process and process related errors resulting from operators making mistakes, equipment (other than laboratory equipment) malfunctions, or a manufacturing process that is fundamentally deficient, such as an improper mixing time, represent product failures.

3. RETESTING

Several opinions about retesting were issued in this decision.  The number of retests performed before a firm concludes that an unexplained OOS result is invalid or that a product is unacceptable is a matter of scientific judgment.  The goal of retesting is to isolate OOS results but retesting cannot continue ad infinitum.

In the case of non-process and process-related errors, retesting is suspect.  Because the initial tests are genuine, in these circumstances, additional testing alone cannot infuse the product with quality.  The court acknowledges that some retesting may precede a finding of non-process or process-based errors.  Once this determination is made, however, additional retesting for purposes of testing a product into compliance is not acceptable.

For example, in the case of content uniformity testing designed to detect variability in the blend or tablets, failing and nonfailing results are not inherently inconsistent and passing results on limited retesting do not rule out the possibility that the batch is not uniform.  As part of the investigation, firms should consider the record of previous batches, since similar or related failures on different batches would be a cause of concern.

A very important ruling in this decision sets for a procedure to govern the retesting program.  The judge ruled that a firm should have a predetermined testing procedure and it should consider a point at which testing ends and the product is evaluated.  If results are not satisfactory, the product is rejected.

Additionally,  the company should consider all retest results in the context of the overall record of the product.  This includes the history of the product ⁴, type of test performed, and in-process  test results.  Failing assay results cannot be disregarded simply on the basis of acceptable content uniformity results being satisfactory.

Retesting following an OOS is ruled appropriate only after the failure investigation is underway and the failure investigation determines in part whether retesting is appropriate.  It is appropriate when analyst error is documented or the review of analyst's work is "inconclusive", but it is not appropriate for non-process or process-related errors.

The court ruled that retesting:

4. RESAMPLING

Firms cannot rely on resampling ⁵ to release a product that has failed testing and retesting unless resampling is in accord with the USP standards (content uniformity and dissolution), or unless the failure investigation discloses evidence that the original sample is not representative or was improperly prepared.

5. AVERAGING RESULTS OF ANALYSIS

Averaging can be a rational and valid approach, but as a general rule this practice should be avoided ⁶ because averages hide the variability among individual test results.  This phenomenon is particularly troubling if testing generates both OOS and passing individual results which when averaged are within specification.  Here, relying on the average figure without examining and explaining the individual OOS results is highly misleading and unacceptable.

Content uniformity results never should be averaged to obtain a passing value for content uniformity.

In the case of microbiological assays an average is preferred by the USP.  Also, the Judge rule that it is good practice to include OOS results in the average, unless an outlier test (microbiological assays) suggests the OOS is an anomaly.

6. REMIXING

The need to remix often is clear indication that the process is invalid and casts doubt on those batches passed through testing without incident.

Remixing is reworking permitted under the GMP regulations.  Occasional remixing is acceptable, but frequent or wholesale remixing is unacceptable.

7. PRODUCT RELEASE

Scientific judgment can play a role when firms decide to release a batch to the public and the court said it cannot articulate specific procedures for release decision making.  However, Judge Wolin said that the USP standards upon which firms release their products are absolute and cannot be stretched.  For example, a limit of 90 to 110 percent of declared active ingredient, and test results of 89, 90,91, or two 89's and two 92's all should be followed by more testing.

It is clear that the release evaluation depends in part on the background of the batch and product.  Secondary factors that affect the actual finished product results as well as their reliability are: 

Judge Wolin ruled that context and history ⁷ inform many final conclusions and that one must consider past problems with the product and batch and evaluate all the data relative to the product and batch.

8.  BLEND TESTING

Blend testing is necessary to increase the likelihood of detecting inferior batches.  Blend content uniformity testing cannot be waived in favor of total reliance on finished product testing because finished product testing is limited.

The court ruled that sample size influences ultimate blend test results and that the sample size should resemble the dosage size.  Any other practice would blur differences in portions of the blend and defeat the object of the test.  The appropriate sample size for blend content uniformity in both validation and ordinary production batches is three times the active ingredient dosage size.

Multiple individual samples taken from different areas cannot be composited.  However, when variation testing is not the object of assay testing, compositing is permitted.

Firms must demonstrate through validation that their sampling technique is representative of all portions and  concentrations of the blend.  This means that the samples must be taken from places that might be problems, weak or hot spots in the blend.

In this case, the firm maintained that samples could be collected from the drums containing the finished blend.  The court ruled that the firm must demonstrate that sampling from drums rather than the mixer is representative.  He also ruled that the firm cannot composite blend samples and that they must take smaller blend content uniformity samples.

9. VALIDATION CRITERIA

A. RETROSPECTIVE VALIDATION

The court ruled that batches meeting the following criteria must be included in retrospective validation studies:

1. All batches made in the specified time period chosen for study must be included unless the batch was made from a non-process related error
2. Only batches made in accord with the process being evaluated  can be included

Only test results determined through an appropriate failure investigation and found to be caused by analyst or operator error can be excluded from the study.  Test results that are explained but merely called into question by successful retesting must be included in the study.  The exclusion of batches and test results must be documented through failure investigation.

The number of retrospective batches chosen for study must be greater than the number used for prospective validation.  Although the court set  no exact number of batches to be chosen, guidelines have been established as follows:

B. CONCURRENT AND PROSPECTIVE VALIDATION

Concurrent and prospective validation requires at least three consecutive batch runs of the process.

Mixing time studies should be included in a prospective validation program and follow any problems that surface in retrospective validation batches.

Particle size distribution specifications are widely accepted industry practice and should be included in validation studies.

10. SIGNIFICANCE OF APPLICATION APPROVALS

The court ruled that a firm cannot rely on the claim that the FDA previously approved their procedures  contained in an approved application.  This approval cannot be used as a defense and cannot be used to shield a process that produces failures.

11.  METHODS VALIDATION

Methods can be validated in a number of ways.  Methods appearing in the USP are considered validated and they are considered validated if part of an approved ANDA.  Also, a company can conduct a validation study on their method.  System suitability data alone is insufficient for method validation.

12.  CLEANING VALIDATION

The court ruled that a firm cannot wait for contamination and other problems to reveal inadequate cleaning procedures.  In order for the cleaning rules to be effective, the specific methods chosen must be shown to be effective.

The court ruled that a milling machine is a major piece of equipment and must be included in the cleaning validation program.

Firms must identify the cleaning agents used in its cleaning process.  When these agents are known to cause residue, the company must check for this residue.

Provided the firm has described its cleaning methods and materials in sufficient detail

AND

Unless the cleaning material is known to cause a residue

THEN

one run through of the cleaning procedure, in the absence of problems, is not insufficient for validation.

Footnotes

1 The court provided explicit limitations on the use of outlier tests and these are discussed in a later segment of this document.

2 The court ruled on the use of retesting which is covered in a later segment of this document.

3 An initial content uniformity tests was OOS followed by a passing retest.  The initial OOS result was claimed the result of analyst error based on a statistical evaluation of the data.   The use of outlier test is inappropriate in this case.

4 The court ordered a recall of one batch of product on the basis of an initial content uniformity failure and no basis to invalidate the test result and on a history of content uniformity problems with the product.

5 The court ordered the recall of one batch of product after having concluded that a successful resample result alone cannot invalidate an initial OOS result.

6 The court ruled that the firm must recall a batch that was released for content uniformity on the basis of averaged test results.