Visual Inspection Query

GOOD MANUFACTURING PRACTICE (cGMP)
1

A Visual inspection machine (For injectable products) is successfully qualified by KNAAP test study using Min / Max and intermediate size of vials during PQ.

But during the challenge test study, it is noted that 1 or 2 defective vials (defects included as challenge) are not detected by the system and are included in the good vials. The acceptance criteria for this challenge test is 100% i.e. all the rejected vials must be rejected and all good vials shall be accepted. The equipment manufacturer is unable to meet the acceptance criteria.

Risk Assessment : This could be a potential risk as any such defect, especially defect like particulate matter could be a risk to the patient.

Queries :
As 100 % detection of particulate matter in the vials is improbable, can we relax the acceptance criteria for the challenge test study as NLT 98 % instead of 100 %. (Considering an AQL testing by production and QA personnel will follow).

What is the industry practice for the acceptance criteria for qualification of automatic visual inspection machine?

What is the challenge test and acceptance criteria followed prior to start of each commercial batch in the industry?

2

Your acceptance criteria is probably too tight. AQL is one suitable way to determine sample size and acceptability levels.

AQL charts all have an accept/reject level. I would use attribute testing (not variable) using Level III, and make it a critical defect. Depending on your sample size, your acceptance/reject level would be set by the tables.

Also, from a bigger picture, keeping the product free from particles is the biggest and most important line of defense. If you don’t have particles in your process to begin with (via correct procedures, airflow, particle monitoring, filtration, etc.) then particulates are less (but not eliminated) concern.

Even from a risk assessment perspective, you have (1) likelihood of failure (2) ease of detection and (3) repurcussion of failure. You can’t control #3, but you can control #2 and #1. Make sure you are discussing #2 and #1 in a risk based summary paper.

With that, you can probably allow for a faiure that is more reasonable than 100% (which rarely if ever happens).