In previous CAPA Corner blogs, we discussed a model for conducting a science-based systematic investigation, and the importance of understanding the problem and ensuring you have all the facts, before conducting a root cause investigation.

We’ll continue with the “widget” example we have been using. The widget is assembled, placed into a pouch, the pouch sealed, and the batch sterilized and released. When the firm performs pull tests on the pouch seal in final inspection, they are finding that some seals fail the minimum seal strength requirements. An investigation is being conducted to find out why the seals are failing this test. In previous steps of our investigation we defined the problem using an IS / IS NOT Diagram and then collected data to verify the veracity of each statement in the diagram and to gather more detailed facts. We also assembled a list of possible causes using multiple techniques.

Technique: Testing

We now test each possible cause against all the facts in the IS / IS NOT Diagram. To illustrate we’ll start by testing one possible cause, a change in the sterilization process. (Perhaps there has been a sterilization cycle change that has led to the rise in defects.) In this example we’ll test this theory against some of the “Where” facts to demonstrate how the testing is done. See Figure A.

We ask the question, “If sterilization is causing the seals to fail the pull test, how do you explain the defect is only seen on Production line 3 and not on lines 1, 2, or 4?” This cannot be explained as the same sterilization cycle is used for products on all four production lines!

We then document the fact not explained. See Figure B. Since the facts do not support the cause we are about to toss out sterilization as a possible cause, but first we ask, “Are there any assumptions we can make, based on our real- life experience, that can explain how sterilization could still be the cause?” In this case we are at a loss to come with any assumption since the defects are being found before sterilization occurs. We can cease investigating sterilization and document that no assumptions can be made.

To continue our illustration we’ll test another possible cause: perhaps the operator on heat seal machine #2 was not adequately trained. We ask the question, “If inadequate training is causing the seals to fail the pull test, how do you explain the defect is only seen on Production line 3 and not on lines 1, 2, or 4?” This is easily explained as the operator on machine #2 works on production line 3 and not on the other lines. Since these facts perfectly support the possible cause, no documentation is required and we test our possible cause against the next set of facts. Refer to Figure C.

“If inadequate training is causing the pouch seals to fail the pull test, how do you explain the defect is found only on heat seal machine #2 and not on machines #1 or #3?” The defect is found on machine #2 because that is where the operator works. However, we cannot perfectly explain why we do not see the defect on the other two machines because all three operators were hired and trained at exactly the same time. If inadequate training is the cause we would expect to see defects from the operators on machines #1 and #3, but we don’t! We document the fact not explained. See Figure C. Since the facts do not support the cause we are about to toss out inadequate training as a possible cause, but first we ask, “Are there any assumptions we can make, based on our real- life experience, that can explain how training could still be the cause?” In this situation we respond that during the training, perhaps only operator #2 wasn’t paying attention. This is a realistic assumption and we document this as well. In this case we keep training as a possible cause. Later we will need to verify this assumption.

We have used abbreviated examples. In actual testing we challenge each possible cause against all the facts. Whenever a possible cause cannot be explained by a fact, and no realistic assumptions can be made, that cause can be eliminated. We document:

• Any fact not explained.
• Any realistic assumptions that we have made to explain how the possible cause could still be the change that caused the issue.

My experience is that the testing process will typically eliminate 80 to 85 percent of the possible causes.

Summary

I have demonstrated how to test the possible causes against the facts to eliminate those causes that simply do not make sense and significantly reduce the number of possible causes under investigation. In the next installment we will further reduce our list of possible cause to identify the technical root cause (the change that precipitated the problem) as well as identify any systemic root causes that allowed the change to occur or to go undetected.