Dear Sean Archer,
Thks yr reply.
IMO, risk matrices are like Mary Poppins, (almost) anything is possible. Subjective is perhaps the polite terminology.
There is a fairly popular view that significant hazards in
haccp require a likelihood of occurrence (LKO), at a minimum, to be “reasonably likely to occur”. For example this pictorial extract –
significant hazard.png 70.23KB
33 downloadsOn such logic, (LxS) is typically non-significant. Hence my “M”. One arguable but quoted definition of “M” is that the user/user's facility has
personally experienced one event of the hazard in a stated time frame, eg one month in a year (8% ?
). But I totally don’t disagree with yr interpretation of “L”, “S” either, there are many, many official 5x5 risk matrices which will agree with you. Just an example of the inherent subjectivity which is possible, especially around the borderlines. And then we might have “Justification Column” methods in Risk Tables, “Regulatory” CCPs, ALOPs, FSOs, Pathogen Reduction Projects. The list is endless. And this is before you consider the blending with, say, different microbiological severities / vulnerable consumers etc etc.
There has been a multi-year case of a Country vs Country row at GATT over the banning of importation of a particular whole fish (hazard of introducing exotic pathogens not then existing in destination) which resulted in teams of national (and co-opted) statistical/fish experts defending the use of
their qualitative risk matrices. The blocker won, on a majority verdict, at a massive cost, essentially because no-one could prove that the reality was not subjective and thus allowed a (wide) range of interpretation. The logic arguments ran into a few hundred pages, and a book.
Clearly epidemiological data should be involved also. But this is not always easy to find in the case of such sensitive issues and may still be difficult to apply in a
specific situation. Regretfully milk is not my own practical area of expertise so I had to rely on literature publications relating to status/control of milk at reception prior to further processing. The “average” conclusion I inferred from my studies was that antibiotic residues is definitely considered an issue in various geographical areas but apparently less so in others. Very little published data of incident frequencies exists AFAIK (could find). Further input is only too welcome.
Similar issues can occur when one tries to assess the “risk status” of approved suppliers, eg everyone (visibly) knows that severe pathogens are being occasionally found in raw, RTE vegetables with harmful consequences of substantial widespread distribution. So what does a COA actually prove ?? or even a 1 year stack of satisfactory COAs. Then again, one has to start somewhere. I recall (from memory) that Canada once simply defined (for a different product which was to be cooked by consumer) that the possession of a supplier audit / COA for every lot was a minimum acceptable official requirement. Fair enough maybe. So the paperwork becomes the critical limit, at a cost +++. But is such logic statistically defendable ?? Seems to defeat one major analytical reason for
HACCP’s elevation in the first place?.
Thks for 1 vote to support my OPRP interpretation. I hv no doubt that there are plenty of dissenters out there also.
Rgds / Charles.C
PS after re-reading yr previous post, i realised that my yoghurt presentation is perhaps slightly ambiguous in that (a) my inclusion of "unapproved supplier" was intended to imply a (generic-type) hazard due to an
error by the raw material reception team in proceeding to evaluate an undocumented lot, not a policy decision to buy in that way, (b) the hazard of antibiotic residues was not intended to be feasible due to an unapproved supplier (although obviously it could be) but more a reflection of the general (known) situation in the industry, ie typical average level of contamination. The problem then passes as to whether a COA for an approved supplier is accepted as an adequate measure of the safety status of the whole batch or not. In practice any option seems to be acceptable to auditors as long as a choice can be validated for the COAs or respectively for another alternative, ie testing for antibiotic residues at reception. For non-liquid materials which actually do contain pathogens like Salmonella, the sampling aspect of a COA should favour the supplier unless gross contamination has occurred. For liquids / chemical hazards, more chance of homogeneity / fairness I guess. In both cases, validation is presumably workable for a typical n=5 / no detection in 25g samples scenario. By definition this result then matches the iso 22000 requirement of achieving an "acceptable level", regardless of conceptual sampling limitations. (a more meaningful statistical result would demand a much heavier sampling).
added - here are links to a couple of recent articles re. antibiotic residues. Predictably, the % positive detections vary substantially-
http://www.nytimes.c...ess/26milk.htmlhttp://www.arpnjourn...bs_1111_329.pdf
