Gone with the wind
In theory.In practice, says our puzzled GMP Agent-in-Place, the production start-up team noticed that a very few capsules were reassembled with a tablet or two missing.
"Now, the encapsulation machine had pin-probes that lowered into the bottom capsule shell to make sure it was full. If the pins didn't 'feel' the right number of tablets in the capsule, the line should have rejected it.
"This wasn't happening. We couldn't imagine what was going wrong, and we were worried. If badly filled capsules had gotten into the market, we would have faced recalls and losses."
Our eagle-eyed Agent-In-Place camped out on the line, watching every move of every operation.
"Its process was really dusty: product and capsule-shell dust would build up particularly fast in the center of the table. Management had told the operators to keep the machine clean...but run it efficiently. So they were, logically, cleaning the table while it was running. With a vacuum. An operator would open up the side-shield doors and stick the snout of the vacuum over the filling table.
"That's where I spotted the problem. As he moved the nozzle through the doors to the middle of the table, he needed to maneuver it between the fill-checking pin station and recapping station. So the vacuum passed, for just an instant, over open capsules that had already registered as correctly filled. The vacuum could, and occasionally did, suck a tablet or two up out of the capsule base.
"Fortunately, this problem was easy to fix once we realized what the problem was—change the cleaning access point. Until then, though, we were tense, I'll tell you."
Too much of a B12 thing
"During a routine stability study of a vitamin product, the B12 (cyanocobalamin) test result was too high by a factor of ten," reports our GMP Agent-In-Place. "Several people looked at the calculations. They figured that someone had made a mistake in the dilution factor, or something similar, but nobody could find the error.
"We retested the batch, with the same out-of-specification (OOS) result. As part of the stability investigation, I looked back at the initial release test results to confirm the B12 value. And, lo, there in the paperwork was a retest for a ten-times-too-high reading.
"Now, that initial release test had been performed by a new analyst, and the quality control folks had assumed that she'd moved a decimal point in one of her calculations. An 'experienced' analyst repeated the release test. Looking at the documentation and the crossed-out calculations, I realized that this test, too, had actually yielded a 10× the OOS result...but that the experienced analyst had treated it as his own calculation error, and made it go away by dividing by his own fudge factor. That result stood, and the batch was released.
"On the evidence, then, the stability test showed ten times too much B12 because the batch had ten times too much B12. We dug back further through the records.