2 replies to this topic

[oseigloria]

Hi All,

What is the minimum number of samples statistically significant for the validation of CCPs, especially in Cocoa Processing.

The roasting step is one of the CCPs to be validated. How many batches should be used for this activity as a minimum to gain statistical significance.

[Scampi]

I would be creating a schedule for a full week, and creating composite samples at intervals that make sense to your process (e.g. every 15 minutes roasting per hour sample)

 

In order to use the data as validation, you have to ensure that you have enough evidence to demonstrate the LOWEST temperature is still effective over time of controlling the hazard, so that needs to be part of your data collection.  You could also use AQL to assist with the sampling plan.   100 data points would be the absolute bare minimum in my eyes

 

Last time I had to validate a process step, I lived in a blast cooler for a month taking temps every 30 minutes and ended up with roughly 30000 data points 

[jfrey123]

More is better, but always within reason.  And time is a consideration too.  I used to work for a dry-steam sterilizer, and for the products we had validated, they often were only a couple of batches but with 20-40 samples taken pre and post treatment.  The reason for high number of samples were to correlate to a cold spot study we did in the steam chamber, telling us which spots in the chamber were likely to be less efficient.  As the chamber handled whole pallets of product, we also prioritized samples from the center of the pallets as it is most difficult to reach those middle boxes with heat.

 

These samples were used by the lab to initially validate our time and temp specifications for those products, create acceptable treatment parameters.  We would sample a few months worth of future in similar ways to verify the initial results.  So you can start heavy, get some good results of a log reduction, and verify consistency over the next few months.  Wrap it up into a nice presentation and your auditors would struggle to find fault with the method.