Back to listing

Date: Fri, 14 Nov 2003 15:48:46 -0000
Reply-To: PSA Forum
Sender: PSA Forum
From: David Vose
Subject: Re: Scientific advice
In-Reply-To:
Content-Type: text/plain; charset="iso-8859-1"

Dear Colleagues

My name is David Vose. I am an independent risk analysis consultant and have been working in the food safety industry for some seven or eight years. I have been a consultant to FAO/WHO and OIE in producing their guidelines, providing training in risk analysis methods, and developing example risk assessments. I have also consulted to governments of twenty or so countries in the field. My independence from any government body or other organisation places me in a somewhat unique position in this field, and my comments reflect what many of my risk assessment colleagues somewhat less free to speak their mind have said to me privately.

I would like to offer my experience as a risk analyst faced with CODEX guidelines on risk analysis, in the hope that this experience will help CODEX improve its ability as a guide in the :

Guidelines can be very useful for clarifying the basic principles of an emerging field like risk analysis. Inevitably, when a field is immature, the people writing such guidelines are taking a stab in the dark at what those basic principles might be. When CODEX approached the task of defining microbial risk analysis guidelines for food safety, it appears that it took its cue from the US National Academy of Science guidelines on risk analysis which were based on chemical risk assessment.

The need to follow these guidelines for the risk assessment to be considered acceptable has led to the development of a number of very large, but nonetheless very crude, quantitative microbial (farm to fork) food safety risk assessment models. The pressure to follow the CODEX approach of separating out an exposure assessment and a hazard characterisation (aka dose-response model) has lead to a great deal of wasted effort. For some bacteria, we have no dose-response data, and for virtually all bacteria, those data are not available for important population sub-groups. We are extremely unclear what the basic pathways really are for most bacteria that constitute the exposure, and certainly don't have the data to be able to populate exposure models with decent numbers. The example models developed by WHO on Campylobacter, Listeria, Salmonella, etc are examples of such models.

On the other hand, a number of alternative approaches have been developed that require much smaller models, require far fewer assumptions, that are readily understandable, and that address specific risk management questions in a useful and timely manner. For example, the US FDA-CVM has produced a risk assessment of antimicrobial resistance that matches fractional changes in the prevalence of contaminated poultry carcasses to fractional changes in the number of campylobacteriosis cases; the Danish veterinary institute has apportioned salmonellosis cases to various food sources by matching strain and phage type proportions; and CDC and others habitually use analysis of case-control studies to apportion disease risk to food sources and to behaviour patterns. Under the CODEX guidelines, none of these methods would be considered risk analyses (because they have omitted the "essential" exposure and hazard characterisation components) and opponents use this argument in an attempt to discredit them. It will only be possible, for example, to manage the increasingly alarming and complicated issue of antimicrobial resistance if risk assessors are sanctioned to embrace any method possible that allows them to form rational arguments from the available data.

The CODEX guidelines and its personnel also place great weight on the separation of the risk assessor and the risk manager. Some countries have taken this separation very seriously, even to the extent of creating different government bodies for each one. I can see the point of such separation, but my experience has been that people's interpretation of the concept has made producing meaningful risk assessments very hard. Risk assessment should be an iterative and fluid process between assessors and managers: as the assessors collects more information, it becomes apparent that some possible risk management options are not feasible, but other possible options are found. For me as a risk analysis practitioner, the greater threat to the separation between assessment and management is when the risk managers have not publicly stated what is an acceptable level of risk. Without that, it seems irrelevant what the assessors produce, and that I think is the key threat that CODEX is attempting to counter.

I am all for establishing standards, but not where those standards erode the flexibility of the risk assessor to properly perform his/her task. CODEX is a political animal, and in my experience of helping produce the guidelines for other organisations, its personnel have applied considerable pressure to see their structure and wording being used, whether or not it is appropriate and useful to the subject. I think that these people have somehow lost the message.

Putting this experience together, it would concern me very greatly to see CODEX establish peer review bodies of risk assessment activities until it has broadened its understanding of the possible approaches to producing rational arguments (aka risk assessments) for evaluating different risk management options available to the decision-maker. I would also very much like to see CODEX review its guidelines on risk analysis by extracting lessons from the decision-supporting performance of completed risk assessments rather than basing them on theoretical concepts.

Sincerely

David Vose

David Vose
Risk Media Ltd
Le bourg
24400 Les Lèches
France
+33 (0)553 82 34 24
www.risk-modelling.com

Risk Media Limited