SEARCH

SEARCH BY CITATION

Keywords:

  • food safety;
  • risk;
  • risk ranking

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Process
  5. Model Components
  6. Risk-Ranking Output
  7. Conclusions
  8. Acknowledgments
  9. References

ABSTRACT:  Through a cooperative agreement with the U.S. Food and Drug Administration, the Institute of Food Technologists developed a risk-ranking framework prototype to enable comparison of microbiological and chemical hazards in foods and to assist policy makers, risk managers, risk analysts, and others in determining the relative public health impact of specific hazard–food combinations. The prototype is a bottom-up system based on assumptions that incorporate expert opinion/insight with a number of exposure and hazard-related risk criteria variables, which are propagated forward with food intake data to produce risk-ranking determinations. The prototype produces a semi-quantitative comparative assessment of food safety hazards and the impacts of hazard control measures. For a specific hazard–food combination the prototype can produce a single metric: a final risk value expressed as annual pseudo-disability adjusted life years (pDALY). The pDALY is a harmonization of the very different dose–response relationships observed for chemicals and microbes. The prototype was developed on 2 platforms, a web-based user interface and an Analytica® model (Lumina Decision Systems, Los Gatos, Calif., U.S.A.). Comprising visual basic language, the web-based platform facilitates data input and allows use concurrently from multiple locations. The Analytica model facilitates visualization of the logic flow, interrelationship of input and output variables, and calculations/algorithms comprising the prototype. A variety of sortable risk-ranking reports and summary information can be generated for hazard–food pairs, showing hazard and dose–response assumptions and data, per capita consumption by population group, and annual p-DALY.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Process
  5. Model Components
  6. Risk-Ranking Output
  7. Conclusions
  8. Acknowledgments
  9. References

Risk analysis is an essential part of science-based policies for food safety and public health protection today (Jaykus and others 2006). Food safety risk assessments completed to date typically focus on a single food product-pathogen pair such as Salmonella in eggs (USDA-FSIS 1998), a single agent such as mercury (Carrington and Bolger 2002), or a pathogen such as Listeria monocytogenes (FDA-CFSAN and others 2003) in one or a few specific food products. Food safety risk assessments today are not typically designed to quantitatively compare and rank risks of different food safety hazards (for example, microbiological hazards compared with chemical ones) because of the complexity of the calculations and comparisons required. A well-conceived strategic approach to public health protection that quickly and accurately identifies different types of hazards, ranks them by level of importance, and identifies approaches with the greatest potential to reduce hazards is critically needed (IFT 2002).

Risk ranking has been applied previously in a variety of settings, but very little activity has been applied to rank different types of risks in food systems. Havelaar and Melse (2003) maintained that to reduce the risk of foodborne illness, the relative risk across the different types of hazards should be compared. The U.S. Food and Drug Administration (FDA) awarded the Institute of Food Technologists (IFT) a 2-year cooperative agreement grant that supported development and implementation of a risk-ranking framework to evaluate potential high-threat microbiological agents, toxins, and chemicals in food. The framework was to include a model for quantitatively or semi-quantitatively comparing and determining potential threats and the ability to evaluate interventions or control points (for example, manufacturing/processing, warehouses, transport, retail) at various places in the farm-to-fork chain. Implementation of the framework would include use of existing and newly developed lists of hazardous agents for systematic ranking. Further, the FDA desired use of criteria in the risk ranking that at a minimum pertained to compatibility of a hazard with food as a vehicle, toxicity (or dose necessary to result in disease), accessibility, and likelihood of effect (illness). While many risk-ranking approaches are possible, the approaches fall into 2 main groups: surveillance-based “top-down” approaches and prediction-based “bottom-up” approaches.

Top-down and bottom-up approaches to risk ranking

With respect to microbial hazards, surveillance-based approaches attempt to infer the level of risk due to foods, hazards, or their combinations based on information gathered by various observation systems such as active or passive disease reporting systems, outbreak databases, and a variety of other observations such as prevalence of pathogens in various commodities. Such information sources may be best for overall ranking of pathogens, but quantitative linkages to particular foods are often very difficult to justify from these sources alone and are typically estimated only for foods that might be attributed to a relatively high percentage of the attributable risk. The Foodborne Illness Risk Ranking Model (FIRRM), initiated in 2003 by the Food Safety Research Consortium, is an example of such top-down approaches to risk ranking (FSRC 2005). The FIRRM integrates data on foodborne illness surveillance; food–pathogen combinations; medical symptoms, complications, and outcomes; economic impact; and social values relevant to judging the significance of a potential hazard to population health.

In most cases, there is no systematic capacity to observe the effects of food-associated chemical exposures in the human population. This is because of a number of challenges, including the many potential causes of symptoms, the sheer number of chemicals that have common outcomes, and the long latency between exposure and outcomes. In addition, many chemical exposures occurring as a consequence of food consumption are at levels believed to be so low that there may not be any readily observable effects for a vast majority of exposed consumers.

The other main group of ranking approaches is based on predictive modeling of the fate of microbes and chemicals in the food supply together with their virulence or toxicity. The FDA's charge to the IFT panel included the capability to deal with a variety of microbial and chemical hazards. Given this and the inherent difficulties associated with top-down approaches for both microbial and chemical hazards noted previously, a bottom-up or predictive model of risk was used as the underlying framework for the ranking application described here. This requires the application of data and expert judgment to assemble sufficient information to predict the fate of the hazards in the food supply, together with their virulence and toxicity characteristics, to generate a prediction (which may be, of necessity, quite crude) of their relative level of risk to human health and the potential for changes to level of risk associated with possible interventions throughout the farm-to-fork chain.

The Process

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Process
  5. Model Components
  6. Risk-Ranking Output
  7. Conclusions
  8. Acknowledgments
  9. References

IFT convened a panel of individuals with expertise in the farm-to-fork food system, food safety, risk assessment and management, microbiology, chemistry, toxicology, predictive microbiology, and computer modeling to develop the risk-ranking framework prototype. IFT staff experts in food safety and project management helped support the initiative. IFT supplemented the panel's expertise and efforts with additional developmental assistance by experts affiliated with risk, food, and chemical consultancies with expertise in food safety, biochemistry, environmental health science, public health, risk analysis, computer programming, and Web technology. The initial concept for the framework, which contributed to deliberations and subsequent prototype development, included an expert elicitation framework, tools, and envisioned information from several sources: expert panel judgment, evidence databases, value models, assessment assumptions, and policy options. This concept would feed into methodological research summary reports that were envisioned to aid the risk-ranking activities of the FDA and other possible users.

Model Components

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Process
  5. Model Components
  6. Risk-Ranking Output
  7. Conclusions
  8. Acknowledgments
  9. References

The panel developed 2 main risk criteria modules: exposure (farm-to-fork) and hazard characterization (health impacts). The exposure module contained questions grouped into 3 food system stages: primary production; processing; and distribution, storage, retail, foodservice, and home. Questions comprising the hazard characterization module addressed agent pathogenicity or toxicity and potential public health burden. Formats for the answers to the explicit questions were qualitative (for example, high, medium, low, likely/not likely), quantitative (metric/scale), objective (available data), subjective (expertise), and rationale based.

Metrics (values assigned to individual risk input criteria) for the factors in the 2 modules were systematically developed. Metrics for levels of consumption of the identified food types of primary concern were compiled using the U.S. Dept. of Agriculture's 1994–1998 CSFII food intake database. The risk criteria comprising the 2 modules were integrated via an algorithm approach.

User inputs

Prototype users are prompted by specific questions for pertinent details on hazard prevalence, concentration, and changes in concentration at each of the 3 food system stages. Monte Carlo simulation computes mean final log concentrations from triangular distributions (minimum, most likely, or maximum log concentration value). To address health impacts, users are prompted to describe and assign importance to health impacts through pseudo-disability adjusted life years (pDALY). The pDALY concept is modified slightly from the general use of DALY (IOM 2005) to allow for a semiquantitative characterization of the disease burden of health impacts. The usual approach to measuring DALY is to assign a severity weight and duration weight to discrete relatively well-characterized health outcomes. The pDALY approach allows for the characterization of a standard health outcome (such as mild illness) without further definition of the exact impact. This was developed primarily to facilitate risk ranking of chemical substances that may present a risk of diverse, poorly characterized outcomes (for example, noncancer toxicity), which may not be easily assigned individual weights and durations.

Users create pDALY templates by assigning a fraction of cases to appropriate health impacts, such as mild, moderate, or severe pathogen, and short-term, adult, elderly, or childhood mortality. Some questions have predefined answers connected with predefined weights for risk-ranking calculations. Guidance exists in the form of help files that facilitate user responses to questions. Users can assign one or more dose–response functions to hazard outcome types, such as cancer or chronic noncancer. Users select the functional form of the dose–response relationship and record appropriate parameters for the chosen dose–response function.

Hazard–food pairs

IFT identified and incorporated into the prototype a number of hazard–food pairs (Table 1) to test the questions developed for the modules and the respective decision logic and to evaluate the metrics, ranking processes, and outcomes. The hazards for the pairs were chosen on the basis of participant knowledge of the hazard. To ensure that the prototype could address the full range of possible outcomes of varying severity and uncertainties, the chemical hazards were also chosen on the basis of conveniently available residue data, comparability to selected microbial hazards, and presence of multiple potential toxic endpoints. The prototype can accommodate additional pathogens and chemical toxicants and other hazard–food pairs, such as combinations involving food canning and post-lethality processing of ready-to-eat (RTE) product or scenarios involving home food storage or preparation (for example, L. monocytogenes and temperature-abused RTE luncheon meat).

Table 1—.  Hazard–food pairs used for prototype testing.
Arsenic and smoked salmon
Bacillus cereus and liquid, extended-shelf-life coffee creamer in individual serving units
Benomyl and apple juice
Clostridium perfringens and beef broth-based gravy prepared in a restaurant
Cyclospora cayetanensis and fresh raspberries
Dioxin and lettuce
Dioxin and fresh green onions
Dioxin and cheddar cheese
Dioxin and whole milk
Escherichia coli O157:H7 and apple juice
E. coli O157:H7 and sprouts
Enterobacter sakazakii and powdered infant formula
Fumonisin and canned corn
Hepatitis A virus and fresh strawberries
Hepatitis A virus and raw oysters
Listeria monocytogenes and whole milk
Methyl mercury and smoked salmon
Nitrate and smoked salmon
Nitrite and smoked salmon
Norovirus and raw oysters
Salmonella spp. and powdered milk
Salmonella spp. and raw oysters
Shigella dysenteriae and fresh green onions
Staphylococcus aureus enterotoxin and natural cheddar cheese

Prototype characteristics and functionality platforms

The prototype exists on 2 platforms: a web-based user interface, implemented in Visual Basic language and an Analytica® model. The web-based platform was developed to provide a user-friendly input/output user interface that facilitates concurrent use and data sharing without significant time delay. More specifically, the web-based platform (Figure 1) allows users to explore the complex ranking hierarchy, view the current evidence, edit evidence, and update assumptions. Calculations are performed in the web-based implementation using Visual Basic. Microsoft Access, a relational database, stores the relationships between variables (foods, hazards, processes, and evidence) that apply to each individually and their many combinations.

image

Figure 1—. Initial view: main page of web-based prototype implementation.

Download figure to PowerPoint

The Analytica model (Figure 2), which complements the web-based prototype application, facilitates visualization of the logic flow and interrelationship of input and output variables. It also allows inspection and auditing of the calculations comprising the prototype. Appropriate consumption measures with census-based population size estimates pulled from the database serve as the basis for risk calculations. Although the Analytica model reproduces the web-based calculations exactly, it allows only calculations based on a single hazard–food pair and does not allow relative risk rankings of different hazard–food pairs. The Analytica model was designed for the initial development of the calculations, given the visualization and computational features of the software, to facilitate further development, discussion, and review of the algorithms. The web-based implementation was then compared with the Analytica-based calculations to ensure that the implementation was sound.

image

Figure 2—. Initial view of Analytica model.

Download figure to PowerPoint

Characteristics and functionality

Two main components make up the key conceptual features of the risk-ranking prototype: computer programming code integrating exposure and hazard characterization modules and risk information data. The framework characterizes the burden of disease for health impacts associated with hazards through illness duration and severity. It also links health impact categories to hazards through the pDALY, a simplified way of addressing burden of disease. CSFII 1994–1998 data were used to estimate the proportion of the population(s) potentially exposed to the hazard and the amount of food eaten.

The prototype generally incorporates empirical evidence (CSFII food intake data, dose response data, and residue data), expert rationale, and module integration algorithms (via Visual Basic language) and provides output in the form of risk-related evidence, assumptions, and risk-ranking reports. Thus, while the product is a prototype for a risk-ranking framework, there is inherent value in the knowledge comprising the prototype.

The framework is not intended to replace or substitute for more complex single hazard–food pair risk assessments since the level of detail is limited in the interest of allowing comprehensive and rapid ranking of many hazard–food pairs. Instead, the framework can provide a comparative risk rank for hazard–food pairs, expressed as annual pDALY. The risk-ranking section of the web-based version uses Monte Carlo simulation to compute a range of doses based on the concentration of the hazard in the food and the average serving size. The doses are used in conjunction with the dose–response model(s) for the hazard to compute a mean probability of illness for each population group. Prevalence values are then used to determine the number of contaminated servings. Triangular distributions were chosen for simplicity and ease of change; other distributions could readily be utilized in future iterations of the model. Combining the number of contaminated servings with the probability of illness and the pDALY template value for the hazard generates a final risk measure (annual pDALY). For chemical hazards, risks that are inferred based on lifetime exposures are prorated to an annual risk estimate by dividing by an arbitrary lifetime value of 70 y (consistent with the value used by the FDA and the Environmental Protection Agency) to allow for compatible timeframes for ranking. Alternatively, acute hazards (primarily microbial hazards) can be multiplied by the same factor to estimate compatible lifetime burden of disease measures. Tables 2 and 3 show the input and output variables of the prototype.

Table 2—.  Risk-ranking prototype input variables.a
  1. aAs shown in the input/output user interface Analytica node.

Initial prevalence
Initial concentration before processing
Change in concentration at primary production
Likelihood of introduction at primary production
Introduced concentration at primary production
Change in prevalence during primary production
Change in concentration at processing
Likelihood of introduction at processing
Introduced concentration at processing
Change in prevalence (processing)
Change in concentration at distribution, storage, retail, foodservice, and in the home
Likelihood of introduction at distribution, storage, retail, foodservice, and in the home
Introduced concentration at distribution, storage, retail, foodservice, and in the home
Change in prevalence at distribution, storage, retail, foodservice, and in the home
Total eating occasions/exposed population
Grams per eating occasions
pDALY per illness
Daily consumption
Dose–response model
 Beta-Poisson
 Exponential
 Linear
 Chemical cancer
 Chemical noncancer
Noncancer method
 Threshold
 Linear model threshold
 Linear model nonthreshold
Hazard
 Microbial or chemical/toxin
 Dose
 RfD
 Threshold
Table 3—.  Risk-ranking output variables.a
  1. aAs shown in the input/output user interface Analytica node.

Final mean concentration in positive lots
Final mean prevalence
Mean probability of illness
Number of illnesses
Annual pDALY

Another advantage of the prototype is its flexibility. For example, one could consider seasonal and geographic impacts on hazard prevalence, contaminated servings, and subsequent risk rank by addressing the appropriate number of suitably defined hazard–food pairs in the web-based implementation. An example of this would be Vibrio vulnificus in raw oysters harvested from the Gulf Coast during summer compared with winter. Similarly, the risk rank of a hypothetical intentional contamination event could be considered by incorporating the hypothetical hazard prevalence, concentration, and locations within the food chain in which contamination occurs.

Exposure module

The panel chose the 3 main food system stages—primary production (includes harvesting); processing (includes post processing); and distribution, storage, retail, foodservice, and home—to enable representation of key points at which hazard prevalence and concentration could change throughout the food system. In the future, the capability exists to address transport of source materials or animals prior to processing or food product subsequent to processing at any of the food system stages. Within each of these 3 food system stages, hazard presence is considered on a bulk lot or truckload type basis rather than by individual consumer or retail units.

The prototype addresses hazard concentration via initial concentration, in log units/g for microbes and g/g for chemicals, at the earliest point of primary production before any known production, processing, distribution/storage-related changes might occur. Subsequent concentration as a result of any increases or decreases or additions (introduction of contamination) occurring during the 3 food system stages is also addressed. The simulation engine examines each possible pathway of contamination explicitly, and the resulting concentrations are weighted by their respective probability of occurrence calculated in concentration weights. As a result, 16 pathways track probabilities for concentration throughout each of the 3 food system stages.

The prototype addresses hazard prevalence more simply by estimating the likelihood of hazard introduction at each of the 3 stages, changes in hazard prevalence during each stage, prevalence at the end of each stage, and final prevalence at the end of the continuum. The calculations for prevalence estimate the concentration of the agent at the end of the farm-to-fork chain based upon the changes in concentration (increases or decreases) and additions that occur throughout the food system as defined by the user. Initial prevalence is expressed on the basis of percentage of total units in which the hazard is present (contaminated units/total units, 0% to 100%). Change in prevalence (occurring independently of initial concentration), change in concentration, or introduced concentration within each of the 3 food system stages is addressed with values between 0 and 1 reducing the prevalence by that factor, values greater than 1 increasing the prevalence by that factor, and a value of 1 leaving the prevalence unchanged.

In allowing the user to address likelihood for introduction or addition of a hazard during each of the stages, the prototype has placeholders for future developmental efforts to address controllability efficacy and controllability compliance. This is based on the understanding that the existence of guidance or regulation to describe how a hazard enters the food chain and the ability to control a hazard is a relevant consideration in risk ranking. For example, if a hazard were controllable, then a risk-rank metric could be used for mitigation, or if not controllable, then the rank could be used in considering the need for research. These considerations, which are managerial in nature, do not currently lend themselves to an obvious numeric or ranking, but this may change with future iterations of the prototype.

Consumption (food intake) submodule

The consumption/food intake submodule addresses the proportion of the population that is exposed to the hazard and the amount of a given food that is eaten. Due to the large number of as-eaten foods in the U.S. Dept. of Agriculture's 1994–1998 CSFII 8-digit food-code database, expert panel members determined that an aggregate approach based on 3- and 5-digit levels of food intake data would be sufficient and effective for developing quantitative metrics for risk-ranking purposes. CSFII data are based on 4 population groups: the entire United States, women 16 y to 49 y of age, children 1 y to 6 y of age, and individuals 65 y of age and older. Users may also specify what percentage of a given population is at risk.

Chemical risks are computed using the mg/kg bw/day consumption measure (in which bw = body weight). Population size based on census estimates for each population group is in the database to compute population risk for chemicals. Microbial risk is calculated using mean serving size and total number of servings. For chemical hazards, risk (probability of illness) is calculated on the basis of 90th percentile for consumption.

Hazard characterization (health impacts) module

Multiple dose responses can be assigned to hazard outcome types (for example, cancer, acute or chronic noncancer [for chemicals] and infectious or toxigenic [for microorganisms]). Each dose response option subcategory offers a subset of appropriate dose–response models. When users address a hazard and corresponding dose–response models, they will encounter the question “What is the strength of judgment that this hazard causes adverse health effects?” for which there are 4 possible responses: no studies available, not well established, moderate evidence, or well established. Because the responses to the question do not readily lend themselves to numeric expression, they are not currently factored into the risk ranks. Nevertheless, the information is pertinent and provides justification which, at some future time, may lead to a more quantitative expression of strength of supporting evidence.

For toxicological dose–response relationships (chemical and toxin-producing microbial hazards), 5 models are available: step threshold, threshold linear, nonthreshold linear, beta-Poisson, and exponential. For infectious dose responses, 4 models are available: beta-Poisson, exponential, threshold linear, and nonthreshold linear. The dose–response templates cannot be changed by users. The dose–response section of the prototype shows appropriate parameters for the selected model; changing the model changes the parameters for the options provided. All dose–response pages allow consideration of probability of illness given response, addressing the question of what proportion of infections would result in illness. All dose–response curves are incorporated into the risk calculations. Users may choose from any number of health impacts, which basically represent a DALY approach (Table 4) and then link them with one or more of the pDALY templates (Table 5).

Table 4—.  Health impacts.
Mild, short-term impacts
Mild, medium-duration impacts
Mild, long-term impacts
Moderate, short-term impacts
Moderate, medium-duration impacts
Moderate, long-term impacts
Severe, short-term impacts
Severe, medium-duration impacts
Severe, long-term impacts
Childhood mortality
Adult mortality
Elderly mortality
Hemorrhagic colitis
Hemolytic uremic syndrome
Enteric fever
Reactive arthritis/Reiter's syndrome
New health impact
Table 5—.  pDALY templates.
Acute (chemicals)
Blood target organ (chemical)
Cancer (chemical)
Escherichia coli O157:H7
Gastroenteritis only (rare fatality)
Hepatitis A virus
Neural tube defect
Neuro-developmental (chemical—below BmD)
Reproductive (chemical)
Salmonella
Severe pathogen
New pseudo DALY template

The pDALY template allows the impact of the hazard to be placed on a relative scale. The results of exposure are captured semi-quantitatively in 2 dimensions: impact severity (mild, moderate, severe, or death) and duration (short, medium, or long), allowing up to 12 ways to describe a health impact. In addition, when selecting a specific health impact, users may indicate and provide support for their choice of health impact, duration, and severity.

Other prototype characteristics

The prototype addresses microbial risk as represented by colony forming units at the point of consumption and does not track toxin production occurring throughout the food chain (for example, staphylococcal enterotoxin formation). Strain-to-strain differences in virulence of microorganisms are not included nor are differences in immunity among individuals because of innate or acquired immunity, such as resistance to certain pathogens (such as norovirus and hepatitis A virus).

Additionally, the model is very sensitive to situations where a microbial hazard has a toxigenic response characterized by a threshold linear model, as observed for C. perfringens and beef gravy. This sensitivity exists because the dose–response model contains a threshold below which a response does not occur and above which it does. Thus, when the predicted concentration of the pathogen is close to the threshold, very slight increases in the concentration of the pathogen can result in very large changes in health effects. The prototype has the capability of accommodating a number of possible modifications:

  • • 
    Inserting additional scientific documentation;
  • • 
    Allowing assignment of a relative estimate of data quality;
  • • 
    Adding more inputs for multiple hazard reductions;
  • • 
    Considering factors that contribute to a decrease or increase of a food hazard (as might occur during in-home preparation or storage);
  • • 
    Integrating the web-based implementation with the Analytica model (allowing users to view and address more than one hazard–food pair at the same time);
  • • 
    Allowing answers to the strength of judgment and hazard controllability questions to be factored into the risk-ranking output to address uncertainty associated with these factors;
  • • 
    Accommodating the input of confidence intervals for input and output estimates;
  • • 
    Considering the benchmark dose lower confidence limit as a risk measure rather than the reference dose;
  • • 
    Standardizing the dose–response modeling for different categories of chemical hazards;
  • • 
    Incorporating consumption data (for example, data from the National Health and Nutrition Examination Survey data); and
  • • 
    Including additional data that would enhance the strength of the exposure and hazard characterization modules (for example, data pertaining to dose response).

Risk-Ranking Output

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Process
  5. Model Components
  6. Risk-Ranking Output
  7. Conclusions
  8. Acknowledgments
  9. References

The prototype provides a basic reporting mechanism that reports selected contents of the database (the evidence) according to foods, hazards, processes, and their combinations. A risk-ranking summary report can be generated, grouped by hazard or food; ordered by total risk or name; and produced in ascending or descending order. Total risk (pDALY) is aggregated by hazard or food depending on the grouping selected. The application sums the pDALY measures as a total risk for a particular food or hazard, depending on the grouping selected. In addition, users have the option to specify foods, hazards, or hazard–food combinations that are to be excluded from rankings due to incompleteness of data or development of assumptions. Checking the pertinent box on the food, hazard, and hazard–food pages determines whether they are included in the ranking. The individual food and hazard settings take priority over the combination of settings.

For the dose–response relationship, the risk-ranking summary report summarizes the type, model, and parameters of the dose–response; grams per eating occasion; total number of eating occasions; mean hazard prevalence; number of contaminated servings from once contaminated lots; mean concentration in food; mean dose; mean probability of illness; number of illnesses; pDALY per illness; and annual pDALY. By default, the risk-ranking summary report prints the 1st dose–response chart, but other charts are included. The “print summary” function produces a summary of the evidence entered and is distributable for discussion and holistic consideration.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Process
  5. Model Components
  6. Risk-Ranking Output
  7. Conclusions
  8. Acknowledgments
  9. References

In cooperation with the FDA, IFT participants in this study developed a functional semi-quantitative risk-ranking framework prototype—a flexible tool that enables relative comparison and ranking of microbial food-related risks with chemical risks via a single metric: annual pDALY. Specific approaches taken in developing the prototype enabled resolution of some broad challenges faced in risk-ranking efforts. The successful production of this risk-ranking prototype holds tremendous potential as a unique tool capable of comparing microbial hazards and chemical hazards not only separately but also comparatively by using a common metric.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Process
  5. Model Components
  6. Risk-Ranking Output
  7. Conclusions
  8. Acknowledgments
  9. References

This study was supported under FDA cooperative agreement grant number FD-U-002255-01 and by additional funding from the Food Risk Analysis Initiative of Rutgers Univ.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Process
  5. Model Components
  6. Risk-Ranking Output
  7. Conclusions
  8. Acknowledgments
  9. References
  • Carrington CD, Bolger MP. 2002. An exposure assessment for methylmercury from seafood for consumers in the United States. College Park , Md. : U.S. Food and Drug Administration/Center for Food Safety and Applied Nutrition: Available from: http://www.cfsan.fda.gov/~acrobat/mehgra2.pdf. Accessed Sept 4, 2008.
  • FDA-CFSAN, USDA-FSIS, CDC. 2003. Quantitative assessment of relative risk to public health from foodborne Listeria monocytogenes among selected categories of ready-to-eat foods. U.S. Food and Drug Administration/Center for Food Safety and Applied Nutrition, U.S. Dept. of Agriculture/Food Safety and Inspection Service, Centers for Disease Control and Prevention. Available from: http://www.foodsafety.gov/~dms/lmr2-toc.html. Accessed Sept 4, 2008.
  • [FSRC] Food Safety Research Consortium. 2005. Prioritizing opportunities to reduce the risk of foodborne illness: a conceptual framework. Available from: http://www.card.iastate.edu/food_national_conference/FSRC_Conceptual_Framework_final.pdf. Accessed Sept 4, 2008.
  • Havelaar AH, Melse JM. 2003. Quantifying public health risk in the WHO guidelines for drinking-water quality. RIVM report nr 734301022/2003. Available from: http://www.who.int/water_sanitation_health/dwq/rivmrep.pdf. Accessed Dec 1, 2008.
  • [IFT] Institute of Food Technologists. 2002. Emerging microbiological food safety issues: implications for control in the 21st century. An expert report. Chicago , Ill. : Institute of Food Technologists. Available from: http://www.ift.org/cms/?pid=1000379. Accessed Nov 20, 2008.
  • [IOM] Institute of Medicine. 2005. Estimating the contributions of lifestyle-related factors to preventable death: a workshop summary. Washington , D.C. : National Academies Press, 80 p.
  • Jaykus L, Dennis S, Bernard D, Claycamp HG, Gallagher D, Miller AJ, Potter M, Powell M, Schaffner D, Smith MA, Ten Eyck T. 2006. Using risk analysis to inform microbial food safety decisions. Issue paper nr 31. Ames , Iowa : Council for Agricultural Science and Technology.
  • Murray CJL, Lopez AD, Mathers CD, Stein C. 2001. The global burden of disease 2000 project: aims, methods and data sources. Global programme on evidence for health policy discussion paper nr 36. Geneva , Switzerland : World Health Organization. Available from: http://www.who.int/healthinfo/paper36.pdf. Accessed Nov 8, 2007.
  • [USDA-FSIS] U.S. Dept. of Agriculture, Food Safety and Inspection Service. 1998. Salmonella Enteritidis risk assessment: shell eggs and egg products. Final report. Washington , D.C. : U.S. Dept. of Agriculture, Food Safety and Inspection Service. Available from: http://www.fsis.usda.gov/ophs/risk/. Accessed Sept 4, 2008.