Agent in Place: Statistically Unstable

Thin are the lines that separate stability, statistics, and chaos.
Nov 01, 2007

Broken charted

"The new stability supervisor came in and rewrote all 40 of the stability system standard operating procedures (SOPs)," recalls our GMP Agent-in-Place. "This occurred in the days of a mostly manual system, when many of the SOPs included forms, which were also redesigned—by hand. The supervisor was very diligent, and had not only flow-charted each SOP, but also the entire stability system on his desk blotter. The coverage of each SOP was marked clearly on the blotter.

"The quality vice-president came from HQ to visit, and was impressed with the flow chart. But there was one box on the chart that was not included in any SOP.

'Why wasn't this item covered, and what is it?' the VP wanted to know.

"'It's those actions and decisions made by you (the quality VP) when there is a confirmed stability failure, and your actions weren't covered in factory SOPs,' replied the supervisor. 'I don't know what you do.'"

A transatlantic stats lesson

"In the days before computers were ubiquitous and affordable, I wrote a statistical linear analysis program for a handheld TI calculator with a memory stick. It was pretty slick, fitting it all in the very limited memory available, and was used by several of our foreign manufacturing sites," notes our GMP Agent-in-Place.

"But then one of the foreign manufacturing sites came up with an error message, and they wanted to know why the program didn't work correctly. Because the memory available was limited, only the normally found cases were considered. If that gave the error, then the confidence bounds were outside the product limits! I had to explain this to someone who was performing the data entry by rote, who didn't understand statistics, and who didn't speak English well. Try giving someone a course in basic statistics over the phone under those conditions."

A BIG study

"We had made the batch, and it was pretty small. At the time I released it, all the data provided met the requirements," says our GMP Agent-in-Place. "But a couple of days later, more data that I hadn't expected came in, and it was clear the batch I had already released was a disaster. But no one else knew about it.

"I asked and found out that the batch was entirely in inventory, so to keep my good name intact I went to the warehouse and requisitioned the entire batch for a stability study. That batch never did see the light of day again."

Rooted

"Early in the computer revolution, we had purchased what amounted to a desk-sized programmable calculator," notes our GMP Agent-In-Place. "It didn't have much memory but could help with certain statistical calculations for our stability studies. We had two different stability groups using it: one group for domestic products and one for international products.

"The domestic group had a lot of turnover, and the new employees didn't know that the international group was also using the calculator, as they were never observed using it even though the calculator was in our office. So when the square root function stopped working, we knew enough not to use the machine, but hadn't told anyone about it. We only realized that the international group was still using it when we came back early from lunch one day and found the international data clerk entering the data.

"We explained the problem to her supervisor and received the expected brow-beating for not following company policy for tagging out defective equipment."



Pharmaceutical Technology's monthly "Agent-in-Place" column distills true-life cautionary tales from the secret files of Control, a senior compliance officer. If you have a story of clueless operators, oblivious management, inopportune lapses of judgment, or Murphy's Law in action, please send it to Control at
We won't use any names, but if we do use your tale of disaster, courage, or just plain weirdness, Control will send you a coveted Pharmaceutical Technology t-shirt.