How Safe Is Our Food?

ow safe is our food? Put another way, how much illness in the United States is caused by foodborne pathogens? It sounds like a simple question. Getting a reasonable answer, however, is far from simple. The basic problem lies in the fact that only a small fraction of foodborne disease cases get reported through offi cial (or unoffi cial) reporting systems. Calculating the " real " rate of foodborne illness requires development of models that use reported cases as a starting point to estimate underlying disease rates. Given the plethora of pathogens that can be transmitted through foodborne routes, this is a complex, and somewhat daunting, process. It is, however, necessary for assessing the safety of foods and developing strategies for disease prevention. The articles by Scallan et al. (1,2) in this issue represent the latest efforts to develop such estimates of the magnitude of foodborne illness in the United States. In 1999, Mead et al. (3) published initial estimates of foodborne disease in the United States. This landmark undertaking was the fi rst to provide a comprehensive compilation of data from a variety of sources, including the Centers for Disease Control and Prevention (CDC) and the medical literature. It resulted in the often-cited estimates that foodborne pathogens cause 76 million episodes of illness , 325,000 hospitalizations, and 5,000 deaths each year in the United States. (Hereafter, episodes of illness are referred to as illnesses.) During the past decade, these numbers have strongly driven ongoing efforts to implement or reform regulatory systems to protect the public from food-borne illness. However, some aspects of the methods have been criticized, particularly the high degree of uncertainty of particular parameters and thus of the results themselves (4–6). These concerns have led to requests for CDC to repeat and update the work of Mead et al., using better methods and parameter estimates that more closely refl ect current realities. Now, ≈11 years later, Scallan et al. have produced " Sons of Mead, " which include substantial improvements to the methods used by Mead et al. and to the quality and timeliness of data (1,2). Scallan et al. should be commended , especially for 2 specifi c improvements: their advanced treatment of statistical uncertainty and variability and their transparent inclusion of voluminous appendixes of data, models, and assumptions. These authors followed the same basic approach as Mead et al. but chose to report their estimates …