Letter Report on the Review of the Food Safety and Inspection Service Proposed Risk-Based Approach to and Application of Public-Health Attribution (2009)

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.

Attachment G FOODBORNE DISEASE ATTRIBUTION6 1.0 Introduction The Food Safety and Inspection Service (FSIS) is proposing a public health risk-based inspection system (PHRBIS) for meat and poultry processing and slaughter establishments. The components of the proposed PHRBIS are science-based and are being designed with input from stakeholder groups and expert peer review. One component of the PHRBIS is a set of criteria for categorizing processing and slaughter establishments with respect to their potential impact on public health. A basic element of prioritizing and allocating resources to reduce the level of foodborne illness is the ability to identify which FSIS-inspected food products are major contributors to human foodborne illness. This report gives an overview of an approach for performing microbial foodborne disease attribution. FSIS acknowledges that no system of estimating foodborne disease attribution is perfect. The best current estimates come from combined consideration of illness outbreak data, illness case-control studies, risk assessments, pathogen serotype data, and expert elicitation (Batz et al. 2005). FSIS has adopted this approach and considered the best information currently available. FSIS, in conjunction with CDC and FDA is investigating methods, such as using serotypes and subtypes of pathogens to improve attribution estimates. FSIS will use these and other advances to improve foodborne disease attribution estimates as better information becomes available. The Centers for Disease Control and Prevention (CDC) in its Healthy People 2010 program, for which Food Safety Inspection Service (FSIS) and the Food and Drug Administration (FDA) are the food safety co-leads, has set to goal of decreasing Salmonella species, Campylobacter species, E. coli O157:H7, and Listeria monocytogenes infections each by 50% by the year 2010 from the period 1996- 1998. It is generally agreed that the best manner of achieving these goals is to focus regulatory attention on those food types that contribute the largest burden of illness for each of these pathogens. This necessitates knowledge of what fraction of foodborne human illness results from consumption of specific food items. This knowledge is called foodborne disease attribution. Estimates of foodborne attribution are pathogen specific, that is, the percentage of disease attributable to a particular food type (i.e., consumption of beef, chicken, eggs or produce) will vary from pathogen to pathogen. The Food Safety Inspection Service (FSIS) has, through contractors, elicited the opinion of experts to rank the contribution of various types of processed meat and poultry products to disease caused by Salmonella, Listeria monocytogenes, and E. coli O157:H7. The last two elicitations (Karns et al. 2005, 2007) produced similar attribution results. Nevertheless, there has been some hesitation in using these attribution estimates in a regulatory framework since they are based on expert opinion rather than empirical data. The purpose of this report is two fold. The first is to compare the attribution estimates obtained from the FSIS 2007 expert elicitation with those from two other sources: an independent expert elicitation performed by Resources for the Future (RFF)/Carnegie Mellon and those derived from a disease outbreak database complied by the Centers for Disease Control and Prevention (CDC). The second purpose is to use these three studies to develop more informed estimates of foodborne disease attribution for 25 meat and poultry food categories of interest to FSIS. 6 E. Dreyling, FSIS, unpublished material, January 6, 2009. 45 

2.0 Foodborne Disease Attribution One frequently used approach to foodborne disease attribution is the use of expert elicitation. During expert elicitation, a group of experts is asked, based on their professional judgment, to either rank food groups as to their relative important as sources of foodborne disease or to estimate the percent contribution of food groups to foodborne disease. The reliability of expert opinion regarding foodborne disease attribution has been questioned since it is based on perception and not quantifiable data (Batz et al. 2005). However, by selecting experts with first-hand knowledge of different aspects of foodborne attribution (e.g. experts working in academia, the food industry, and public health) it is possible to obtain an informed and integrated judgment of the impact of different food types of human illness. Moreover, expert judgment is often the best source for guidance when scientific and epidemiologic data are sparse (Batz et al. 2005, National Academy of Sciences 2003). We briefly review the results of two recent expert elicitations. 2.1 FSIS Expert Elicitation Karns et al. (2007) conducted an expert elicitation for FSIS to determine foodborne disease illness attribution for 25 meat and poultry food categories. In what follows this study is referred to as the FSIS expert elicitation. The expert panel consisted of 12 experts equally divided among scientists from the public health community, industry, and academic institutions. The expert panelists were asked to attribute foodborne illnesses of Salmonella, E. coli O157:H7, and Listeria monocytogenes to handling and consuming foods in 25 processed meat and poultry product categories. The attributions obtained for the Karns et al. (2007) study are presented in Table 2-1. TABLE 2-1 Attribution of Foodborne Illness (Percentages) for 25 Processed Meat and Poultry Product Categories Based on the 2007 FSIS Expert Elicitation Finished Product Type Salmonella E. coli O157 Listeria M Raw ground, comminuted, or otherwise nonintact chicken 8.9 0.4 1.3 Raw ground, comminuted, or otherwise nonintact turkey 6.8 0.3 1.2 Raw ground, comminuted, or otherwise nonintact poultry—other 2.8 0.4 0.9 than chicken or turkey Raw ground, comminuted, or otherwise nonintact beef 8.4 57 1.9 Raw intact chicken 22.0 1.1 1.3 Raw intact turkey 14.1 0.3 0.8 Raw intact poultry—other than chicken or turkey 3.7 0.7 1.4 Raw otherwise processed poultry 5.6 0.6 1.4 Raw ground, comminuted, or otherwise nonintact meat—other 2.7 13.8 0.8 than beef or pork Raw otherwise processed meat 3.5 2.9 1.5 Raw ground, comminuted, or otherwise nonintact pork 4.3 1.4 0.9 Raw intact beef 4.6 8.4 1.4 Raw intact meat—other than beef or pork 2.2 2.6 0.4 Raw intact pork 2.8 1.3 0.6 (Continued) 46 

TABLE 2-1 Continued Finished Product Type Salmonella E. coli O157 Listeria M RTE acidified/fermented poultry (without cooking) 1.6 0.3 4.4 RTE acidified/fermented meat (without cooking) 1.0 4.2 6.4 RTE fully cooked poultry 1.0 0.2 25.0 RTE salt-cured poultry 0.6 0.2 4.0 RTE salt-cured meat 0.5 0.8 3.6 RTE dried meat 0.9 1.3 3.2 RTE dried poultry 1.0 0.2 3.2 RTE fully cooked meat 0.5 1.1 30.2 RTE meat fully cooked without subsequent exposure to 0.3 0.3 2.1 the environment RTE poultry fully cooked without subsequent exposure to 0.3 0.3 2.0 the environment Thermally processed, commercially sterile 0.0 0.0 0.1 Source: Karns et al. 2007. 2.2 Resources for the Future Expert Elicitation Resources for the Future in conjunction with Carnegie Mellon University conducted an expert elicitation attribution study to determine the relative contribution of different foods to foodborne illness in the United States (Hoffmann et al. 2007). In what follows this study is referred to as the RFF expert elicitation. The authors of the study used a panel of 42 food safety experts to perform a separate food attribution relative ranking for each of 11 pathogens. For each pathogen, respondents were asked to provide their best estimate of the proportion of cases of foodborne illness caused by a specific pathogen in a typical year associated with consumption of each of 11 food categories. While the RFF study (Hoffmann et al. 2007) looked at 11 different pathogens, we present their results for only three pathogens: Salmonella, E. coli O157:H7, and Listeria monocytogenes. A valuable contribution of the Hoffmann et al. study is that it includes both FSIS- and FDA- inspected food categories. It thus provides a more complete picture of disease attribution than the FSIS expert elicitation. However, the FSIS expert elicitation provides much more detail on specific meat and poultry food categories. Thus, both elicitation studies provide slightly different perspectives on the food attribution problem. Table 2-2 presents data from the RFF elicitation of the percent contribution (attribution) of 11 food types to foodborne illness in the United States. 2.3 Foodborne Disease Outbreaks Data on foodborne disease outbreaks can provide a useful source of information concerning some aspects of the food attribution problem. An outbreak is defined as the occurrence of two or more cases of a similar illness resulting from the ingestion of a food in common. The CDC maintains a database of foodborne illness outbreaks that covers the years 1990 to 2006 (CDC 2008). Reported data on foodborne disease outbreaks can be valuable in establishing a link between foodborne illness and the specific food sources that cause them. As pointed out above, while only a small fraction of total foodborne disease is caused by outbreaks, this does not automatically mean that attribution estimates derived from outbreak data disagree with those derived from sporadic disease data. As will be seen below, attribution estimates for the major FSIS-inspected food categories of beef, poultry, pork, and deli derived from CDC outbreak 47 

data agree closely with estimates from the two above expert elicitations which account for sporadic illness. This increases confidence in using the outbreak data for these pathogens. In addition, outbreak data represent the largest epidemiological dataset available for attribution studies and provide an important source of information linking foodborne illness with specific food sources. Table 2-3 presents attribution information related to outbreaks of E. coli O157:H7, Salmonella, and L. monocytogenes. With respect to FSIS-inspected products, the RFF and CDC studies considered the general food categories of beef/meat, poultry, pork, and deli meats, while the FSIS expert elicitation covered 25 specific FSIS- inspected food categories. To compare the results of all three studies with respect to meat and poultry food categories, we collapse the 25 food categories to four meat and poultry food categories. Table 2-4 presents the correspondence used to compare studies. TABLE 2-2 Attribution of Foodborne Illnesses (Percentages) from RFF Expert Elicitation Food Type Salmonella E. coli O157 Listeria M Beef 10.9 67.9 1.6 Poultry 35.1 0.9 2.7 Pork 5.7 0.6 1.3 Deli meats 1.9 1.8 54 Eggs 21.8 0.03 0.3 Seafood 2.04 0.05 7.1 Produce 11.7 18.4 8.7 Breads and bakery 0.03 0 0.2 Dairy 7.3 4.0 23.6 Beverages 1.7 3.2 0.2 Wild game 1.6 3.2 0.3 Source: Hoffmann et al. 2007. TABLE 2-3 CDC Outbreak Data for Salmonella, E. coli O157:H7, and L. monocytogenes by Specific Food Category Salmonella E. coli O157:H7 Listeria M Food Type Cases Percent Cases Percent Cases Percent Meat 2,444 9.6 2,030 54.1 0 0.0 Poultry 5,681 22.3 0 0.0 3 0.8 Deli Meats 284 1.1 49 1.3 251 69.9 Pork 1,121 4.4 0 0.0 0 0.0 Seafood 791 3.1 14 0.4 0 0.0 Produce 6,096 23.9 1190 31.7 0 0.0 Eggs 4,309 16.9 0 0.0 0 0.0 Dairy 2,748 10.8 301 8.0 105 29.3 Breads, Bakery 1,154 4.5 0 0.0 0 0.0 Game 0 0.0 15 0.4 0 0.0 Beverages 841 3.3 153 4.1 0 0.0 Total 25,469 100 3,752 100 359 100 48 

Using the mapping in Table 2-4, food attribution for the four meat and poultry food categories can be calculated. It is necessary to normalize the percentages so they add to 100 percent for these four food categories. Normalization is necessary because the FSIS study only considered FSIS regulated meat and poultry categories, while the RFF and CDC studies considered both FSIS and FDA food categories. Table 2-5a presents a comparison of the three studies. TABLE 2-4 Correspondence between FSIS Expert Elicitation Categories and General Meat and Poultry Categories FSIS Food categories Meat and Poultry Categories Raw ground, comminuted, or otherwise nonintact beef Meat Raw intact beef Raw ground, comminuted, or otherwise nonintact meat—other than beef or pork Raw otherwise processed meat Raw intact meat—other than beef or pork Raw ground, comminuted, or otherwise Poultry nonintact chicken Raw ground, comminuted, or otherwise nonintact turkey Raw ground, comminuted, or otherwise nonintact poultry—other than chicken or turkey Raw intact chicken Raw intact turkey Raw intact poultry—other than chicken or turkey Raw otherwise processed poultry Raw ground, comminuted, or otherwise nonintact pork Pork Raw intact pork All RTE categories Deli meats TABLE 2-5a Comparison of Normalized Attribution (Percentage) Developed by the FSIS, RFF, and CDC Studies Finished Product Type Salmonella E. coli O157 Listeria M a a FSIS RFF CDC Av FSIS RFF CDC Av FSIS RFF CDC Ava Meat 21.4 20.4b 25.7 22.5 84.7 95.3 97.6 92.5 6.0 2.7 0.0 2.9 Poultry 63.9 65.5 59.6 63.0 3.8 1.2 0.0 1.7 8.3 4.5 1.1 4.6 Pork 7.1 10.6 11.8 9.8 2.7 0.8 0.0 1.2 1.5 2.2 0.0 1.2 Deli meats 7.7 3.5 2.9 4.7 8.9 2.5 2.4 4.6 84.2 90.6 98.9 91.3 a Average of three studies. b Beef only. Note: As can be seen from Table 2-5a, the three attribution studies (one of which is an actual count of CDC outbreak illness) produce very similar estimates of attribution for FSIS-inspected beef, poultry, pork, and deli meat products. This result provides an independent validation of the attribution results of the FSIS 2007 expert elicitation (Karns et al. 2007). 49 

The RFF and CDC studies provide attribution estimates for both FSIS and FDA-inspected foods. These can be used to estimate the average contribution of FSIS-inspected food categories to the total illness impact of Salmonella, E coli O157, and Listeria M in the United States. Table 2-5b presents these estimates. 3.0 Attribution for 25 FSIS Meat and Poultry Product Categories We are now in a position to use the above foodborne disease attribution results to estimate attribution for the 25 meat and poultry product categories defined by FSIS in the Karns et al. (2007) study. We accomplish this in a two-step process:  First, the average normalized attribution estimates in Table 2-5b are adjusted by the percent contribution of FSIS-inspected foods to U.S. foodborne illness rates (Table 2-5b) to arrive at an estimate of the percent contribution of each of the four food type categories to U.S. foodborne illness rates.  Second, attribution estimates in Table 2-1 for each of the four food type categories are normalized so that they total the percent contribution for that specific food type. Figure 3-1 illustrates the process for estimating the percent contribution of meat products to total foodborne disease Salmonella illnesses. TABLE 2-5b Estimate of Average Percent Contribution of FSIS-Inspected Products to U.S. Foodborne Illness Salmonella E. coli O157 Listeria M RFF CDC Av RFF CDC Av RFF CDC Av FSIS Inspected 53.6 37.4 45 71.3 55.4 63 59.6 70.7 65 Foods FDA Inspected 46.4 62.6 55 28.7 44.6 37 40.4 29.3 35 Foods 45% FSIS Contribution to Total Illness Adjust for % contribution of Individual Food Groups 3 Normalized Attribution (%) Attribution Major Food Final Attribution Estimate Studies Types Product Salmonella Product Salmonella Type Type Meat Salmonella Meat 22.5 Raw intact beef 2.2 Meat 10.0 Raw intact meat – 1.0 Poultry 63.0 Poultry 28.4 other than beef or pork Pork 9.8 Pork 4.4 Raw ground beef 3.9 RTE 4.8 Raw ground, meat – 1.3 RTE 2.2 other than beef or Total 100 pork Total 45 Raw otherwise 1.6 processed meat Sum Meat 10.0 FIGURE 3-1 Example of process for estimating attribution for 25 FSIS food categories. 50 

The results are presented in Table 3-1. TABLE 3-1 Foodborne Disease Attribution Estimates for 25 FSIS Food Categories Finished Product Type Salmonella E. coli O157 Listeria M Meat Raw intact beef 2.2 5.8 0.2 Raw intact meat—other than beef or pork 1.0 1.8 0.1 Raw ground, comminuted, or otherwise nonintact beef 3.9 39.2 0.3 Raw ground, comminuted, or otherwise nonintact meat—other 1.3 9.5 0.1 than beef or pork Raw otherwise processed meat 1.6 2.0 0.2 Sum Meat 10.0 58.3 0.8 Poultry Raw intact chicken 9.8 0.5 0.27 Raw intact turkey 6.3 0.1 0.15 Raw intact poultry—other than chicken or turkey 1.7 0.3 0.30 Raw otherwise processed poultry 2.5 0.2 0.30 Raw ground, comminuted, or otherwise nonintact chicken 4.0 0.2 0.27 Raw ground, comminuted, or otherwise nonintact turkey 3.0 0.1 0.27 Raw ground, comminuted, or otherwise nonintact poultry—other 1.3 0.2 0.19 than chicken or turkey Sum Poultry 28.4 1.6 1.75 Pork Raw intact pork 1.7 0.4 0.31 Raw ground, comminuted, or otherwise nonintact pork 2.7 0.5 0.47 Sum Pork 4.4 0.9 0.78 RTE RTE acidified/fermented poultry (without cooking) 0.4 0.1 3.1 RTE acidified/fermented meat (without cooking) 0.3 1.4 4.5 RTE fully cooked poultry 0.3 0.1 17.5 RTE salt-cured poultry 0.2 0.1 2.8 RTE salt-cured meat 0.1 0.3 2.5 RTE dried meat 0.3 0.4 2.2 RTE dried poultry 0.3 0.1 2.2 RTE fully cooked meat 0.1 0.4 21.2 RTE meat fully cooked without subsequent exposure to the 0.1 0.1 1.5 environment RTE poultry fully cooked without subsequent exposure to the 0.1 0.1 1.4 environment Thermally processed, commercially sterile 0.0 0.0 0.1 Sum RTE 2.2 2.9 59 51 

4.0 Discussion The CDC outbreak database and two expert elicitations were used to derive foodborne disease attribution estimates for meat and poultry products. The three different approaches produce consistent estimates of attribution. 5.0 References Batz, M.B., M.P. Doyle, J.G. Morris, Jr., J. Painter, R. Singh, R.V. Tauxe, M.R. Taylor, and D.M. Lo. Fo Wong. 2005. Attributing illness to food. Emerg. Infect. Dis. 11(7):993-999 [online]. Available: http://www.cdc. gov/ncidod/EID/vol11no07/04-0634.htm CDC (Centers for Disease Control and Prevention). 2008. Outbreak Surveillance Data. Karns, S., M. Muth, and M. Cogliaiti. 2005. Relative Risks of Meat and Poultry Products: An Expert Elicitation. Research Triangle Institute, Research Triangle Park, NC. Karns, S.A., M.K. Muth, and M.C. Coglaiti. 2007. Results of an Additional Expert Elicitation on the Relative Risks of Meat and Poultry Products. Research Triangle Institute, Research Triangle Park, NC. Mead, P.S., L. Slutsker, V. Dietz, L.F. McCaig, J.S. Bresee, C. Shapiro, P.M. Griffin, and R.V. Tauxe. 1999. Food- related illness and death in the United States. Emerg. Infect. Dis. 5(5):607-625. NAS (National Academies of Sciences). 2003. Scientific Criteria to Ensure Safe Food. Washington, DC: National Academies Press. 52