The Scientific Basis for Visible Particle Inspection

Abstract

The following paper was presented at the 1999 PDA International Conference in Tokyo, Japan February 24, 1999. The paper was written in response to the recent well-publicized incidents in Japan in which the presence of visible particulates in injectable products was observed. These incidents should be considered a wake-up call to all concerned with injectable pharmaceutical products. The overview of visible particle inspection presented at the Tokyo PDA Meeting is reprinted here as a reminder that a global marketplace requires well-defined parameters and evaluation methods. In the absence of these basic requirements continuation of the present conflicting particle contamination evaluations will continue by default. At this time, visible particle inspection around the world lacks both direction and an accepted common language. The adoption of the probabalistic concepts introduced by Knapp and co-workers in 1980 can supply these basic requirements.

The use of the probabalistic model and the statistically defined particle quality regions defined by Knapp and co-workers in 1980 was, at that time, a concept foreign to pharmaceutical quality. In the period since 1980, the methodology then introduced has been expanded and simplified. The analysis used by these workers was similar to that used in life insurance calculations and is just as reliable.

The results obtained following the adoption of the framework and probabalistic concepts introduced by Pflug in sterile products since 1973 supports the use of tested and verified probabalistic concepts in the production of injectable products. Pflug's work has provided a basis for the secure communication of sterile production methods and results. This improvement in communication has potentiated a surge in sterile product improvement and a consequent reduction in product cost. The use of the probabalistic model and statistically defined particle contamination quality regions introduced by Knapp and co-workers in 1980 is believed to have a similar potential to improve product quality and reduce inspection costs in visible particle inspection procedures. Application of the probabalistic model results in a numerical evaluation of both the inspection security achieved in rejecting visible defects and the excess cost introduced by the false reject rate of good product. It has the ability to translate the U.S.P. language “essentially free” of particle contamination into objective, reproducible numerical results.

Visible particle contamination data can be successfully analyzed only when its probabalistic nature is understood. The effect of disregarding the probabalistic nature of visible inspection data is well illustrated by the results of the landmark 1940 court case brought by Bristol Labs against the F.D.A.. The F.D.A. contention that the containers were contaminated with visible particles was challenged by Bristol in a court case. When the experienced inspector, who reported the contamination, was asked to differentiate between good and contaminated ampules on the witness stand he failed to distinguish between them. Consequently, the case was dismissed. Any attempt to devise or analyze visible particle quality limits that is not based on the probabalistic nature of the accept/reject decision process can only result in the frustration and failure experienced by the F.D.A. expert in 1940.

In the GMP philosophy, the introduction of any new device or system must be preceded by a validation demonstration. This is to ensure that the new device or system functions at least as well as the device or system to be replaced. The validation of semi- and fully automated systems for the detection of particle contamination rests upon the statistically replicable assessment of the security achieved with human inspection performance. Any attempt to validate an alternative inspection system without a knowledge of its probabalistic nature and the probabalistic nature of the human inspection benchmark performance cannot succeed. Any presently “validated” system in which the validation demonstration is not based on the probabalistic nature of the inspection data is operating in violation of GMP requirements.

The need for the probabalistic visible inspection model introduced by Knapp and co-workers was the validation, in rigorous GMP terms, of a semi-automated inspection system at Schering-Plough. The system had been designed and built to eliminate inspection backlogs of a life-saving antibiotic. In this validation, the performance of a skilled group of human inspectors to reject serious rejects was evaluated within 0.05 significance limits. The methodology translated the rule-of-thumb quality categories then in use (good-good, gray, bad-bad) into probabalistically defined quality zones. The capability of the alternative inspection system was then adjusted to reject with equal probability those containers identified by the human inspection to contain the serious “visible” defects within the 0.05 significance level limits desired. This procedure has been unchallenged since its 1980 introduction.

The data that resulted in each evaluation is a random distribution. To validate the new systems, these workers developed and published methodology to evaluate and compare the inspection security of the alternative system in terms of the human inspection performance benchmark. This methodology has been used around the world and is in present use to define the performance of most of the automated particle inspection systems in use today.

In this methodology, the human inspection results of a test group of containers with, ideally, a uniform distribution of rejection probability is used as a working standard. The inspection security and economic effectiveness of an alternate manual method, a semi- or fully automated inspection system is then compared to the human performance benchmark. Validation has been demonstrated when the average reject rate of the manually identified Reject Zone containers is matched or exceeded. A collaboration with the Optics Institute of Rochester University succeeded in reducing the rejection probability data developed in the standard Schering-Plough manual inspection into an objective physical particle size.

On two occasions, 1987 and 1990, the PDA provided pro-active leadership out of a morass of particle quality measurements. This was accomplished by convening international meetings to examine and evaluate the full range of particle contamination measurement and control. It is now nine years since the last such conference met. There have been major advances in methodology and technology that affect the full range of contaminating particle measurements and assays, including those for visible and sub-visible particles. Harmonization of particle measurement concepts and language will assist both the national and international information exchange upon which our global pharmaceutical marketplace is built. Will the PDA provide this needed leadership in the new millennium?