Sampling size for validation of control measures for a food industry
Dear Co-Food Safety Practitioners,
I would like to ask for your kind help, I'm currently revising an Internal Procedure for Validation of Control Measures for a food industry.
I've been looking for the minimum sampling size for validation study but there is no specific quantity indicated on the standards.
Can you help me to find a reference material for this? Your help would be highly appreciated.
Many thanks!
Olive Realubit
Dear Co-Food Safety Practitioners,
I would like to ask for your kind help, I'm currently revising an Internal Procedure for Validation of Control Measures for a food industry.
I've been looking for the minimum sampling size for validation study but there is no specific quantity indicated on the standards.
Can you help me to find a reference material for this? Your help would be highly appreciated.
Many thanks!
Olive Realubit
Hi Olive,
Is there a particular FS Standard involved ?
In general the scope of yr query is rather "enormous".
Can you be more specific as to the Procedures requiring validation ?
Hi Charles,
The procedure that I'm pertaining to is our internal procedure for validation of control measures in compliance to the requirements of ISO 22000:2005 8.2.
Currently we are following a minimum of 30 data points for validation and it has a great impact to the cost. I'm trying to find a reference material / standard that will help me to establish a lower sampling size that is statistically correct and acceptable.
Thanks,
Olive
Hi Charles,
The procedure that I'm pertaining to is our internal procedure for validation of control measures in compliance to the requirements of ISO 22000:2005 8.2.
Currently we are following a minimum of 30 data points for validation and it has a great impact to the cost. I'm trying to find a reference material / standard that will help me to establish a lower sampling size that is statistically correct and acceptable.
Thanks,
Olive
Hi Olive,
Do you mean 30 points to validate one control measure ? Or ??
Normally IMEX "validation" is only required to be carried out at intervals, not continuously. Verification follows up Validation. (the interpretation sometimes varies with Standard).
Generally sample size relates to the specific activity/its variation/objective/accuracy required.
Yes 30 data points for 1 control measure.
The current practice is to identify the hazard and control measure, after that the control measure will be validated if it can consistently deliver the same output.
To give you an example the team identified B. cereus as hazard. Control measure to be validated is Pasteurization 85degC for 15mins., we need to collect atleast 30 batches for microbiological test and we will relate the result to the time and temperature log and other intrinsic factors to measure the effectiveness of control measure in controlling the hazard. Are we doing it right or is there any other approach that can minimize the testing of microbiological aspects?
Can you give me some references to further enlighten me on how to effectively execute validation study aside from codex?
Thanks,
Olive
Yes 30 data points for 1 control measure.
The current practice is to identify the hazard and control measure, after that the control measure will be validated if it can consistently deliver the same output.
To give you an example the team identified B. cereus as hazard. Control measure to be validated is Pasteurization 85degC for 15mins., we need to collect atleast 30 batches for microbiological test and we will relate the result to the time and temperature log and other intrinsic factors to measure the effectiveness of control measure in controlling the hazard. Are we doing it right or is there any other approach that can minimize the testing of microbiological aspects?
Can you give me some references to further enlighten me on how to effectively execute validation study aside from codex?
Thanks,
Olive
I assume you have a reference which validates that a specific lethality related to, say, achieving 5-6 log reduction for B.cereus, is being delivered at 85degC/15min. (unusual target species).
I assume you are aware of the (thermally) most demanding variety of product involved (eg largest size).
The usual requirement is to do a few runs, eg 3-5, measure time vs (core) temperature, and calculate the integrated lethality to demonstrate it matches the requirement. (Excel programs are available free).
Above is for an arbitrary process, the specifics will relate to what you are actually working with. There are various threads on this forum discussing aspects of this and including references.
30 points sounds like a number taken from a statistics book to measure sigma assuming a normal distribution. :smile:
If you have an effective process in place that ensures a 5 log reduction, then all you really need to do daily is verify time and temperature in process....unless of course you do not have an established process which would explain your sampling plan
The whole idea behind thermal process is that you've achieved commercial sterility based on a set thermal process. So Heat transfer, internal temperature, pasteurizer temperature and line speed all need to have been tried and tested. FOR EACH DIFFERENT PRODUCT you make
Then micro becomes redundant as your process is designed and approved by a competent authority to produce a safe product with no spoilage or pathogenic bacteria left
What are you processing?
Hello Scampi,
Thank you for providing your insights. We are processing cakes, biscuits, seasonings and wafer products.
I'm currently documenting the validation procedure that will cover the CCPs, PRP and OPRPs and I want to establish a minimum sample size for validation only.
There will be a separate sample size for verification and routine monitoring of parameters related to the identified control measures.
I have read so many references including statistical books. I was thinking if I can use a minimum sample of 10 (benchmarked from a minimum sample for method validation). Some of the references suggests 3-5 production run only, but I'd like to use 10 to test reproducibility/repeatability.
Please let me know your thoughts about my insights. Your feedback will be highly appreciated.
Many thanks!
Olive
Oh ok, now I completely understand
You need to take samples based on what you can manage CORRECTLY. When I validated a blast freezer it took me an entire month to complete because ALAS we all have other work that needs done LOL
That validation ended up being 45 data sets that were repeated based on location and height in the blast freezer
The data sets included;
room temperature where product was trayed
product temperature pre freeze
time to less than 10C
time to less than 4 C
time to less than 0C
Wind velocity
Average temperature in blast freezer
Temperature and relative humidity of outside air
So to validate your bake
you need to run data loggers in product in at least these positions
X X
X
X X
depending on your oven, you may need more (like every other row if it's a rack push in oven) and then send those samples out for micro in sterile bags. YOu need to also make sure you've got worst case scenario covered (rolls a bit big, oven temp low etc)
Any your thought pattern sounds right. Repeat repeat repeat, then you know the measures ARE covering the hazards and then you can just monitoring things like line speed, oven temperature and dough ball weight/thickness during operations
I have read an article about establishing sample sizes for process validation when destructive or expensive testing is required, he recommend the use of AQL and risk level.
He used the C=0 sampling plan (0 defect = accept, 1 or more defect reject), to provide more protection to the consumer especially if it involves health and human welfare.
It's good to apply the C=0 sampling plan in designing a validation study. The initial sample size is 10 but if it will not provide consistent result the control measure will not be implemented unless it will be re-designed/revised. The new control measure will undergo process of re-validation.
Hi Olive,
I think you can deduce from previous posts that "sample size" is not readily able to be generalized.
Sample sizes in statistical texts aim to be mathematically rigorous. But they also often require that very large resources / time are available for the implementation. Unfortunately such options are rare in the food business and particularly where destructive sampling is involved.
It is likely that the minimum sample size you can use will often depend on yr desired accuracy / intrinsic variance of the target to be assessed.
Dear Scampi / Charles,
Thank you for your inputs. Already submitted my proposal to the core team.
I used the following:
C=0 sampling Plan
Risk Levels incorporated with ANSI Based sampling (sample size will depend on the batch size, lot, etc and off course the risk)
WHO Technical Guidelines
Codex (Validation of Control Measures)
Warm regards,
Olive