Public Health Surveillance

Public health surveillance is conducted to define the magnitude and burden of a disease that needs public health action, to identify and investigate outbreaks so that control measures can be rapidly implemented and issues in need of further research swiftly identified, and to measure the impact of control and prevention efforts. The public health surveillance of infections that are likely to be foodborne now includes a substantial list of pathogens whose diagnosis is to be reported to public health authorities, and a new set of national networks for characterizing the pathogens and the illnesses they cause. The recent improvements in surveillance have been summarized in detail in a recent Institute of Medicine (IOM) publication (IOM/NRC, 2003). The following is a brief sketch of some of the improvements.

The primary authority for surveillance rests with the state health departments, which gather information from cities and counties and operate most of the public health laboratories. State and local notifiable diseases laws request or require clinicians and laboratories to report specific infections and to refer isolates of some pathogens to the public health laboratory for further characterization. These laws also typically require the reporting of unusual clusters or outbreaks of disease. In addition, many jurisdictions maintain complaint lines, to which concerned citizens may directly report illnesses or observations they think may need public health attention. Some food testing occurs in the course of routine inspections and as part of process monitoring within food production. This testing may also provide some information about the status of the food supply, though its purpose is usually the ongoing verification of process control, not safety testing of each lot.

Since 1996, the public health surveillance system for foodborne diseases has been strengthened in several ways. Several diseases were added to the standard notifiable disease reporting system, including non-O157 Shiga toxin-producing E. coli, hemolytic uremic syndrome, Cyclospora cayetanensis, and Listeria monocytogenes. The routine public health serotyping of Salmonella and Shigella was strengthened by the production and distribution of new antisera and training in their use; now new DNA sequence-based methods are being developed for more rapid identification of the serotype of Salmonella (McQuiston et al., 2004). Public health monitoring of antimicrobial resistance in several enteric bacterial pathogens has been implemented in parallel with monitoring of resistance in the same pathogens isolated from animals and foods, leading to the identification of such hazards as fluoroquinolone-resistant Campylobacter jejuni and multi-drug resistant strains of Salmonella enteriticas serotype Typhimurium and Salmonella enteriticas serotype Newport (Holmes and Chiller, 2004).

The reporting of outbreaks of foodborne diseases from local and state health departments has been improved by standardized and rapid reporting via the Internet and the Electronic Foodborne Outbreak Reporting System (CDC, 2005d). Enhanced surveillance, including a new collection form and improved close-out procedures doubled the number of foodborne outbreaks reported to more than 1,200 outbreaks each year (Figure 3-1). Now the Electronic Foodborne Outbreak Reporting System has changed an old and slow paper-based system into a more rapid reporting that makes it likely that a cluster of similar outbreaks occurring in several parts of the country at once will be detected and flagged, and also increasing the utility of the surveillance data to track trends in specific foodborne outbreak categories.

FIGURE 3-1

Reported outbreaks of foodborne diseases, 1990–2004, United States. SOURCE: Adapted from CDC (2006b).

PulseNet, CDC’s national network for subtyping foodborne bacterial pathogens, has been implemented in all 50 states and a growing number of large city health departments, as well as in the laboratories of the food regulatory agencies at the U.S. Department of Agriculture (USDA) and the FDA (Gerner-Smidt et al., 2006). This network relies on the submission of isolates of E. coli O157:H7, Listeria monocytogenes, Salmonella, and other bacterial pathogens from clinical laboratories to the public health laboratory, where the DNA “fingerprint” is determined using pulsed-field gel electrophoresis. Automated comparison of the digitized DNA pattern with the growing state and national database can swiftly identify strains (and therefore cases) that might be related, detecting clusters spread across multistate jurisdictions that might otherwise have been missed completely. In the 1960s, Salmonella serotyping transformed surveillance for that organism by increasing the signal-to-noise ratio and making it possible to pick out outbreaks of one serotype from the background noise of all salmonellosis (CDC, 1965). Now PulseNet provides an additional specificity, with a generally applicable tool for identifying clusters of infections that are likely to be related, even within a single closely-related serotype such as E. coli O157:H7, or within individual Salmonella serotypes. PulseNet test protocols have now been developed for seven bacterial foodborne pathogens, as well as for Yersinia pestis and F. tularensis.

PulseNet protocols have now been adopted in Canada, Japan, Australia and other countries and are the heart of international networks for surveillance in Europe, Asia and the Pacific, and Latin America (Swaminathan et al., 2006). This will enhance our own prevention capacity. For example, in 2004, public health laboratories in Japan detected a small cluster of E. coli O157:H7 infections in Okinawa that they linked to consuming ground beef from the commissary at a U.S. military base there, and an indistinguishable E. coli was detected in ground beef in Japan, which came from the United States (CDC, 2005b). The notification by Japan led to recall of 90,000 pounds of ground beef shipped to the military and other institutions in the United States. The same strain was also identified in two persons in the United States who did eat beef the origin of which was not traceable, and who would not otherwise have been linked.

In the future, routine usage of multilocus variable number tandem repeat assays or other sequence based-methods in state health department laboratories will further refine the speed and precision of the network. However, the promise of real-time results is more dependent on resources, rather than technology, including the vital participation of the private clinical laboratory sector to refer strains rapidly to the public health laboratory and on the laboratory support within the state health department to run the tests swiftly.

Another major advance in foodborne surveillance has been FoodNet, the active sentinel site surveillance system for foodborne illness (Allos et al., 2004). While PulseNet enhances the ability of all states to detect clusters and investigate outbreaks, FoodNet is focused on developing standard and detailed surveillance data on sporadic (nonoutbreak-associated cases) in 10 sites around the country, now representing 14 percent of the U.S. population. Though sporadic cases are far more common than those that are associated with outbreaks, they receive far less attention in general. Active surveillance means that the health department regularly contacts the clinical laboratories to collect reports of what has been diagnosed, rather than relying on the laboratories to report them. In addition FoodNet conducts specialized surveys of the clinical laboratories, of the general population, and of other groups to obtain measures of the frequency of gastroenteritis in general, of specific diagnostic tests, and other measures important to interpreting surveillance data. Data from FoodNet have been critical to refining the overall estimates of the burden of foodborne disease and to tracking trends in specific infections over time. For example, between 1996 and 2004, FoodNet documented a 42 percent decline in diagnosed E. coli O157 infections, decreasing to 0.9 per 100,000 in the year 2004; a 40 percent decline in Listeria infections; and a 31 percent decline in Campylobacter infections (CDC, 2005c). With case-control and other studies, FoodNet also defines the association between infections and specific foods, contributing to the attribution of the burden of specific infections to foods. Increasingly, FoodNet serves as a platform for developing and evaluating improved public health surveillance and investigative and prevention strategies.

Estimating the Burden of Foodborne Diseases

The health burden of an infection includes the morbidity it causes, the hospitalization and other medical care that results, and the mortality, among other measures. Estimating this burden for a given pathogen means going beyond the reported cases. To contribute a reported case, the person must become ill, must seek medical care, the physician must ask for a laboratory test, the patient must provide a specimen for diagnostic study, the specimen must yield evidence of the pathogen, and the case must be reported. Slippage at each point means that the diagnosed cases are likely to represent only a small fraction of the cases that actually occur. Other measures of severe infection, such as hospital discharge summary records and cause of death as reported on death certificates, may be used to estimate the total number of hospitalizations and deaths due to acute enteric disease, but these measures significantly underreport specific infections, as laboratory diagnoses may often not be reflected in the discharge or death certificate coding. In 1999, we published a report estimating the actual acute health burden of foodborne disease in the United States (Mead et al., 1999).

These estimates were assembled from a variety of data collected by FoodNet and other sources. FoodNet population surveys measure the number of cases of acute gastroenteritis that actually occur and the proportion of these that seek care and are cultured (Herikstad et al., 2002). The FoodNet clinical laboratory surveys measure the likelihood that a specimen will be routinely tested for say, Salmonella or Campylobacter or E. coli O157 (Voetsch et al., 2004a). This information can then be used to amplify the number of cases that are diagnosed and reported; in this way FoodNet estimated that there are actually 38 cases of salmonellosis for every one that is diagnosed and reported (Voetsch et al., 2004b). FoodNet data also provide the number of diagnosed salmonellosis cases that lead to hospitalization and the number that lead to death. Doubling that number to account for cases that were not cultured provides a conservative estimate of the total number of hospitalizations and deaths. Using similar data and assumptions, the incidence of other infections under surveillance by FoodNet can also be estimated, and by use of a uniform set of assumptions and expert opinion it is possible to estimate the overall burden of known enteric infections at some 39 million infections per year (Mead et al., 1999).

The next step was to estimate the proportion of these infections that are transmitted through food, rather than through water, direct contact with ill children, or other pathways. The estimated proportion of infections that are transmitted through foods varied by pathogen, and in sum was 38 percent. Thus, of 39 million enteric infections estimated to be caused by the known enteric pathogens, 16 million were attributed to food. A curious observation is that the estimate of acute enteric illness developed pathogen by pathogen (annual incidence of 39 million cases) is substantially less than the total amount of acute gastroenteritis in the population estimated by population survey (annual incidence of 211 million cases) (Mead et al., 1999). This “diagnostic gap” suggests that there are more pathogens yet to be discovered (Tauxe, 2002). The fraction of these other cases not accounted for by known pathogens that might be attributed to food is not directly measurable. The authors of the 1999 estimate chose 38 percent, the summary statistic for the known pathogens, as the best point estimate of what it might be for the other acute illnesses not accounted for by known pathogens. The final estimate, 76 million illnesses, 323,000 hospitalizations, and 5,000 deaths, refers to the year 1997. This comprehensive estimate is now being revised in a similar stepwise approach, starting with the measurement of the overall burden of acute gastroenteritis and with more refined and pathogen-specific approaches to the estimates of unreported illness.

There are other ways of measuring the burden of unreported illness. In the United Kingdom, the Intestinal Infectious Diseases study empanelled a group of citizens who recorded their symptoms prospectively and provided stool specimens for even mild cases of diarrheal illness (Wheeler et al., 1999). The Dutch SENSOR study followed a similar strategy, working with a group of sentinel general practitioners and their patients (de Wit et al., 2001). Both European approaches depended on the national healthcare system itself to provide a population-based framework, and both were sufficiently expensive that they have not been repeated. There are also measures of burden other than simple counts of cases, hospitalizations, and deaths. For example, the health-related costs for the principal bacterial foodborne pathogens (Salmonella, Campylobacter, E. coli O157, other Shiga-toxin-producing E. coli, and Listeria monocytogenes) have been estimated to be $6.9 billion (ERS, 2000). The cost to society associated with the estimated number of deaths that were not attributed to known etiologies could be as high as $17 billion, underlining the need for further refinement of this sector of the estimate (Frenzen, 2004). Inclusion of the postinfectious sequelae in the estimate can also greatly increase the economic burden. A detailed model developed for Campylobacter in the Netherlands included the burden of postinfectious arthritides and Guillain-Barre syndrome and measured the burden in disability-adjusted life years; this estimate indicated that a greater burden was due to the sequelae, rather than the acute illness (Havelaar et al., 2005). The industry costs of disrupted trade and development that can be occasioned by foodborne illness can be enormous, though they usually do not appear on the public health ledger. The costs of antimicrobial resistance associated with foodborne exposures have also not been estimated, but they might include the cost of illness caused by resistant foodborne pathogens and the costs related to the spread of transmissible resistance genes that are present in commensal organisms in the food supply, from which they may transfer to human pathogens.

Prevention of Emerging Foodborne Threats: The Importance of the Outbreak Investigation

The prevention of foodborne diseases in general is a complex effort, involving many different actors along the chain of production from the farm to food service. There are many different pathogens involved, almost none of which are vaccine preventable in the final consumer. Educating consumers, food handlers, and producers about their role in preventing foodborne disease is important, but not sufficient. Contamination of food can occur at many points from farm to table. Often the key to prevention is to understand those mechanisms of contamination well enough to prevent them, before the food reaches the final consumer. Investigation of contamination events, and especially investigation of outbreaks, is critical to understanding the mechanisms of contamination. Prevention often means reengineering food processes and policies for safety, usually with a focus on a specific food and/or pathogen.

The foodborne outbreak investigation is thus a major driver for enhancing overall food safety. When an outbreak is detected, the first priority is to learn enough to prevent further cases from occurring in the current outbreak. However, it is also an opportunity to learn something new, and to open research agendas with impact far beyond the one outbreak. Many foodborne pathogens were first identified in the course of an outbreak investigation. A new combination of pathogen and food may be revealed that needs further study by food scientists, animal and plant pathologists, as well as medical researchers. Just as the National Transportation Safety Board investigates a plane crash thoroughly after the fact to learn how to prevent similar ones, careful investigation even after an outbreak is over can define gaps in the system, stimulate further specific research, and indicate the needs for new processes or regulations. New combinations of specific pathogens and foods identified by outbreak investigations have been critical to guiding research and prevention (see Table 3-2).

TABLE 3-2

Some New Pathogen-Food Combinations Characterized During Outbreak Investigations in the United States.

As the surveillance systems that we use in the United States have been enhanced in the last 10 years, we have observed a change in the number and nature of outbreaks detected. This is a paradox of surveillance: making surveillance better often reveals more of the problem, so that the actual public health burden appears worse. For example, as noted above, the number of foodborne outbreaks reported through the Electronic Foodborne Outbreak Reporting System doubled following relatively simple improvements in process and participation. PulseNet has caused a more substantial change in the nature of the outbreaks detected. By increasing the signal-to-noise ratio for specific pathogen subtypes, PulseNet makes it far more likely that geographically diffuse outbreaks will be detected. Those diffuse outbreaks are particularly instructive.

PulseNet has had a profound impact on the kind of outbreaks that have been detected because the nature of the outbreaks detected depends critically on the methods used to detect them. If outbreaks only come to the attention of public health when concerned citizens, physicians, or healthcare facilities report them, then only large and locally apparent outbreaks are likely to be detected. These classic point source outbreaks often affect a single group of people in a single town or city, following a single meal, with a substantial attack rate. Investigating this outbreak proceeds with local authority, and the food-handling problems that are identified are often local in scope. Although important, these investigations may have greatest impact at the local level.

The use of molecular subtyping to compare strains across many jurisdictions has revealed an entirely different kind of outbreak in which a dispersed group of persons who do not know each other are affected at the same time with the same infecting organism in many different locations. In this scenario, no local listening post may perceive more than a few cases, and the local increase is often not apparent against the background of cases. Although each individual case may appear to be sporadic, the outbreak may in fact be very large but dispersed. Investigating these dispersed scenario outbreaks requires the coordinated efforts of many health authorities acting in concert and pooling the information. Though difficult to detect and to investigate, the findings of these outbreaks can be of particular importance. The dispersion may well reflect a contamination event high in the food’s chain of production, not just a problem in the final kitchen. Identifying that event can instruct the industry and regulatory authorities about a flaw in the system that was previously unappreciated. Correcting it can lead to lasting and generalized protection across the country.

This means that improved detection and investigation can serve to probe the safety of the food production system at several levels. These investigations, providing information about gaps in the food safety system, drive the cycle of prevention faster and reduce the overall number of infections. The results of enhanced prevention can be seen in the recent declines in the incidence of infections with Listeria monocytogenes and E. coli O157, the two pathogens tracked most intensively by PulseNet. Following the institution of PulseNet surveillance for Listeria monocytogenes, there was an important increase in the number of outbreaks detected (Figure 3-2). Many of these were related to processed meats, focusing prevention efforts on that sector; incidence declined to an all time low of 2.7 per million in 2004, a drop of 40 percent since the baseline period 1996–1998 (CDC, 2005c). The incidence of E. coli O157 infections began to decrease sharply after 2002, as the repeated investigations of pulsed-field gel electrophoresis clusters focused attention on more specific controls at the level of ground beef. By 2004, the incidence of E. coli O157 infections as measured in FoodNet had dropped 42 percent since the baseline period of 1996–1998, and was 0.9 per 100,0000, below the goals set by Healthy People 2010. It is doubtful that such progress would have been made without PulseNet.

FIGURE 3-2

Reported incidence of Listeria monocytogenes infections and reported outbreaks of listeriosis, United States, 1986–2004. SOURCES: Adapted from Tappero et al. (1995); CDC (2006a,).

The most recent outbreak surveillance information published for E. coli O157:H7 also illustrates how improved surveillance can first produce a sharp increase in reported outbreaks, followed by a drop as better prevention strategies take effect (Figure 3-3) (Rangel et al., 2005). In the 1980s, E. coli O157 outbreaks of infection were rare, perhaps because the pathogen itself was less common, but also because it was not likely to be diagnosed or reported. Washington, the first state to make it a notifiable condition, did so in 1988. After the large Western states outbreak of 1993, centered in Washington, many other states made it notifiable, and it became nationally notifiable in 1994. At the same time, an education effort targeting clinical laboratories promoted simple laboratory screening for the pathogen. It is not surprising that the number of reported outbreaks jumped to more than 30 in 1994, and then began to decline, as many in the fast-food industry and homes changed burger cooking procedures to avoid undercooking. In 1996–1997, in the first FoodNet case-control study of sporadic E. coli O157:H7 infections, eating burgers at a fast-food restaurant was no longer associated with illness, though it had been in earlier studies (Kassenborg et al., 2004). Following the launch of PulseNet, the number of reported outbreaks more than doubled again in 1998, and since then has generally trended downwards as other prevention measures have been enacted. As noted above, the biggest decrease may have happened after 2002, after new procedures to reduce the contamination of ground beef were implemented, though the impact of those on E. coli O157:H7 outbreaks has not yet been summarized in print.

FIGURE 3-3

Reported outbreaks of E. coli O157 infections, United States, 1982–2002. SOURCE: Rangel et al. (2005).

Constraints and Limitations on Using Outbreaks to Drive Control and Prevention

Detecting and investigating foodborne outbreaks triggers public health intervention, but as a prevention system, this has built in delays. Most obviously, it is not activated until after people are exposed, become ill, and the outbreak is detected. Often the outbreak is actually over by the time it is detected, making the outbreak investigation itself a blunt instrument for achieving control over single brief exposures to contamination. There are biology-dependent delays, like the incubation period between the moment of exposure and the onset of symptoms, which can vary from less than a day for some pathogens to several weeks for others. Delays in diagnosis may depend on when the typical patient feels sufficiently ill to consult a physician and on how long the laboratory tests take to yield a diagnosis. Signal delay depends on the time it takes to accumulate enough cases in one database to be detectable as a distinct cluster and may be longer if cases are more dispersed. There are also the resource-dependent delays that depend critically on support that surveillance gets in the private and public sectors. Clinical laboratories may batch isolates for shipment to the public health laboratory to cut shipping costs, adding delays. The speed of testing strains in public health laboratories also depends on the available resources. Interviewing cases and tracing implicated foods back to their sources depends on the availability of trained and supported investigative staff, for whom this is part of their core work duties, not a distraction from the “real” work they must accomplish.

Many state and local health departments lack sufficient capacity to effectively investigate outbreaks. For example, in a recently published survey of state health departments, 88 percent reported there were barriers to more active case finding, and 30 percent reported that they lacked adequate epidemiological staff to conduct investigations. Outbreaks go uninvestigated for a number of reasons, the most common of which are delayed notification (83 percent of those respondents) and lack of staff (67 percent). Many state public health laboratories are also understaffed and under supported, making real-time testing of submitted strains difficult and leaving them with little surge capacity for emergencies (APHL, 2003).

Pooling Resources

The fundamental ability to detect and investigate outbreaks is critical to the response to any new threat, be it intentional or natural in origin. In the only two intentional foodborne attacks involving bacterial infectious agents in recent experience in the United States, local and state health departments detected and responded to outbreaks as a matter of course; it was not recognized at the outset that the events were intentional in origin (Torok et al., 1997; Kolavic et al., 1997). Some states have recognized that robust public health surveillance is a fundamental part of protecting the public health against both natural and intentional events, and the infusion of new resources to strengthen the response to bioterror threats has improved surveillance and response capacity in general. However, in many others, bioterror response resources have lured staff and attention away from the traditional activities of public health, leaving those systems weaker than before. Yet actual foodborne bioterror events remain remarkably rare. During the last 25 years, two have been documented, while during this same time, applying the recent Electronic Foodborne Outbreak Reporting System number of 1,200 food-borne outbreaks a year, there were an estimated 30,000 nonintentional foodborne outbreaks. Like firefighters in a firehouse restricted to arson fires, the dedicated squad will have to practice its skills in drills and table top exercises, while their less well-equipped and perhaps less well-paid colleagues put out all the fires, which must occur before the fire can be determined to be arson or not.

Enhancing Foodborne Outbreak Investigations

Despite the many constraints, the response to foodborne outbreak investigations can be improved substantially. Expanding routine subtyping to a greater number of isolates and different pathogens may help drive prevention faster and more broadly. Faster and more automated subtyping methods, including sequenced-based methods for both bacteria and viruses, need to be deployed as the cost and complexity of sequencing equipment decreases. Cluster detection can be speeded by providing the resources for swift transport and real-time testing rather than batch processing. Routine case interviews can be made more swift and standardized, and they can be integrated with control interviews to make investigation rapid. Improving the record keeping necessary for accurate tracing of foods to their sources, and increasing the skilled capacity in the state and national regulatory agencies to perform such investigations would speed that critical phase of investigations. Fingerprinting the pathogens isolated from foods and food animals in real time and linking those data with the human isolate database would make it possible to rapidly generate hypotheses about potential sources. Expanding the capacity for surveillance in other countries around the world and expanding regional and global surveillance networks to detect and investigate outbreaks can enhance the detection of foodborne threats at home and abroad (Tauxe and Hughes, 1996).

Conclusion

Enhancing foodborne surveillance, outbreak detection and response means better public health. Outbreaks will continue to occur, and people will continue to get ill, but with effective response, these unfortunate events can drive prevention. FoodNet is providing better data to measure trends and burden of illness and to determine sources of sporadic infections. PulseNet, the new network for molecular subtyping, is blurring the line between obvious outbreaks and apparently sporadic cases, and it is probing more deeply into the safety of the entire food system. Investigating and learning from the outbreaks is critical to achieving continuous improvement in food safety. As the public health system is likely to be the first responders in the event of an intentional attack, as well as for the far more frequent unintentional outbreaks, having a more robust and effective response is better for the public protection in either event.

Prologue: An Ongoing Tale of Two Settings, Viewed from Multiple Perspectives

Cyclospora cayetanensis infection is endemic in many resource-poor and middle-income countries (Bern et al., 1999, 2000; Herwaldt, 2000; Lopez et al., 2003; Ortega et al., 1993). The United States, a resource-rich consumer country, has unwittingly imported this foreign, enigmatic parasite along with fresh produce that has been linked to outbreaks of cyclosporiasis. From the U.S. perspective, the unforeseen emergence of this microscopic pathogen has evolved into an ongoing tale of large foodborne outbreaks, which have entailed crossing jurisdictional borders and working with foreign governments, produce growers, and trade organizations (CDC, 1998, 2004; Herwaldt, 2000; Herwaldt et al., 1997, 1999; Ho et al., 2002; Lopez et al., 2001).

The parasite C. cayetanensis, which was christened in 1994 (Ortega et al., 1994), and the precedent-setting series of outbreaks of cyclosporiasis continue to be sources of fascination, frustration, challenges, and learning opportunities for the persons, agencies, industries, and governments they have directly or indirectly affected. The parasite and various aspects of the outbreaks (e.g., coordination of multisite investigations, interactions with foreign sources of implicated produce, regulatory responses when the mode of contamination is unknown, impacts on international trade) have been and continue to be subjects of case studies (Calvin, 2003; Jackson, 2006; Powell, 1998), including a U.S. Senate hearing in 1998 entitled The Safety of Food Imports: From the Farm to the Table: A Case Study of Tainted Imported Fruit (U.S. Senate, 1998) and this case study, which was presented in part at the IOM’s October 2005 workshop on food safety.⁵

This case study—in which the parasite and the U.S. outbreaks are viewed with a wide-angle lens that includes public health, societal, and historical contexts—focuses on food for thought (e.g., perspectives, principles, and issues for the public health community to ponder) rather than detailed commentary about foodborne threats to health or comprehensive review of the outbreaks of cyclosporiasis. In the text of the article, selected details about the parasite and the outbreaks are included for illustrative purposes, such as to underscore lessons learned, relearned, or yet to be learned and to highlight common themes (e.g., challenges intrinsic to emerging pathogens). Supplemental details and perspective are provided in figures (Figure 3-4, Figure 3-5) and tables (Tables 3-3 through 3-6, which may be found at the end of this article in Annexes 3-1 through 3-4). Table 3-4 and Table 3-5 represent attempts to list and dissect the elements of foodborne outbreaks and investigations to demonstrate what various ingredients add to the mix (e.g., the challenges, opportunities, and approaches if the etiologic agent is an enigmatic parasite).⁶

FIGURE 3-4

Generic and Cyclospora-specific challenges in bridging the chasm between disease and prevention and control. SOURCE: B. Herwaldt and D. Juranek, CDC, Division of Parasitic Diseases, April 2006 (see Acknowledgments).

FIGURE 3-5

Sporulation of Cyclospora cayetanensis oocysts. Unsporulated and sporulated C. cayetanensis oocysts are shown in graphic illustrations (top) and photographs (bottom), as viewed by differential interference contrast (DIC) microscopy, a specialized type (more...)

TABLE 3-3

Cyclospora cayetanensis and Cyclosporiasis: Perspectives and Status as of April 2006.

TABLE 3-4

Characteristics of Foodborne Outbreaks and Investigations: Challenges, Opportunities, Approaches, and Advances, in General and Specific to Outbreaks of Cyclosporiasis Associated with Imported Fresh Produce.

TABLE 3-5

Goals of Investigations of Foodborne Outbreaks: Challenges, Opportunities, Approaches, and Advances, in General and Specific to Outbreaks of Cyclosporiasis Associated with Imported Fresh Produce.

TABLE 3-6

Factors that Have Complicated Efforts to Communicate About Foodborne Outbreaks of Cyclosporiasis.

Investigating the initial outbreaks in the 1990s would have been even more difficult than it was if Cyclospora and cyclosporiasis had emerged in the United States as complete unknowns. The fact that they did not reflects the contributions and astute observations of relatively few persons with expertise in parasitology and tropical medicine, in diverse places such as Papua New Guinea, Peru, and Nepal (Ashford, 1979; Hoge et al., 1995; Ortega et al., 1993, 1994). Their efforts culminated in fundamental scientific and medical advances, described in articles published seemingly just in the nick of time. Through studies in Peru, the confusion about the identity of the organism was resolved: it is not a species of blue-green algae (cyanobacteria); it is a coccidian parasite, the first and only species in the Cyclospora genus known to infect humans (Ortega et al., 1993, 1994). In a placebo-controlled clinical trial in Nepal, the antimicrobial combination of trimethoprim-sulfamethoxazole was demonstrated to be highly effective treatment of cyclosporiasis (Hoge et al., 1995), the first and only such therapy to have been documented (Table 3-3, Annex 3-1).

Unfortunately, the parasitologists were not prophets: the experts were as surprised as the novices by the unpredicted U.S. emergence of C. cayetanensis and by the unprecedented association between a parasite and large, common-source foodborne outbreaks. Although other enteric parasites are known to be transmissible by contaminated food, nothing remotely comparable to the widespread, recurrent outbreaks of cyclosporiasis has been documented in the United States for any other parasite. The first of the series of major eruptions of C. cayetanensis on the international scene occurred in the spring of 1996, after premonitory rumblings earlier in the decade (Herwaldt, 2000; Huang et al., 1995; Koumans et al., 1998). The eruption in 1996 took the form of a large (>1,000 reported cases⁷), multinational outbreak of cyclosporiasis in two countries, the United States and Canada, that was linked to a third country, Guatemala, where the epidemiologically implicated raspberries were grown (Herwaldt et al., 1997).

The modus operandi of this pathogen in the United States as the etiologic agent of outbreaks has not changed during the subsequent decade, although the repertoire of food vehicles and sources has expanded beyond raspberries from Guatemala to include assorted types of fresh produce from several middle-income countries (Table 3-4, Annex 3-2; Table 3-5, Annex 3-3). The saga of outbreaks appears to have evolved into an interminable tome, with no end in sight. Its inscrutable chief character, C. cayetanensis, a unicellular (protozoan) parasite, continues to wreck havoc, surprise, outsmart, baffle, and bewilder us (Table 3-3, Annex 3-1). As discussed in this article, Cyclospora epitomizes the challenges intrinsic to addressing obscure pathogens that appear, seemingly out of nowhere, including how and why the challenges translate into difficulties investigating and preventing outbreaks and communicating among health professionals, the general public, and the produce industry. The scientific unknowns and political overtones are among the factors that have complicated efforts to communicate and collaborate (see Tables 3-4 through 3-6, Annexes 3-2 through 3-4).

Cyclospora, the U.S. outbreaks of cyclosporiasis, and their aftermaths have affected physical, economic, and communal health in exporter and importer nations (Calvin, 2003; Herwaldt, 2000; Jackson, 2006; Powell, 1998) (Table 3-6, Annex 3-4). The need to invest resources to investigate the outbreaks has resulted in increased recognition of and interest in this parasite and its effects on the persons, products, and places where Cyclospora infection is endemic. The extent to which the heightened awareness will stimulate long-term investments in multi-disciplinary, multilingual research activities; the research will solve remaining mysteries about this elusive pathogen, its quirky human hosts, and their micro-and macrohabitats; and the increased knowledge will be translated into wisdom, vision, and sustainable, effective, transnational prevention and control measures remains to be seen and recorded. The potential for additional scientific advances to have positive ripple effects that extend beyond Cyclospora and cyclosporiasis is high.

Challenges in Addressing Public Health Issues in General and Foodborne Outbreaks of Cyclosporiasis in Particular

Addressing public health issues, even ostensibly straightforward matters, can be difficult in part because of competing demands for scarce resources. The challenges are compounded for chronic issues such as foodborne cyclosporiasis that are associated with confounding complexities and unknowns. Figure 3-4 and the text in this section of the article represent attempts, replete with mixed metaphors, to place the challenges in perspective by depicting and describing the elements of a support structure for a public health bridge between disease and prevention and control. The goal (i.e., to bridge the chasm between disease and prevention and control), the base on which the support structure for the bridge is built (i.e., the public health infrastructure), and the societal and historical contexts for public health activities are not unique to cyclosporiasis.

Building and maintaining a structurally sound, science-grounded bridge are challenging (i.e., are uphill struggles, as depicted by the angle of the bridge). The instability and vulnerability of the base (soil) are major impediments. The infra-structure is portrayed as underground (unseen), overburdened, stretched, atrophied, eroded, and diverted. The impoverished infrastructure includes the residual resources of all types, at all tiers (e.g., local, state, federal) of the public health system. Core infrastructure constraints can be appreciated through the lens of an analogy, in which the public health system is viewed as an internal combustion engine, subject to the laws of thermodynamics (see footnote for details).⁸ Revitalizing the infrastructure will require commitment, concerted effort, and coordination from and among all tiers of the public health system, as well as great wisdom and vision, archetypical eroded, diverted elements.

Addressing foodborne cyclosporiasis entails adding more loads to the overburdened infrastructure in the form of the support elements for the public health bridge between disease and prevention and control. The elements are depicted as:

• a foundation stone, which symbolizes the importance of addressing fundamental constraints intrinsic to emerging pathogens; and
• pathogen and vehicle pillars (support columns), which symbolize the needs to address the superimposed challenges specific to the type of pathogen (the unique peculiarities of the coccidian parasite C. cayetanensis [Table 3-3, Annex 3-1]) and the types of food vehicles (not just fresh produce, but imported produce served in inconspicuous ways, such as garnishes [Table 3-4, Annex 3-2; Table 3-5, Annex 3-3]).

A multidimensional, complex web of interactions is depicted by the checkerboard pattern under the bridge. Although the concept of pathogen-vehicle interactions is highlighted, additional types of interactions (synergisms, antagonisms, collaborations, feedback loops) among the elements of the support structure for the bridge, the infrastructure, and society at large are germane (Table 3-4, Annex 3-2; Table 3-5, Annex 3-3). The fact that the parasite C. cayetanensis is emerging while parasitologists are becoming an endangered species is an ironic example of a negative pathogen-infrastructure interaction.

The public health infrastructure is further stretched and strained if the challenges associated with emerging pathogens must be addressed in the context of emergencies (the extraordinary demands associated with large, multisite outbreaks [Table 3-4])—i.e., if a base of scientific knowledge about the pathogen and databases for outbreaks must be created concurrently and on the fly. If the infrastructure withstands the cumulative burden of many converging stresses, outbreak investigations, in concert with basic and applied research activities, provide opportunities to solve mysteries through scientific advances and to identify and reinforce weak elements in the public health system (Buchanan, 1997; Hall, 1997; Tauxe, 1997) (see Table 3-4, Annex 3-2; Table 3-5, Annex 3-3). The yin of outbreaks can be partially converted into yang, by translating challenges into opportunities into advancements in science and the public good.

Challenges in Addressing Emerging Pathogens, Parasites, and Cyclospora cayetanensis

Challenges Entailed in Discovering and Classifying Cyclospora cayetanensis (1977–1994)

Challenges Associated with Detecting and Investigating the First Three Documented U.S. Outbreaks (1990 and 1995)

The accounts of the first three documented U.S. outbreaks of cyclosporiasis, which occurred in 1990 and 1995, underscore the understandable difficulties intrinsic to investigating outbreaks when little is understood about the etiologic agent (e.g., potentially pertinent questions and hypotheses are not considered or rigorously explored). All three of these outbreaks were detected because of unusual circumstances: stool specimens from the index case-patients were examined in clinical microbiology laboratories in which techniques that facilitate detection of Cyclospora were used routinely in parasitologic examinations (Table 3-3, Annex 3-1).

Evolving and Persisting Challenges Associated with the Next Era of Outbreaks of Cyclosporiasis: From Harbingers (in 1990 and 1995) to Sonic Booms (1996–2005)

The ongoing saga of cyclosporiasis has evolved from harbingers to serial, large outbreaks, detected during many of the seasons and most of the years in the subsequent decade (1996–2005). Despite the preparedness exercises with the harbinger outbreaks in 1990 and 1995, we¹⁶ knew little about the biology of Cyclospora or the epidemiology of cyclosporiasis in 1996, when confronted with a large outbreak (1,465 reported cases), which included 55 clusters of cases (i.e., 55 minioutbreaks associated with social and other events). In retrospect, the outbreak in 1996 can be viewed as the first course of what has become an ongoing curriculum of difficult challenges and learning opportunities: a progressive dinner with assorted types of fresh produce, nontraditional exports from several middle-income countries (e.g., raspberries and snow peas from Guatemala, mesclun lettuce from Peru, basil from Peru and Mexico¹⁷).

Selected details about some of the outbreaks are provided, for illustrative purposes, in the text and tables. As noted in the prologue, Table 3-4 and Table 3-5 provide matrices for listing and dissecting the ingredients of outbreaks and investigations. Examples of challenges, opportunities, approaches, societal contexts and trends, scientific advances, and lessons (re)learned, codified, and yet to be learned are included as well. The outbreaks and their impacts, particularly the series of outbreaks linked to Guatemalan raspberries, are discussed in detail in other articles and case studies (Calvin, 2003; CDC, 1998, 2004; Herwaldt, 2000; Herwaldt et al., 1997, 1999; Ho et al., 2002; Jackson, 2006; Lopez et al., 2001; Powell, 1998; U.S. Senate, 1998).

Epilogue: The Saga Is Ongoing, More Work Needs to Be Done, and the Workers and Resources Are Few

The work is not done just because the organism (i.e., C. cayetanensis) has been classified by genus and christened, one highly effective therapy for cyclosporiasis has been identified, our abilities to detect and investigate cases and outbreaks have markedly improved, and specific produce vehicles and sources have been identified in some of the outbreak investigations. Cyclospora did not gracefully exit after its dramatic entrance, nor has it fully emerged from obscurity into clarity. Despite scientific advances at the margins, Cyclospora remains largely veiled in mystery; we remain plagued by fundamental unknowns and by fundamental constraints as we seek to convert unknowns into knowns (Table 3-3, Annex 3-1). Our ability to prevent contamination of produce and thereby to prevent what we have been investigating (i.e., foodborne outbreaks) would be markedly improved by increased understanding of the biology of Cyclospora and the epidemiology of cyclosporiasis.

The lack of commitment to and the paucity of resources for the long-term investments—i.e., marathons not sprints—needed to solve the many remaining mysteries are disconcerting, especially because the ramifications of our ignorance and the collateral benefits of the advances to date are palpable. If for no other reason than self-interest (e.g., to avoid the actual and opportunity costs of investigating serial outbreaks), the need to decipher Cyclospora’s enigmatic code is evident. Cracking the code of a microbe as hardy and resilient as Cyclospora—e.g., identifying decontamination strategies that kill or remove the pathogen without damaging delicate fresh produce—almost assuredly would have far-reaching positive ripple effects for food safety.

One of the reverberating themes in the saga of this parasite is that the investments of “few” (e.g., the parasitologists’ mite) can result in large proceeds. Fortunately, some doggedly persistent scientists are continuing to poke and prod Cyclospora, despite the paucity of resources and oocysts, with the expectation that ultimately it will excyst its secrets. These resilient investigators emulate the parasite’s uncanny abilities to survive in austere microniches and to resist the pressures and stresses in harsh macroenvironments. Relatively few persons deserve the credit for the fact that Cyclospora and cyclosporiasis did not emerge in the United States as complete unknowns. In the future, which, by definition, is unknown, relatively few persons may deserve the credit for identifying and implementing an exit strategy for this inscrutable pathogen and for writing the last chapter of the chronicle of cyclosporiasis, which now appears to have no end in sight.

Acknowledgments

I am indebted to the countless persons who have contributed in various ways to the marathon investigations of the U.S. outbreaks of cyclosporiasis. I reluctantly acknowledge that Cyclospora, which has defined the seasons of more than a decade of my life, has been a worthy opponent. Special thanks to my CDC colleague, Dr. Dennis D. Juranek, for his innumerable, invaluable insights and for his artistic contributions, most notably to Figure 3-4 and Figure 3-5.

Background

In mid-April 2005, a private laboratory reported a dozen cases of cyclosporiasis to the Florida Department of Health (FDH), Bureau of Epidemiology Surveillance Section. The total number of cases reported in 2004 for Florida was nine. For reporting week 14 ending April 16 (the week the positive results were received from the private lab) the average number of cyclosporiasis cases from 2003 to 2005 was 1.67. The number of cases up to week 14, 2005, was approximately 20 percent higher than normally expected (FDH, 2005a). By reporting week 17, the percent increase was 162 percent, a clear indication of a possible outbreak. Cases were reported from numerous counties with no initial apparent geographical or temporal pattern. This article will discuss the methods of the ensuing complex investigation, epidemiological findings, and recommendations for future investigations involving outbreaks of disease resulting from intentional or unintentional contaminants of the United States food supply.

Methods

Epidemiology teams were formed at the state and affected county levels of the FDH. The core statewide team consisted of epidemiologists from the Division of Disease Control and the Division of Environmental Health. A statewide lead investigator was established for this investigation using the established organizational position of Statewide Coordinator of Food and Waterborne Diseases. The statewide team also consisted of information management personnel, public communication experts, and administrators. Each assembled county health department team comprised county epidemiologists, environmental health personnel, and epidemiology nurses. Each of these teams also included regional food and waterborne disease epidemiologists and/or Florida Epidemic Intelligence Service epidemiologists. County health department administrators were also a very supportive part of the county teams. As the investigation progressed, owners and management of the numerous affected food service establishments became integral parts of the investigation, supplying the investigators with critical information pertaining to product and patrons. The FDA, the CDC Division of Parasitic Diseases, the Florida Department of Agriculture and Consumer Services, and the Department of Business and Professional Regulation also provided valuable assistance with consultation, formal traceback, investigation of food service facilities, and farm investigation activities.

Laboratory analysis of clinical specimens originated with private laboratories. Florida has had numerous previous experiences with Cyclospora outbreak investigations and, due to prior misidentification issues, understood the necessity of confirming the findings of private laboratories. Table 3-7 depicts a listing of previous outbreaks in Florida, other states, and other countries. Thus a system was set up whereby private laboratory results were sent to a single FDH coordinator. The private laboratories were also asked to send their slides to the FDH Bureau of Laboratories in Jacksonville for confirmation. Laboratory results were provided daily, sometimes up to 30 per day or more. This process also allowed for the opportunity to ensure completeness of information on cases. The FDH’s lab confirmation coordinator provided information for out-of-state cases to the CDC laboratory.

TABLE 3-7

Selected History, Cyclospora Outbreaks, and Vehicles: Florida, National, and International.

A web-based data collection system was inaugurated in order for individual county health department epidemiology staff to directly enter case information into a database for quick, real-time analysis. The web-based system is a module designed for outbreak investigations that is part of Florida’s statewide electronic reportable disease management system. The outbreak database included demographic and exposure variables and was monitored in real time by the lead investigator. Florida residents who were confirmed cases and entered into the web-based outbreak module automatically were also entered into the disease reporting system.

Communication of outbreak investigation progress and current descriptive data were electronically mailed regularly to appropriate FDH administration staff and investigation team members as well as to other state agency partners. Routine and timely updates were also posted on the FDH disease alert notification system, EpiCom, which is accessible by external partners in addition to all department staff. The CDC equivalent system, EpiX, was utilized for solicitation of out-of-state cases and communication nationwide. Press releases were also employed as needed, discussing the statewide number of cases, organism ecology, and methods of reducing risk of illness to the public.

The case definition for this outbreak investigation was a probable or confirmed case of Cyclospora infection, using the surveillance case definition, with onset since March 1, 2005, in a resident of or visitor to Florida. The FDH surveillance case defines a confirmed case as a clinically compatible case that is laboratory confirmed; the FDH defines a probable case as a clinically compatible case that is epidemiologically linked to a confirmed case.

Investigation Summary

Dates of exposure in the clustered cases ranged from March 19 to May 15, 2005. Dates of onset in the clustered cases ranged from March 24 to June 24, 2005 (see Table 3-8). Dates of onset of both sporadic and clustered cases ranged from March 1 to July 10, 2005 (see Figure 3-6). Predominant symptoms included diarrhea (78.5 percent), fatigue (64 percent), and abdominal pain (61.8 percent) (see Table 3-8). Over 75 percent of the cases were older than 40 years of age, 81 percent of the cases were Caucasian, 79 percent non-Hispanic, and 57 percent were female. Each case was asked a series of risk factor questions including a long list of various raw fruits and vegetables, other foods, and travel histories. The widespread nature of the cases and the lack of any readily apparent common food item was a strong indicator of a widely distributed food. The only weakly significant preliminary risk factors were iceberg lettuce (OR = 2.94, 95% CI 1.17–7.42; P < .02) and limes (OR = 8.54, 95% CI 1.13–64.79; P < .02). Initially all the cases appeared to be sporadic, but eventually some clusters emerged (see Table 3-8). Investigation of three of these clusters, from Pinellas, Flagler, and Sarasota Counties were used to determine the implicated food product. The Palm Beach County and the Orange County cluster investigations had inconclusive results.

TABLE 3-8

Cyclospora Clusters, Florida 2005: Range of Dates of Exposure, Dates of Onset, Confirmed and Probable Cases.

FIGURE 3-6

Epidemiology curve by week of onset, Florida 2005 Cyclospora outbreak. NOTE: The epidemiology curve is by week of onset, thus the first case of March 1 occurred during the week of February 27–March 5, 2005. SOURCE: Florida Department of Health (more...)

The first cluster to emerge was the Pinellas County cluster, associated with consuming food at chain restaurant A. In this cluster, there was a total of 42 cases (17 laboratory confirmed, 25 probable). The range of exposures was from April 1 to 2, 2005. The range of dates of onset was March 25–April 23, 2005. The implicated menu item was herb-flavored oil used for bread dipping with the following ingredients: olive oil, fresh basil, Italian parsley, rosemary, and fresh garlic (OR = 52, 95% CI 8.99–300.78; P < .0000001411 Fischers exact). During the investigation of the cluster associated with chain restaurant A, another small cluster became apparent at a different chain restaurant owned by the same company. A total of eight cases (four confirmed, four probable) was linked to this second cluster. The implicated item in the second cluster was bread dipping oil mixed with pesto. Both restaurants from different chains receive Italian parsley and fresh basil from the same distributor.

The second cluster, in Flagler County, was associated with consuming food at an independent restaurant. This cluster had a total of 20 cases (16 confirmed, 4 probable) with exposures ranging from April 1 to 12, 2005. The Flagler County cluster investigation also implicated a flavored bread dipping oil with the following ingredients: olive oil, fresh basil, fresh garlic, and parmesan cheese (OR = 27, 95% CI 2.29–534.3; P = .002).

The Sarasota County cluster is really five separate doctor’s offices whose staffs were provided catered lunches from the same independent restaurant by drug company representatives. There was an additional sporadic group associated with eating at the same independent restaurant. Exposures ranged from March 19 to April 17, 2005, and dates of onset from March 24 to April 21, 2005. While no single, statistically significant food item was identified, an ingredient can be implicated through the food histories. All five medical groups were served a lunch of meat wraps, vegetable wraps, and Greek salad, all with sun-dried tomato vinaigrette. The sporadic cases ate at the restaurant where Greek salad, Moroccan salad, and cucumber salad were on the menu. The Greek salad, meat wrap, and veggie wrap all contained sun-dried tomato vinaigrette with the following ingredients: olive oil, balsamic vinegar, sun-dried tomatoes, fresh onions, salt and pepper, and fresh basil. The inability to generate a statistically significant food is attributed to the lack of controls available for the case-control study, the suspected food ingredient being in multiple menu items and lack of recall for food histories.

A short questionnaire was administered to 35 confirmed Cyclospora cases picked at random from the sporadic outbreak cases in various areas of the state to assess fresh basil consumption habits. Five cases were selected from each of seven areas. Questions were asked pertaining to exposure to herbed green salads, basil, herbs, bruschetta, pesto, and pasta salads. There were also three questions related to visiting Italian, Thai, and gourmet restaurants that commonly serve dishes with fresh basil or fresh basil garnish. The frequencies of response to the questions included two questions that had more than 50 percent of respondents answering affirmatively. These were eating at Italian restaurants (64.7 percent) and bread dipped in oil with fresh herbs (68.8 percent). An analysis of these two variables showed significance in going to an Italian restaurant and having bread dipped in olive oil with fresh herbs. The Fisher exact value was P = .03. Eighty-one percent of the 31 cases who responded to both questions had visited an Italian restaurant where the practice of dipping bread was customary and where they ate bread in this manner. The FDH, in consultation with epidemiologists at CDC and FDA, requested a formal traceback of the fresh basil based on the significance of the findings of the three disease cluster investigations and the random case-control study.

Results

This disease outbreak was caused by Cyclospora cayetanensis, a single-celled protozoan with symptoms of watery diarrhea, nausea, loss of appetite, abdominal pain, fatigue, and weight loss. The case fatality rate is very low. The incubation period is one to seven days, usually about one week, and the ensuing illness can last anywhere from one to three weeks. Typical vehicles include raspberries, basil, lettuce, snow peas, and water. Though water has been implicated, 90 percent of outbreaks of cyclosporiasis are foodborne. Cyclosporiasis is endemic in many developing countries and is often associated with diarrhea in travelers to Asia, the Caribbean, Mexico, and Peru (Heyman, 2004).

The implicated food item in this outbreak was fresh basil imported from Peru, a widely distributed food ingredient used raw in many salads, sauces, and garnishes (Food Track Inc., 2005). It has been called a “stealth” ingredient by many because unless one knows the ingredients of a particular menu item, one might not remember having eaten it. Anecdotal evidence from a visit to the implicated farm in Peru indicates that farm conditions could have been conducive to opportunities for contamination of the basil. There was a total of 592 cases with 365 confirmed and 227 probable. The investigated illness clusters accounted for 71 confirmed and 210 probable cases (see Table 3-8). A total of 493 cases were residents of Florida with 10 cases in Canadian residents and 89 residents of other states, all having visited Florida during their exposure period. Refer to Figure 3-7 and Figure 3-8 for details on geographical distribution of cases nationwide and in Florida. All out-of-state cases were visitors to Florida who were exposed in Florida during their incubation period.

FIGURE 3-7

Laboratory-confirmed and probable cases of cyclosporiasis by state of residence, March–June 2005, Cyclospora outbreak, Florida. SOURCE: FDH (2005b).

FIGURE 3-8

Laboratory-confirmed and probable cases of cyclosporiasis by Florida county of residence, March–June 2005, Cyclospora outbreak, Florida. SOURCE: FDH (2005b).

Conclusion

Due to the nature of this widely distributed stealth ingredient used raw in many common foods, this outbreak was large and diffuse and the investigation thereof was exceedingly complex, involving the entire Regional Environmental Epidemiology Strike Team, the FDH Bureau of Laboratories, staff from the Bureau of Epidemiology and all county health departments who reported cases. The FDH also collaborated with multiple partners in this outbreak investigation including private laboratories who reported cases, the Florida Department of Business and Professional Regulation, the Florida Department of Agriculture and Consumer Services, the FDA, and the CDC’s Division of Parasitic Diseases.

It can be expected that similar or more spectacular disease outbreaks will be seen in the future due to increased global distribution of foods (particularly stealth ingredients such as basil), shifts in consumption towards raw consumption of these ingredients, unusual ingredients and recipes, and the increased expectation for availability for out-of-season produce from other countries. The importation of foods from underdeveloped countries possibly with insufficient potable water supplies and processing sanitation standards is also a significant factor in these types of outbreaks. The potential for large outbreaks of this kind is great in Florida, given the large population (18 million) and the estimated annual number of visitors (74.5 million). The FDH continues to conduct surveillance for Cyclospora cases along with other emerging and reportable pathogens in order to discover outbreaks early in their occurrence so that their cause can be discovered and further spread of illness can be prevented. FDA continues its ongoing efforts in working with produce-exporting countries to ensure that produce exported to this country is safe and free from disease (FDA, 1998, 2001, 2003c, 2004; DOT and HHS, 1999).

Recommendations

Accurate laboratory analysis is critical in determining the etiological agent and scope of any foodborne outbreak. It should be noted that in this particular investigation only one clinical sample was initially misidentified as a cryptosporidium. It is imperative that public and private laboratories have the capability to accurately detect and quantify emerging pathogens and threats to our food supply as quickly as possible. These analytical and technical capabilities must include all biological, chemical, natural, and intentional threats. Public health laboratory systems need to facilitate and lead in this endeavor. Public funding allocations must reflect these high priorities for detection of food safety threats. It is also important for owners/managers and personnel at private laboratory concerns to be educated on their important role with disease surveillance and outbreak investigations and “buy in” to the investigation and critical communication processes and keep all staff apprised of this responsibility.

While the web-based system used for the data collection for cases for this outbreak investigation was somewhat helpful in rapidly collecting data from a wide geographical area, it was determined to have limited capabilities to collect control data, perform multiple variate analysis easily, and be conducive to easy manipulation of data for analysis. The limited flexibility of the design resulted in duplication of investigation and analytical efforts during some phases of the investigation. When designing and testing such elaborate systems, information technologists and software designers should include epidemiologists and other technical experts on the design and use. Scientists should also welcome and understand that programmers and data system designers should also be a part of the planning process of responding to natural and man-made biological events. Resources also need to be devoted and planned to include training of end users of these types of systems.

It should be noted that Florida was able to successfully investigate this extremely large and complex outbreak using epidemiology, nursing, laboratory, and environmental health personal within the existing organizational structures. Many resources at the county and state level in multiple scientific disciplines have been developed in the past 10–12 years that permitted this to happen. State and local municipalities who have the responsibility to conduct the surveillance, investigation, and reporting of diseases, both natural and intentional, that negatively impact public health must obtain funding to secure the human resources and develop the expertise to respond to large-scale threats to human health.

Acknowledgments

The following people provided their extensive skills and expertise with the successful investigation and reporting of this extremely large foodborne disease outbreak: Kathleen Ward, R.S., M.S.E.H., Bureau of Community Environmental Health; Mike Friedman, M.P.H., Bureau of Community Environmental Health; Robin Terzagian, Bureau of Community Environmental Health; Janet Wamnes, M.S., Bureau of Community Environmental Health; Juan Suarez, Bureau of Community Environmental Health; Carina Blackmore, Ph.D., D.V.M., Bureau of Community Environmental Health; Richard Hopkins, M.D., M.P.H., Bureau of Epidemiology; Joann Schulte, D.O., M.P.H., Bureau of Epidemiology; David Beall, Ph.D., Bureau of Laboratories; Doc Kokol, Public Information Officer; Lindsay Hodges, Public Information Officer; Maria Donnelly, M.S.P.H., Pinellas County Health Department; Sue Heller, R.N., B.S.N., Pinellas County Health Department; Hunter Zager, Pinellas County Health Department; Joe Zwissler, Pinellas County Health Department; Rick Barrett, Pinellas County Health Department; Kelly Granger, M.P.H., Hillsborough County Health Department; Aimee Pragle, M.S., Nassau County Health Department; Andre Ourso, M.P.H., Volusia County Health Department; Quintin Clark, Sarasota County Health Department; K. Eric Stutz, M.P.H., R.S., Sarasota County Health Department; Maria Teresa Bonafonte, Ph.D., Palm Beach County Health Department; Dawn Ginzl, M.P.H., Orange County Health Department; Bill Toth, M.P.H., Orange County Health Department; Barbara Herwaldt, M.D., M.P.H., CDC, Division of Parasitic Diseases.

Features of Hepatitis A Virus Infection

Several key features of HAV infection provide the context for the outbreaks and are relevant to any consideration of prevention strategies (Fiore, 2004). After an incubation period that averages 28 days but can range from 15–50 days, HAV infection can present in a number of different ways, ranging from asymptomatic infection; to nonspecific symptoms such as nausea, abdominal pain, and fatigue; to jaundice and other classical symptoms of acute hepatitis. The likelihood of symptomatic infection and of jaundice is directly related to age; young children with HAV infection are unlikely to have a clinical illness recognizable as acute hepatitis. The period of communicability extends from about two weeks before until about one week after jaundice occurs, and HAV can be excreted in very high concentration in the stool of infected people (e.g., about 1 billion particles per gram). HAV in organic material is stable in the environment at least for a period of weeks.

Molecular Surveillance

In the investigations described here, we used molecular subtyping to characterize outbreak-related HAV strains as they became available and to explore their relationship to each other and to strains identified in the context of ongoing molecular surveillance projects in the United States and Mexico. We use nested reverse transcription polymerase chain reaction (RT-PCR) to amplify a 315 nucleotide segment at the VP1–2a junction of strains from persons with hepatitis A onset between January 2002 and August 2003, collected through the six counties comprising the Sentinel Counties Study of Acute Viral Hepatitis, and through the Border Infectious Disease Surveillance (BIDS) Project, which operates along the U.S.-Mexico border (Amon et al., 2005)

At the time the outbreak investigations began, this database included over 100 distinct sequences from over 500 individuals. Approximately 95 percent of the distinct sequences, representing 99 percent of specimens, were genotype 1A (Figure 3-9) (Amon et al., 2005). The majority of these distinct sequences formed a single cluster, in which all sequences were >96 percent similar to each other (cluster X). This cluster included sequences from all individuals identified through the BIDS project, as well as from travelers to Mexico. Particular risk factors predominated in other clusters, such as being a homosexual man or using illicit drugs (Figure 3-9).

FIGURE 3-9

Comparison of hepatitis A viral sequences among individuals with hepatitis A from northern Mexico (Border Infectious Disease Surveillance [BIDS] Project), 2002–2003; outbreak-related surveillance, October–December 2003; and six U.S. sentinel (more...)

Outbreaks in Tennessee, North Carolina, and Georgia

The first series of outbreaks involved over 400 cases and occurred in Tennessee, North Carolina, and Georgia during August to September 2003 (Amon et al., 2005). In Tennessee and North Carolina, investigations indicated that cases were associated with one restaurant in each state, but the restaurants were unrelated to each other. Ill food service workers identified in each outbreak had illness onset concurrent with other cases, indicating that they could not have been sources of the outbreaks. Most cases in Tennessee reported an onset of illness during the first week in September. The peak in North Carolina was about a week later. In Georgia, at least three restaurant-associated clusters were identified, but information about many of the cases that shared the outbreak strain was incomplete. An epidemiologic investigation of the largest of these clusters showed that dates of onset were similar to those in North Carolina (Figure 3-10). Exposures at each restaurant occurred primarily during a 10-day period in August, and green onions were implicated as the source in case-control studies among restaurant patrons in each state. In the case-control study in Tennessee, for example, green onions were eaten by 98 percent of 57 case-patients and 46 percent of 204 control subjects (OR = 65.5, 95% CI 8.9–482.5). The epidemiologic investigations were conducted during late September and early October, and the FDA initiated traceback investigations on October 9.

FIGURE 3-10

Date of illness onset among restaurant patrons, September–November 2003; n = 590. The dotted bars represent Tennessee (TN) cases, the hatched bars North Carolina (NC) cases, the diagonal striped bars Georgia (GA) cases, and the solid bars Pennsylvania (more...)

Sera obtained from outbreak-related cases in the three states yielded two distinct outbreak strains (sequence A and B) (Figure 3-9). Sequence A was found among restaurant patrons from Tennessee, and sequence B was found among Georgia and North Carolina cases. Both sequences fell into cluster X, the same cluster of strains that also included most strains identified among persons who acquired illness in Mexico.

Pennsylvania Outbreak

In early November, as the epidemiologic investigations were winding down and the traceback investigations were underway, an outbreak in Pennsylvania occurred, eventually involving a total of more than 600 cases among patrons of a single restaurant in Beaver County, including 13 employees who became ill at the same time as patrons (Figure 3-10) (Wheeler et al., 2005). Over 80 percent of the 425 case-patients who reported eating only once at the implicated restaurant ate there between October 3 and October 6, including 67 percent who dined on October 4 or October 5. The estimated attack rate for the four-day period was 17.9 percent, including an estimated 25 percent of diners on October 4 and 29 percent of diners on October 5.

A case-control study among patrons at the restaurant included 181 cases and 89 controls. Five of the 121 menu items were associated with illness. Mild salsa was eaten by 91 percent of case-patients and 35 percent of controls (OR 19.6, 95% CI 11.0–34.9) and was the only item eaten by more than 25 percent of case-patients. Eating green onions, an ingredient in over 50 menu items, was reported by 98 percent of case-patients and 58 percent of controls and was strongly associated with illness (OR 33.3, 95% CI 12.8–86.2). The final multivariate model included age, eating mild salsa, and eating any other menu item containing green onions.

Green onions arrived at the restaurant in bundles of six to eight onions each, packed on ice in boxes. After unpacking into metal pans, they were stored in the refrigerator for up to five days. When needed, bundles were rinsed with tap water, the rubber band around the bundle was removed, and the onions were chopped using an electric dicer. After chopping, they were refrigerated in plastic containers for up to two days.

Mild salsa, the menu item most strongly associated with infection, was prepared in 40-quart batches. The restaurant prepared up to two batches each day, and stored them for up to three days in the refrigerator. Each quart contained six ounces of raw chopped green onions, equivalent to 10–16 whole green onions.

Of course a pressing question was the relationship, if any, of this outbreak to the outbreaks that had occurred earlier in the fall. Molecular surveillance results obtained to date had pointed to the likelihood of two separate instances of contamination accounting for these earlier outbreaks. The outbreak strain from cases associated with the Pennsylvania outbreak (sequence D, Figure 3-9) turned out to be distinct from but closely related to the other outbreak strains, and fell into the same cluster of Mexico-related strains. These findings established that the four geographically separate but temporally related outbreaks represented at least three distinct events.

Other States

Not all hepatitis A is foodborne, and a common question that arises in the context of many foodborne hepatitis A outbreaks is the extent to which available surveillance methods are sensitive enough such that outbreak-associated cases or small clusters can be distinguished from “background” cases. This is particularly relevant for outbreaks, such as those described here, that are associated with a distributed food item, but in which the majority of cases are associated with exposure at a restaurant. Another “first” accomplished in the context of these investigations was an improvement in the sensitivity of surveillance by incorporating molecular methods. Comparison of strains identified during the outbreak period provided evidence that some apparently “sporadic” hepatitis A cases were indeed foodborne. Specimens were requested from any cases that did not have an identified source of transmission. Of over 50 specimens submitted, a number were identical to outbreak strains (Table 3-9) (Amon et al., 2005).

TABLE 3-9

Source and Distribution of Cluster X Hepatitis A Virus Sequences, September–December 2003 (Outbreak Surveillance Specimens) and January 2002–August 2003 (Sentinel Counties and Mexico [BIDS] specimens); n = 478.

Investigation of Farms

Findings of molecular surveillance were consistent with sources in Mexico, as sequences matching each of the outbreak strains were identified from among BIDS specimens (Table 3-9). Four farms, all located in northern Mexico, potentially supplied the implicated restaurants, but no single farm could explain all four outbreaks. These traceback results were consistent with the results of sequencing—three distinct strains were identified from outbreak-associated cases in the four states.

Representatives from the FDA and CDC visited the farms in question (FDA, 2003b). The harvesting procedure included a lot of handling of the onions, which were pulled from the ground by hand, after which the outer layer was peeled off, the roots were removed, the onions were cut to a consistent size, and they were banded into bunches. Packing involved spraying bunches with chlorinated water as they passed on a conveyor belt, followed by loading into a cardboard box which was topped with chipped ice. In the distribution network, boxes generally were not handled between the farm and the restaurant destination. A number of conditions on the farm were identified as areas of concern, including poor sanitation, inadequate hand washing facilities, worker health and hygiene, the quality of the water used in the fields at packing sheds, and the ice-making process. However, no single practice or event was identified that could have explained the outbreaks.

Because HAV has no animal host, the original source of green onion contamination was a human infected with HAV and excreting the virus in stool. This fecal contamination could occur in a number of ways. Adults with contaminated hands could have touched the green onions during harvest or processing. Hepatitis A is endemic in Mexico, which means that the vast majority of the population is infected during childhood, and most adults are immune (Tanaka, 2000). Hence the majority of infections at any given time are occurring among children. Thus likely sources of contamination of hands include sewage or feces from workers’ HAV-infected children. It is also possible that HAV-infected children were present in the fields and contaminated the green onions directly. Direct contamination of the growing areas by sewage is also possible.

Discussion

The outbreaks described here were investigated rapidly and tracebacks were initiated early. However, a number of features of hepatitis A make detection and control of foodborne hepatitis A difficult, and the results of these investigations illustrate important areas of progress and remaining challenges (Fiore, 2004). Because HAV contamination of foods can be focal and the virus remains viable in the environment for months, cases can be both geographically and temporally dispersed. These investigations demonstrate the benefits of wider and faster use of molecular epidemiologic methods, both in outbreak investigations and in the context of routine hepatitis A surveillance. Viral sequencing showed that four geographically separate outbreaks that occurred in the fall of 2003 represented at least three distinct events. Sequencing activities also improved the sensitivity of surveillance to define the scope of the outbreaks, distinguish outbreak-related from nonoutbreak-related cases, and identify evidence of sporadic unrecognized foodborne transmission. Finally, viral sequencing supported the results of the first traceback investigations and accelerated control efforts related to the outbreak in Pennsylvania.

The investigations also exemplify challenges in foodborne hepatitis A outbreak response that stem from characteristics of routine hepatitis A surveillance and inherent aspects of the infection itself. Because cases reported through routine surveillance are not typically asked about foodborne exposures, the recognition of an unusual increase in the number of cases or of cases occurring among those in an unusual demographic group serves to alert authorities to begin asking about foodborne exposures as one potential common link among cases. However, even with the most rapid response and investigation of clusters of cases, the long incubation period of hepatitis A and inevitable delays in diagnosis and reporting necessitate a considerable lag time between exposure and the earliest possible detection of a foodborne outbreak. For example, the exposure that resulted in the outbreak in Pennsylvania was occurring as cases associated with the previous outbreaks were just being reported in the other states. Thus, even if a farm implicated in the earlier outbreaks had been linked to a farm implicated in the Pennsylvania outbreak, it is unlikely that even the most rapid of responses to the earlier outbreaks could have averted the subsequent outbreak in Pennsylvania.

Green onions are emerging as a potential “problem” food, having been implicated in at least two previous restaurant-associated hepatitis A outbreaks (Dentinger et al., 2001; Datta et al., 2001). The vast majority of green onions consumed in the United States are imported from Mexico, a country in which hepatitis A is endemic (Calvin et al., 2004). They require extensive handling during harvest and may be particularly difficult to clean. A pattern of focal, low-level contamination in which possibly very few bunches were contaminated, may make it difficult to detect transmission when it occurs.

The outbreak in Pennsylvania illustrates how conditions at the point of sale can amplify an outbreak. A combination of factors probably contributed to this outbreak’s size and high attack rate. The large number of diners who ate at the restaurant during the days of peak exposure were all offered mild salsa, the food item most strongly associated with illness. Preparation practices, such as rinsing green onions while they are still bundled and chopping and storage methods that allowed for cross-contamination, could also have contributed to the size of the outbreak. Because of the high concentration of HAV in stool and the likely low infectious dose, even a small amount of fecal contamination might result in many hundreds of infectious doses. Although the 2005 Food Code includes a requirement that vegetables that are not subsequently cooked be washed, it does not offer guidance about specific methods to prevent cross-contamination of produce (HHS, 2005).

Progress has been made in developing methods to detect HAV in food, including reproducible methods to detect the virus in “spiked” food samples and produce washes (Shan et al., 2005). Although theoretically attractive, there are a number of difficulties inherent in attempting to detect HAV in produce. The virus does not multiply in foods, and the concentration may be quite low. However, viral culture is not feasible, so there is a need to rely on RT-PCR techniques, which may not perform consistently in the presence of complex food mixtures. Further, RT-PCR cannot distinguish infectious HAV from noninfectious HAV RNA. Even if these technical problems were solved, HAV detection in food is unlikely to be of much practical use in the context of outbreak investigations for a number of reasons. Perhaps most important, particularly in the case of produce, is that the implicated item has almost invariably been consumed or discarded by the time illness is occurring. Further, methods are not at a level of development as of yet such that they can be scaled up to volumes needed to be reasonably sure that contamination is not present, particularly given the low infectious dose. Finally, these currently available methods take days to complete.

Perhaps most important is prevention of HAV (and other enteric pathogens) contamination of produce in the first place by preventing fecal contamination of produce on the farm. Hepatitis A is endemic in Mexico, and while the precise mechanism of transmission in the outbreaks described here could not be determined, control measures can be implemented that could prevent such outbreaks. These include ensuring that field workers are healthy and have access to adequate sanitary facilities and ensuring that water used to irrigate and rinse produce is not contaminated with feces. Children are the source of most transmission of HAV in rural communities in Mexico and much of the developing world, and children should not be present in areas where food is harvested. Reduction in HAV transmission among children in areas where produce is grown would further reduce opportunities for contamination.