The FDA has posted the report on its inspections of the PCA facility on the FDA website, here. I'm hoping to have time this week to work through the FDA's comments one at a time and see what we can learn. As PCA has noted on its website, the FDA's findings are not final, and PCA will have the opportunity to respond.
"Observation 1" covers testing and re-testing of peanut paste, peanut butter, and peanuts for salmonella. The allegations differ a bit, but most involve instances in which PCA received positive salmonella tests on certain lots of product which subsequently re-tested negative and then were shipped. This practice is generally described as "lab-shopping," but it is difficult to tell exactly what is going on here.
I can think of four different scenarios, each troubling in its own way, that might allow for a negative test to be generated after a positive test on a given lot of product.
First, however, we need a little background on testing. When testing a given "lot" of peanut products you can't test every bit of product. You have to pull a sample according to a strategy that will give you a representative sample of the full lot. Rather than do the statistical analysis necessary to work out a sampling strategy, most manufacturers follow the plan that the FDA would use in the case of an investigation. (You can find that plan here.) After the samples are pulled and sent to the lab, the lab combines them and then takes a sub-sample from which it attempts to culture salmonella.
So, how can you get a positive and a negative test on the same product?
One possibility is that certain labs are corrupt and that a negative test can be bought. I have seen no evidence that this is the case here.
Another possibility is that a lab could re-test the same sample and get a negative result. That raises a difficult question. Is there a false positive problem in salmonella testing such that poor handling of samples or lab procedures regularly generates false positive tests? Or is there a false negative problem, in which a sample with salmonella in it could test negative? Or both?
A third possibility is that the manufacturer could pull a new sample before re-testing, and hope for a negative outcome. Here, it seems that the manufacturer would be banking on getting what amounts to a false negative on a given lot based on a second sample (though maybe there is a false positive problem in salmonella testing).
The troubling thing about scenarios two and three is that if there is a serious "false negative" problem in salmonella testing, then PCA (and other manufacturers) will be shipping product in good faith that is in fact positive for microbiological contamination. In any testing scenario relying on sampling there is always the possibility of false negatives. The question is how likely is this? I'm not enough of a statistician to say. It is worth noting, however, that PCA has recalled all the product from the Georgia plant, including product which, I presume, passed micro testing with flying colors.
The fourth scenario is that a manufacturer could submit a false sample if the re-test is performed on a new sample from the lot. The FDA report did not suggest this was a problem in this case, but it would be easy to do.
The bottom line: testing by itself is not 100% reliable in preventing manufacturers from shipping bad product. Never has been, never will be.