FSMA Compliance: Water Usage for Produce

There is an evolving literature surrounding cross-contamination control, however it is nowhere near as developed as heat penetration studies for thermal processing. Most of the experiments are bench scale without steady state control of the antimicrobial level. It is tempting to generalize based on these limited studies to all systems. Very few studies have been done at commercial scale, so recommendations are largely extrapolations to large scale systems. Good studies will insure steady state control of the product flow, the contaminant flow, and the water chemistry including antimicrobial level and pH. The use of any wash adjuvants should also be noted as they can improve wash chemistry performance. There are indications that the sequestering power of some adjuvants, such as SmartWash, mitigate the problems of high organic loading.

Returning to our diligent processor, he or she must make some decisions about how to proceed with incomplete and imperfect information. For the balance of this discussion, let’s assume that he or she has digested the available literature, has generated some of his or her own data, and has pushed his or her suppliers for information. Based on this information, the diligent processor can make a best effort to define his or her process, thus answering the first question. This process will allow cooling and foreign material removal and avoid cross-contamination to the best of their ability. The quality of this process will depend on the quality of the information inputs. The diligent processor can expect to have his or her reasoning questioned during an FDA inspection. To date, FDA has not made an inspection of this type. Until such an inspection occurs, the industry does not know what FDA will find acceptable. However, months and years of safe operation with good control should weigh into this discussion, taking us to the third question, how do you know that you did your process? Contrary to many wishes, it is unreasonable to expect the FDA to accept a 10 ppm control point for free chlorine as sufficient to define an acceptable process given that FDA has not provided a safe harbor, as already discussed.

To prepare for this third question, it is recommended that the diligent processor use an automated control system that automatically logs the key operational parameters of the defined process. Antimicrobial level and pH are the two most fundamental parameters. Temperature should also be considered if it is part of the defined process that the processor developed. However, generating and logging this data is insufficient. The data must be vetted and used.

The data vetting process involves loaded terms such as precision, accuracy, verification, validation, calibration, standards, reference, and many more. Vetting is the process for insuring that the generated and logged data are useful and meaningful. The diligent processor needs to be able to articulate why he or she trusts the logs of operational parameters. “Garbage in yields garbage out” is an oft-cited aphorism that clearly applies in this situation. If the processor is using the suggested controller, the supplier of this controller should be able to demonstrate the utility of the generated data. Utility of the data is considered because all measurements inherently include error and therefore are to some extent wrong. A measurement can be expected to be no more accurate than its reference. A reference should have a traceable pedigree that instills confidence.

The inherent error is normally discussed in terms of variance and statistical probabilities. These tools can be used to establish the probability or confidence that all of the product received the defined process. At present, there is no regulatory guidance as to how this should be accomplished. A best practices approach will need to address at least four types of variance including instrumental variance, process variance due to inhomogeneity in the processing system, process variance associated with product, product feed rate, etc., and reference variance. This is a big task.