The first problem concerns the role of microbiological sampling and has been raised by the industry. It is a cogent argument. The meat and poultry industry argues that microbiological sampling by the government is ineffective and not germane to processing operations. In at least one of these areas, the industry has been supported by a Federal District Court judge, who has ruled that the presence of Salmonella spp. does not in and of itself indicate a failed process as USDA FSIS has asserted.

In its waning days, the Clinton Administration proposed two new rules for ready-to-eat (RTE) products. One requires product testing by establishments of all RTE products and the other requires environmental testing in the establishments where RTE products are produced; environmental testing includes both food-contact and non-food-contact surfaces. The Bush Administration put both rule-making initiatives on hold pending review but the new Secretary of Agriculture, Ms. Veneman, interceded and the rules have been published in the Federal Register for public comment. Industry argues logically that environmental sampling is unnecessary because RTE products have a lethality step in their processing that is designed to kill pathogens.

Salmonella Enteritidis

Industry also argues that HACCP is not science-based because requirements for acceptable pathogen levels were established using a national baseline sample that was subject to several sampling biases, including selection bias. The claim is made that baselines resulting from this biased sampling were then used to establish regulatory standards. USDA FSIS argues in return that the baselines are simply baselines and the regulatory standards are based on scientific principles. The courts probably will ultimately settle this issue.

A trained risk analyst might point out that attempting to establish a standard dose-response curve for a microbial pathogen is futile. It is possible to quantify a dose-response within a range, but the range would be so wide as to be meaningless because of the wide variances in Lowest Observed Effect Level among ages and health statuses, races and ethnic groups, socioeconomic status groups, ad infinitum.

The industry's challenge of HACCP on scientific grounds has some merit. Some of the problems with exposure assessments and mathematical models are outlined below.

• Lack of necessary data - because many of these predictive tools require both input and validation data to ensure that the model reflects reality, this option may not be feasible. In many cases, laboratory costs, manpower utilization, time constraints, and political agenda will prohibit both businesses and government regulatory agencies from obtaining more than a one-time vertical slice sample.
  
  
• Bacterial growth models may not reflect reality - it must be constantly borne in mind that, with microbial pathogens, the unit of infection is the single cell and, while the risk of infection from a single cell is exceedingly small, it is not zero. For example, as few as ten individual vegetative Campylobacter jejuni cells have been shown to cause morbidity.
  
  
• Bacterial dose-response levels are undefined - because bacterial levels are not constant, exposure by consumers can only be estimated. In order to estimate exposure, one must be able to estimate the probability that the pathogen is present in target foods, the initial level of the pathogen if it is present, and how this initial level varies as a result of food processing operations.
  
  
• Model inaccuracy - as shown in the risk analysis example, many models use the beta Poisson equation to model bacterial growth, but many growth patterns are more closely described by Gomperz's equation, while still others evade precise mathematical modeling.
  
  
• Hazard analysis is not risk analysis - when establishments determine where potential biological, chemical, and physical hazards exist in their food processing operations, they are relying on historical data and not scientific validation. USDA FSIS has recently been attempting to make the claim that absence of evidence is not acceptable as proof of absence of potential hazard.
  
  
• Individual susceptibility is variable - individuals with healthy immune systems are seldom at risk for most foodborne toxins. There is even relative sensitivity to staphylococcal enterotoxin.

All of the points raised indicate that risk characterization is an uncertain process. It must be stated in terms of probabilities of pathogen presence, exposure odds, probable average dose-response, and so forth. We cannot categorically state that, because a particular ground beef lot contains 5,000 colony-forming units per milliliter (cfu/ml) of coliform bacteria, there is a specific percentage of the population exposed to this lot that will become ill. In addition to all of the uncertainties posed by the nature of microbial pathogens, we must factor in the vagaries of food processing operations; home food preparation efficacy; food safety practices in the home; post-preparation handling; retail establishments' and distribution agents' methods of handling, storing, and processing.