1972 Journal of Food Protection, Vol. 67, No. 9, 2004, Pages 1972 1976 Copyright, International Association for Food Protection Food Consumption Data in Microbiological Risk Assessment LEILA M. BARRAJ* AND BARBARA J. PETERSEN Exponent, Inc. (formerly Novigen Sciences, Inc.), 1730 Rhode Island Avenue N.W., Suite 1100, Washington, D.C. 20036, USA MS 03-701: Received 25 August 2003/Accepted 12 March 2004 ABSTRACT The 1st International Conference on Microbiological Risk Assessment: Foodborne Hazards was held in July 2002. One of the goals of that conference was to evaluate the current status and future needs and directions of the science of microbial risk assessment. This article is based in part on a talk presented at that meeting. Here, we review the types of food consumption data available for use in microbial risk assessments and address their strengths and limitations. Consumption data available range from total population summary data derived from food production statistics to detailed information, derived from national food consumption surveys, about the types and amounts of food consumed at the individual level. Although population summary data are available for most countries, detailed data are available for a limited number of countries and may only be available in summary format. Despite the relatively large amount of detailed information collected by these national surveys, information crucial to microbial risk assessments, such as the specific types of foods, the eating patterns of susceptible populations, or an individual s propensity for consuming high-risk foods (e.g., eating undercooked hamburgers, raw shellfish, or temperature-abused foods), are not collected during these surveys. Microbial risk assessment (MRA) fits the general risk assessment paradigm used for chemical and environmental contaminants and traditionally involves the following four elements (excerpted from the procedural manual of the Codex Alimentarius Commission (5)): 1. Hazard identification. The identification of biological, chemical, and physical agents that are capable of causing adverse health effects and that may be present in a particular food or group of foods. 2. Hazard characterization. The qualitative or quantitative evaluation of the nature of the adverse health effects associated with biological, chemical, and physical agents that may be present in food. For chemical agents, a doseresponse assessment should be performed. For biological or physical agents, a dose-response assessment should be performed if the data are obtainable. 3. Exposure assessment. The qualitative or quantitative evaluation of the likely intake of biological, chemical, and physical agents via food and exposures from other sources if relevant. 4. Risk characterization. The qualitative or quantitative estimation, including attendant uncertainties, of the probability of occurrence and severity of known or potential adverse health effects in a given population based on hazard identification, hazard characterization, and exposure assessment. Here, we focus on one component of MRAs, the exposure assessment component, specifically on the consumption data that are used in MRAs. We review the types of consumption data available for use in MRAs, how these data are used in MRAs, and the data gaps that need to be * Author for correspondence. Tel: 202-772-4909; Fax: 202-772-4979; E-mail: lbarraj@exponent.com. filled. This report is based in part on a presentation given at the 1st International Conference on Microbiological Risk Assessment: Foodborne Hazards (24 to 26 July 2002). MRAs differ from chemical risk assessments in that it is often possible to validate the various components of the MRA through a comparison between the predicted number of illnesses and the actual number of reported cases. After intake estimates are derived and dose-response models are applied to these estimates, the estimated potential risk (typically an estimate of the number of illnesses) can be compared with an estimate of the number of cases based on other surveillance programs, such as the Centers for Disease Control and Prevention Foodborne Diseases Active Surveillance Network (FoodNet), which collects data on foodborne diseases in the United States (2). The process is generally iterative and can serve as a validation step in MRA (3). In general, such surveillance databases are not available for validating results of chemical risk assessments. Another difference between the two types of risk assessment stems from the fact that unlike chemical contaminants, bacteria can grow in food as it moves from farm to fork, and because cross-contamination can occur during that process, more detailed information about the history of the food before it is consumed is required by MRAs than by chemical risk assessments. CONSUMPTION DATA IN THE MRA PROCESS Exposure is basically a function of two components: (i) the concentration of the pathogen on or in the food of interest and (ii) the amount of food consumed. In current MRA, most of the attention is focused on the concentration of the contaminant, in part because the task of estimating the amount and distribution of microbial contamination on the food as consumed is more complex than that of esti-
J. Food Prot., Vol. 67, No. 9 FOOD CONSUMPTION DATA IN MICROBIOLOGICAL RISK ASSESSMENT 1973 mating the amount of food consumed. Several factors must be considered in estimating whether the contaminant is present on the food, how much is present, and how it is distributed in the food. These factors include the likelihood, amount, and distribution of initial contamination, the likelihood of cross-contamination or recontamination at later stages, how the food is handled, stored, and cooked, at what temperatures it is stored or cooked, and the likelihood of growth or inhibition during storage, transport, handling, and cooking. Estimating the consumption component of exposure requires information about the frequency of consumption of the foods of interest, the forms in which they are consumed, and the amounts consumed. The term frequency of consumption has been used in MRA to mean number of meals per day or per year, number of servings per day or per year, or number of people consuming the food in a given day, year, etc. Similarly, the term amount consumed refers to amount by meal, by serving, by day, etc., and it can be for eaters of the foods or per capita. The estimate of consumption used can be a point estimate (typically but not necessarily an average amount) or the entire distribution, because all people do not eat the same amount of food. Which of these estimates is used in an MRA depends on the purpose of the assessment and on what data are available. Factors that may influence food consumption patterns or the risk of illness from foodborne pathogens (whether increasing or decreasing the risk) should be considered in the estimation of dietary exposure. Food consumption patterns will likely differ based on population demographics (age, gender, ethnicity, socioeconomic group) and seasonal and regional (both national and international) differences in food availability. For MRA, consideration of food consumption patterns for sensitive subpopulations (e.g., young children, pregnant women, the elderly, and the immunodeficient) and of high-risk consumer behavior (e.g., consuming unpasteurized dairy products or undercooked meat products) are also particularly important. Differences in food consumption databases. Food consumption data sets differ in how the information is collected and reported, the form of the foods for which data are collected (i.e., raw agricultural commodities or foods as consumed), and whether information on consumption by population subgroups is available. Two types of food consumption data are frequently used for characterizing food consumption patterns for microbiological risk assessments: food production statistics and food consumption surveys. Other sources of information such as retail food purchase data may be useful in filling data gaps in either food production or food consumption survey data. Food production statistics provide an estimate of the amount of food available to the total population. Examples of this type of data include the amounts of foods available for human consumption, which are derived from national statistics on food production, disappearance, or utilization such as those derived by the U.S. Department of Agriculture (USDA) Economic Research Service (6) or the Australian Bureau of Statistics (1). The amount of food available for consumption is calculated as the difference between available supplies, including production, beginning inventories, and imports, and nondomestic food uses, such as exports, farm use, and industrial uses. The Food and Agriculture Organization FAOSTAT database is a compilation of similar statistics for more than 250 countries. The data are compiled, or estimated when official data from member countries are missing, from national food production and utilization statistics (4). Food production statistics are generally compiled and reported for raw or semiprocessed agricultural commodities, and they represent the total annual amount of a commodity available for domestic consumption. Dividing these amounts by the country s population size results in estimates of the total annual quantity of food available for each person in the total population (per capita amount). The daily per capita amount may then be calculated by dividing the annual amount by 365. Because these data are available for most countries and because they are compiled and reported fairly consistently across countries, they can be useful in conducting exposure assessments at the international level. Food consumption surveys can be conducted at the national level or for specific subpopulations, such as people residing in particular regions, people of certain age groups, or people with certain occupations, e.g., fishermen. The surveys can provide information regarding the types and amounts of foods consumed by individuals and, in some surveys, the long-term frequency with which the foods are consumed. The surveys usually encompass a short period of time (one to several days for each survey participant), and they generally provide information about the types of food consumed and the time of day and place that foods are consumed. Food frequency data are typically collected for a predefined list of foods or food groups and encompass longer periods of time. Nationwide food consumption surveys are conducted by a limited number of countries, each using a different method for collecting and reporting the data. Thus, it is difficult to rely solely on food consumption survey data when conducting international exposure assessments. These surveys are usually conducted to assess the nutritional status of the population and thus may not include information that is of interest to food safety managers, such as whether the milk consumed was pasteurized or whether the hamburger was well cooked versus undercooked. Because of the heavy burden on respondents, national consumption surveys usually cover a short period of time. For instance, the Continuing Survey of Food Intakes by Individuals, conducted by the USDA (8), collects data over two nonconsecutive days, whereas the National Diet, Nutrition and Dental Survey of Children Aged 1½ to 4½ Years conducted in the United Kingdom (7) collects data over four consecutive days. Thus, these surveys do not allow for the estimation of usual or long-term intakes. Retail food purchase data provide detailed information about specific food products, which is often lacking from food consumption surveys. Such data are available though household surveys of foods purchased over a given period (e.g., a week) or through data collected by grocery stores
1974 BARRAJ AND PETERSEN J. Food Prot., Vol. 67, No. 9 that track purchases at the household levels. Because these data are typically collected for households, they do not describe the amount of food actually consumed by specific individuals nor do they provide information on which household member is consuming the food products. However, the information can be combined with other food consumption data to refine the characterization of food consumption. For instance, information about relative sales of raw versus pasteurized milk can be used to adjust milk consumption estimates derived from national consumption surveys. Amount of food consumed and frequency of consumption. To characterize the risk from exposure to microbiological hazards in food, it is necessary to know both the amount of food consumed and the frequency with which the food is consumed. There are many ways by which both consumption amount and frequency of consumption may be calculated, each resulting in different values and representing a different characterization of food consumption. When modeling food consumption, it is important for risk assessors to understand the specifics of how the food consumption data were collected and analyzed, to clearly describe how these data on food consumption amount and frequency were used in the model, and to state the assumptions used in arriving at the estimates. One important aspect of estimating the amount of food consumed, particularly when using results from food consumption surveys, is whether the estimate applies to the total population (per capita) or to the consumers of the food (per user). For foods that are consumed regularly by the majority of the population (e.g., bread), the per capita and per user amounts will be nearly equal. For foods that are consumed less frequently (e.g., fish) or by fewer individuals (e.g., soft cheese), the per capita and per eater amounts will be quite different. The choice of which of these estimates to use depends on both the purpose of the assessment and the data available. When using production statistics, it is not possible to get per user estimates. However, other data such as household purchase data may be available to supplement the information. Another important aspect is whether the consumption amount estimate available represents the amount consumed per year, per day, or per eating occasion. Depending on how the food consumption data are collected and reported, consumption may be calculated as the amount per year, per day, or per eating occasion. The basis for the consumption period is particularly important in microbiological risk assessments and other situations in which acute rather than chronic exposure is of concern. Another issue to consider when choosing a consumption estimate is whether to use a point estimate or a distribution. People obviously do not all eat the same amounts of foods nor do they eat the same amount of food every day. Therefore, risk assessors must determine whether point estimates or distributions will better characterize the dayto-day (i.e., the intraindividual) variability in types and amounts of foods consumed and the person-to-person (i.e., interindividual) variability in types and amounts of foods consumed. When a point estimate is used, the risk assessors must determine whether it should be used to represent the average consumer or the high-end consumer. A point estimate of the average consumption is calculated by dividing the total amount of food available to (or consumed by) the population by the total number of people in the population (per capita) or by the number of people who reported eating the food (per user). These estimates represent the amount of the food available to the total population or to the consumers of the food but not necessarily the amount of the food actually consumed by an individual. The frequency of consumption may refer to the proportion of the population that consumes a food or how often an individual consumes a food over a specific period of time. This information can be derived from food consumption surveys and other sources, such as food frequency questionnaires. In previous MRAs, frequency of consumption has been expressed as an annual measure in the following ways: number of days annually on which the food is consumed; number of eating occasions over a year, i.e., annual number of meals; number of 100-g portions consumed in a year; and percentage of the population who ate the food in a specific time period (e.g., a year). The number of days of consumption during the survey period can be determined directly from the survey results, and an annual number of days of consumption may be extrapolated from that estimate. The number of days can be determined for the total population or for any subgroup, such as eaters only or the elderly. Frequency of consumption may also be expressed as the number of meals, eating occasions, or individual food items. In some instances, it is possible to refine the estimated frequency of consumption by combining food consumption data with other industry information such as annual sales volume or market share information. For example, when the food consumption data include the frequency of consumption of a broad category such as milk, market share data may be used to predict the frequency of consuming particular types of milk, e.g., pasteurized or nonpasteurized. An underlying assumption in that case is that the proportion of nonpasteurized milk sales out of all milk sales applies to the consumption of milk and that the amount of milk consumed is similar across types of milk and across the population. USING FOOD CONSUMPTION DATA Using food production statistics. Food production statistics apply to the total population and report an amount of food per year. They also can be used to arrive at per capita estimates. Often a daily consumption amount is estimated by dividing the total annual amount by 365. When these estimates are used in MRA, important assumptions underlie their use, i.e., that consumption is uniformly distributed across the population (all age groups, regions, etc.). Alternative estimates can be derived based on what is known about the food. For example, are there seasonal or day-of-the-week differences in food availability and consumption? Because these data are typically collected for raw or semiprocessed commodities, another issue to con-
J. Food Prot., Vol. 67, No. 9 FOOD CONSUMPTION DATA IN MICROBIOLOGICAL RISK ASSESSMENT 1975 sider when using these data in MRA is how to adjust the amounts for processing and cooking. Using consumption survey data. When using data from consumption surveys, it is important to know whether the survey participants are representative of the population. If they are not, statistical weighting could be used to make the survey results representative. Because these surveys are usually conducted over a limited number of days, consideration must be given to whether the information from the 2 to 3 days of surveys are representative of a person s eating pattern and whether they can be used to estimate dayto-day variability. For MRAs, it is important to estimate consumption by sensitive population groups such as immunocompromised individuals and elderly people living in institutions. Often, this subpopulation information is not available, and it is assumed that consumption patterns in these groups are the same as those for the healthy population of the same age and gender. When using food frequency data that refer to food groups, it is important to decide how to move from the food groups to the individual foods within these groups. The average consumption amount for a food category is affected by the number of foods it represents and how similar the foods are in terms of the usual amount and frequency of consumption. If the foods are too dissimilar, the average amount and frequency of consumption may be misrepresented. In addition, it may not be possible to adjust for potential differences in contamination amounts and growth patterns for foods within groups. When estimating the daily consumption from food consumption survey data, it is important to note whether the amount reflects all days of the survey or only the days on which a food is consumed. As an example, in the 1994 to 1998 USDA CSFII (8), 2 days of dietary records were collected for individuals participating in the survey. From those data it is possible to calculate the average consumption of a food on the days the food was actually consumed or the average over the 2 days for which each person participated in the survey. Thus, for instance, if individuals who ate fish reported consuming 6 oz (170 g) of fish on only 1 of the 2 days of the survey, the average daily consumption would be 3 oz (85 g) if calculated as the average of all days or 6 oz if calculated as the average amount on days when fish was consumed. Some foods may be consumed both as discrete items and as components of combination foods or food mixtures. For example, ground beef may be consumed as a single food item or as a component of a beef casserole. As another example, milk may be consumed as a beverage or as an ingredient (often in very small amounts) in many food items. When modeling food consumption, it is important to know whether the consumption estimate includes all sources of the food or only the amount of food consumed as a discrete item. If the consumption estimate should include the food from all sources, it may be necessary to create a generic recipe for combination foods to account for all sources of the food. It is important also to decide whether a total intake estimate should be derived for multiple foods or whether multiple intake estimates should be derived for individual foods. If the purpose of the assessment is to identify risks associated with different foods for development of a risk ranking, as in the recent Listeria monocytogenes risk assessment by the U.S. Food and Drug Administration (FDA) (9), then the latter approach may be more appropriate. Some surveys include additional information, such as where the food was obtained. This information is particularly important to the risk assessor interested in a specific contamination source, e.g., the meat supplier for a fast-food chain. Other potentially relevant information sometimes available in food consumption surveys refers to the place and time of consumption and in the case of drinking water to the source of the water, i.e., a private well or a community water system. LIMITATIONS OF CURRENTLY AVAILABLE FOOD CONSUMPTION SURVEY DATA Food consumption surveys for individuals are frequently conducted to assess the nutritional status of a population rather than to characterize consumption of specific foods. Although these surveys may provide information about consumption by specific age and gender groups, they may not describe foods in sufficient detail or include enough participants from sensitive subpopulations or even collect the information necessary to identify these subpopulations. The food description may not indicate whether foods such as milk or juices are pasteurized or the degree to which foods such as eggs or ground beef are cooked. Raw data from the surveys are not always available, and the risk assessor must rely on aggregated data, which may not be sufficiently detailed or targeted to the foods of concern. Information on consumer behavior that may increase or decrease the risk of foodborne illness is absent from most food consumption surveys. The ability to link food consumption data to information about an individual s propensity for consuming high-risk foods (e.g., eating undercooked hamburgers, raw shellfish, or temperature-abused foods) would be extremely useful in estimating exposure to microbiological hazards in foods. ACKNOWLEDGMENTS The information summarized in this presentation was obtained in part from our experience at Exponent with food consumption data and how they can be used with dietary exposure assessments and from discussions and interactions with various scientists at the FDA, USDA, and Environmental Protection Agency. In particular, we acknowledge Ms. S. Kathleen Egan (FDA) for her valuable input. REFERENCES 1. Australian Bureau of Statistics. 2000. Australian apparent consumption of food stuff, 1997 98 and 1998 99. Available at: http://www.abs.gov.au/ausstats/abs@nsf/lookupmf/ 123FCDBF086C4DAACA2568A9001393A. Accessed 24 February 2004. 2. Centers for Disease Control and Prevention. 2004. Foodborne diseases active surveillance network (FoodNet). Available at: http:// www.cdc.gov/foodnet/default.htm. Accessed 24 February 2004.
1976 BARRAJ AND PETERSEN J. Food Prot., Vol. 67, No. 9 3. Food and Agriculture Organization. 1999. Principles and guidelines for the conduct of microbiological risk assessment. Available at: http: //www.fao.org/docrepy1579e/y1579e05.htm. Accessed 24 February 2004. 4. Food and Agriculture Organization. 2004. FAOSTAT database. Available at: http://apps.fao.org/lim500/wrap.pl?foodbalancesheet& Domain FoodBalanceSheet&Language english. 5. Joint FAO/WHO Food Standards Programme. 2001. Procedural manual of Codex Alimentarius commission, 12th ed. Food and Agriculture Organization of the United Nations, Rome. 6. Putnam, J. J., and J. E. Allshouse. 1999. Food consumption, prices, and expenditures, 1970 97. Statistical bulletin no. 965. Food and Rural Economics Division, Economic Research Service, U.S. Department of Agriculture, Washington, D.C. 7. United Kingdom Office of Population Censuses and Surveys Social Survey Division. 1995. National diet, nutrition and dental survey of children aged 1½ to 4½ years, 1992 1993. 13 December 1995. SN 3481. The Data Archive, Colchester, Essex, UK. 8. U.S. Department of Agriculture. 2000. CSFII data set and documentation. The 1994 96, 1998 continuing surveys of food intakes by individuals. Food Surveys Research Group, Beltsville Human Nutrition Research Center, Agricultural Research Service, Washington, D.C. 9. U.S. Food and Drug Administration, Center for Food Safety and Applied Nutrition. 2003. Quantitative assessment of relative risk to public health from foodborne Listeria monocytogenes among selected categories of ready-to-eat foods. Available at: http://www.foodsafety. gov/ dms/lmr2-toc.html. Accessed 24 February 2004.