National Academies Press: OpenBook

Scientific Criteria to Ensure Safe Food (2003)

Chapter: 3 Food Safety Tools

« Previous: 2 The Science of Public Health Surveillance
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 69
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 70
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 71
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 72
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 73
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 74
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 75
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 76
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 77
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 78
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 79
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 80
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 81
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 82
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 83
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 84
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 85
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 86
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 87
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 88
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 89
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 90
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 91
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 92
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 93
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 94
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 95
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 96
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 97
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 98
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 99
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 100
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 101
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 102
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 103
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 104
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 105
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 106
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 107
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 108
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 109
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 110
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 111
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 112
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 113
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 114
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 115
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 116
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 117
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 118
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 119
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 120
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 121
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 122
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 123
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 124
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 125
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 126
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 127
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 128
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 129
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 130
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 131
Suggested Citation:"3 Food Safety Tools." Institute of Medicine and National Research Council. 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: The National Academies Press. doi: 10.17226/10690.
×
Page 132

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

~2 Food Safety Tools This chapter describes some of the major modern tools available to regula- tory agencies for use in developing food safety criteria and standards. Some of these techniques or concepts are widely known and extensively used, whereas others are still in the developmental stage. The description of these tools and the discussion of their current or potential uses and applications to enhance food safety have been organized as a progression from the better known to the novel. In addition, the committee strived to circumscribe the material on each tool to that which is relevant to food safety, recognizing that some of the sections, such as "Statistical Process Control" and "The Economics of Food Safety Criteria," are not only foreign to many food processors and food safety regulators, but are technical and scientific fields that only recently have been brought into play in the food safety arena. Thus, in view of the limitations in space and time facing the committee, the reader is referred to specialized treatises that expand on these areas when additional information is desired. HAZARD ANALYSIS AND CRITICAL CONTROL POINTS Introduction The Hazard Analysis and Critical Control Point (HACCP) system is a meth- odology that constitutes the foundation of the food safety assurance system in the modern world. Although a detailed history and description of HACCP principles and applications are beyond the scope of this report, the invaluable contribution that this food safety tool is making to improve public health, its central role in 69

70 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD present-day food processing, and its inseparable relationship to the issues dis- cussed in this report demand a short introduction and description of it. HACCP history goes back to 1959, when the National Aeronautics and Space Administration (NASA) commissioned the Pillsbury Company to manu- facture food products for use by astronauts during space missions. The stringent safety requirements imposed on these foods were a reflection of deep concerns in NASA about the potential consequences of foodborne sickness among astronauts in space, as well as of food particles interfering with flight systems (Stevenson and Bernard, 1995~. Although HACCP made its debut at the 1971 National Conference of Food Protection (Stevenson and Bernard, 1995), analogous systems (not yet designated as HACCP) had been in existence and had been applied in practice in some food-processing operations, notably in the canning of low-acid foods and in milk pasteurization. These operations included: (1) identification and assessment of the hazards: Clostridium botulinum spores in canned low-acid foods and milk-borne pathogens such as Mycobacterium tuberculosis, Brucella spp., and Coxiella burnetii; (2) identification of the critical control point for these hazards: heating at specified temperatures and for similarly specified times in either of these operations; and (3) a system to monitor the critical control point: time and temperature recorders. Despite the fact that these food-processing operations had built-in notions of HACCP, the efforts of the Pillsbury team in articulating the fundamentals of present-day HACCP and testing its effective- ness, followed by additional contributions from the U.S. Army's Natick Labora- tories, are nothing short of landmarks in food safety history. HACCP is well established in the food-processing regulations of the United States. However, its introduction proceeded slowly, beginning in the 1970s and accelerating only until the mid-199Os. The migration of HACCP from textbooks into the U.S. Code of Federal Regulations came about, in part, as a result of a National Academies report (NRC, 1985a) that recommended the adoption of HACCP ". . . universally in food protection programs . . ." and of subsequent, instrumental efforts by the International Commission on Microbiological Specifi- cations for Foods (ICMSF, 1988) and the National Advisory Committee on Microbiological Criteria for Foods (NACMCF, 1998~. Other reports of the National Academies (IOM, 1990, 1991; TOM/NRC, 1998; NRC, 1985a, 1985b) have further endorsed the introduction or expansion of HACCP into the process- ing and inspection of meat, poultry, seafood, and, in general, throughout the food industry. Implementation of HACCP by the food industry has been a slow and at times painful process that still is in progress. To facilitate implementation of HACCP by the food industry and help standardize HACCP training, a coalition of industries and trade organizations in the United States formed the International Meat and Poultry HACCP Alliance in 1994. This group has since endeavored to "train the trainers" by conducting training courses and certifying HACCP trainers who can further train personnel at the processing-plant level. In addition, the

FOOD SAFETY TOOLS 71 International HACCP Alliance has contributed to the development of generic HACCP plans for use by regulatory agencies in facilitating the preparation of specific HACCP plans by food processors. There is also a Seafood HACCP Alliance and a Juice HACCP Alliance. The committee recognizes the multiple technical, financial, and educational efforts made by the food industry to imple- ment HACCP, including the development and adoption of various interventions to enhance the microbiological safety of the food supply often in anticipation of regulations and commends such efforts. National food safety regulatory agencies and international institutions have published procedures for the development and implementation of HACCP plans. Some of these are established national food regulations, such as those mandated by the Food and Drug Administration (FDA) (21 C.F.R. part 114) and the U.S. Department of Agriculture (USDA) (FSIS, 1996), while others, such as the Codex Alimentarius guidelines on HACCP (CAC, 1997), play a central role in inter- national food trade despite the fact that their adoption by Codex Alimentarius member countries is voluntary. There are numerous HACCP training manuals, including a few that are international in nature (WHO, 1999), as well as a wealth of information on HACCP from various sources. An example of these sources is a joint USDA/ FDA website that offers a variety of training materials (USDA/FDA, 2002~. Continued training in HACCP principles to attain proper implementation by industry personnel and consistent interpretation and monitoring of compliance by inspectors from the regulatory agencies is necessary. The Principles of HACCP Unlike the traditional model for food safety assurance that has been used for decades, HACCP does not rely on end-product testing to ensure the safety of food batches, but on continuous control and monitoring of Critical Control Points (CCPs) along the production and processing continuum. It is, therefore, a preven- tive food safety assurance system in that it focuses on ensuring control of known potential hazards before the product reaches the end of the line, as opposed to the traditional corrective system that focuses on examining the final product and determining whether any hazard of concern is present. CCPs, in general, are defined in HACCP language as "those points where loss of control would result in an unsafe food product," and more specifically as "those points where the identified hazardous) may be prevented from entering the food, eliminated from it, or reduced to acceptable levels" (Stevenson and Bernard, 1995~. It is noteworthy, however, that an intrinsic weakness of HACCP is that it does not provide information on what these acceptable levels are or a guide on how to set them. Linkage between public health goals and HACCP, through a developing concept of Food Safety Objectives (described later in this chapter), may enable regulators in the future to define numerical levels of tolerance for

72 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD foodborne hazards in foods at the point of consumption that could be translated into "acceptable levels" at CCPs in food-processing plants. The methodology for developing a HACCP plan calls for the systematic application of seven principles: 1. Hazard analysis 2. Identification of CCPs 3. Establishment of critical control limits for each CCP 4. Establishment of monitoring procedures for each CCP 5. Establishment of corrective actions 6. Establishment of record-keeping procedures 7. Establishment of verification procedures The process begins with the formation of a team that includes plant manage- ment and personnel, as well as individuals who have expertise in foodborne hazards and the particular product and process being used. The team prepares a flow diagram of the production process and physically examines each of its steps in the actual premises where production takes place. Points along the flow dia- gram where the hazard may be prevented, eliminated, or reduced to acceptable levels, and for which a control exists that can be established and monitored, are designated as CCPs. Critical limits are then set for the parameters that can be measured to determine that the control at each CCP is being effectively applied. Monitoring procedures are then established, and corrective actions are predeter- mined to be taken if a loss of control is indicated by a deviation from the critical limits. The HACCP plan, along with records demonstrating that the controls at each CCP have performed successfully and have been continuously monitored during processing, are organized for ease of access by the processor and by inspectors from the regulatory agency charged with ascertaining compliance with the regulations. Finally, internal and external verification procedures are defined to periodically assess the performance of the system and to revise the HACCP plan whenever changes are introduced in the production process that could com- promise the effectiveness of the system. Internal verification procedures may involve such activities as instrument calibration, periodic product testing, and records review, while external verification may involve expert audits and exter- nal product testing. Full compliance with Good Manufacturing Practices (GMPs) and the pre- existence of Standard Operating Procedures for plant sanitation are assumed to be in place when introducing HACCP into a food-processing plant. Therefore, HACCP is not a stand-alone methodology, but part of a larger set of manufactur- ing practices that include these preconditions. In addition, the HACCP plan is specific for each processing plant, processing line, and product manufactured in each line. As a result of discussions held during information-gathering meetings, the committee has been made aware that inappropriate identification of CCPs and

FOOD SAFETY TOOLS 73 inappropriate HACCP plans have caused problems in complying with HACCP regulations. Similarly, the committee recognizes that inconsistency in the inter- pretation and enforcement of HACCP rules between and within regulatory agencies has hampered a smooth transition to the new food-processing inspection model and monitoring of compliance with HACCP rules. HACCP has revolutionized food safety assurance by bringing about a radical change in the roles of regulators and regulated industries regarding food safety responsibilities, as described in Chapter 1. The committee believes that despite some continued disagreements between these sectors and some widely publi- cized failures of the system notwithstanding the balance of progress in food safety after implementation of HACCP in various sectors of the food industry is decidedly favorable and commendable. The committee, therefore, endorses the recommendations made by previous reports of the National Academies (IOM, 1990, 1991; TOM/NRC, 1998; NRC 1985a, 1985b) and strongly recommends that the regulatory agencies continue to introduce and audit the implementation of HACCP in all sectors of the food industry as appropriate. RISK ASSESSMENT Various techniques have been examined for their potential to provide a sci- entific basis for improving public health and to address emerging foodborne diseases. Risk assessment has surfaced as one key method to embark upon these challenges. The use of quantitative and qualitative risk assessments for biological issues has emerged from the use of quantitative risk assessments for chemical and environmental toxicology (Dourson et al., 2001; IFT, 2002; Neubert, 1999; Paustenbach, 2000~. In simple terms, quantitative risk assessment uses math- ematical equations, numerical data, and expert opinion to create a computer simulation of reality. These computer models allow interested individuals to explore various risk-management options. Quantitative risk assessment is useful because it allows risk managers to see the entire situation related to a hazard without being an expert on each one of the component factors. Risk managers can rapidly examine various technical solutions to a problem using computer-based models, while using their expert judgment on the social, political, and economic factors that also influence how policies are perceived. Risk assessment is usually presented as part of the overall risk analysis paradigm, where risk analysis consists of risk assessment, risk communication, and risk management (Figure 3.1) (dose, 2000~. Quantitative risk assessment is a scientific process that addresses the magnitude of the risk and identifies factors that control it. Risk communication is a social and psychological process that promotes dialogue among different affected individuals regarding the risk. Finally, risk management is a process that combines science, politics, economics, and proper timing to arrive at a decision regarding what to do about the risk.

74 \ FIGURE 3.1 Components of a risk analysis. SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD - - Risk / \ Risk Management - ~ Assessment Risk Communication - Differences and Similarities Between Chemical and Microbial Risk Assessment Chemical risk assessment is a relatively mature field compared with that of microbial risk assessment. This is due in part to the requirement for drugs and chemicals to be approved or registered by either FDA or the U.S. Environmental Protection Agency (EPA) prior to human exposure. Rigid guidelines have been established and quantitative approaches to assessing adverse effects in humans have been developed. Despite the differences in maturity, the overall paradigm of chemical risk assessment has remarkable similarities to the emerging practice of microbial risk assessment. A comparison of key differences and similarities may benefit both fields. In both fields, risk assessment is a component of the larger field of risk analysis that also encompasses risk management and risk communication. A variety of diagrams have been used to explain the interaction of these compo- nents, including that shown in Figure 3.1. Chemical (and microbial) risk assess- ments are typically divided into four parts: hazard identification, dose-response assessment (or hazard characterization), exposure assessment, and risk character- ization (Lammerding and Paoli, 1997; Neubert, 1999; Paustenbach, 2000~.

FOOD SAFETY TOOLS Hazard Identification 75 Hazard identification involves assessing whether the agent (chemical or microbial) produces adverse effects in biological systems. Historically, this was assessed for chemicals through the use of animal bioassay screens, but now it is largely accomplished using in vitro systems and, recently, by techniques targeted to advances in genomic sciences. Microbial risk assessments are typically initi- ated in response to a public health concern, and hazard characterization in micro- bial risk assessment typically uses epidemiological or outbreak data (Escherichia cold 0157:H7 Risk Assessment Team, 2001; Salmonella Enteritidis Risk Assess- ment Team, 1998~. The hazard characterization step in microbial risk assessment includes iden- tifying the organism that caused the public health concern and summarizing the details regarding the exposure pathway and the microbial ecology of the particular hazard (see Chapter 2~. Dose-Response Assessment Once an agent is identified as potentially injurious, the next phase is to define the dose-response relationship. The techniques for chemical dose-response assessments are well defined, while the same cannot be said for their microbial counterparts. Studies conducted in laboratory animals form the basis of the field of toxi- cology and are readily used in chemical risk assessment. There is an extensive experimental database of well-designed laboratory animal studies, all conducted under agreed upon Good Laboratory Practice (GLP) guidelines (40 C.F.R. §160.1~. GLP guidelines ensure that all tests conducted for regulatory action on a drug or for chemical registration are conducted according to acceptable practices and generate an auditable paper trail. The validity of this approach to chemical risk assessment has a proven track record: FDA uses essentially these same techniques in preclinical studies of human drugs. The determination of dose for a human drug is based on knowledge of the dose-response relationship for both beneficial and adverse effects. The extensive pro- and postmarketing drug approval process validates the accuracy of these approaches. Tolerances for man-made chemicals introduced into the food supply are based on extrapolation of no-effect data from laboratory animal studies. Experi- ences with FDA drug approval would indirectly support the validity of this approach, as stated above. Microbial risk assessment is qualitatively quite different, for microbial hazards are not man-made and usually are introduced into the food supply only naturally or accidentally. Because of the host-pathogen specificity differences, animal studies are of only limited use in microbial risk assessment. Additionally, no microbial equivalent of the FDA human-drug approval process

76 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD exists to validate any proposed dose-response relationships, although if properly collected, outbreak data may help in this regard. Experimental designs in chemical risk assessment are specific for different toxicological endpoints (e.g., acute, subacute, chronic, reproductive, carcino- genic). The mathematical form of the dose-response relationship is assessed based on the biological mechanism of action of the chemical being studied. The end result is a definition of a dose that does not produce adverse effects in laboratory animals: the no-observed-adverse-effect level (NOAEL). There are many variations on how this is determined and on how data from multiple studies are combined (Neubert, 1999~. However, for the purpose of this discussion, the key point is that in chemical risk assessment, the end product (derived from standard toxicological testing protocols) is a defined dose considered safe by the scientific community. Microbial dose-response relationships have been derived from human feed- ing trials (many done on volunteer prisoners in the early part of the twentieth century), animal studies, and, increasingly, data from foodborne disease out- breaks, as noted. As with chemical risk assessment, various endpoints can be used, ranging from mild diarrhea to death; also, data from multiple studies can be combined (Holcomb et al., 1999~. A variety of mathematical forms for microbial dose-response has been proposed. Microbial dose-response equations do not have as clear a link to a biological mechanism as in chemical risk assessment, due in part to the complexity of the underlying biology. The committee believes that defining microbial dose-response relationships for foodborne pathogens is important if more accurate risk assessment results are desired. Allocation of resources to fund basic research studies defining these relationships would help to remedy this deficiency. The host side of the dose-response relationship may also be different for microbial and chemical risk assessments. Some researchers have suggested that in the case of microbial risk assessment, a population's response to an infectious pathogen is more variable than it is to acutely toxic chemicals and rivals the complexity seen with carcinogens. This variability is due to altering immune status as a function of genetics, environment, age, concurrent diseases, and a host of other factors (ICMSF, 1998~. However, the response of an individual to a chemical exposure is also variable based on many of the same factors and indi- vidual differences in the inherent receptor sensitivity, pharmacokinetics (includ- ing metabolism), and simultaneous exposure to a myriad of drugs and chemicals. In both scenarios, the large degree of interindividual variability makes the risk assessment process prone to large degrees of uncertainty. In the drug arena, the development of population pharmacokinetic tech- niques has partially reduced this uncertainty by identifying subpopulations that vary significantly from the norm. Perhaps the most important difference is that microbial dose-response assessment for infectious pathogens does not produce any concept analogous to the NOAEL, since a single microbial cell may (under

FOOD SAFETY TOOLS 77 the right circumstances) produce illness. It may, however, be possible to use a risk assessment term analogous to the NOAEL for organisms like Staphylococ- cus aureus or Bacillus cereus that cause illness through formation of a toxin in the food, or for Listeria monocytogenes in healthy adults. Because microbial dose-response assessment does not typically produce a NOAEL, the key point in microbial risk assessment is that for many pathogens there is no safe dose. Even if a microbial NOAEL could be determined, it might not be adopted. USDA's Food Safety and Inspection Service (FSIS) has taken the position with respect to Escherichia cold 0157:H7 that it is an adulterant, and hence, it is not allowed in raw ground beef in any number (see Chapter 4~. While the agency could change its position in this regard, it might be difficult to explain such a change to the public, and so it might hesitate to do so. If a firm scientific basis for determining no-effect levels for some pathogens existed, along with appropriate detection and enumeration methods to ensure that microbial NOAELs are not exceeded, it would still be necessary to convince the public that their safety would be suffi- ciently assured by the implementation of the microbial NOAELs. Exposure Assessment The next step in either microbial or chemical risk assessment is to estimate human exposure to the agent. For chemicals such as pesticides, environmental compounds, and food additives, potential modes of exposure must be assessed. These include assessing whether the primary routes are inhalation, dermal, or, in the case of food chemicals or microorganisms, oral. Aggregate exposure must be determined where multiple routes may contribute to human exposure. This often occurs in the case of pesticides, where exposure may occur by inhalation after spraying in a home or place of work, orally in food, or dermally by physical contact with a sprayed surface. For chemicals, a major task of exposure assess- ment is to determine the fraction of the dose that is actually absorbed into the body, that is, the bioavailability. Additionally, it is important to determine if this absorbed dose is metabolized, either to an inactive moiety or to an active and potentially toxic metabolite. An arena where risk assessment is routinely applied to chemicals is in the drug approval process. Pharmaceutical drugs are somewhat different in this respect than other chemicals because hazard characterizations and dose-response assessments are conducted in the preclinical phases of drug development in order to estimate a tolerable dose for humans. Hazard identifications for pharmaceuti- cals are essentially validated in the first phase of human testing. The appropriate dose is finally determined after the second and third phases of human testing, which seek to determine effectiveness and obtain additional safety information. When the drug's sponsor applies to FDA for approval of its application to market the drug, a determination is made on whether it is safe and effective and may be released to the marketplace. The approval process necessitates balancing the

78 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD potential benefits of the drug to the patient population against the risks that it might pose. Through the initial testing or postmarket surveillance, information may arise that suggests that certain specific patient populations are more at risk than others for adverse effects or treatment failures; such information may be reflected in labeling information that guides proper drug use. If information is developed later that changes the risk/benefit ratio significantly, FDA may require that the drug be withdrawn from the market. Exposure assessment in quantitative microbial risk assessment (QMRA) involves modeling movement of the pathogen through the production system. Both temporal (in time) and spatial (in space) exposure data are relevant to this step. Exposure assessment results in an estimate of the likelihood of pathogen ingestion by the consumer. Exposure assessment for microorganisms is quite different from that for drugs or other chemicals, primarily because (at least with bacterial pathogens) some microorganisms can increase or decrease in number in the food under suitable conditions. Aggregate exposure to multiple chemicals is often consid- ered, especially with carcinogens. Although each chemical exposure to an indi- vidual in a given time period might not produce illness, such exposures may produce subclinical organ damage, induce metabolic changes, or result in accu- mulation that could modify subsequent responses. In contrast, if repetitive expo- sure to low levels of infectious microbes occurs, host immunity may decrease risk (ICMSF, 1998), but counterexamples also exist (Maijala et al., 2001~. Unlike a chemical that has a constant potency (unless degraded), a microbe is dynamic and adaptable. Virulence factors acquired from other organisms could change the inherent infectivity and pathogenicity of a foodborne microorganism (ICMSF, 1998~. In food-processing operations that combine raw materials from multiple sources, microbial or chemical contamination in some of these raw materials would have differing effects on contamination in the resulting product. While a chemical contaminant would be diluted during mixing, similar dilution of bacte- rial contaminants would mean that the bacteria are spread throughout the mix (e.g., by breakup of microbial colonies that initially may be highly localized into what is referred to as "point source" or "hot spots" in the incoming raw material). For example, consider the mixing of meat trimmings in a grinding operation where a point source of either a chemical or a bacterial pathogen occurs. Dilution of the chemical from a point source to a larger mass of product would be expected to reduce the hazard by decreasing the concentration of the chemical a consumer would ingest. In the case of bacteria, mixing meat trimmings from multiple sources (animals, producers, packing plants, states, countries) would increase the volume of contaminated ground product and, because of bacterial growth, the potential number of consumers that might be affected. The spread of bacterial contaminants would also seriously confound attempts to trace back the source of contamination to a specific supplier of raw material.

FOOD SAFETY TOOLS 79 This effect is well known in the dairy industry, where milk that contains antibiotic residues from an individual cow will be diluted in the tank truck after mixing with milk containing no antibiotic residues. Thus, because of similar dilution effects, ground-meat products would be expected to raise no major concerns regarding chemical residues; but, unlike the situation in whole-muscle meat, chemical hot spots would likely be spread in ground meat. Therefore, the microbiological risk in ground meat may be expected to be greater than any chemical risk. The same logic could be extended to processing food from multiple sources or to consump- tion of a contaminated item in a multi-ingredient meal (e.g., vegetables, meat, and sauces). There are also some differences in the analytical detection of microbes versus chemicals that may impact data used in exposure assessment calculations. Concerns about sampling strategies are fairly similar for both chemicals and microbes, although the latter may be more prone to localization from hotspots of point- source microbial contamination. In the chemical residue arena, the development of multiple drug-class residue screening assays that would detect and quantify multiple contaminants in a single assay has been the focus of recent research efforts. Once considered cost prohibi- tive, these techniques are based on gas chromatography/mass spectrometry and are now feasible. Similar developments have begun to occur in the microbiological arena (see Chapter 1~. A similarity between chemicals and microbial pathogens is that all chemicals and pathogens do not have, qualitatively or quantitatively, the same propensity for causing human illness. Chemicals may exert a number of different types of toxicological reactions, including allergenicity, immunotoxicity, mutagenicity, carcinogenicity, and "classic" chemical toxicity (renal, hepatic, etc.) seen with many pesticides and drugs. A single chemical may exhibit the full spectrum of effects depending on the dose and length of exposure. Quantitative structure- activity relationships have also been developed that help in the prediction of these chemical effects. For microbes, a similar diverse spectrum of potential adverse effects can be observed depending on the species, serotype, strain, or host differ- ences. For example, ingestion of foods contaminated with some strains of E. cold may produce a transient gastrointestinal disturbance, while exposure to strains such as 0157:H7 may be fatal for some individuals. Finally, detection of a chemical allows one to estimate whether the sample exceeds tolerance. Tech- niques such as polymerase chain reaction (PCR), which amplifies deoxyribo- nucleic acid (DNA), can detect and in some cases can also quantify patho- gens (Hein et al., 2001a, 2001b; Li and Drake, 2001~. However, rapid tests that determine microbial viability and infectivity are just becoming available (see Chapter 1~. The issue of multiple points of contamination within a food-processing estab- lishment is also different for some chemical classes versus microorganisms because of the ability of some of the latter (e.g., bacteria, molds) to multiply and

80 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD cross-contaminate. Antibiotics or pesticides that occur either in animal or plant products will not likely result in cross-contamination in a processing plant. Control of the raw product at the producer or harvest level is essential. Approved chemical or drug tolerances in meat or produce serve as effective performance standards to control these hazards. However, the same cannot be said for microbial contamination because bacteria can be transferred from one to other parts of a production line. Data on microbial cross-contamination rates suitable for quantitative risk assessment are only now becoming available. Precise localization of where such cross- contamination occurs would require multiple sampling points in the production system. The committee calls on USDA and FDA to undertake or fund studies on food-pathogen combinations for which insufficient knowledge has prevented intervention to characterize the points in the production system where control would be most effective and could have the greatest impact on reducing foodborne disease. Such information is essential in the application of appropriate controls at critical points and for the development of future microbiological criteria for foods. Risk Characterization The risk characterization phase of a chemical risk assessment differs depend- ing on the type of chemical involved and on the regulatory agency that has jurisdiction (e.g., EPA vs. FDA for an animal drug). However, all chemical risk characterization approaches are conceptually similar, and can be quite different from microbial risk characterizations. Chemical risk characterization involves determining the dose of a chemical that is essentially not harmful to humans, based on the dose-response data from laboratory animal studies and exposure assessments. In contrast, most microbial risk assessments have been undertaken with full knowledge that a particular pathogen is harmful. Microbial risk charac- terization involves estimating the risk to the consumer population (or in some cases a subset of the consumer population) and prioritizing effective control strategies. Chemical risk characterization is used to determine some "risk value," which is a point on a dose-response curve with some probability of occurrence. Data such as the NOAEL or a benchmark dose from laboratory animal studies are reduced to adjust for uncertainty (e.g., species to species extrapolation, experi- mental shortfalls, increased sensitivity of the young) through the use of safety or uncertainty factors ranging from 100 to 1,000. For many pesticides and environ- mental compounds, the result is a reference dose or reference concentration. For a drug used in food-producing animals, an allowable daily intake is computed. Alternate endpoints, such as those related to allergenicity or inducement of microbial resistance, may be employed. The potential amount of food consump-

FOOD SAFETY TOOLS 81 tion is then estimated and the allowable daily intake or reference dose is parti- tioned across all food items to arrive at a tolerance or a maximum contaminant level goal below which food consumption or exposure is assumed to be safe. In the European Union and in the Codex Alimentarius, a similar process is used to calculate a maximum residue level. These are all variants of a theme of accept- able exposure or tolerable intake. Recent work has attempted to directly deter- mine these endpoints using human data that would eliminate the uncertainty of interspecies extrapolations. A threshold of toxicological concern approach that uses a threshold based on chemical structure-activity relationships in an attempt to integrate all adverse effects has recently been proposed (Kroes and Kozianowski, 2002~. If the compound is a potential carcinogen, the allowable concentration in food may be restricted to that which can be detected analytically using the most sensitive method. Finally, when the exposure is widespread, the question is often related to estimating the risk to the human population from this ubiquitous exposure (e.g., dioxin, mercury). In this case, exposure and the dose- response data are used to estimate risk to the human population of exposure to specific concentrations, which are then employed in remediation and risk- management strategies to reduce exposures to an acceptable level (Dourson et al., 2001). Microbial risk characterization is not as well defined as its chemical counter- part. The goal of finding a risk value endpoint is similar and, in some cases, the methods by which this is obtained are also similar. In the absence of human- or animal-feeding models, a number of dose-response models based on epidemio- logical data, animal studies, expert opinions, or combinations thereof are evalu- ated to determine an endpoint or risk value. The highly variable nature of the microbial dose and the human response, as well as the fact that each model is based on different biological endpoints, make it extremely difficult to find one model that fits every situation. For example, the Joint FAD/WHO Expert Consul- tation on Risk Assessment of Microbiological Hazards in Foods suggests that it is not possible to endorse a single dose-response model for L. monocytogenes in ready-to-eat foods (FAD/WHO, 2000~. A variety of data gaps have been identified that must be addressed before microbial risk characterization will be as effective as chemical risk characteriza- tion. As more accurate dose-response models become available, it should be possible to identify the risk-value endpoint needed to achieve a desired public health outcome. Compliance with chemical residue tolerances in meat, poultry, and eggs in the United States is monitored through the FSIS National Residue Program (FSIS, 1999~. This dynamic residue surveillance program monitors domestic, as well as imported, food-animal carcass and egg products for a number of drug, pesticide, and environmental residues. This surveillance, based on a random statistical sam- pling protocol for a list of target drugs determined by a multidisciplinary, inter- agency working group, is designed to assess prevalence and define areas that

82 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD need further attention. In addition, the National Residue Program undertakes a number of special projects to target specific residue concerns. Samples for these programs are usually collected from healthy animals to provide surveillance data. Because the surveillance sampling is conducted to develop databases for future reference and the product is not traceable, if a violating residue is found, recall does not occur. The final component of the National Residue Program is enforcement test- ing, where samples are collected from individual animals or lots that appear suspicious to FSIS inspectors. This program is also used to follow-up on produc- ers who have a history of violations or to verify HACCP performance. Violative products detected using this system are removed from the food supply because they are considered adulterated. If the product has been distributed into com- merce, it may be subject to market recall. It should be noted that the analytical techniques used for these programs are not the same. Enforcement testing may use rapid screening methods that, if positive, force the carcass to be held until confirmatory tests are conducted at an approved laboratory. FSIS maintains a record of such violations in its Residue Violation Information System that is shared with FDA for follow-up investigations. The results of these investigations are then stored in the Tissue Residue Information Management System (Paige and Fell, 1997~. No data from a system analogous to the National Residue Program exist for use in microbial risk assessments. Typically, the results of each microbial risk assessment are validated based on a comparison with current Centers for Disease Control and Prevention estimates for the pathogen of interest. The National Resi- due Program may represent a useful working model on which a national pathogen system could be based. Just as the National Residue Program can be used to validate chemical risk assessments, such a national pathogen program would be invaluable in validating microbial risk assessments. The strength of the chemical risk assessment approach is that there is a defined process whereby an acceptable exposure or tolerable intake of a chemi- cal, based on a public health endpoint, can be defined and calculated from either experimental animal or human data. A specific dose-response relationship is defined for the chemical and adverse effect being modeled. In food safety appli- cations, this allows definition of a tolerance below which lifetime human expo- sure is not deemed to be of concern to public health. In a HACCP environment, this tolerance can be directly employed as a performance standard (Taylor, 2002~. Microbial risk assessment currently suffers from a lack of a standardized process and from a perception that such a process would be expensive and very time consuming. The form of the dose-response relationship is not known and thus is difficult to quantify. Microbial risk assessment is also hampered by the infectious nature of microorganisms, such that some exposure almost always poses some risk. The current level of exposure of the population to a pathogen may be tolerated by most of the population because most people do not experi-

FOOD SAFETY TOOLS 83 once adverse consequences from the foods they consume every day. In light of current morbidity and mortality statistics, however, the level of exposure should be less than it is today. Microbial risk assessment may provide the tools needed to help identify the most effective solutions for lowering consumer exposure to foodborne microbio- logical hazards. In fact, this is the philosophy behind setting microbiological performance standards as a percentage reduction of baseline data that should reduce overall levels of microbial contamination. From the above discussion, it is clear that QMRA can benefit from accom- plishments in chemical quantitative risk assessments in that the lessons learned from the latter can be applied to the new challenges of developing the former. Risk assessment offers a systematic approach to estimating the impact of patho- genic microorganisms in the food chain. In this way, risk assessment may assist public health decision-making and thus help improve overall public health by reducing the burden of foodborne illness. Dealing with Microbial Risk Assessment Data Gaps Several areas where data gaps exist in current microbial risk assessments have been identified by various groups studying this technique (Cassin et al., 1998; FAD/WHO, 2001; IOM, 2002; Whiting and Buchanan, 1997b). During hazard identification, gaps in data can significantly impact the resulting risk assessment. These gaps include, but are not limited to, microorganism variability regarding pathogenicity and infectivity in human hosts; variability of human hosts' susceptibility to illness; complete epidemiological data from outbreak studies, including organism dose and environmental factors of both organism and host; and data on the prevalence of pathogenic microorganisms throughout the food chain. Exposure-assessment data gaps, in turn, include information on routes of animal infection; prevalence in animal groups (e.g., flocks); dynamics of within- animal group transmission of organisms; microbial stress adaptation; and cross- contamination within the production, processing, and consumption segments of the food chain. There are also data gaps in dose-response assessment. These include data on the number of cells of particular microorganisms required to constitute an infec- tive dose, as well as detailed information concerning the dose and the correspond- ing response of human hosts who are infected. Finally, risk characterization data gaps include association of risk with human health effects, identification of potential risk mitigation strategies, and costs and benefits of mitigation strategies once the strategies are identified. Some of the information listed above is available for a few microorganisms, whereas for others the data gaps are more significant. Nevertheless, despite these data gaps, there have been and will continue to be advances in the development of

84 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD microbial risk assessments in foods (Cassin et al., 1998; FAD/WHO, 2001; Whiting and Buchanan, 1997a). Each new risk assessment adds to the informa- tion already in place and increases our understanding of the issues, while further defining what information is still lacking. The list of identified data gaps available at the completion of a microbial risk assessment can assist government and industry in targeting funds to generate missing information. If data are not available for part of a food production chain, it may be possible to simplify the QMRA model such that this part of the chain is excluded. For example, if data on prevalence of a particular pathogen in a live food-animal population were not available, a QMRA could be constructed such that the start of the process was postslaughter. This assumes, of course, that pathogen prevalence and concentration data are available for the carcasses. If a QMRA were constructed in this way, important factors that affect pathogen prevalence and concentration in the live animal population obviously would be accounted for in the final assessment results. Predictive models for the growth and inactivation of pathogens as influenced by environmental conditions have gained increased visibility in the last decade (Whiting and Buchanan, 1997b). If information on the behavior of a pathogen in a particular part of the food chain is not available, and a predictive model exists that could represent that part of the chain, then model predictions, rather than actual data, could be used. For example, data are seldom available on the levels of pathogens in a food just prior to consumption, but if data are available from an earlier part of the chain, and temperature and food composition data are available, predictive models could be used to estimate pathogen levels just prior to con- sumption. Limitations of predictive models include the use of models that have not been fully validated and a lack of information on prediction uncertainty. It may be possible to use surrogate data if neither actual data nor predictive models are available. Surrogate data are data from a related organism that experts believe to be "close enough" to the unknown behavior of the actual pathogen to stand in its place. Examples might be the use of cross-contamination data for generic E. cold as a surrogate for E. cold 0157:H7 cross-contamination and the use of dose-response data on Shigella dysenteriae as a surrogate for E. cold 0157:H7 (Escherichia cold 0157:H7 Risk Assessment Team, 2001; IOM,2002~. Data gaps may not mean just the lack of a point estimate (e.g., mean, mode, or median), but also a lack of knowledge regarding the uncertainty and/or vari- ability associated with the point estimate. The amount of effort needed to ade- quately fill a data gap either by combining data from a multitude of sources or conducting original research can make the elimination of data gaps a long process. Another method to reduce and eliminate some of the existing data gaps in QMRAs could be stochastic simulation using probabilistic distributions to replace the data-gap information. In published risk assessments, probability distributions have been used to estimate the parameters associated with various parts of a QMRA, for example, the dose-response curve (Cassin et al., 1998; FAD/WHO,

FOOD SAFETY TOOLS 85 2001; Whiting and Buchanan, 1997a). It follows that in places where data gaps exist, probabilistic models could be useful in providing information that helps to fill the data gap. In order to accomplish this, one of two conditions would need to be met. One requires the modeler to make an assumption about the shape of the probability distribution from estimates based on somewhat qualitative previous experience or other more quantitative data (FAD/WHO, 2001~. The other condi- tion relies on the use of probability distributions where variance which arises from both uncertainty and variability is large (e.g., exponential or beta distribu- tions) to accommodate for the unknown information in the data gap. If either of these conditions were met, then the use of a probability distribution would be a valid method to fill a data gap. Some data gaps can be filled through the use of expert opinions and consults (sometimes referred to as qualitative risk assessment) (IFT, 2002~. Some oppo- nents of using qualitative risk assessment as a component of a QMRA state that the former dilutes the latter's effectiveness, scientific basis, and end use of the resulting risk estimate. However, without the use of these qualitative expert consults, it is likely that some of these data gaps would continue to exist for some time. Waiting for "hard" scientific data would postpone the development of QMRAs that could be instrumental and effective in public health decision-making despite their qualitative or "soft" expert opinion content. Those involved in quali- tative consults often have a qualitative feel for the data needed that is based on previous experience that has a foundation in quantitative research (Busta, 2002; IFT, 2002~. Therefore, to include qualitative information from expert consults in a QMRA where data gaps exist and are difficult to fill seems both reasonable and scientifically sound. It should be noted that it is best to use standardized methods for eliciting expert opinion to enhance transparency and avoid introducing any potential bias into the process, and that techniques are available for pooling different opinions from a range of experts (dose, 2000~. As noted above, most QMRAs will have data gaps. These data gaps should not prevent a risk assessment from being initiated and completed and from serv- ing a useful purpose. However, these data gaps must be communicated to those requesting the QMRA, so that they will be aware of its limitations. The inherently iterative nature of risk assessments allows continual updating as more and better- quality data become available, thereby increasing their effectiveness as a qualita- tive tool for policy-making. Using Microbial Risk Assessment as a Policy Tool Each of the large QMRAs commissioned by the United States has been initiated with the objective of guiding policy. Table 3.1 provides the relevant quotation from each of these risk assessments. Since the field of microbial risk assessment as applied to food is relatively new, there are few case histories that detail how QMRA can successfully impact

86 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD TABLE 3.1 Quantitative Microbial Risk Assessments Commissioned by the United States Government and Their Policy-Guiding Objectives Risk Assessment Objective Reference Salmonella "The risk assessment results detailed in this Final Salmonella Enteritidis Risk Report will be used by the agency, working in Enteritidis Assessment. conjunction with economists from within and from Risk Shell Eggs and outside the agency, to conduct cost-effectiveness Assessment Egg Products studies and cost-benefit analysis in order to set forth Team, recommendations for policy." 1998 Preliminary Pathways "The baseline risk assessment is intended to inform E. cold and Data for a a distinct FSIS policy analysis that will identify Risk Risk Assessment feasible risk mitigation options for further Assessment of Escherichia cold comparative analysis." Team, 0157:H7 in Beef 2001 Draft Assessment of "The scientific evaluations and the mathematical CFSAN/ the Relative Risk to models developed during the risk assessment, provide FSIS/CDC, Public Health from a systematic assessment of the scientific knowledge 2001 Foodborne Listeria needed to assist both in reviewing the effectiveness monocytogenes of current policies, programs, and practices, and Among Selected new strategies to minimize the public health impact Categories of of foodborne L. monocytogenes." Ready-to-Eat Foods Draft Risk "FDA anticipates that periodic updates to the risk Posnick Assessment on the model will continue to reduce the degree of et al., Public Health uncertainty associated with risk estimates, and that 2001 Impact of Vibrio these updates will assist FDA in making the best parahaemolyticus possible decisions and policies for reducing the in Raw Molluscan risk posed by V. parahaemolyticus in raw Shellfish molluscan shellfish." The Human Health "The modeling approach we have used has been CVM, Impact of designed to address the effect of specific risk 2001 Fluoroquinolone management actions [i.e. policies], while also Resistant providing the facility to take into account the effect Campylobacter of the most important future changes in the Attributed to the physical system . . ." Consumption of Chicken continued

FOOD SAFETY TOOLS TABLE 3.1 Continued 87 Risk Assessment Objective Reference Draft FSIS Risk "By changing in-plant practices such as the frequency Gallagher Assessment for of testing and sanitation of food contact surfaces, the et al., Listeria in effectiveness of pre- and post-packaging interventions, 2003 Ready-to-Eat the effectiveness of growth inhibitors, effectiveness of Meat and enhanced sanitation, etc., including combinations, this Poultry Products risk assessment can provide numerous outputs to address specific risk management questions. This risk assessment model was also developed with user-friendly interfaces to allow users to change scenario conditions and assumptions. As a result, this risk assessment model can be used as a tool to explore a variety of risk management scenarios beyond those developed for this report." policy-making. In a few short years, QMRA has become the new way of organiz- ing and interpreting data to enhance food safety. The definitive example of a "full-blown" QMRA for the U.S. food supply was the USDA Salmonella Enteritidis risk assessment for shell eggs and egg products (Salmonella Enteritidis Risk Assessment Team, 1998), although an example from Canada was published earlier the same year (Cassin et al., 1998), and an example from water microbiology predates these by several years (Rose et al., 1991~. The Salmonella Enteritidis Risk Assessment (SERA): Shell Eggs and Egg Products was the first of the major government-commissioned QMRAs, so it has the longest history that can be used to track any possible policy impact. Follow- ing the publication of the SERA in 1998, the President's Council on Food Safety (1999) published the document Egg Safety from Production to Consumption: An Action Plan to Eliminate Salmonella Enteritidis Illnesses Due to Eggs. The action plan attributes the FSIS final rule on shell eggs storage, transportation, and con- sumer labeling (Salmonella Enteritidis Risk Assessment Team, 1998) and the FDA proposed rule for shell egg safe handling statements and retail refrigeration requirements (FDA, 1999a) to the SERA. The action plan also states that the SERA predicts that multiple interventions could achieve a more substantial reduc- tion in S. Enteritidis illnesses than could any one intervention alone, and then goes on to lay out such a broad-based policy approach. Finally, the action plan also indicates that the research needs identified in the SERA have been incorpo- rated into the plan.

88 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD The policy implications of the Draft Assessment of the Relative Risk to Public Health from Foodborne Listeria monocytogenes Among Selected Categories of Ready-to-Eat Foods (CFSANIFSIS/CDC, 2001) were laid out in a U.S. Depart- ment of Health and Human Services-USDA Joint Action Plan (FDAIFSIS, 2001) based on, and released concurrently with, the risk assessment (FDAIFSIS,2001~. This plan includes a number of areas for policy change, including redirection of enforcement and microbial product sampling strategies; proposal of new regula- tions and revisions to existing regulations; and support of additional research on exposure assessment, treatment strategies, safety-related date marking, and im- proved detection and quantification (FDAIFSIS, 2001~. Internationally, the World Health Organization/Food and Agriculture Orga- nization of the United Nations has led the microbial risk assessment effort and has many projects underway, including risk assessment for Salmonella spp. in eggs and broilers, L. monocytogenes in ready-to-eat foods, Vibrio spp. in seafood, and Campylobacter spp. in broiler chickens. Various European countries have also developed risk assessments suited to various products, pathogens, and pro- cessing systems. Plans are underway to catalog and index European risk assess- ments through a European concerted research effort known as COST Action 920 (COST Action 920, 2002~. Clearly, each of these major risk assessments was undertaken to help make sound policy. In some cases, policy decisions have been made or proposed in the United States that are based on QMRA results. If one considers the pace with which QMRAs are being conducted around the world, the next decade should provide some interesting examples of their impact on the promulgation of sound science-based food safety policies. FOOD SAFETY OBJECTIVES Food Safety Advances with No Quantitative Measure of Impact on Public Health Historically the major advances in consumer protection have resulted from the development and implementation of selected, targeted control measures at one or more steps along the food continuum. However, more often than not, the goal of such control measures has not been expressed in a numerical value (e.g., a specified reduction in the prevalence of a particular foodborne infection), or the relationship between hazard and risk has not been determined. This does not mean that control measures cannot be taken. Some examples of measures that might result in safer food without quantitative performance criteria include binomial slaughtering, where pathogen-free herds are slaughtered before those that are infected to prevent cross-contamination; vaccination programs to prevent infection in animals; and consumer information programs that target high-risk populations.

FOOD SAFETY TOOLS 89 Efficient communication to all stakeholders of the reason for, and expected outcomes of, food safety control measures has been an important aspect of the acceptance of the measures. Any food safety criterion, the effectiveness of which is not readily observable, should be coupled with some sort of verification measure to ensure that the criterion actually has an effect. The Need for Regulatory Flexibility Because the pace of the regulatory process seldom matches that of innova- tion and scientific advancement, regulatory policies should ideally be designed with this understanding in mind. Good science-based policies should allow flex- ibility and encourage innovation, with minimal regulatory revisions. This implies a regulatory framework that specifies results, but not the methods used to achieve these results. It also implies a flexible, moving "results target" that can be easily changed in response to changing public health goals. Food Safety Objectives and Traditional Microbiological Criteria One approach that could provide this changeable regulatory structure is that of Food Safety Objectives (FSOs). FSOs are also important because they allow translation of public health goals (e.g., reduce the incidence of foodborne disease x by 50 percent in a specified period of time) into measurements that food processors are directly able to effect (e.g., ensure that no more than y cells per gram of the microorganism causing foodborne disease x are present in product z at the time of consumption). This is a novel approach that may allow regulators to address the inherent weakness of HACCP, that defines a CCP as any point, stage, or step along the food production and processing chain where a hazard can be prevented, eliminated, or reduced to an acceptable level, but it leaves the acceptable level undefined. An FSO provides the basis for defining this level. An FSO is a statement of the maximum frequency or concentration of a microbiological hazard in a food at the time of consumption that provides the appropriate level of protection (ICMSF, 2002~. FSOs are specified at the point of consumption, and they provide flexibility to food processors because various means of meeting an FSO may be practical and available for the same product. FSOs are quantitative and verifiable, are limited to food safety, and do not address concerns for quality. Regulatory agencies could use FSOs to define the level of control of a hazard expected in a food product at the time of consumption. They could also be used to subsequently evaluate the adequacy of a facility's control system to achieve the FSO given all the relevant assumptions about transporta- tion and retail and consumer handling of the product. FSOs differ from the microbiological criteria that have been traditionally used to determine the acceptance of food products. Traditional microbiological criteria specify details such as a sampling plan and the method of sample prepa-

9o SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD ration and analysis, whereas FSOs do not prescribe a particular analytical method. Microbiological criteria are typically used to determine the safety or quality of a batch of food products, and as such provide a snapshot limited to the time the food was produced, but are not typically used in such a way as to provide infor- mation on process stability and capability. A review of any individual plant's food safety management system using an FSO approach could provide an assess- ment of long-term control. How Are Food Safety Objectives Established? Regulatory agencies may find that FSOs represent a useful concept for estab- lishing a theoretical framework to relate performance standards to public health objectives. Conceptually, an FSO would be established on the basis of a quantita- tive risk assessment of the hazard of interest and would be consistent with the level of consumer protection that the regulatory agency deems appropriate to fulfill the public health objective. The reasoning followed in setting the FSO would be: no more than x mg/kg (chemical hazard) or no more than y cfu/g (microbial hazard) can be present in a given food product at the time of consump- tion to keep the number of illnesses attributable to the hazard below the preset public health objective. From there, the regulatory agency could establish a per- formance standard that would ensure control of the hazard at the processing plant so that the product would be consistent with the FSO when it reached the con- sumer. It would then be the processor's decision what process or combination of processes to apply and what additional parameters (e.g., antimicrobial food addi- tives, packaging, and refrigeration and cooking protocols) to introduce or modify to ensure that the performance standard is met at the processing plant, and through it, that the product meets the FSO at the time of consumption. FSOs offer one practical, if yet unproven, means to convert public health goals into values or targets that can be used by regulatory agencies and industry. For example, a public health goal may be to reduce the incidence of foodborne illness attributed to pathogen a by 50 percent (e.g., from 30 to 15 cases per 100,000 people per year). A regulatory agency or manufacturer could not design a control system that would be certain to meet such a goal. However, if this goal were translated into a numerical measure of the microbial hazard's frequency or concentration at the time of consumption (e.g., less than 100 cfu/g of pathogen a or less than 15 mg/kg of aflatoxin), industry could design control processes at the plant necessary to achieve this FSO and the regulatory agency could then estab- lish inspection procedures at the plant to ensure processes are under control. For newly emerging food safety concerns, however, there may be so little information available that it is difficult or impossible to relate the public health objective to an eventual FSO. In such a situation, qualitative risk assessments and, in some cases, simple dose-response estimates, could be used to set an FSO.

FOOD SAFETY TOOLS 91 In this manner, depending on the urgency or the complexity of the situation, an FSO may be derived from a quantitative risk assessment or from expert opinion. The FSO may be based on a realistic estimate of the risk. However, if time is short, it could also be based on a detailed examination of the frequency or levels of a hazard that can be expected to protect consumers. FSOs should be considered interim standards that could be adjusted to be more or less stringent as more information becomes available. Examples of criteria that are continually updated include the International Organization for Standardization (ISO) standards, which are reviewed every five years. Following review, these standards are accepted, revised, or eliminated (Cianfrani et al., 2002~. Another example is FDA's model Food Code, which is revised every two years by the Conference for Food Protection (FDA, 2002~. FSOs can play an important role in modern food safety management by linking information from the risk assessment processes with measures to control the identified risk. As more information becomes available, risk assessments should be updated and FSOs adjusted accordingly. Thus, the FSO concept may be a useful tool for developing policies that are consistent with current science and could offer an alternative approach to food safety management focusing on the protection of human health, while offering flexibility in achieving that goal. The Food Safety Objective Equation The level of a microbial contaminant in a food at the point of consumption is related to (1) the initial level of that contaminant in the food, (2) the sum total of contaminant reductions occurring up to the point of consumption, and (3) the sum total of contaminant increases up to the point of consumption. A simple equation summarizes the relationship between these three concepts and FSOs: Ho-~;R+~I< FSO Here, FSO = Food Safety Objective, Ho= initial level of the hazard, ); R = cumulative (total) decrease (reduction expressed as positive) in the level of the hazard up to the point of consumption, and ); I= cumulative (total) increase in the level of the hazard up to the point of consumption. It is very important to note that FSO, Ho, R. and I are expressed in logic units, so if the initial level of a hazard is 100 cfu of a microorganism per gram of product, this is represented as Ho = logic 100 = 2. It also should be noted that controlling initial levels, preventing an increase in levels, and reducing levels of the hazard are all important in meeting the FSO, and that increases can occur from growth as well as from recontamination. Hypothetical examples of FSOs are the following:

92 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD . . The level of a potential bacterial pathogen in a certain food must not exceed x cfu/g at the time of consumption. · The concentration of a certain enterotoxin in a certain food must not exceed y ,ug/100 g at the time of consumption. The concentration of a certain mycotoxin in a certain food must not exceed z ,ug/kg at the time of consumption. Integrating Food Safety Objectives into the Food Safety Management System The FSO is a new concept that builds on, rather than replaces, existing food safety terminology and concepts. FSOs have been discussed by a number of countries around the world, and internationally within Codex Alimentarius, spe- cifically by the Codex Committee on Food Hygiene (Woteki, 2000~. ICMSF recently proposed an approach to food safety management that in- volves a series of seven steps that incorporate Codex Alimentarius principles (ICMSF, 2002~. This approach, outlined below, integrates risk assessment and current hazard-management practices into a framework that could be used to achieve public health goals in a science-based, flexible manner. This approach also shows how FSOs relate to many existing food safety concepts: 1. Assemble epidemiological information indicating a need for improved control. 2. Conduct a qualitative or quantitative risk assessment, as appropriate. 3. Assess possible risk-management options, including an appropriate level of protection (ALOP). 4. Establish an FSO. 5. Confirm that the FSO is achievable through Good Hygienic Practices (GHP, GMP in the United States) and HACCP. 6. Establish process/product requirements. 7. Establish acceptance procedures. Food Safety Objectives and Appropriate Level of Protection The FSO concept was first introduced because of the difficulty in using public health goals (e.g., an ALOP) to establish control measures. An FSO is an intermediate step in the conversion of the ALOP into other parameters (i.e., performance standards) that can be controlled by food producers and monitored by government agencies. The ALOP is an expression of a public health risk that is, the achieved or achievable level proposed following consideration of public health impact, technological feasibility, economic implications, and comparison with other risks in everyday life while an FSO expresses the level of a hazard in relation to this risk.

FOOD SAFETY TOOLS 80 60 Q 0 40 ILL ~ o = 0 ~ ° 20 0 ~ Z O -20 93 Food Safety Objective ./ 0.001 0.01 0.1 1 10 100 Potential Pathogen cfu/g FIGURE 3.2 Relating a food safety objective and a hypothetical dose-response curve for a pathogen. A hypothetical dose-response curve for a certain infectious pathogen is shown in Figure 3.2. In this figure, the estimated number of foodborne illness cases per 100,000 individuals increases as the concentration of the causative pathogen in the food exceeds 1 cfu/g. The FSO has been established at 100-fold less than this dose (i.e., 0.01 cfu/g at the time of consumption). This example could be representative for E. cold 0157:H7 in products submitted to heat treat- ment or other processing steps. Food Safety Objectives, Good Manufacturing Practices, Good Agricultural Practices, and HACCP Once an FSO has translated a public health goal into a quantifiable standard, hazard control and monitoring practices must be developed. The ICMSF scheme recognizes that it is most effective to emphasize the design and control of food operations through the application of GHPs (or GMPs in the United States) and HACCP. However, it is important to note that other food safety concepts can be

94 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD combined with this scheme to achieve the desired results in the farm-to-table approach to food safety; for example, implementing good agricultural practices may provide microbiologically safer foods. GMPs, in turn, are important to minimize the hazard and prevent recontami- nation after processing. HACCP manages the application of control methods, ensuring that the process is effective. As mentioned earlier, one of the long-standing limitations of HACCP is that the actual level of hazard control may not be clearly stated in the HACCP plan. Additionally, there is little or no guidance on the level of hazard control expected in an adequately designed and implemented HACCP plan. As is currently done with performance standards, use of the FSO concept could help remedy this problem by clearly indicating the level of control needed for adequate GMP and HACCP systems. Table 3.2 provides examples of how the FSO approach might be used to address a specific microbiological food safety issue. Because the FSO must be met at the time of consumption, but regulatory action must take place at other locations in the food production and distribution chain, it may be necessary to introduce additional terms that represent various microbiological objectives throughout the food-processing chain. These examples might include slaughter safety objectives, processing safety objectives (analogous to the current Salmonella performance standard), transportation safety objectives, or retail safety objectives. For example, if the FSO is less than 100 cfu/g of a certain potential pathogen at the point of consumption and 1 logic cycle of growth is projected during transportation, retail, and home storage, a hypothetical pro- cessing safety objective is calculated as no more than 10 cfu/g of the pathogen. Alternatively, if no growth of the pathogen is projected, the processing safety objective would be the same as the FSO. The processing safety objective can then be used to develop the performance and process/product criteria and to establish verification and acceptance procedures in the HACCP plan. . Food Safety Objectives and Performance Criteria or Standards At certain points in the processing of a food, control measures can be applied to prevent an unacceptable increase in a hazard, eliminate it, or reduce it to an acceptable level. Each CCP must include parameters with defined critical limits. For example, pasteurization of milk at 72°C for 15 seconds inactivates recog- nized pathogens. Similar critical limits would define the degree of hazard control necessary to meet a processing safety objective (i.e., a performance standard) derived from an FSO. Process or product criteria, respectively, would define the process variables or product characteristics that will achieve the performance criteria or standard. Default criteria also play a very important role in the food safety system by providing one or more "safe harbor" sets of criteria (processes) for food operators lacking either the resources or the desire to develop a HACCP plan suited to their specific operation or product. Finally, microbiological criteria

FOOD SAFETY TOOLS TABLE 3.2 Framework for Food Safety Management 95 Action Process Formulate Public Health Goal to establish targets for improvement in the food safety system. Perform Risk Assessments (RAs) to apportion risk across food groups and estimate risk associated with various levels of contamination for specific foods. Establish Food Safety Objectives (FSOs) for specific foods needed to reach public health goals given apportionment of risk across food groups. Establish Transportation and Retail Safety Objectives (TRSOs) for specific foods needed toreach food safety objectives, orto reach public health goals in the absence of FSOs. Establish Processing Safety Objectives (PSOs) for specific foods, needed alone or in combination with available TRSOs to reach food safety objectives, or to reach public health goals in the absence of FSOs. Establish Farm Safety Objectives (FarmSOs) for specific foods, needed alone or in combination with available PSOs and TRSOs to reach food safety objectives, or to reach public health goals in the absence of FSOs. Example: Healthy People 2010 Objectives Reduce infections caused by key foodborne pathogens. Cases per 100~000 1997 Baseline 2010 Target Escherichia cold 0157:H7 2.1 1.0 Salmonella spp. 13.7 6.8 Establish expert panels representing public health, regulatory agencies, industry, and academia to identify issues, available data sources, and knowledge gaps. Assemble scientific teams to conduct qualitative or quantitative RA and to develop surveillance and monitoring plans to address knowledge gaps. Assemble scientific teams to establish FSOs for specific foods at the point of consumption, to develop monitoring plans for compliance with the FSOs, and to identify foods for which FSOs cannot be reasonably formulated due to the nature of the food. Assemble scientific teams to establish TRSOs for specific foods at the point of distribution or retail sale, to develop monitoring plans for compliance with the TRSOs, and to identify foods for which TRSOs cannot be reasonably formulated due to the nature of the food. Assemble scientific teams to establish PSOs for specific foods at the point of processing, to develop monitoring plans for compliance with the PSOs, and to identify foods for which PSOs cannot be reasonably formulated due to the nature of the food. Assemble scientific teams to establish FarmSOs for specific foods at the point of production or harvest, to develop monitoring plans for compliance with the FarmSOs, and to identify foods for which FarmSOs cannot be reasonably formulated due to the nature of the food. NOTE: This framework for food safety management establishes relationships between public health goals and measures or indicators of microbial contamination at each level of the food system from farm to table. The framework recognizes the wide variety of production, processing, and marketing practices that exist for different foods and can accommodate a range of different risk management options. The monitoring plans required for verifying compliance with the various Safety Objectives should be compatible with the development of Hazard Analysis and Critical Control Point systems, and will provide feedback for the periodic re-evaluation of the public health goals and the specific Safety Objectives needed to achieve these goals. SOURCE: Adapted from IFT (2002).

96 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD and testing may be used to further verify that a processing safety objective has been met. Examples Relating Performance Criteria to Food Safety Objectives As many of these concepts are relatively new, there is clearly a need for further discussion relating to the terminology to be used in this area. The follow- ing examples show various ways in which FSOs can be related to performance . . cntena. Example 1 Although FSOs should be quantitative and verifiable, this does not always imply that they must be verified by microbiological testing. For example, an FSO for low-acid canned foods could be established in terms of the probability of a viable spore of C. botulinum being present (< 0.000000001 per can). It is obvi- ously impossible to verify this by end-product testing, and therefore it is done by measuring time/temperature protocols that are based on a performance criterion. Example 2 A performance criterion could be used to limit recontamination and growth of a particular pathogen at any point after processing. Assume that the FSO for a certain potential pathogen in a food product is < 100 cfu/g (see Figure 3.3~. Also assume that the greatest expected concentrations postslaughter and on arrival are both 1 cfu/g. If the heating step produces a 3-log reduction, the greatest expected concentration after heating will be 0.001 cfu/g or 1 cfu/1,000 g. The criterion (limit) for recontamination could be less than 0.1 cfu/g and the limit for growth could be less than a 3-log cfu/g increase, thereby meeting the FSO. Example 3 In this example, the initial bacterial population (Ho) in the raw material is estimated to be as high as 103 cfu/g, but growth (I) can be prevented (e.g., ~ I= 0~. The FSO is 1 cfu/100 g of product. The required performance criterion would be expressed as: Ho - ~ R + ~ I < FSO 3 -I R + O <-2 -I R <-5 Therefore, based on these calculations, the process must result in an overall reduction of greater than or equal to 5 logic (i.e., 5-D reduction) to meet the FSO.

FOOD SAFETY TOOLS 100 10 - ~ 1 o it Do Go 0.1 ,_ ~C,) o 0.01 to 0.001 0.000 1 FSO . <~' I\ ~ blow FIGURE 3.3 Relating a Food Safety Objective (FSO) and a performance criterion. 97 This corresponds to a performance criterion of a 5-D reduction of the pathogen and could be achieved by one control hurdle (measure) or a combination of hurdles. Example 4 A 5-D reduction is currently required for the control of enteric pathogens such as salmonellae and E. cold 0157:H7 for nonshelf-stable juice in the United States. It might be useful to consider what an appropriate FSO for such a product might be. If the initial level of salmonellae or E. cold 0157:H7 could be as high as 100 cfu/mL of juice, then a 5-D reduction step theoretically would result in a level of 0.001 cfu/mL of juice or 0.1 cfu/100 mL of juice (100 mL is an assumed normal serving size). This would not be adequate to ensure the safety of the juice considering the total quantity of juice consumed on a daily basis by a diverse population of consumers, including some who may be at higher risk. The alterna- tives would be either to control the incoming juice to maintain a lower initial pathogen level or to apply a reduction step that would achieve greater than a 5-D reduction. The question then is, "What level of microbial hazard would be considered tolerable for juice?" NACMCF (1997) has suggested that a level of < 1 cfu of

98 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD salmonellae or E. cold 0157:H7 per 10 L of juice (< 1 cfu /10,000 mL or < 0.00001 cfu/mL) would be considered adequate to provide an appropriate level of protection. Returning to the FSO-based scheme for the management of food safety, an FSO for fresh juice could be expressed as: "The level of enteric pathogens, such as salmonellae end E. cold 0157:H7, must not exceed 1 cfu/10 L of juice." This value should be considered when assessing the adequacy of a 5-D process and establishing control measures through the application of GMPs and HACCP. When to Use Food Safety Objectives On an Interim Basis In the case of a new or emerging pathogen, establishment of an interim FSO could be an initial step to communicate to the food industry or to countries exporting food products to the United States the acceptable maximum level of a hazard. As further knowledge about the hazard, the food, and conditions leading to illness become available, and effective control measures can be determined, the interim FSO can be adjusted. To Promote Industry Change In the past, governments have used various mechanisms to bring about the changes necessary to reduce or eliminate the risk of disease. In some cases, modifications in commercial practices are necessary, including the adoption of new or more reliable technologies. These approaches are not inconsistent with the use of FSOs. As is currently done with some performance standards, FSOs also could be used to promote change in an industry and enhance the safety of certain products. Many examples could be cited where epidemiological data have linked certain foods to foodborne illness. Government risk managers could use an FSO to communicate the level of control expected and thereby compel change on the part of the industry. A particular FSO may require some processors to modify their operation, implement more effective technologies, adopt tighter control systems, or even cease operation. Limitations of Food Safety Objectives FSOs are not a panacea, much in the way that HACCP, GMPs, novel pro- cessing technologies, or improved consumer education have not been able to solve all food safety problems. FSOs are simply the latest tool available in a growing food safety toolkit. There may be situations where FSOs are not appro- priate. Such would be the case if the potential microbiological hazards associated with a food represent so little risk that an FSO is not needed (e.g., granulated

FOOD SAFETY TOOLS 99 sugar, most breads, carbonated drinks). In other cases, the sources of a pathogen are so variable that identifying the foods for which FSOs should be set is not possible. An example of the latter is shigellosis, which can be transmitted by many routes, many of which are more important than food (e.g., waterborne, person-to-person). Further, if a particular industry has been operating success- fully for many years without FSOs, their introduction may offer no significant public health advantage. Examples of such industries include the pasteurized milk industry and the low-acid canned food industry. The introduction of FSOs may lead to additional regulatory confusion, as FSOs for different products, developed at different points in time or by different expert groups, are compared. For example, if one set of FSOs were developed from the USDA Salmonella performance standard for raw meat and poultry- which allows some level of contamination while another set of FSOs were developed from the FDA Salmonella performance standard for raw seafood- which does not allow any contamination these two FSOs for the same pathogen in different products would be different. There are also examples of foods recently regulated by performance stan- dards, such as the 5-D process performance standard for fresh juice and the Salmonella performance standard for raw meat and poultry. It is reasonable to expect that both these performance standards have resulted (or will result) in improved public health, even though the interventions at the processing plant- are separated in time and space from the effect at the point of consumption. If these products were to be processed in ways that achieve an FSO, which is by definition at the point of consumption, this would introduce an additional layer of complexity. Consider the following example with two fresh juice producers, both trying to meet a fresh juice FSO of < 1 cfu of salmonellae or E. cold 0157:H7 per 10 L of juice (< 1 cfu/10,000 mL, < 0.00001 cfu/mL, or -4 logic cfu/mL). Juice producer 1: This producer squeezes the juice on site using tree-picked apples. Historical data collected by the processor over a number of years indi- cates that generic E. cold is occasionally present but always at levels of less than 10 cfu/mL (i.e., Ho = logic = 1~. The pH of the juice is always 4.0 or below and he knows from published research that E. cold will not multiply in the juice at any storage temperature. The juicer applies a 5-D (i.e., ~ R = 5) thermal process. Ho-I R +ZI<-4 1 - (5) + 0 < -4 Juice producer 2: This juicer is producing a melon juice with a pH of 6.0. Although the juice is refrigerated, he has data demonstrating that, with tempera- ture abuse (25°C), a maximum increase in Salmonella of 1 logic (i.e., ~ I = 1) prior to spoilage of the product is possible. Historical data collected by the com- pany over a number of years indicates that generic E. cold is occasionally present

100 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD but always at levels of less than 100 cfu/mL (i.e., Ho= logic 100 = 2~. He achieves a 7-log treatment using a 3-log thermal process combined with a 4-log ultraviolet treatment (i.e., ~ R = 3 + 4 = 7~. Ho-I R +ZI< -4 2-~3+4~+ 1 < -4 While the net effect is identical, the additional complexity makes the regula- tory verification of compliance significantly more difficult. The trade-offs between encouraging innovation and managing regulatory complexity will need to be evaluated carefully if FSOs are to be used successfully. Additional limita- tions to the rapid adoption of FSOs include the lack of definitive data on the initial level of the hazard (Ho) for many foods, and the lack of familiarity of many food-processing companies, particularly small- and medium-sized ones, with the FSO concept. Definitive instructions for food processors on what is needed to document achievement of an FSO are also lacking. The current situation regard- ing FSOs might be likened to that of HACCP 10 or 15 years ago. Another limitation is that the measurements required to define whether an FSO is in fact working are rarely obtained directly. In order to validate or verify that a product meets an FSO or that overall progress has been achieved, the FSO needs to be linked to a contamination level in production, such as a processing safety objective, and that is where the level of contamination should be moni- tored. Government enforcement necessarily must focus on compliance at the level of production or retail sale because inspection is not possible literally at the point of consumption. One of the major benefits of the FSO concept is the flexibility it affords to producers to utilize different means of achieving the same ultimate level of food safety at the time of consumption. However, the practical need for government to measure compliance earlier in the product cycle than the point of consumption necessarily limits this flexibility. FSOs may also be problematic because they introduce additional computa- tional complexities and because the databases needed to calculate microbial con- centrations at the point of consumption may not be adequate. For example, it is part of the definition of an FSO that it specifies pathogen concentrations at the point of consumption, yet very little data on pathogen concentrations at this point actually exist. Furthermore, the data on transportation, retail, and home storage and preparation practices needed to estimate pathogen concentration at the point of consumption are extremely limited and variable. Concentration at this point would typically be estimated using the techniques of QMRA described earlier, which introduces uncertainty and variability with every calculation. When all of the sources of variation are included in calculations of pathogen concentrations at the point of consumption, the overall range of possible concentrations can be quite large. Improvements in data quantity and quality may be needed to calculate useful estimates of expected FSOs.

FOOD SAFETY TOOLS 101 In summary, the FSO concept may prove useful to both regulatory authorities and the food industry. FSOs could help to: Translate a public health goal into a measurable level of control upon which food processes can be designed. Validate food-processing operations to ensure that they will meet the expected level of control. Assess the acceptability of a food operation by regulatory authorities or other auditors. · Highlight food safety concerns as separate from quality and other concerns. · Compel change on the part of the food industry to improve the safety of a particular food commodity. Serve as the basis for establishing microbiological criteria for individual lots or consignments of food when the source or conditions of manufac- ture are uncertain. . . To be used successfully, FSOs must: · Be used only where appropriate. · Be based on definitive data on the initial level of the hazard and be supported by sufficient data on transportation, retail, and home storage and preparation practices. · Become familiar to food companies of all sizes. Include definitive instructions for food processors on what is needed to document achievement. · Gain acceptance from the public, consumer organizations, and the public health community as a method to ensure safer food. STRATEGIES FOR DEVELOPING CRITERIA AND PERFORMANCE STANDARDS There are several strategies a regulatory agency can use to develop regula- tions, and care must be used in selecting the proper one to maximize the efficacy of food safety regulations. This procedure should be a transparent process in the gathering and analysis of data and in the development of the regulations. This section identifies some benefits and limitations of basic statistical approaches that may be used in developing food safety criteria and standards. Food regulation should always be based on science. The President's Council on Food Safety, which was established in 1998 (E.O. 13040, 1998), directed regulatory agencies to use science-based approaches to develop new regulations. Therefore, a science-based approach must be used also in developing perfor- mance standards.

102 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD Depending on the quality of the data that are available or that can be gener- ated in a pilot study, the committee defines a science-based approach as using one of the following strategies: 1. A statistically valid, controlled study in the laboratory or field, which might include risk assessment modeling (laboratory-based strategy). 2. The expertise and derived opinions from the best understanding of risk assessment, pathogenesis, and current food-processing techniques (expertise- based strategy). 3. A combination of a controlled study and expertise (combination strategy). These strategies should not be seen as separate and individual, but as a continuum of a science-based approach. Each one has benefits and limitations, and lack of time is one limitation that recurs in the development of new regula- tions because many regulations are developed in response to a crisis. The actual approach will be dependent on the quality of data that are available or that can be generated through a pilot program or modeling approach. In most cases, strategy three (combination approach) presents the most effective and practical strategy to develop performance criteria. To improve the process of developing regulations, it is important to understand the limitations of each approach and select the best one. Strategies to Develop Food Safety Criteria, Including Performance Standards Laboratory-Based Strategy A statistically valid, controlled-study strategy to develop regulations is one that applies the pure scientific method to develop regulations. An example of this strategy is the design and data analysis in standard laboratory experiments. This strategy can be summarized as follows: 1. Development of a hypothesis. 2. Design of a study (laboratory or field) to test the hypothesis. 3. Analysis of the results using appropriate statistical methods (such as analysis of variance) and use of these statistical results to determine if the hypothesis is accepted or rejected. After these steps are followed, regulations may be developed on the basis of whether the hypothesis is accepted or rejected. For efficiency in conducting a study, scientists study a sample of behavior or product and then generalize the results to the entire population or phenomenon. The following assumptions must be met to ensure validity of the statistical

FOOD SAFETY TOOLS 103 methods used to analyze the data from the study and apply the results to the entire population (Steel et al., 1997~: . The variation (either measured as standard deviation or variance) must be constant. · There must be a defined population from which a sample can be selected. · The sample size is a function of the standard deviation. The larger the standard deviation, the larger the sample size. · The sampling method must assure a random sample. When all of these criteria are met, it is possible to calculate the necessary statistics with a mathematically defined probability and known confidence, and then analyze the results. These statistics will define a probability that the hypothesis is correct at a known confidence level a statistic that provides the experimenter with the probability that the answer is correct. In addition to the requirements stated above, if the data collected in the study are going to be extrapolated either beyond the population or into the future, the mean and the variation must remain stable and predictable over the period of time of the extrapolation. If any of the requirements listed above is not met, numerical values can still be calculated; however, it is not possible either to accurately estimate the prob- ability level or to determine the confidence level of the statistical data and result- ing outcome. It can be safe to assume that most experiments conducted never meet all of the criteria and thus regulatory agencies must use some expertise in evaluating the results. Furthermore, the committee feels that there are probably no clear-cut examples of food safety regulations created under the strictest sense of the statistically valid, controlled study strategy as described. The greater the violation of the statistical criteria, the greater the likelihood that the statistics do not represent either the population or the sample. Such data, collected in a labo- ratory under a controlled environment, can still be used to develop regulations provided statistical gaps in the data are filled with scientific knowledge and derived scientific expert opinions. However, limitations exist if the pure science- based strategy is applied to the development of regulations to govern the food- processing industry. During the development of regulations, the actual laboratory is the field, and because of limited time and resources, there is often not enough data gathered to ensure statistical accuracy with a known certainty. In addition, it is not known whether the mean and standard deviation of the performance stan- dard or criterion that is measured will remain stable over the period of time the regulation is enforced. Thus, it is not possible for regulatory agencies to rely solely upon the statis- tically valid, controlled study strategy, as described above, to develop regula- tions. There will always be gaps in the knowledge, with subsequent gaps in the experimental design. In response to these gaps, the regulatory agencies must use expert knowledge to satisfy assumptions and develop knowledge bases.

104 Expert-Based Strategy SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD This approach can best be described as exclusively using expert opinions to develop the new regulation or performance standard, and this can be depicted in the following way: a group with broad-ranging expertise on a given food product, including regulators and outside experts, comes together, deliberates, and develops the performance standard or regulation. In this manner, the performance standard is developed using only the scientific-based expert knowledge present during the deliberations. As with other strategies, this process has a number of limitations. First, the standard will be only as good as the knowledge of the experts who are gathered to develop the standard. Second, although experts may have an excellent knowledge of the situation, rarely do they have all of the needed knowledge to develop a robust performance standard. Therefore, the experts will have to fill in the knowl- edge gaps with assumptions, including how the regulation will perform in the future. Third, each expert works within a personal and professional paradigm. These are difficulties associated with the expertise-based method if the new standard requires a novel approach. Thus, if the standard requires thinking beyond the conventional framework and all of the participating experts have the same professional paradigm, it is likely that the expert group will not be able to develop an effective standard that is valid beyond that framework. In addition, if the assumptions change or if one assumption is slightly incorrect, a poor perfor- mance standard will result. (A poor performance standard can be defined as either not being effective in meeting the public health objective or generating a needless economic burden to one or more sectors of the food system.) A fourth limitation of this strategy is that the process is rarely transparent to the public, which may then question the validity of the performance standard. The advantage of the expert-based strategy is that it requires the use of a minimal amount of resources such as time, money, and personnel. Therefore, regulations may be developed rapidly in response to a public health crisis. In summary, although regulations derived exclusively using expert opinions require minimal resources, their success depends highly on the expert knowledge used to develop them. Combination Strategy The combination strategy uses both the laboratory-based and the expert- based strategies. It is a hybrid that includes the strengths of both strategies, while minimizing their weaknesses. Regulatory agencies must strive to develop regula- tions using the best available data. The general precept is that the more data (laboratory-based), the better; however, assumptions will always be made because rarely or never will there be enough appropriate data available to fully develop a regulation. Assumptions will need to be made using expert opinion. Consequently, this approach recognizes that expert knowledge will always be used to fill in the

FOOD SAFETY TOOLS 105 data and knowledge gaps. Currently, this approach is being used to some extent by regulatory agencies in developing new regulations and performance standards; for example, FSIS is using it in developing the performance standards for the HACCP-Based Inspection Models Project (HIMP) (FSIS, 2001~. This approach is needed because it is impossible for a regulatory agency to utilize a pure labora- tory-based approach in developing a regulation for the field.

106 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD Several dilemmas may be encountered during the development of regula- tions. For example, although the development of scientific knowledge is acceler- ating, corresponding advances in new technology development to either prevent or reduce the likelihood of a food safety hazard from occurring lag behind. In addition, the regulatory environment is such that it is exceedingly costly and time-consuming for the food safety regulatory agencies to implement new and innovative regulatory strategies to reduce the risk of foodborne illness. Once regulations are finalized, modifying them is time consuming and tedious. In addition, approval of new technologies for controlling pathogens (e.g., an additive or a new method for killing or reducing the numbers of a pathogen) is a very slow process. To remedy this lack of flexibility and as previously recommended in the National Academies report, Ensuring Safe Food from Production to Consump- tion (TOM/NRC, 1998), Congress should grant the regulatory agencies the legal authority to develop, and the administrative process flexibility to update, food safety criteria, including performance standards. This flexibility includes incor- porating new processing or assessment techniques and allowing the agencies the ability to improve a performance standard to align it with the best contemporary scientific knowledge. Appropriate Data for Developing Performance Standards Another dilemma that may be encountered during the development of regula- tions is that regulatory agencies, by mandate, must use a science-based approach (Presidents Council on Food Safety, E.O. 13100, 1998), and must usually do so within a very short time frame. Unfortunately, it normally takes time, in addition to other resources, to collect the appropriate data to make scientific decisions. One way to overcome this dilemma is to develop and maintain databases on critical information. Regulatory agencies can develop and maintain databases on the prevalence of specific contaminants for critical commodities (e.g., ground meat). In addition, regulators can conduct or fund pilot studies to collect appropriate data if these data are not available. (Chapter 4 describes the particular need and justifications to maintain current databases on the major animal species that supply the majority of the meat consumed in the United States.) In addition to maintaining these databases, regulatory agencies must continually analyze these data using basic time series analysis (e.g., control charts, histograms, and capability analysis). Congress, in turn, should provide adequate resources to develop and maintain these databases. Pilot studies are the preferred method for gathering the appropriate data to develop science-based regulations because they are designed to provide the spe- cific data needed to develop a new regulation. A study of this type was conducted as part of the HIMP project (FSIS, 2001~. In contrast, data analysis problems

FOOD SAFETY TOOLS 107 were identified when an old database was used to justify establishing a perfor- mance standard for stabilization of ready-to-eat meat (FSIS, 1998~. Chapter 4 provides details of the analysis used to develop this performance standard. Once the appropriate data are available through pilot studies or databases, there are two ways to proceed in developing a performance standard, depending on the desired outcome. The first assumes that all food-processing companies would be complying, that is, producing food of a predetermined acceptable level. If this strategy is used, the performance standard should be set at a level such that the lowest compliant processor will pass, while all of the noncompliant plants will fail. A second way is to set the performance standard at a level where only a portion of the plants will pass. This strategy is used to allow the regulatory agency to raise the bar of what is classified as acceptable performance. An example of the latter strategy was used in developing the HIMP performance standards, which were set at the 75th percentile of the plants that participated in the pilot study (FSIS, 2001~. When this strategy is used, the regulatory agency must balance the benefits of raising the bar to meet the nation's public health goals with the economic consequences of strengthening the performance stan- dard. Furthermore, flexibility must be incorporated into the development of performance standards so that the regulatory agencies may adjust a performance standard to meet future public health goals; that is, the regulatory structure should allow for review process flexibility. In the absence of appropriate data or when only limited data are available, the only way to set a performance standard is to build in a safety factor of sufficient magnitude to ensure that any current or future process variation is of no public health significance. Such a safety factor may force the food processor to overprocess (e.g., cook excessively) a product to ensure that the performance standard is met, and thus may have a negative effect on the product. An example of this type of performance standard is the 12-D reduction of C. botulinum for low-acid canned foods (Karel et al., 1975~. STATISTICAL TOOLS TO VERIFY PROCESS STABILITY AND CAPABILITY Manufacturing processes tend to vary over time. For example, in a canning operation, the temperature of the retort may vary by a degree or two from the target temperature. In a chicken processing operation, in turn, the weight of a dressed carcass can vary by as much as 20 percent. To assure that the outcome of the processing operation is predictable, it is critical for both processors and regu- lators to understand whether this variation is predictable or not. (Statistical Process Control [SPC] terminology uses the term "common causes of variation" when processes show only predictable variation, and the terms "special causes of varia- tion" or "assignable causes of variation" when the process shows nonpredictable variation.)

108 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD Food-processing regulations should require that food processors and regula- tory agencies analyze performance data to assure that the variation is stable. This can be done by using simple time-series analyses such as control charts, histo- grams, and process capability analysis, which are all tools to measure the stability of variation. Capability indices are statistical calculations that relate the performance stan- dard with both the amount of product variation and the relation of the process mean to the performance standard. There are three major types of performance indices: Cpk, Cal, and Cpu. These indices provide the regulatory agency with information to determine if the food processor has the capability of meeting the performance standard. The Cpk is calculated when there is both a maximum and a minimum limit specification (i.e., performance standard); the Cpu is calculated when a performance standard has only an upper limit; and the Cp~ is calculated when a performance standard has only a lower limit. SPC texts written by Kane (1989), Bothe (2001), and Montgomery (2001) provide details on the use and calculations of these indices. The indices provide a science-based approach for processors to demonstrate compliance with the performance standard and capa- bility of their process, and for the regulatory agency to monitor such compliance. SPC is a very robust scientific analysis that uses control charts and capability analysis to monitor process performance. When using SPC, all the tests that monitor the manufacturing process are linked into an appropriate process control plan that includes control charts, a simple but effective form of time series analysis. The charts are designed to measure process variation over time and to verify that the variation is stable and predictable. An example of the use of control charts in determining regulatory compliance is the current Pathogen Reduction (PR)/HACCP regulation in which the food processor has the choice of reporting generic E. cold carcass data either on a control chart or in tabular form (FSIS, 1996~. SPC processes are easy to audit by a trained investigator, which enables efficient regulatory oversight. In addition, it is difficult to falsify analytical data gathered through an appropriately designed system based on SPC, for the same analytical techniques that are used to control the process can be used by regula- tory agencies to determine if the data accurately described the production process and, therefore, the safety of the product. SPC provides the signals that processors need to effectively improve their processes. Continuous improvement is a strategy that focuses on using a systematic process to identify and remove the root causes of variation in products and critical processes. The international community recognizes the importance of continuous im- provement as part of a quality management system. Section 8.5 of ISO 9001:2000 (Ketola and Roberts, 2001) requires that "organizations that are compliant to the standard must have a process that continually improves the quality management system." Continuous improvement is interpreted as both incremental improvements (small continuous improvement accomplishments) and breakthrough improve-

FOOD SAFETY TOOLS 109 meets (large, technology-driven improvement gains). Thus, it has been widely recognized that an effective SPC program must be linked to an effective continu- ous improvement program. Kume (1985) described a number of simple statistical tools that can be used to continuously improve manufacturing processes. An example may help illustrate these concepts. Suppose a processor, to eliminate or reduce the population of a pathogen, is required to heat each unit of food product to x temperature and hold it at that temperature for y minutes. To assure the safety of each unit, the processor must plan to heat all units to a somewhat higher temperature to ensure that, with normal variation, no individual unit falls below that temperature. Using SPC techniques, the processor can map the variations in the process, thereby determining the optimal temperature at which to operate. It is in the processor's interest to minimize the amount of variation, because that will save energy costs and will also produce a more consistent product as no unit would be subjected to more heat than that necessary to ensure that each unit is adequately heated. If the process variation is +3° and an adjustment of the cooking equipment could reduce that variation to +1°, the processor could save money and deliver better products by investing in such an adjustment. This is an example of continuous improvement. The use of SPC linked to continuous improvement creates a situation where all involved parties consumer, regulatory agencies, and industry benefit: con- sumers will have safer food, industry will have lower production costs, and regulatory agencies will observe better regulatory compliance. It provides a logi- cal, methodical way to establish process stability and capability analysis that is economically efficient for industry and is easy to review. In addition to its poten- tial for facilitating regulatory compliance, the systematic, continuous process improvement focuses on eliminating the causes of foodborne disease and thus contributes to enhancement of food safety. Moreover, the actions taken to reduce a foodborne hazard will usually reduce waste and decrease product rework or loss in the plant, thus reducing production costs. It is generally believed that if a company does not have an active systematic continuous improvement process, the projected cost attributed to poor quality is at least 20 percent of the sales dollar amount (Breyfogle et al., 2001~. Therefore, food safety regulations should incorporate the concepts of SPC linked to continuous improvement, and require that food processors analyze and maintain records to ensure that their processes exhibit (1) stable and predictable variation (rather than unpredictable variation) and (2) are capable of meeting performance standards. The regulatory agencies, in turn, must ensure that their professional staff assigned to either inspecting or auditing food-processing plants are appropriately trained so that they can determine if a processing plant is properly using SPC techniques to monitor performance standards and whether the plant is capable of meeting the performance standards.

0 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD Statistical Process Control: A Science-Based Approach to Ensure Regulatory Compliance There are two methods by which food processors and regulators can deter- mine conformance to a performance standard. The first method is to inspect either 100 percent of the product or a sample of the product. The second method is to rely on SPC. This section provides an overview of process control and process control as a tool for use in ensuring food safety, including a comparison between process control methods and the traditional inspection method used to verify compliance with food safety criteria and standards. This section does not cover in detail the statistical nuances of process control, an understanding of which, however, is required for proper development and implementation of SPC and continuous improvement procedures in the plant, and for the incorporation of SPC principles in regulations. Interested readers should refer to the numerous texts that have been published on the subject (Kane, 1989; Kume, 1985; Montgomery, 2001; Wheeler and Chambers, 1992~. Inspection Inspection may be conducted on 100 percent of products or on a sample of the products. Neither strategy is practical or effective. One hundred percent inspection cannot guarantee that the product either meets specifications or is safe because no inspection technique is perfect (Konz et al., 1981~. Many inspection techniques for food safety require the use of a destructive test. For example, if one wanted to use 100 percent inspection to ensure that all milk in a specific lot is free of pathogens, the only way this could be accomplished would be to open each container of milk, thus breaking the seal, remove a portion of the milk for microbiological analysis, conduct the analysis, and report the results. If this proce- dure were used, no package of milk tested would be acceptable for sale to the public. An example of the ineffectiveness of 100 percent inspection was documented by the Research Triangle Institute (RTI) when conducting the baseline study for HIMP (RTI, 2000~. RTI conducted a study that measured the effectiveness of the traditional inspection process used in poultry slaughter facilities. In this process, 100 percent of the chicken carcasses are inspected by an FSIS official to determine whether food safety defects or other consumer protection defects are present. RTI found that even after the FSIS inspector step, 1.9 percent of the carcasses con- tained a food safety defect or fecal contamination. In addition, a larger number of carcasses contained "Other Consumer Protection" (OCP) defects such as ingesta (13.8 percent), sores and scabs (16.0 percent), or pathological lesions (1.3 per- cent). The carcasses that contained OCP defects should have been removed from the line and reworked to remove the unacceptable tissue. In addition, RTI found that 12 percent of the carcasses that were condemned did not have either a food safety or OCP defect and thus should not have been condemned.

FOOD SAFETY TOOLS 111 Because 100 percent inspection is often impossible, food processors and regulators instead use sampling techniques. In this method, a sample of the product is obtained and analyzed, and the test results are used to determine if the entire production lot is acceptable or unacceptable. This approach is called accep- tance sampling. Acceptance sampling assumes that the product characteristic that is being measured exhibits relatively stable variation or consistent variation within the lot. Thus, even using a true random sampling technique, acceptance sampling proce- dures are not designed to identify "hot spots" (i.e., when microorganisms or toxins are concentrated in a very small portion of the lot), sporadic food safety hazards, or food hazards that occur at very low levels in a production lot (like many microbial foodborne hazards). The following example illustrates how acceptance sampling may be used to test for product safety when the hazard appears at a very low level. A person may need to know how many eggs must be sampled from a lot to be reasonably confident that the lot is not contaminated with S. Enteritidis. It can be assumed that the level of S. Enteritidis contamination in eggs is 1 egg in 20,000 (Salmo- nella Enteritidis Risk Assessment Team, 1998~. An acceptable guideline to deter- mine the sample size is to take a sample large enough that there is a chance that 8 contaminated eggs will be selected (LSRO, 1995~; this guideline gives the investigator statistical confidence in the results of the test. Thus, the individual would have to sample 160,000 eggs and test them using an analytical method sensitive enough to detect one S. Enteritidis cell per egg. Obviously, a sample size of 160,000 eggs corresponds to a very large testing rate and is not practical in the food-processing industry. This sample size is independent of the size of the lot. Therefore, if the lot contained 120,000 eggs, each egg would have to be sampled and destroyed, making this sampling system very expensive; the cost to sample and analyze this number of eggs would be in excess of several million dollars and there would be no eggs left to sell at the end of testing. When it is not possible to inspect 100 percent of a production lot, regulatory agencies may establish statistical criteria as an indication of the acceptable level of control of a potential food hazard. An example of this is the low-acid canned food performance standard, which requires an intervention capable of reducing the population of C. botulinum by 12 logic in the final product (Karel et al., 1975~. Process Control Process control is based on four premises: (1) product quality or product safety must be built into the manufacturing process, (2) the manufacturing process must be monitored and the data must be analyzed using appropriate measurement and statistical techniques, (3) the process must be managed to ensure its variation remains stable and predictable, and (4) the process is capable of delivering product

2 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD that meets the performance standard. As described earlier, SPC relies on the appropriate generation and analysis of data by using control charts, histograms, and capability studies (Kane, 1989; Montgomery, 2001~. When this is done, the data can be used to predict the performance of a process and the safety of a product. SPC is the combination of these analytical procedures and it allows for the following assumption to be made: if the process exhibits stable variation or if the process is in statistical control, it will result in a product within a set of math- ematically defined, predetermined limits. These limits are known as control chart limits and are calculated from process or product data. The control limits cannot be set by expert opinion. Process Variation, Stability, and Capability Control charts are used to determine if the variation of the process or product is predictable (stable variation) or nonpredictable (nonstable variation). When the variation is stable, the process is said to be in statistical control. If a process is in statistical control, it is possible to determine whether the process is capable of meeting performance standards by using process capability analysis. Then, if the process is found to be both capable and in statistical control, end-product inspec- tion may become unnecessary. On the other hand, if a process is not in statistical control, it is not possible to statistically determine the extent of the variation of the product or whether the product will meet a performance standard. To ensure product safety in this case, a very conservative performance standard must be developed, namely one that has a very large safety factor (which, as discussed earlier, may have a negative impact on the product). If the process is not in statistical control, the food processor must take appro- priate action to identify the causes of the problem (called in the literature "assign- able causes") and eliminate them (Kane, 1989; Kume 1985~. When properly designed, these actions can be taken in advance so that the risk of producing unsafe products is minimized. Regulatory agencies, in turn, need to monitor food processors to ensure that this task has been accomplished. Therefore, SPC can be used to show process stability and, once the process is in statistical control, to show whether the process is capable of meeting a perfor- mance standard. A number of texts on SPC provide the details on creating control charts and evaluating the stability of a process (ASTM, 1976; Montgomery, 2001; Wheeler and Chambers, 1992~. The stability of a process is paramount in determining whether a process is in statistical control. Capability analysis, in turn, provides a statistical tool to deter- mine if a process that is in statistical control can deliver product that meets the performance standard.

FOOD SAFETY TOOLS 113 It must be noted that SPC requires conducting appropriate tests or measure- ments to predict the performance of a product. (These tests must show a correla- tion between process performance and specific product attributes.) If process variables measurable on the processing line do not exist for a particular process, then standard tests (microbiological or other) are necessary for process control evaluation and should be used as measures of performance. The results of these tests can then be used in control charts and capability analysis to evaluate process control. Control charts are used to demonstrate that a manufacturer has main- tained control of the process. Histograms and capability indices are used to demonstrate that the product meets the performance standards. The committee recommends that performance standards incorporate the analysis of appropriate data on process and product characteristics using SPC, and that the regulatory agencies define what constitutes the minimal acceptable process capability. In addition, it is recommended that performance standards link the SPC requirement to continuous improvement. Examples of Other Process Control Approaches Other methods may be appropriate to assure control of food-manufacturing processes. These control strategies include automation, education and training, procedures and check sheets, checks on incoming product quality (raw input acceptance sampling), or a combination of these. An automated process control has been used successfully in assuring that milk is properly pasteurized in accor- dance with the pasteurized milk ordinance (FDA, 1 999b). Measuring and control- ling a process parameter (i.e., temperature/time) using an electronic feedback control system accomplishes this objective (e.g., safe milk). Another example is the combined set of process control strategies that has been successfully incorporated into the low-acid canned food regulations (21 C.F.R. part 114~. The following is a summary of this process control strategy: 1. The processor determines the critical measures for the process. 2. The processor validates the thermal process. 3. The product is processed in accordance with the validated process. 4. The critical parameters are monitored. 5. The food processor verifies that the process was conducted in accordance with the validated process. When all of the above steps are properly carried out, FDA declares that the low-acid canned food is safe and no final product testing is necessary to deter- mine if C. botulinum is present (Gavin and Weddig, 1995~. The development of the Juice HACCP Final Rule is another example of a science-based approach that used both expert opinion and statistical studies to determine a sampling plan that provided the basis for the rule (the aforemen-

4 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD tioned combination strategy). One of the rule's supporting documents referenc- ing the generic E. cold levels most likely to be found in juice states: "Data in this area are limited so certain assumptions were made" (Garthright et al., 2002~. At the time the rule was being developed, there were no laboratory data that could substantiate the levels of E. cold 0157:H7 that were found in juice. In the absence of such data, several assumptions had to be made using the best available exper- tise (expert-based strategy) of NACMCF. The resulting requirement of a 5-D reduction in pathogen numbers was a consensus value arrived at by NACMCF after reviewing comments received from the public (Personal communication, W. Garthright, Center for Food Safety and Applied Nutrition, FDA, October 2002~. In addition to the 5-D pathogen reduction performance standard subsequently established by FDA, producers of raw citrus juices that use surface decontamina- tion to achieve the standard must conduct end-product testing to ensure that generic E. cold is absent. To this effect, a sampling protocol to be implemented by processors as part of their HACCP plan was developed. The two issues central to the sampling plan were sample size and testing window. The sample size for juice was determined to be 20 mL for each 1,000 gal of juice produced or, if the processor produces less than 1,000 gal in 5 days, a 20-mL sample must be taken every 5 days. This sampling procedure was developed using computer simulation techniques (Personal communication, W. Garthright, Center for Food Safety and Applied Nutrition, FDA, October 2002; Garthright et al., 2002), which is a science-based mathematical approach. The next step in the development of this performance standard was the design of a sampling plan that would ensure absence of generic E. coli. An evaluation problem occurs in developing a performance standard when the vari- able to be tested, such as the presence of generic E. cold in juice, rarely occurs in the product, and yet the processor must determine whether there is a failure in its HACCP plan. A solution to this problem is to analyze the data using the moving window technique, which requires counting the number of positive samples within a specific time frame. FDA set the juice performance standard at no more than one generic E. coli-positive sample in any consecutive seven samples. If two or more samples test positive, FDA considers that there is a loss of process control and immediate corrective actions are necessary (FDA, 2001~. FDA validated these statistics by means of a mathematical technique known as Monte Carlo Simulation that is used in many industrial analysis situations (Law and Kelton, 2000~. This example demonstrates the successful use of the combination strategy in developing a regulation. FDA used a combination of the best science-based expert opinion (expert-based strategy) and mathematical studies (laboratory-based strategy) to develop the sampling plan for the Juice HACCP Final Rule; however, the process could have benefited from more transparency regarding access to information on the assumptions that were made. The document stated that "as

FOOD SAFETY TOOLS 115 additional data become available, the agency [FDA] will consider those data and propose adjustments to the HACCP regulation and to the juice hazards guide as necessary" (FDA, 20011. The committee commends FDA's willingness to consider adjustments to performance standards as data become available, and recommends that the food safety regulatory agencies routinely conduct periodic, mandatory reviews of all performance standards. Collecting the Appropriate Data Any performance standard requires that monitoring and/or testing be con- ducted on the process or product. Ensuring that the monitoring and testing methods are validated and deliver the best data is essential when developing standards or verifying processes. This need for adequate data is recognized at the international level and has brought about the development of international norms that describe standard and approved analytical techniques and GLPs (Singer, 20011. These principles require that a number of critical issues be addressed and controlled to ensure good analytical results, including sample collection, storage, and analysis; data management (collection, storage, analysis, and reporting); laboratory and testing facilities; calibration of equipment; and training of personnel. It is also critical that the proper test methods (i.e., having adequate specificity, sensitivity, precision, accu- racy, and reproducibility) be used. This is ensured through validation of sampling and testing methods. Details on the application of GLPs are described in texts such as that by Singer (2001) and in federal regulations (40 C.F.R. §160.11. When zero tolerance is used as a performance standard (see Chapter 1), unique methodology issues need to be considered. The concept of a zero toler- ance performance standard is inextricably linked to the sensitivity of the method employed to detect the offending hazard, as well as the sampling strategy em- ployed. Sampling protocols must take into account that a large sample is needed to ensure the absence of the hazard; also, the sample must be representative of the material being tested. The level that can be detected is a function of the sensitivity of the method as well as of the sample volume. An assay's limit of quantitation and limit of detection are defined on the basis of measured performance of the specific assay being used and on agreed statistical criteria. When zero tolerance is applied in this context, zero is opera- tionally defined as the limit of detection applied to the specific sample. It must be stressed that the hazard, be it chemical, microbial, or other, may still be present in the sample but not be detectable with the assay method being used. The limit of detection is a function of the precision of the analytical methodology. In conclusion, regulatory agencies should use a science-based approach both to develop regulations and to measure compliance. Performance standards need to be based on appropriate data, to be possible to implement, and to be linked to public health objectives. This approach will require that the regulatory agencies

6 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD use well-defined and current databases and/or conduct pilot studies. When data are limited, regulatory agencies need to make assumptions to fill any gaps. Subject- matter experts, using the best available knowledge, should make these assump- tions. During the development of performance standards, regulatory agencies need to use a transparent process that publicly reports the data used, the statistical methods used to analyze the data, and the assumptions made to fill any data gaps. The committee recommends that regulatory agencies adopt a transparent approach that uses a combination of controlled studies and expertise to develop science-based food safety criteria, including performance standards. Similarly, for flexibility, the periodic evaluation and updating of performance standards by the regulatory agencies is highly recommended by the committee. The committee recognizes the value of SPC as a scientific method that can be used to (1) verify the control of a food-processing system, (2) provide a source of information to the food processor for properly controlling the manufacturing process, and (3) provide information that can be used to critically examine the food-processing system so that appropriate actions can be taken to reduce the likelihood of manufacturing unsafe food products. The committee also recog- nizes the potential benefit that could be derived from the use of SPC principles linked to continuous improvement by food processors, to continually reduce the risk of producing unsafe food products, and possibly also to reduce production costs. In addition, the committee concludes that the most effective procedure to determine whether a food processor is complying with a performance standard is to analyze process and product data using control charts, histograms, and process capability indices; therefore, the committee believes that SPC, linked to continu- ous improvement, provides a very robust methodology that is easy to monitor from a regulatory perspective. Accordingly, the committee recommends the adoption by food processors of SPC principles linked to continuous improvement, as well as incorporation of such principles by the regulatory agencies into food safety regulations and into the agencies' compliance monitoring procedures. THE ECONOMICS OF FOOD SAFETY CRITERIA Any evaluation of food safety criteria needs to consider the costs and benefits incurred by government, companies, and consumers as a result of the regulation. Proposed new regulations are required to include a Regulatory Impact Assessment to evaluate their costs and benefits. Consequently, the charge to the committee included a request to examine the economics of food safety criteria. This section compares the effectiveness, efficiency, and equity of two broad sets of not- always mutually exclusive tools: process criteria and performance standards. When regulation is deemed necessary, a target level (such as a performance standard) provides companies with flexibility in the manner of compliance. Sur- prisingly, however, the application of food safety policies based on this approach

FOOD SAFETY TOOLS 117 has become fashionable only over the last decade or so. Prior to this period, command-and-control or process criteria were more commonly adopted. Effectiveness Key questions that must be asked when evaluating the economics of food safety criteria include: Can a performance or process criterion be constructed to exactly fit with the stated aim? Because food safety regulations can be explicitly stated in terms of their quantitative public health goals (e.g., reduction in illnesses due to a particular foodborne pathogen by 10 percent over a number of years), is one form of criterion more naturally fitted to this goal than another? How can the effectiveness of a regulation be assessed once in place? Indirect and direct mea- sures of effectiveness are being collected to assess risk reductions achieved by food safety policies, including performance standards. Examples include trends in foodborne illnesses and microbial sampling results, as described in Chapter 2. However, as stated earlier in the present chapter, without an understanding of attributable risk and clear links between hazard reductions at one particular stage (e.g., slaughter or processing for meat and poultry) and the reduction in illnesses, determination of the effectiveness of a regulation becomes complicated or even impossible. Efficiency How has a regulation been implemented? What are the monitoring/inspection costs surrounding it? Such questions focus on the technical efficiency of the manner of implementation. With process criteria, compliance is assessed by determining whether com- panies are using the particular piece of equipment mandated by the standard and whether they are doing so correctly. With a performance standard, compliance may be more difficult to assess. Government costs involved with implementation of a performance standard (e.g., from sampling for verification) may prove to be higher than with process criteria. For companies, performance standards may confer flexibility and re- duced costs. However, if these savings are small and do not offset higher govern- ment costs, the overall societal costs may be lower for process criteria. Lower company costs may be seen over time given new technologies that achieve the process criteria. The determination of these cost reductions for use in cost-benefit analyses is a challenge. This may bias findings towards process standards. Equity When comparing performance and process criteria, the issue of equity centers on the incidence of costs and benefits placed upon, or derived by, a particular

8 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD section of society as a direct result of the regulation under consideration. Theo- retically, performance standards are more likely to be scale neutral compared with process criteria. However, a performance standard may cost large com- panies less to comply with because of economies of scale or scope. There may be a limited amount of research and development dedicated to providing interven- tions for smaller operations. The "safe harbor" strategy, whereby smaller opera- tors are provided a set of validated interventions from which they may select, provides an example of how equity can be built into a regulation. However, there is an implicit inflexibility with the reliance on such safe harbor processes, and there may be a concern over the lack of plant-specific adaptation of the under- lying HACCP plan. This echoes concerns stated earlier about the use of generic HACCP plans without a full appreciation of how appropriate these may be for the individual plant, line, and product. The range of food safety criteria discussed in this report includes those that rely solely on performance standards (e.g., 5-D pathogen reduction in juice), mixed regulations that combine process and performance standards (e.g., the PR/ HACCP rule), and process criteria (e.g., pasteurization of milk). Such broader aspects of the equity of food safety regulations as potential regional dimensions, distribution of costs along the supply chain, and equity dimensions on the benefits side in terms of greater risks incurred by particular subpopulations, are beyond the scope of this report. Given each of the economic concerns listed above, the individual regulation must be assessed for its impact on the balance of costs and benefits for each section of society (companies, consumers, government, and ideally subgroups of these such as small versus larger companies and immuno- compromised populations versus the healthy), as well as the remaining incentives to innovate and therefore improve quality. Costs and Benefits of Food Safety Regulations The evolving field of food safety economics has focused significant attention on the tools necessary to first forecast and then track costs and benefits of regula- tions. This has led to many refinements in the methodology for forecasting benefits, including such impacts on specific populations as age-based morbidity and mortality calculations and early efforts to incorporate disability or quality adjusted life-year measures. Depending on the empirical method adopted for valuing such reductions in foodborne illness (e.g., cost-of-illness or willingness- to-pay [Kuchler and Golan, 19991), large ranges in the estimates of a policy's hypothesized benefits generally result. Similarly, costs of compliance must be estimated ahead of time, often with limited knowledge of current industry prac- tices or likely adoption of response strategies. It should be noted that the bulk of food safety economics research does not focus on the impact of individual performance standards isolated from the overall food safety regulation or program under review (mostly HACCP-based regula-

FOOD SAFETY TOOLS 119 lions). As such, it is difficult to quantify the unique costs and benefits of perfor- mance standards implemented as part of broader regulatory change. In order to complete such evaluations it would be necessary to have representative, detailed cost data linked to actual microbiological improvements solely due to the particu- lar performance standard under review. In this way, one could avoid (or at least minimize) incorrectly assigning costs and benefits to regulations (or parts of a regulation) that are more correctly due to a general trend in food safety enhance- ments that the plant, company, or industry may have performed in the absence of the regulation (MacDonald and Crutchfield, 1996~. For example, if pathogen reduction resulted from an investment in a new piece of equipment purchased in response to customer demands and was not required by the regulation per se, then it would be incorrect to attribute this cost and the resultant food safety benefit- to the regulation. The scale of pathogen reductions used as inputs in benefit estimations also needs to be considered. Clearly, it is desirable for such food safety gains to be calculated from real-world changes in specified bacterial populations observed at the plant level. Based on this information, some form of aggregation would then provide a measure of the societal gain derived from the regulation. These pathogen reductions should not be laboratory-level performance evaluations of a strategy unless they have been validated in real-world applications. Challenge experi- ments often use inoculated samples with elevated populations of microorganisms and can bias results in favor of certain interventions, suggesting large pathogen reductions that may not be achieved in the processing plant. Issues related to maintaining reductions in pathogens beyond the point or stage of application of a performance standard (e.g., the slaughterhouse or pro- cessing facility for meat and poultry), and to the optimal stages where these reductions were attained, remain understudied in the field of food safety econom- ics. Thus, the benefits of large reductions in microbial loads on freshly slaugh- tered or processed meat and poultry may be diminished or even completely lost by downstream recontamination and thereby provide no risk reduction. When stage-specific risk-management strategies are assessed without the chain-wide determination of all economic implications, it is possible that, at best, an ineffi- cient criterion may be selected and, at worst, that significant disincentives for companies to adopt proven food safety strategies will result. Similarly, cost shifting among segments of the chain (transfers), as opposed to true cost reductions, may arise through the application of food safety criteria. An example could be a performance standard that leads to a requirement placed on input suppliers via a CCP at receiving that may drastically increase suppliers' costs and yet have limited public health benefits when compared with an end- point performance standard.

120 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD Innovation: Lessons from Environmental Regulations The degree of innovation for food-processing companies can be thought of as a continuum between the two endpoints: (1) target standards (low degree of government intervention, such as sanitation guidelines), and (2) process criteria (a high degree of government intervention, such as pasteurization of milk). Per- formance standards lie somewhere in the middle of this continuum and are much less intrusive than process criteria. Most performance standards give require- ments in terms of results (e.g., a 5-D reduction in bacterial numbers) and do not specify particular production or process methods. Therefore, they are more flex- ible than criteria. This flexibility should allow innovation and result in reduced costs. The thesis that flexibility allows innovation has been borne out in the area of environmental regulations. In fact, in the most general sense, successful com- panies innovate to fight pressures from competitors and customers, so this thesis may be amenable to extension into the food safety regulatory environment. If any lesson may be taken from environmental economics, it might be that properly designed regulations and standards can trigger innovations that lower the total cost of a product and improve its value (Porter and van der Linde, 1995~. A highly competitive company may see compliance with a performance standard as a challenge and respond by creating innovative solutions to meet the standard. The lessons learned from environmental regulations provide a basis for some general guidelines for setting performance standards (and subsequent regula- tions) in food safety systems, as described by Golan (2002~: Regulate as close to the end user as possible, thus encouraging upstream inno- vation; choose strict, not simply feasible, standards to encourage efficiency and innovation; regulate along international trends; and select criteria for compli- ance verification that [are] informative, reliably measurable, and flexible. These proposals are valid only if industry and regulators remove the conten- tious belief that regulations erode competitiveness (Porter and van der Linde, 1995~. Therefore, if viewed as a challenge, a performance standard at the appro- priate point could result in cost-reducing innovations that accrue for the entire food industry sector, while making food safer. Innovation and Performance Standards No regulation should be static. Every industry, regardless of its maturity, should be constantly challenged to innovate to reduce costs and improve quality. There is nothing implicit about either a process or a performance standard that either encourages or constrains innovation, so long as these standards are dynamic. This point was acknowledged by FSIS in the PR/HACCP rule (FSIS, 1996) and related collection of updated baseline data (see discussion in Chapter 4~. Process

FOOD SAFETY TOOLS 121 criteria, however, by their nature as a preapproved form of production, may suffer from more "institutional friction" than performance standards. A great effort is thus required when implementing new, stricter, process criteria following an innovation such as the invention of a new piece of equipment designed to reduce pathogens in a food product. Evidence of the impact on innovation of the introduction of food safety performance standards is unclear. There have been significant efforts placed on pathogen reduction strategies targeting carcasses (e.g., steam pasteurization, hot water and acid rinses, steam vacuum systems), meat products (e.g., irradiation), and other food products (e.g., high pressure and ultraviolet light treatment of juices). Further, many rapid pathogen tests have been developed to service the market created by performance standards and contractual specifications. Some of this research and development is subsidized by the public sector (e.g., universities, FDA, and USDA's Agricultural Research Service), whereas other efforts are solely in the private domain. In the United States, some of these innovations would likely have emerged without the implementation of performance standards, either because of international market demands or because these innovations lend themselves to becoming validated strategies for use in future processing of food and perhaps even in process criteria. In relation to the PR/HACCP rule, the time period for evidence is still quite short (full implementation of HACCP in the meat and poultry industry is only three years old). Therefore, it is difficult to determine if innovation has been promoted by performance standards. Based on these simple economic principles, the remaining challenge is how to design food safety regulations that help within the framework of risk analy- sis to link public health goals to scientifically valid and economically feasible performance standards. Risk management clearly serves the role of evaluating alternative food safety criteria to determine if they attain a prestated public health goal. Risk-Management Economics One economic approach that may highlight the connections among a public health goal, a specific food safety objective, and a performance standard consists of determining the relevant marginal social costs changes in costs or benefits for the whole economy (companies, government, and consumers) as the level of food safety changes and benefits uniquely due to the regulation (see Figure 3.4~. This approach demonstrates that as the level of safety increases, so do social costs (borne by companies, the government, and consumers together). A 100 percent safe food supply is unachievable, and movement towards this goal leads to higher costs. Similarly, the benefits of additional increases in food safety decrease as the control of the food supply is progressively strengthened. However, most econo- mists agree that without some form of government intervention, the market alone would not achieve the optimal level of food safety seen when marginal costs and

SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD \ Marginal social benefit Marginal social cost / I. #~d i.. - id i.. - id i.. - ad i.. - ~ I Higher level of safety ~ 00°/0 FIGURE 3.4 Toward a public health goal: relating an appropriate level of protection (ALOP) to marginal social benefit and cost. benefits coincide (Figure 3.4~. (As stated above, it is unlikely that definitive values can be provided of costs and benefits, and therefore such curves convey the most likely values around which confidence intervals must be built.) The inability of consumers to fully identify a product level of safety compared with the greater knowledge that processors have of the ability of a process to deliver safety (termed "imperfect and asymmetric" information problems in the litera- ture) suggests that the market will fall short in providing the socially optimal level of protection for the particular product or pathogen under review. Economic efficiency requires that the ALOP to aim for be at the point where marginal social costs equal marginal social benefits (Figure 3.4~. Away from this equilibrium, either society desires a safer product and would benefit more than the additional costs of the stricter regime (points to the left of ALOP), or society is expending too many resources compared with the additional safety gains real- ized (to the right of ALOP). The ALOP can be related to the particular public health goal of the regulator because the model is stated in dollar terms but is partially based on population measures (benefit estimates). It is important to note that marginal social costs and marginal social benefits may change given the form of a regulation, the particular population and food product under assess-

FOOD SAFETY TOOLS 123 meet, and, over time, with a change in available technology or changing con- sumer demands or consumption patterns. Therefore, the ALOP and the most efficient food safety criteria are likely to be dynamic, given changing consumer tastes and preferences, risk tolerances, industry capabilities, and government oversight functions. An example of how such marginal social costs can be calculated, highlight- ing costs to companies from the adoption of particular food safety strategies, is shown in Figure 3.5. Four possible strategies or combinations of efforts having various levels of effectiveness and cost are shown. Various interventions (single- or multiple-hurdle strategies) can be assessed based on their cost of implementa- tion (possibly reported for various sizes or types of plants) and the most likely effectiveness (e.g., ability of the process to reduce the presence of a particular pathogen by x logic) and, therefore, on their ability to attain a performance standard (S) with a certain probability. Similarly, if S were a food safety objec- tive, then the technique could be used to assess sets of interventions adopted by various companies throughout the supply chain. The horizontal line in Figure 3.5 indicates points associated with the concept that multiple strategies may meet the necessary effectiveness (S) but with different varying costs. Effectiveness __— _c B _- C ma_ ~- c1 Cost FIGURE 3.5 Relationship between the effectiveness (i.e., pathogen reduction) and cost of hypothetical food safety strategies available to food-processing companies. SOURCE: Jensen et al. (1998), Markarian et al. (2001~.

24 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD Strategies such as point D (Figure 3.5) are dominated by each of the other options (A, B. and C) in the figure; these other options have either lower cost (like point A) or higher effectiveness (points B and C), or both. The curved line passing through points A, B. and C links all of the most favorable strategies and therefore provides an optimal path of technical food safety effectiveness. The area to the right of the curved line also suggests that there are marginal costs for various levels of food safety (for example, consider moving from point B to C). The standard S in Figure 3.5 will result in a cost of at least Cat based on where the optimal curved line and horizontal line intersect. Technical effectiveness (the frontier) is dynamic; innovations shift the curve up, allowing enhanced effectiveness for the same cost. Process criteria essentially dictate the particular strategy that must be followed by the industry (for example, strategy at point D). However, this may not result in the lowest cost (compare A with D). Furthermore, process criteria likely prevent the selection of more effective interventions (like B or C). The strategies that meet (and in this case exceed) standard S are both B and C. The particular intervention that would be selected by industry is less clear when facing a performance standard (which is considered more flexible, since many options to meet the standard may be available) as opposed to process criteria. This situation illustrates the difficulty in forecasting costs in response to a performance standard. Certain companies may decide to exceed the standard by a long measure, while others may choose to meet the standard and no more. Resulting from these different decisions, an array of potential costs can be estab- lished creating a large range (with a well-defined lower bound Car, Figure 3.5) of estimates for the related economic impact assessment of performance standards. This wide range of impact-assessment estimates would also be related to a broad range for the marginal social cost estimate (recall the marginal social cost curve in Figure 3.4), with the lower bound relating to the minimal cost (C~ in Fig- ure 3.5) of achieving standard S. This illustrates the difficulty of performing economic impact assessments. Because of the complicated situation presented above, the committee con- cluded that uncertainty still exists with respect to the economics of food safety regulations. The following are examples of questions that need to be answered: Has the correct balance of incentives to innovate, benefits, and costs been achieved? From an economic standpoint, are performance standards or process criteria better for food safety? Which economic sector benefits most from perfor- mance standards? What about performance criteria? In economic terms, what are the consumer, government, and industry responses to performance standards and performance criteria? Traditional economics suggest that performance standards should lead to a no-higher set of industry (company) costs, yet performance standards may cause the government sector to incur additional costs. Therefore, the specifics of a particular performance standard should be assessed to deter- mine this balance. Further research in these areas is required to better answer the questions above and similar ones not yet raised.

FOOD SAFETY TOOLS 125 THE IMPACT OF CHANGING TECHNOLOGY: NEW DIAGNOSTIC TOOLS Any regulatory system is heavily dependent on the technology available to detect deviations from regulatory performance standards. For that matter, the performance standards themselves may be influenced by available diagnostics, with the requirement for nondetectable levels as established by regulations having less meaning when it is possible to detect problems (such as the presence of specific pathogens) with a 10-, 100-, or 1,000-fold increase in sensitivity. Current regulatory standards for foodborne pathogens, in almost all instances, assume use of traditional culture techniques to determine the presence and number of pathogens or indicator organisms in a product. However, culture techniques tend to be slow, with two or three days often required for initial isolation of a microorganism, followed in many instances by several days of additional testing to confirm that the microorganism isolated is indeed pathogenic or that it carries the necessary virulence genes to represent a hazard to humans. There has been increasing movement toward the use of immunological assays in diagnostics which, when combined with traditional culture techniques, can provide results in less time and with greater accuracy. However, it is genetic techniques that have the greatest potential for revolutionizing these more traditional approaches. There is now increasing experience with PCR, and PCR and probe-based methods are being used with increasing frequency. Examples in work with seafood include the use of DNA probes for V. vulnificus (Wright et al., 1996) and pathogenic (tdh-, trh-, or tlh-containing) strains of V. parahaemolyticus (DePaola et al., 2000), and use of PCR assays for the tdh gene in assessing possible virulence of clinical and environmental V. parahaemolyticus strains (Young et al., 2002~. Further rapid advances in molecular diagnostics may be anticipated, includ- ing the development of some microarray assays for pathogenic microorganisms. Microarrays, as currently formulated, are multiple assay arrays on glass slides on which hundreds or thousands of probes are spotted, permitting a test sample to be screened against all probes simultaneously. Currently, the most common applica- tion of microarrays is to measure the presence and quantity of up to 20,000 messenger ribonucleic acid (mRNA) transcripts from mammalian cells (Schena et al., 1996~. However, genomic microarrays to distinguish among species of bacteria using the 16S ribosomal RNA gene have also been reported (Bavykin, 2001), with each probe on the microarray selected to identify a species of bacteria. In addition, microarrays have been used to identify genes lost between different strains of E. cold (Ochman and Jones, 2000), Helicobacter (Salama et al., 2000) and Staphylococcus (Fitzgerald et al., 2001~. With microarrays it is theoretically possible to immediately and quantitatively identify many, if not all, potential pathogens in a sample; to identify strains carrying specific virulence genes or strain subsets that have been linked with increased transmission potential (i.e., superclones); and to identify other genes of interest, including resistance genes.

26 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD While such microarray systems are not currently available commercially, they represent a very promising technology for food safety applications. The rapid advances being seen in this field of diagnostic technology under- score the need for flexibility in any regulatory approach or development of per- formance standards. This includes a need for flexibility at several levels. Currently, there is a perception on the part of regulatory agencies that iden- tification of a pathogen for regulatory purposes is not "real" unless a micro- organism is isolated. Regulations need to be changed to recognize that molecular and other rapid methods can produce results of comparable or greater accuracy than those obtained with traditional culture techniques; there must be provisions in regulatory actions for the use of data obtained with such methods. Any regulatory approaches, including the establishment of performance stan- dards, must have built into them sufficient flexibility to take advantage of the improvements in diagnostics that will almost certainly occur. THE LIMITS OF SCIENCE Some portion of the public surely is skeptical about most scientific pro- nouncements because of the seemingly conflicting advice, over time, from studies conducted in areas such as nutrition and health. However, the committee recog- nizes that many people believe that science and technology, given time and money, can fix everything. While this expectation may not hold for vexing problems deemed to be natural in origin (e.g., in respect to diseases such as cancer and acquired immune deficiency syndrome), man-made problems seem amenable to man-made solutions. Pathogens in store-bought foods are likely perceived by many as a man-made problem (e.g., E. cold in juices). When the committee held an open meeting to hear testimony from families that had suffered tragic losses from foodborne illness, the speakers (on the record as well as in private pleas in hallways after the session) urged committee members to "do something" to prevent others from suffering as they had. Eminent scientists, it was their heartfelt belief, could solve the problem. Scientists and engineers have developed skills and made discoveries that do enable the solutions to numerous problems of human origin. One example is the carnage done over the years because of vehicle accidents. Technological and legal changes that have made cars and their passengers safer have reduced the vehicular death and disability toll. While increased enforcement could further reduce the problem, this toll could be dramatically reduced through technology by designing all vehicles much like military tanks, but such a drastic step would dramatically increase the costs of vehicular travel and, through greater fuel needs, their environmental impact. Even where science and technology have solutions, their costs may be greater than society is willing to pay to achieve the projected benefits. In these cases, society must determine the trade-off between costs and benefits by tackling the question: What is the optimum level of safety we should

FOOD SAFETY TOOLS 127 seek to achieve? To pick an extreme example, it soon will be possible to test food for all pathogens and toxins of concern; all food could, in theory, be sampled prior to consumption. Such a system would of course be entirely impractical, both financially and logistically, although it would make the food almost thoroughly safe for the consumer. For our society, ensuring food safety is certainly an important goal that has not yet been adequately achieved. Policymakers who wish to improve the food safety system need to ensure adequate government financial resources for the creation and enforcement of safety rules. Food safety requirements imposed upon the food industry have financial consequences that may result in higher food prices. For example, significant changes could be made in animal husbandry and slaughter practices that would reduce the level of pathogens in food sold to the public. Science might be able to discover better, less expensive means to deal with pathogens in the food supply. Vaccines might be created that prevent food animals from being colonized by pathogens that, while harmless to the animals, are a danger to people. Simple, safe methods to kill pathogens on produce might be developed. Some scientific advances that their proponents claim will lead to a net benefit in food safety such as food irradiation and changes involving genetic modification are opposed by some members of the public because of concerns that one set of risks is being exchanged for another, to the frustration of many in the scientific community (Henderson, 2002~. Although there are limits to what science can achieve in consumer protec- tion, a more significant limit in the food safety system may well be the willing- ness of the public to accept the costs of implementing the measures that are available. Given the high costs to our society of morbidity and mortality that are related to foodborne illness, it would be sensible to require investment in food safety that yields a positive return. That is, to the extent that expenditures to improve food safety overall exceed the costs of the harm, these expenditures should definitely be made (and prices allowed to rise to cover the extra costs). Making such changes might interfere with consumer expectations about the low- cost availability of food. Some of the least-expensive interventions (such as hand washing by food handlers and improving retail worker and consumer compliance with safe food handling and cooking guidelines) are the most difficult to attain because they necessitate changing behaviors of vast numbers of people. How- ever, while everyone must purchase food and eat (and thus everyone has an interest in keeping down the cost of food), the harm from serious foodborne illness falls on a small fraction of the population. Are the many willing to devote resources to prevent serious harm to the few? Those who have lost loved ones (many of whom have been young children) to foodborne illness answer this question loudly in the affirmative; others are far less certain. While science and technology will continue to search for and discover answers to problems involv- ing foodborne illness, inexpensive answers are often unavailable or impractical. Where to draw the line between requirements that should be implemented and

28 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD that are reasonably cost-effective, and those that would be beneficial but would have too great an impact on food pnces, is a question for politics rather than for science. REFERENCES ASTM (American Society for Testing and Materials). 1976. Manual on Presentation of Data and Control Chart Analysis Committee E on Quality and Statistics. ASTM MNL7. Philadelphia: ASTM. Bavykin SO, Akowski JP, Zakhariev VM, Barsky VE, Perov AN, Mirzabekov AD. 2001. Portable system for microbial sample preparation and oligonucleotide microarray analysis. Appl Environ Microbiol 67:922-928. Bothe DR. 2001. Measuring Process Capability. Cedarburg, WI: Landmark Publishing. Breyfogle FW, Cupello JM, Meadows B. 2001. Managing Six Sigma: A Practical Guide to Under- standing, Assessing and Implementing the Strategy that Yields Bottom-Line Success. New York: John Wiley & Sons. Busta F. 2002. Application of Science to Food Safety Management. Presentation to the Institute of Medicine/National Research Council Committee on the Review of the Use of Scientific Criteria and Performance Standards for Safe Food, Washington, DC, April 3. CAC (Codex Alimentarius Commission). 1997. Hazard Analysis and Critical Control Point System and Guidelines for its Application. Annex to CAC/RCP 1-1969, Rev. 3-1997. Rome: Food and Agriculture Organization of the United Nations. Cassin MH, Lammerding AM, Todd ECD, Ross W. McColl S. 1998. Quantitative risk assessment of Escherichia cold 0157:H7 in ground beef hamburgers. Int J Food Microbiol 41 :21-44. CFSAN/FSIS/CDC (Center for Food Safety and Applied Nutrition/Food Safety and Inspection Service/Centers for Disease Control and Prevention). 2001. Draft Assessment of the Relative Risk to Public Health from Foodborne Listeria monocytogenes among Selected Categories of Ready-to-Eat Foods. Online. Food and Drug Administration (FDA), U.S. Department of Agri- culture (USDA). Available at http://www.cfsan.fda.gov/~dms/lmrisk.html. Accessed May 8, 2003. Cianfrani CA, Tsiakals JJ, West JE. 2002. The ASQ ISO 9000:2000 Handbook. Milwaukee: ASQ Quality Press. COST Action 920. 2000. Foodborne Zoonosis: A Co-ordinated Food Chain Approach. Online. Available at http://www.cost920.com. Accessed December 12, 2002. CVM (Center for Veterinary Medicine). 2001. The Human Health Impact of Fluoroquinolone Resis- tant Campylobacter Attributed to the Consumption of Chicken. Online. FDA. Available at http:/ /www.fda.gov/cvm/antimicrobial/revisedRA.pdf. Accessed August 1, 2002. DePaola A, Kaysner CA, Bowers J. Cook DW. 2000. Environmental investigation of Vibrio parahaemolyticus in oysters after outbreaks in Washington, Texas, and New York (1997 and 1998). Appl Environ Microbiol 66:4649-4654. Dourson ML, Andersen ME, Erdreich LS, MacGregor JA. 2001. Using human data to protect public health. Reg Toxicol Pharmacol 33:234-256. Escherichia cold O 157:H7 Risk Assessment Team. 2001. Draft Risk Assessment of the Public Health Impact of Escherichia cold 0157:H7 in Ground Beef. Online. FSIS, USDA. Available at http:// www.fsis.usda.gov/OPPDE/rdad/FRPubs/00-023N/00-023NReport.pdf. Accessed September 12, 2001. FAD/WHO (Food and Agriculture Organization of the United Nations/World Health Organization). 2000. Hazard Characterization, Exposure Assessment of Listeria monocytogenes in Ready-to- Eat Foods (RTE). Joint FAD/WHO Expert Consultation on Risk Assessment of Microbiological Hazards in Foods. Rome: FAO

FOOD SAFETY TOOLS 129 FAD/WHO. 2001. Hazard Identification, Exposure Assessment, and Hazard Characterization of Campylobacter spp. in Broiler Chickens and Vibrio spp. in Seafood. Joint FAD/WHO Expert Consultation on Risk Assessment of Microbiological Hazards in Foods. Geneva: WHO. FDA. 1999a. Food labeling: Safe handling statements: Labeling of shell eggs; Shell eggs: Refrigera- tion of shell eggs held for retail distribution; Proposed rule. Fed Regist 64:36491-36516. FDA. l999b. Grade "A" Pasteurized Milk Ordinance. 1999 Revision. Online. Available at http:// www.cfsan.fda.gov/~acrobat/pmo99-l.pdf. Accessed July 19, 2002. FDA. 2001. Hazard Analysis and Critical Control Point (HAACP); Procedures for the safe and sanitary processing and importing of juice; Final rule. Fed Regist 66:6137-6202. FDA. 2002. FDA Food Code. Online. Available at http://www.cfsan.fda.gov/~dms/foodcode.html. Accessed December 12, 2002. FDA/FSIS. 2001. Relative risk to public health from foodborne Listeria monocytogenes among selected categories of ready-to-eat food risk assessment document and risk management action plan. Fed Regist 66:5515-5517. Fitzgerald JR, Sturdevant DE, Mackie SM, Gill SR, Musser JM. 2001. Evolutionary genomics of Staphylococcus aureus: Insights into the origin of methicillin-resistant strains and the toxic shock syndrome epidemic. Proc Natl Acad Sci USA 98:8821-8826. FSIS. 1996. Pathogen reduction; hazard analysis and critical control point (HACCP) systems; Final rule. Fed Regist 61: 38805-38855. FSIS. 1998. Lethality and Stabilization Performance Standards for Certain Meat and Poultry Prod- ucts: Technical Paper. Online. USDA. Available at http://www.fsis.usda.gov/oa/haccp/ lethality.pdf. Accessed August 1, 2002. FSIS. 1999. 1999 National Residue Program: Residue Data. Online. USDA. Available at http:// www.fsis.usda.gov/OPHS/red99/intro.pdf. Accessed August 1, 2002. FSIS. 2001. HACCP-Based Inspection Models Project (HIMP): Young Chicken Inspection. Online. USDA. Available at http://www.fsis.usda.gov/OPPDE/Nis/HIMP/Docs/YNGChk_Drf6.pdf. Accessed July 19, 2002. Gallagher DL, Ebel ED, Kause JR. 2003. Draft FSIS Risk Assessment for Listeria in Ready-to-eat Meat and Poultry Products. Online. FSIS, USDA. Available at http://www.fsis.usda.gov/OPHS/ lmrisk/DraftLm22603.pdf. Accessed May 5, 2003. Garthright WE, Chirtel S. Graves Q. 2002. Derivation of Sampling Plan to Meet the Testing Require- ment in the Juice HACCP Final Rule for Citrus Juices that Rely Solely or in Part on Surface Treatments to Achieve the 5-Log Reduction Standard. Washington, DC: Office of Plant, Dairy Food and Beverages, CFSAN, FDA. Gavin A, Weddig L. 1995. Canned Foods: Principles of Thermal Process and Control, Acidifica- tion, and Container Closure Evaluation, 6th ed. Washington, DC: Food Processors Institute. Golan E. 2002. Performance Standards and the Economics of Compliance and Innovation. Presented at the USDA Symposium on Pathogen Reduction A Scientific Dialogue, Economic Research Service, USDA. Washington, DC, May 6. Hein I, Klein D, Lehner A, Bubert A, Brandl E, Wagner M. 2001a. Detection and quantification of the iap gene of Listeria monocytogenes and Listeria innocua by a new real-time quantitative PCR assay. Res Microbiol 152:37-46. Hein I, Lehner A, Rieck P. Klein K, Brandl E, Wagner M. 200 lb. Comparison of different approaches to quantify Staphylococcus aureus cells by real-time quantitative PCR and application of this technique for examination of cheese. Appl Environ Microbiol 67:3122-3126. Henderson M. 2002, September 9. Public "must allow scientists to take risks." Online. The Times (London). Available at http://www.timesonline.co.uk/article/0,,2-408587,00.html. Accessed October 10, 2002. Holcomb DL, Smith MA, Ware GO, Hung YC, Brackett RE, Doyle MP. 1999. Comparison of six dose-response models for use with food-borne pathogens. Risk Analysis 19:1091-1100.

30 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD ICMSF (International Commission on Microbiological Specifications for Foods). 1988. Microorgan- isms in Foods 4. Application of the Hazard Analysis Critical Control Point (HA CCP) System to Ensure Microbiological Safety and Quality. Oxford: Blackwell Scientific Publications. ICMSF. 1998. Potential application of risk assessment techniques to microbiological issues related to international trade in food and food products. J Food Prot 61:1075-1086. ICMSF. 2002. Microorganisms in Foods 7. Microbiological Testing in Food Safety Management. New York: Kluwer Academic/Plenum Publishers. IFT (Institute of Food Technologists). 2002. IFT Expert Report on Emerging Microbiological Food Safety Issues: Implications for Control in the 21st Century. Chicago: IFT. IOM (Institute of Medicine). 1990. Cattle Inspection. Washington, DC: National Academy Press. IOM. 1991. Seafood Safety. Washington, DC: National Academy Press. IOM. 2002. Escherichia cold 0157:H7 in Beef. Review of a Draft Risk Assessment. Washington DC: National Academy Press. IOM/NRC (National Research Council). 1998. Ensuring Safe Foodfrom Production to Consumption. Washington, DC: National Academy Press. Jensen HH, Unnevehr LJ, Gomez MI. 1998. Costs of improving food safety in the meat sector. J Agric Appl Econ 30:83-94. Kane VE. 1989. Defect Prevention: Use of Simple Statistical Tool. Milwaukee, WI: ASQ Quality Press. Karel M, Fennema OR, Lund DB. 1975. Physical Principles of Food Preservation. New York: Marcel Dekker. Ketola J. Roberts K. 2001. ISO 9001:2000 Management Responsibility in a Nutshell. Milwaukee, WI: ASQ Quality Press. Konz S. Perterson G. Joshi A. 1981. Reducing inspection errors. Quality Progress 14:24-27. Kroes R. Kozianowski G. 2002. Threshold of toxicological concern (TTC) in food safety assessment. Toxicol Let 127:43-46. Kuchler F. Golan E. 1999. Assigning Values to Life: Comparing Methods for Valuing Health Risks. Agricultural Economics Report No. 784. Washington, DC: Economic Research Service, USDA. Kume H. 1985. Statistical Methods for Quality Improvement. Tokyo: The Association for Overseas Technical Scholarship. Lammerding AM, Paoli GM. 1997. Quantitative risk assessment: An emerging tool for emerging foodborne pathogens. Emerg Infect Dis 3:483-487. Law AM, Kelton WD. 2000. Simulation Modeling and Analysis, 3rd ed. Boston: McGraw-Hill. Li WL, Drake MA. 2001. Development of a quantitative competitive PCR assay for detection and quantification of Escherichia cold 0157:H7 cells. Appl Environ Microbiol 67:3291-3294. LSRO (Life Sciences Research Office). 1995. Third Report on Nutrition Monitoring in the United States, Appendix III. Washington, DC: U.S. Government Printing Office. MacDonald JM, Crutchfield S. 1996. Modeling the costs of food safety regulation. Am JAgric Econ 78: 1285-1290. Maijala R. Lyytikainen O. Johansson T. Autio T. Aalto T. Haavisto L, Honkanen-Buzalski T. 2001. Exposure of Listeria monocytogenes within an epidemic caused by butter in Finland. Int J Food Microbiol 70:97-109. Markarian V, Hooker NH, Murano EA, Acuff GR, Carroll S. 2001. Comparative costs of pathogen reduction strategies for Australian beef slaughter plants. In: Hooker NH, Murano EA, eds. Interdisciplinary Food Safety Research. Boca Raton, FL: CRC Press. Pp. 25-41. Montgomery DC. 2001. Introduction to Statistical Quality Control, 4th ed. New York: John Wiley & Sons. NACMCF (National Advisory Committee on Microbiological Criteria for Foods). 1997. Recommen- dations on Fresh Juice. Online. FDA. Available at http://cfsan.fad.gov/~mow/nacmcf.html. Accessed January 23, 2003.

FOOD SAFETY TOOLS 131 NACMCF. 1998. Hazard analysis and critical control point principles and application guidelines. J Food Prot 61:1246-1259. Neubert D. 1999. Risk assessment and preventative hazard minimization. In: Marquardt H. Schafer SS, McClellan RO, Welsch F. eds. Toxicology. New York: Academic Press. Pp. 1153-1190. NRC (National Research Council). 1985a. An Evaluation of the Role of Microbiological Criteria for Foods and Food Ingredients. Washington, DC: National Academy Press. NRC. 1985b. Meat and Poultry Inspection: The Scientific Basis of the Nation's Program. Washing- ton, DC: National Academy Press. Ochman H. Jones IB. 2000. Evolutionary dynamics of full genome content in Escherichia colt. EMBO J 19:6637-6643. Paige JC, Pell F. 1997. Drug residues in food-producing animals. Online. FDA Veterinarian News- letter. Available at http://www.fda.gov/cvm/index/fdavet/1997/july.htm#res. Accessed April 11, 2003. Paustenbach DJ. 2000. The practice of exposure assessment: A state-of-the-art review. J Toxicol Environ Health 3: 179-291. Porter ME, van der Linde C. 1995. Green and competitive: Ending the stalemate. Harv Bus Rev 73: 120-134. Posnick L, Burr D, Bowers J. Walderhaug M, Miliotis M. 2001. Draft Risk Assessment on the Public Health Impact of Vibrio parahaemolyticus in Raw Molluscan Shellfish. Online. CFSAN, FDA. Available at http://www.cfsan.fda.gov/~dms/vprisksu.htm. Accessed August 1, 2002. President's Council on Food Safety. 1999. Egg Safety from Production to Consumption: An Action Plan to Eliminate Salmonella Enteritidis Illnesses Due to Eggs. Online. Available at http:// www.foodsafety.gov/~fsg/ceggs.htm. Accessed July 19, 2002. Rose JB, Haas CN, Regli S. 1991. Risk assessment and control of waterborne giardiasis. Am J Public Health 81 :709-713. RTI (Research Triangle Institute). 2000. HACCP-Based Inspection Models Project: Baseline Results for Young Chickens. Online. FSIS, USDA. Available at http://www.fsis.usda.gov/oa/haccp/ base_broilers.pdf. Accessed January 13, 2002. Salama N. Guillemin K, McDaniel TK, Sherlock G. Tompkins L, Falkow S. 2000. A whole-genome microarray reveals genetic diversity among Helicobacter pylori strains. Proc Natl Acad Sci USA 97:14668-14673. Salmonella Enteritidis Risk Assessment Team. 1998. Salmonella Enteritidis Risk Assessment: Shell Eggs and Egg Products. Online. FSIS, USDA. Available at http://www.fsis.usda.gov/OPHS/ risk/index.htm. Accessed August 1, 2002. Schena M, Shalon D, Heller R. Chai A, Brown PO, Davis RW. 1996. Parallel human genome analysis: Microarray-based expression monitoring of 1000 genes. Proc Natl Acad Sci USA 93: 10614-10619. Singer DL. 2001. A Laboratory Quality Handbook of Best Practices and Relevant Regulations. Milwaukee, WI: ASQ Quality Press. Steel RGD, Torrey JH, Dickey DA. 1997. Principles and Procedures of Statistics: A Biometrical Approach, 3rd ed. New York: McGraw-Hill. Stevenson KE, Bernard DT, eds. 1995. HACCP. Establishing Hazard Analysis Critical Control Point Programs. A Workshop Manual, 2nd ed. Washington, DC: Food Processors Institute. Taylor M. 2002. Microbiological Performance Standards for Food Safety. Presentation to the Insti- tute of Medicine/National Research Council Committee on the Review of the Use of Scientific Criteria and Performance Standards for Safe Food, Washington, DC, April 3. USDA/FDA. 2002. HACCP Training Programs. Online. National Agricultural Library. Available at http://www.nal.usda.gov/foodborne/haccp/training.html. Accessed October 16, 2002. Vose D. 2000. Risk Analysis: A Quantitative Guide, 2nd ed. New York: John Wiley & Sons. Wheeler DJ, Chambers DS. 1992. Understanding Statistical Process Control, 2nd ed. Knoxville, TN: SPC Press.

32 SCIENTIFIC CRITERIA TO ENSURE SAFE FOOD Whiting RC, Buchanan RL. 1997a. Development of a quantitative risk assessment model for Salmo- nella Enteritidis in pasteurized liquid eggs. Int J. Food Microbiol 36:111-125. Whiting RC, Buchanan RL. 1997b. Predictive modeling. In: Doyle MP, Beuchat LR, Montville TJ, eds. Food Microbiology: Fundamentals and Frontiers. Washington DC: ASM Press. Pp. 728- 739. WHO. 1999. HACCP Principles and Practice. Teacher's Handbook. CD-ROM. Geneva: WHO. Woteki CE. 2000. Is There a Place for Food Safety Objectives in USDA's Food Safety Program? Online. FSIS, USDA. Keynote Address to the CERES Forum On Food Safety Objectives, Washington, DC, December 4. Available at http://www.fsis.usda.gov/OA/speeches/2000/ cw_ceres.htm. Accessed May 20, 2002. Wright AC, Hill RT, Johnson JA, Roghman M-C, Colwell RR, Morris JO Jr. 1996. Distribution of Vibrio vulnificus in the Chesapeake Bay. Appl Environ Microbiol 62:717-724. Yeung PSM, Hayes MC, DePaola A, Kaysner CA, Kornstein L, Boor KJ. 2002. Comparative pheno- typic, molecular, and virulence characterization of Vibrio parahaemolyticus 03:K6 isolates. Appl Environ Microbiol 68:2901-2909.

Next: 4 Scientific Criteria and Performance Standards to Control Hazards in Meat and Poultry Products »
Scientific Criteria to Ensure Safe Food Get This Book
×
Buy Paperback | $68.00 Buy Ebook | $54.99
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

Food safety regulators face a daunting task: crafting food safety performance standards and systems that continue in the tradition of using the best available science to protect the health of the American public, while working within an increasingly antiquated and fragmented regulatory framework. Current food safety standards have been set over a period of years and under diverse circumstances, based on a host of scientific, legal, and practical constraints.

Scientific Criteria to Ensure Safe Food lays the groundwork for creating new regulations that are consistent, reliable, and ensure the best protection for the health of American consumers. This book addresses the biggest concerns in food safety—including microbial disease surveillance plans, tools for establishing food safety criteria, and issues specific to meat, dairy, poultry, seafood, and produce. It provides a candid analysis of the problems with the current system, and outlines the major components of the task at hand: creating workable, streamlined food safety standards and practices.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!