SQF 2.4.4.2 proficiency testing - water activity
Hello, I have a question that I didn't see in the forums or in the SQF GUIDANCE:
We are using a water activity meter to test water activity of ingredients and products. We have been doing proficiency testing for water activity by splitting a sample and testing part in-house for water activity and sending part to a third-party laboratory and comparing the results from the two.
The part I don't understand is about comparing results. For example, one dried botanical (cut & sifted, i.e., pieces about 6 mm long) has in-house and third-party water activity results of 0.60 and 0.474, respectively. What acceptance criteria should we be using, e.g., a maximum relative standard deviation (RSD) percentage? These results seem quite far apart, but I have a difficult time finding acceptance criteria for such proficiency testing.
We manufacture dietary supplements in the United States, are compliant with 21 CFR Parts 111 and 117, and have an SQF Edition 9 certification. We have already read the SQF guidance for proficiency testing, and this type of detail is not in the guidance.
Thank you,
Matthew
From my understanding of proficiency testing, it is up to each company to decide what the acceptance criteria should be, based on risk.
Still working on becoming compliant with this part of the code, so hopefully others can share their experiences.
Hello, I have a question that I didn't see in the forums or in the SQF GUIDANCE:
We are using a water activity meter to test water activity of ingredients and products. We have been doing proficiency testing for water activity by splitting a sample and testing part in-house for water activity and sending part to a third-party laboratory and comparing the results from the two.
The part I don't understand is about comparing results. For example, one dried botanical (cut & sifted, i.e., pieces about 6 mm long) has in-house and third-party water activity results of 0.60 and 0.474, respectively. What acceptance criteria should we be using, e.g., a maximum relative standard deviation (RSD) percentage? These results seem quite far apart, but I have a difficult time finding acceptance criteria for such proficiency testing.
We manufacture dietary supplements in the United States, are compliant with 21 CFR Parts 111 and 117, and have an SQF Edition 9 certification. We have already read the SQF guidance for proficiency testing, and this type of detail is not in the guidance.
Thank you,
Matthew
Hi mattthew,
So what is the estimated +/- confidence in the results ?
Hello Charles.C, I am not sure. The accuracy of the unit we use for testing for water activity is ±0.02 aw according to the manual. However, that is not the only factor affecting uncertainty or confidence, which might be most clearly evident in how much variability we are seeing in replicate results for a given sample type. I am trying to get data on replicate analyses from our lab and from the third-party lab.
Matthew
Was the sample treated in the same way?
Are you doing 2-3 readings on the same sample?
Are you calibrating your machine each time?
Seems like too much of a swing to me. I would expect results within 0.05 of each other....if it's the same sample.
Looking forward to hearing how it goes
Hello Charles.C, I am not sure. The accuracy of the unit we use for testing for water activity is ±0.02 aw according to the manual. However, that is not the only factor affecting uncertainty or confidence, which might be most clearly evident in how much variability we are seeing in replicate results for a given sample type. I am trying to get data on replicate analyses from our lab and from the third-party lab.
Matthew
Hi matthew,
Sorry for delay, I missed yr reply.
As per AJL, you should be using reference samples for calibration ?
Differences can be a question of sampling or analytical variation, or both.
You need to do a little "blind" replicate sampling at both ends to get an idea of accuracies involved (+ ideally include a (anonymous) reference sample for ext.lab.
Offhand, I agree with AJL, the results look simply incompatible. My guess is non-homogeneous sample if yr calibration/Procedure is ok. Should be readily checkable though.
Thanks, I'm trying to get replicate data from our lab to see how close they are to one another. Unfortunately, not much of an investigation was done into the differences & I am pushing for more info.
Remember it's temp dependent so the temperature you have det your chamber to is important.
Try and find some literature as well for expected valued in your product.
An external lab isn't always right
Ask to see their test method if you can.