17 replies to this topic

[bpoon]

I know this is a definite no-no, because it would be testing into complacency. I've been scouring the FDA and USDA for explicit mentions against re-testing of final products (or any component items) to show, but I only found two sentences in a guideline by FSIS about trusting your lab and advising away from re-testing. I'm hoping someone may know where I can find very specific statements. By the way, I'm not trying to re-test to complacency- I'm against re-testing in general without an extremely just reason. 

 

 

[FurFarmandFork]

The explicit citation you're looking for will be deep inside old warning letters. FDA routinely states that "your answer is inadequate" when suppliers test food to get out of adulteration.

[bpoon]

Can I search these old warning letters on FDA's website or is it something that FDA sends personally to those companies?

[Bo16]

I have looked into this in the past.  Here are some of the references we used when we wrote our OOS SOP.   Retests are not always bad, especially if you are running manual assays, but you must have a plan.  Investigation, multiple samples, multiple analysts etc.   

 

1.1.  Guide to Inspections of Pharmaceuticals Quality Control Laboratories, http://www.fda.gov/o.../igs/pharm.html

1.2.   Investigation of Out-of-Specification Results, Hoinowski et al. Pharmaceutical Technology, January 2002.

1.3.  Federal Register / Vol. 61, No. 87 / Friday, May 3, 1996 / Proposed Rules.

1.4.  Guidance for Industry:  Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical production:  October 2006:  Pharmaceutical CGMPs

[bpoon]

Thanks for the advice. You're right- I'm going to have ours go into an investigation plan that requires isolation of affected hours, then multiple sampling of pallets. That should be a reasonable plan. 

[FurFarmandFork]

FDA warning letters are public and can be browsed here: https://www.fda.gov/...ers/default.htm. Sign up for the FDA daily digests and get new warning letter notifications delivered directly to your inbox, it's the best way to get a gauge on how FDA is actually enforcing the regulations.

 

Frankly, the best example of why you can't "retest until negative" is the PCA indictment. That's called fraud. FDA's drug regs are more specific as to why microbial results cannot be invalidated unless a specific lab error can be found to blame for the false positive.

 

A recent example: https://www.fda.gov/...7/ucm565516.htm

• In your firm’s correspondence dated 5/18/17, it was stated that your corrective action includes conducting random pentobarbital tests of finished products prior to shipment into the market to ensure that the raw materials are unadulterated. FDA has concerns about using random finished product testing as evidence that all lots of your finished products are unadulterated. The samples collected by FDA during this investigation demonstrate that pentobarbital contamination is not homogeneous throughout all units in a lot. Therefore, random testing of finished product may not be representative of all units of your products. Furthermore, finished product testing cannot mitigate the risk of pentobarbital in your raw material.

[bpoon]

Wow! Thanks FurFarmandFork! Your sites on the FDA's warning letters and the PCA indictment is exactly what I needed. Your recent example was an awesome read as well.

 

Just out of curiosity though, the FDA said that :

 

 

 

 The samples collected by FDA during this investigation demonstrate that pentobarbital contamination is not homogeneous throughout all units in a lot.

 

This is mainly because it is a chemical adulterant and that the introduction was specific in the food making process? I'm just wondering because a microbiological contamination would normally affect the entire lot unless there was a sanitation process in between or an event that blocks the flow of contamination. The investigation plan we have now is supposed to sample every pallet of the affected production hours with a composite sample composed of units from the beginning, middle, and end of each pallet. That would technically not be "random", right?

[FurFarmandFork]

Wow! Thanks FurFarmandFork! Your sites on the FDA's warning letters and the PCA indictment is exactly what I needed. Your recent example was an awesome read as well.

 

Just out of curiosity though, the FDA said that :

 

 

This is mainly because it is a chemical adulterant and that the introduction was specific in the food making process? I'm just wondering because a microbiological contamination would normally affect the entire lot unless there was a sanitation process in between or an event that blocks the flow of contamination. The investigation plan we have now is supposed to sample every pallet of the affected production hours with a composite sample composed of units from the beginning, middle, and end of each pallet. That would technically not be "random", right?

False. Charles C. will probably be here in a second to add some clarification to my micro sampling comments below.

 

Depending on the source of the contamination, it migh tbe homogenous throughout the lot, but I would consider this the exception to the rule. If micro contamination arises from a cross-contamination event or an employee or environmental hazard, it will only present where it was introduced.

 

E.g a condenser pan is dripping condensate on a conveyor, the condensate has l. mono. The l mono will only be in a sample which received a drip. But with a drip every few seconds and 1% of product affected, you could have tons of food that is contaminated. Your micro samples only tell you that the material that was actually sampled is free of the pathogens tested.

 

If your contamination event was truly randomly distributed throughout your lot, and 5% of the product was contaminated, a single sample has only a 1/20 chance of actually coming up positive! (generalization, but makes the point).

 

One last example, think of the last time you had a piece of moldy bread or cheese, was that mold evenly distributed throughout the loaf, or could you have sampled some from the good side or inside and given it to someone who didn't see the whole thing. What would they conclude about whether the cheese was spoiled or not?

[bpoon]

FurFarmandFork- In this case, would implementing an AQL scheme for the pallets be an acceptable criteria for investigating an OOS lot? I was thinking of an AQL scheme that will require under 1% total OOS after sampling more than 5% of the total lot in order to pass the lot. The OOS found, along with surrounding pallets, would be destroyed. 

[Charles.C]

Hi bpoon,

 

Sampling Plans are available for a wide range of situations.

 

The selection of any sampling scheme is unavoidably linked to yr criterion(a) for accept/reject of the "Lot".

 

You need to consider this aspect first, eg what is the justification for yr max. 1% average(?) "contamination" level (actual meaning= ?) suggestion in Post 9 ?

 

Some possible types of criteria which have develped sampling schemes are (a) statistically valid (allowing for comments such as Post 8) estimation of an average  acceptable (max/min) value of "XYZ" for a whole lot, (b) an atttribute sampling scheme like Mil.Std based on A(acceptable)QL (obviously requires a quantitative definition of a significant defect), (c) a scheme specifically focused on a situation where a zero-tolerance is assigned to the target contaminating entity (again allowing for comments such as Post 8". Or other variations.

 

The choice may also be pre-defined by relevant Regulatory Requirements of course.

 

More info. here -

 

http://www.ifsqn.com...-sampling-plan/

[bpoon]

I'll definitely do some more digging into the regulatory requirements. Thanks for the breakdown on what to look at for the criteria. The biggest challenge for me is creating a criteria that can apply to all the types of products we make for different customers whilst limiting too many individual investigation procedures that further decrease efficiency. However, I'm thankful for everyone that helped answer my questions! This forum is seriously awesome. 

[Charles.C]

Hi bpoon,

 

Just as a generic comment, the common problem is that the ideal recommended sampling quantity is typically logistically impossible for manufacturing facilities. (Although this may depend on precisely what you have to measure of course, eg ner weights of cans are easy/automatable).

 

This is why many sampling plans you will see in the literature are "pragmatic". For example the typical FDA sampling plan for a lot for salmonella uses IIRC a minimum of 13 samples with a -ve result. Very few factory labs with a range of products can approach such quantities. So a compromise is required. The typical compromise is 5, ie nmMc = 5/0/-/0

[bpoon]

I totally understand what you mean Charles. When I brought up testing enough samples to have a statistical significance, the issue was that we don't have enough labor to isolate, transport, and plate the samples for testing. My company doesn't do pathogens on finished products, but instead use indicator tests. However, having all the samples done for all the petrifilms is impossible for the amount of staff available. Instead, we send all the investigation samples out to a third-party lab (in which case the cost is what hits us the most).

 

The issue of compromise raises an interesting question: How would one know what is considered a "typical compromise" for the type of testing and why? Personally, I don't like to compromise too much. I've performed Salmonella USDA-MLG testing and having 13 samples vs 5 samples tested is completely different. Although it was less work for me, it also lowered the ability to catch typical colonies significantly, even on selective plates. I guess what I'm trying to say is that from a business perspective, I need to make a compromise; but from a quality perspective, I don't want to compromise that much. 

[Charles.C]

I totally understand what you mean Charles. When I brought up testing enough samples to have a statistical significance, the issue was that we don't have enough labor to isolate, transport, and plate the samples for testing. My company doesn't do pathogens on finished products, but instead use indicator tests. However, having all the samples done for all the petrifilms is impossible for the amount of staff available. Instead, we send all the investigation samples out to a third-party lab (in which case the cost is what hits us the most).

 

The issue of compromise raises an interesting question: How would one know what is considered a "typical compromise" for the type of testing and why? Personally, I don't like to compromise too much. I've performed Salmonella USDA-MLG testing and having 13 samples vs 5 samples tested is completely different. Although it was less work for me, it also lowered the ability to catch typical colonies significantly, even on selective plates. I guess what I'm trying to say is that from a business perspective, I need to make a compromise; but from a quality perspective, I don't want to compromise that much. 

 

Hi bpoon,

 

Regarding "compromise" i suggest you study the ICMSF publication Microorganisms in Foods Vol 7 which contains a lengthy discussion on Lot Acceptance Criteria.

[bpoon]

Thanks for the direction, Charles. Now, I just need to get my boss to sign off on getting this! 

[Charles.C]

Thanks for the direction, Charles. Now, I just need to get my boss to sign off on getting this! 

 

Hi bpoon,

 

This (fairly advanced) "starter"  may not all directly relate since this is micro. rather than chemical contamination oriented but I anticipate there is a close parallel from an attribute sampling POV - 

 

http://www.fao.org/d...7-2f2b1c7a0f92/

[bpoon]

Charles, the link you provided has been an exceptionally good read so far! Thanks for sharing it!

[FurFarmandFork]

Came across this today during some other research, explicit guidance direct from FDA:

 

 

D.9 [Amended May 2010] I received a confirmed positive microbiological test result indicating the presence of a pathogen in a food. Based on this test result, I determined the food to be “reportable”. However, I retested the food for the pathogen and the second test result did not indicate the presence of the pathogen in the food. Should I still consider the food to be reportable?

Yes. There are a number of explanations why a food may test positive for a pathogen in one test and negative in one or more additional tests although the food continues to be contaminated. For example, the distribution of a pathogen in the food may not be homogeneous. Therefore, absent other circumstances clearly demonstrating the inaccuracy of the first test result, the first test result upon which the reportable food determination was made should be considered valid.

https://www.fda.gov/...212793.htm#repo