Implementation and automation of a particle detection system for parenteral products.

摘要

The availability of the probabilistic model of particulate inspection in sealed sterile containers, first reported in 1979 (/), potentiated both a study of human inspection capability and the rapid objective validation of a high speed, automated inspection system. The U.S. Pharmacopeia XIX implies a deterministic human particulate inspection standard. The data presented earlier and in the present paper demonstrate that the detection of particulates is probabilistic not deterministic. The resolution of the conflict between an invalid deterministic particulate inspection standard and a demonstrated probabilistic inspection is a prerequisite to the wider utilization of automated inspection devices. Specially addressed in this paper is the problem of data replicability in probabilistic validation trials. A demon-stration of the utility of 95% confidence limits for critical validation parameters is presented. The evaluation of detection probability for identified containers is the basis for relative determinations of both inspection security and discrimination. This methodology is equally applicable to the evaluation of comparative performance in an inspector group and to the validation of an automated device against the performance of selected inspectors. A brief overview of robotic inspection considerations is included.