Risk Metrics Quantifying the Impact of Adverse Health Effects


Due to the complexity and changing nature of the food supply, ensuring its safety has been identified as a wicked problem - i.e., a problem that arises in complex and interdependent systems and that is difficult or impossible to solve because of incomplete, contradictory, changing, or incomprehensible requirements (Institute of Medicine 2012). Indeed, the food system is multi-faceted, with a large number of stakeholders having diverse interests. The international food production and distribution systems play a major role in the global economy, with significant impacts on population health, income, employment, rural and urban economies, and the environment. Historically, the approach to ensuring food safety has been reactive - responding to crises as they occur - rather than preventive (Koutsoumanis and Aspridou 2016). Globally, many countries lack the infrastructure needed to meet international food safety standards, which in turn, impacts trade and local access to safe food. To address the food challenges of the 21st century, this paradigm is slowly shifting to an integrated, multi-disciplinary, systems-based approach that is informed by the best available science and focuses on prevention. At the same time, there is an increasing need to utilize limited resources so that they effectively address the most important issues and provide the greatest benefits to the most people. As outlined in Chapter 1, risk analysis, which consists of risk assessment, risk management, and risk communication (Figure 3.1),

Components of risk analysis

Figure 3.1 Components of risk analysis.

provides a framework for supporting decision making; it is internationally accepted as the best approach to food safety (Food and Agriculture Organization of the United Nations 2006).

A risk-based food safety system is one that uses "a systematic means by which to facilitate decision-making to reduce public health risk in light of limited resources and additional factors that may be considered" (Havelaar et al. 2007; National Research Council 2010). Central to the risk- based framework (Figure 3.2) is an understanding of the risks and burden of disease (i.e., the impact of a disease in the population). Understanding the burden can be complemented by quantifying, ranking, and attributing

Framework for a risk-based food safety system. (Adapted from National Research Council, Enhancing Food Safety

Figure 3.2 Framework for a risk-based food safety system. (Adapted from National Research Council, Enhancing Food Safety: The Role of the Food and Drug Administration, The National Academies Press, Washington, DC, 2010.) the risks to the responsible sources. From there, public health goals can be established - such as the United States Healthy People 2020 goals or the United Nations' Millennium Development Goals - and potential prevention and control interventions can be determined. An evaluation of potential interventions and/or policies allows the determination of their ability to positively impact public health at a reasonable cost in a fair manner. After having identified potential prevention and control strategies, priorities need to be set and resources allocated to those that will have the biggest public health impact. Finally, the effectiveness of the efforts in meeting public health goals and objectives must be quantified.

Risk-based food safety systems are steadily but increasingly being implemented to replace the historical reactive food safety systems. In 2002, the Council of the European Union and the European Parliament adopted Regulation (EC) No. 178/2002, known as the General Food Law of 2002 (http://eur-lex.europa.eu/legal-content/EN/ALL/?uri=CELEX: 32002R0178). One of the key principles of the food law is that "measures adopted by the Member States and the Community governing food and feed should generally be based on risk analysis". The Regulation further created the Rapid Alert System for Food and Feed (RASFF) and the European Food Safety Authority (EFSA), an independent agency that provides scientific advice and risk assessments to relevant bodies in the European Commission, the European Parliament, and Member States. In 2010, recognizing the importance and infrastructure needs around food safety, the United States Congress passed the Food Safety Modernization Act of 2010 (FSMA), the first comprehensive reform of the Food and Drug Administration's (FDA) food safety oversight since 1938 (U.S. Food and Drug Administration 2017). FSMA mandates FDA to adopt a science- based, risk-informed approach to food safety and holds the food industry more accountable for producing safe products.

Although the concept of "risk" is fundamental to these systems, its definition is not entirely standardized. According to the Codex Alimentarius, risk is defined as "a function of the probability of an adverse health effect and the severity of that effect, consequential to a hazard(s) in food" (Codex Alimentarius Commission 2018). However, severity can be viewed in different ways - it may, for instance, be defined as the health or economic impact of the adverse health effects. Many risk assessments do not take severity into account and restrict themselves to the probability of illness (e.g., Bollaerts et al. 2009) or even more distant, the probability of infection (e.g., Hamilton et al. 2006). In toxicology, for instance, the definition of risk by the International Programme on Chemical Safety (2004) is often used, i.e., "the probability of an adverse effect in an organism, system, or (sub)population caused under specified circumstances by exposure to an agent", which thus does not take severity into account.

When quantifying both probability and severity, the function to combine both dimensions can take many different shapes - ranging from a qualitative integration to intricate weighted averages. A risk matrix is, for instance, one of the most basic ways of combining probability and severity (Figure 3.3). Here, both dimensions are expressed in a semi-quantita- tive way, and risk is expressed as a combination of the two scores - e.g., ranging from low-low to high-high. Despite their intuitive nature, they

Risk matrix integrating probability and severity of adverse health effects in a semi-quantitative way to obtain an overall risk level

Figure 3.3 Risk matrix integrating probability and severity of adverse health effects in a semi-quantitative way to obtain an overall risk level.

pose several limitations for risk-based decision making (Cox, 2008). Their semi-quantitative nature and poor resolution make them less suitable for quantitative risk assessments and risk rankings (see also Chapter 2). Furthermore, the categorization of probability and severity, as well as the definition of the integrated risk levels, is in essence subjective - e.g., what is "medium" risk to one person might well be considered "low" or even "high" risk to another person.

In this chapter, we will focus on the more advanced risk metrics that quantify the health impact or economic impact of foodborne disease, building on Devleesschauwer et al. (2018). Fundamental to these metrics is the burden of illness, i.e., the quantification of the number of foodborne illness cases. We will illustrate each method's utility, data requirements, and output by developing an example on the burden of salmonellosis. We also discuss how these risk metrics can be used to perform risk ranking and how they can be integrated and extended to accommodate further risk ranking criteria. Finally, we introduce the concept of risk-benefit assessment as an extension of burden of disease studies.


Bottom-Up versus Top-Down Approaches

A fundamental input to both health and economic impact metrics is the number of foodborne illness cases that exist in the population or that arise through a given transmission route. Two general approaches, based on the data sources used in model construction, are used to assess the burden of illness (National Research Council 2010): a bottom-up approach following the risk assessment paradigm and a top-down approach following an epidemiological paradigm (Figure 3.4). In theory, both the top-down and

Bottom-up and top-down approaches for assessing risk. (Adapted from the EFSA Panel on Biological Hazards (BIOHAZ), EFSA /., 10, 2724, 2012 and Devleesschauwer, B. et al., in Food Safety Economics

Figure 3.4 Bottom-up and top-down approaches for assessing risk. (Adapted from the EFSA Panel on Biological Hazards (BIOHAZ), EFSA /., 10, 2724, 2012 and Devleesschauwer, B. et al., in Food Safety Economics: Incentives for a Safer Food Supply, 2018.) bottom-up approaches should result in similar estimates for likelihood and severity; in reality, significant data gaps and biases and uncertainty in the metrics make that unlikely (Bouwknegt et al. 2014). The approach selected will probably depend on the risks under consideration and available data. For example, epidemiologic data are typically less specific for assessing risks of exposure to specific food products, such as a particular brand of raw milk cheese, making the bottom-up approach more appealing. Alternatively, epidemiological data are typically more reliable to estimate the total incidence of disease due to a foodborne pathogen, such as campylobacteriosis, making top-down more appealing. EFSA has proposed a strategy to integrate top-down and bottom-up approaches in a Scientific Opinion about risk ranking (EFSA Panel on Biological Flazards (BIOHAZ) 2015).

The bottom-up approach, which derives estimates using the classic risk assessment paradigm that assesses risk using exposure and dose- response information, has been the subject of previous chapters. The focus of this section will therefore be on the top-down approach.

The top-down approach uses information on human disease gathered from public health surveillance and other epidemiological systems to estimate risk at the point of consumption. This can be accomplished according to two main models.

The first approach starts from pathogen-specific surveillance data collected through national surveillance systems. These data typically provide an underestimation of the true burden of illness because of under-ascertainment (i.e., not all patients seek healthcare) and underreporting (i.e., not all healthcare seeking cases will be diagnosed and reported to the national surveillance system). To estimate the true burden of illness, it is therefore necessary to reconstruct the surveillance pyramid. This allows the quantification of multiplication factors, which need to be multiplied with the number of reported cases to obtain an estimate of the true number of cases. For instance, in the United States, Scallan et al. (2011) used data from the Foodborne Diseases Active Surveillance Network to estimate the true number of cases of several foodborne infections in the population. For salmonellosis, the multiplier for under-diagnosis was defined as 29.3 (i.e., 1/29.3 = 3.4% of cases seek healthcare), while the multiplier for underreporting was set at 1.0 (i.e., no underreporting). By applying these multiplication factors to the number of 41,930 reported cases, the true number of cases could be estimated at 41,930 x 29.3 x 1.0 = 1,228,549. Of these cases, 11% were assumed to be travel-related, resulting in an estimated 1,228,549 x (1 - 0.11) = 1,093,409 domestically acquired salmonellosis cases.

The other approach starts from burden of illness envelopes, e.g., the total number of diarrhea cases in the population, and attributes these to specific foodborne hazards using population attributable fractions, which may be derived from surveillance, cohort, or cross-sectional studies. For instance, Pires et al. (2015) estimated in a meta-analysis that 2.2% of all diarrhea cases in children <5 worldwide were attributable to salmonellosis. Multiplying this etiological fraction with the total number of diarrhea cases worldwide thus yielded an estimate of the global number of salmonellosis-associated diarrhea cases in children <5.

In the next step, the estimated incidence of a foodborne illness can be attributed to specific transmission routes using results from source attribution studies (Pires et al., 2009). For instance, in an expert elicitation study conducted by Plaid et al. (2016), 73% of salmonellosis cases in the AMR A region (World Plealth Organization [WPIO] Region of the Americas - A; including the United States, Canada, and Cuba) were estimated to be foodborne; Ploffmann et al. (2017) further calculated that of those estimated to be foodborne, 22% were attributed to eggs, 22% to poultry meat, and 12% to pork. Continuing our example, this would result in an estimate of 1,093,409 x 0.73 = 798,188 salmonellosis cases obtained through foodborne transmission in the United States, of which 175,601 could be attributed to eggs, 175,601 to poultry meat, and 95,783 to pork.

< Prev   CONTENTS   Source   Next >