Background. Listeriosis, a life-threatening foodborne illness caused by Listeria monocytogenes, affects ∼ 2500 Americans annually. Between July and October 2002, an uncommon strain of L. monocytogenes caused an outbreak of listeriosis in 9 states.
Methods. We conducted case finding, a case-control study, and traceback and microbiological investigations to determine the extent and source of the outbreak and to propose control measures. Case patients were infected with the outbreak strain of L. monocytogenes between July and November 2002 in 9 states, and control patients were infected with different L. monocytogenes strains. Outcome measures included food exposure associated with outbreak strain infection and source of the implicated food.
Results. Fifty-four case patients were identified; 8 died, and 3 pregnant women had fetal deaths. The case-control study included 38 case patients and 53 control patients. Case patients consumed turkey deli meat much more frequently than did control patients (P =.008, by Wilcoxon rank-sum test). In the 4 weeks before illness, 55% of case patients had eaten deli turkey breast more than 1–2 times, compared with 28% of control patients (odds ratio, 4.5; 95% confidence interval, 1.3–17.1). Investigation of turkey deli meat eaten by case patients led to several turkey processing plants. The outbreak strain was found in the environment of 1 processing plant and in turkey products from a second. Together, the processing plants recalled >30 million pounds of products. Following the outbreak, the US Department of Agriculture' Food Safety and Inspection Service issued new regulations outlining a L. monocytogenes control and testing program for ready-to-eat meat and poultry processing plants.
Conclusions. Turkey deli meat was the source of a large multistate outbreak of listeriosis. Investigation of this outbreak helped guide policy changes designed to prevent future L. monocytogenes contamination of ready-to-eat meat and poultry products.
Listeriosis, a life-threatening, primarily foodborne illness caused by Listeria monocytogenes, affects an estimated 2500 people in the United States annually . Immunocompromised persons, elderly persons, and pregnant women and their fetuses or newborns are at highest risk for infection . Listeriosis most commonly presents as an invasive illness with bacteremia or meningitis and has a case-fatality rate of 20% . In pregnancy, listeriosis may lead to fetal death, preterm labor, or invasive listeriosis in the newborn child . We describe the investigation of a large, multistate listeriosis outbreak in 2002 that was linked to turkey deli meat and prompted new changes in US food regulatory policy.
Case finding. During July and August 2002, there were 22 cases of listeriosis reported in Pennsylvania, a nearly 3-fold increase over baseline. Clinical isolates were sent to the Centers for Disease Control and Prevention (CDC; Atlanta, GA) for molecular subtyping by PFGE, which identified a cluster of cases caused by a single L. monocytogenes strain. The CDC asked health departments in the northeast United States to conduct active case finding, prompt reporting of listeriosis cases, and retrieval of clinical isolates for rapid PFGE testing via PulseNet. PulseNet, the National Molecular Subtyping Network for Foodborne Disease Surveillance, is the CDC-coordinated network of public health laboratories that perform PFGE on foodborne bacteria . The network permits rapid electronic comparison of PFGE patterns through a database at the CDC. PulseNet staff compared the electronically posted PFGE patterns from all public health laboratories to the PFGE pattern from the Pennsylvania cluster.
Definitions. Listeriosis was defined as illness in a person from whom L. monocytogenes was isolated (from any clinical specimen). The 2-enzyme PFGE pattern from the Pennsylvania listeriosis cluster was designated as the outbreak pattern. Infection with the outbreak strain was defined as having listeriosis with an L. monocytogenes isolate yielding a PFGE pattern indistinguishable from the outbreak pattern. For the purposes of case counting, a mother-neonate pair was considered to be 2 cases if both mother and newborn were ill and had culture-confirmed L. monocytogenes infection; if only one of the pair was ill with culture-confirmed L. monocytogenes infection, only 1 case was counted. However, for the purposes of the case-control study, a mother-neonate pair was always considered to be a single case. For an ill pregnant woman with an associated fetal death (stillbirth or miscarriage), only the mother' illness was counted as a case.
Case-control study. We conducted a case-control study to identify the contaminated food source. A case patient was defined as a person with culture-confirmed listeriosis between 1 July and 30 November 2002, whose infection was caused by the outbreak strain. A control patient was defined as a person with culture-confirmed listeriosis between 1 July and 30 November 2002, whose infection was caused by any other nonoutbreak strain of L. monocytogenes, and who was from a state with at least 1 case patient. Case patients and control patients were interviewed with a standard questionnaire that addressed medical and food histories during the 4 weeks preceding culture for L. monocytogenes (all patients except mother-neonate pairs) or delivery date (mother-neonate pairs). The questionnaire included >70 specific food items.
Statistical methods. Statistical analyses were performed with SAS software, version 8.2 (SAS). Food exposures of case patients and control patients were compared using the Wilcoxon rank-sum test. ORs and exact 95% CIs were calculated, controlling for state or city by conditional logistic regression.
Traceback investigation. Health and regulatory officials traced back the food implicated in the case-control study. A purchase site was defined as any delicatessen, grocery store, or restaurant where the implicated food consumed by a case patient was purchased. Inventories and invoice information were obtained from purchase sites to determine which processing plants had supplied the food.
Plant investigations. Plants A, B, C, and D were investigated by teams led by the US Department of Agriculture' Food Safety and Inspection Service (FSIS). Noncompliance reports, production and construction records, and records of product and environmental sampling for Listeria species maintained by plant staff were reviewed. Environmental samples were collected aseptically by cellulose sponge. Unopened food products were collected from each plant' “shelf-life” stock and from current production. Recalled products from plant A were also collected.
Laboratory investigation. PFGE testing was performed in accordance with the standard PulseNet protocol using restriction endonucleases AscI and ApaI . Study isolates were confirmed at the CDC to be L. monocytogenes by AccuProbe (GenProbe), ribotyped using the RiboPrinter microbial characterization system (Qualicon), and serotyped using standard methods .
Food samples were transported at 4°C and cultured for L. monocytogenes using standard methods . Environmental sponges were transported at 4°C in Dey-Engley neutralizing media and tested using the standard methods mentioned above. L. monocytogenes isolates were subtyped by PFGE in the same manner as were human isolates. PFGE patterns were electronically compared with human isolate patterns through PulseNet.
Case finding and description. Between 1 July and 30 November 2002, there were 188 patients with L. monocytogenes infection identified in 9 states: Pennsylvania, New York, New Jersey, Delaware, Maryland, Connecticut, Massachusetts, Michigan, and Illinois. of these patients, 54 were infected with the outbreak strain, 1 was infected with a strain that resembled the outbreak strain but differed by 1 band on the ApaI PFGE analysis, and 122 were infected with unrelated L. monocytogenes strains; no isolates were available for 11 patients. The dates of specimen collection for the 54 case patients ranged from 18 July to 26 October ( figure 1). of the ∼ 2000 human L. monocytogenes isolates in the PulseNet database from the 6 years before the outbreak, only 20 were indistinguishable from the outbreak PFGE pattern (PulseNet pattern designation GX6A16.0235-GX6A12.0003). All outbreak strain isolates tested were serotype 4b and ribotype DUP-1044.
Characteristics of the 54 case patients and 122 control patients are shown in table 1. Eight case patients (15%) were pregnant women, and 4 (7%) were neonates. Prior medical information was available for 41 of the 42 non—maternal-neonatal case patients. Thirty patients (71%) had the following medical conditions or factors, which are considered to be immunocompromising or predisposing for listeriosis: hematologic malignancy (5 patients), corticosteroid use (4 patients), HIV infection and/or AIDS (4 patients), solid malignancy (3 patients), diabetes (3 patients), liver disease (3 patients), inflammatory bowel disease (3 patients), organ transplantation (2 patients), end-stage renal disease (1 patient), rickets and malnutrition (1 patient), and sickle-cell anemia (1 patient). Most elderly case patients also had an underlying immunocompromising condition; only 4 (7%) of the 54 case patients were aged ⩾ 65 years and were not immunocompromised. Seven patients (13%) were aged 1–64 years and were not pregnant or immunocompromised. Case patients were more likely than control patients to be pregnant or young and healthy (i.e., aged 1–64 years without predisposing medical conditions [P <.001]).
Eight (15%) of 54 case patients died. of these, 4 patients were aged 1–64 years and were immunocompromised, 3 were aged ⩾ 65 years and were also immunocompromised, and 1 was a neonate. The case-fatality rate among case patients was lower than that among control patients (15% vs. 26%), although not significantly (P =.14). However, among immunocompromised and neonatal patients, the case-fatality rates were similar: 24% for case patients versus 29% for control patients. of the 8 pregnant case patients, 6 (75%) had complications: 3 had fetal deaths, 2 gave birth to infants with listeriosis (1 birth was premature), and 1 woman gave birth prematurely to an infant without listeriosis. Complications among the 4 pregnant control patients were similar, with 2 fetal deaths and 1 birth of an infant with listeriosis.
Case-control study. Questionnaire responses were obtained from 91 (91%) of 100 eligible case patients and control patients. We excluded patients identified after the announcement of case-control study findings on 4 October 2002 (n = 73). Three mother-neonate pairs were counted as single patients. The final epidemiologic analysis included data obtained from 38 case patients and 53 control patients (table 2). Infection with the outbreak strain was strongly associated with consumption of precooked turkey breast products sliced at the deli counter of groceries and restaurants (P =.008, by Wilcoxon rank-sum test). The Wilcoxon rank-sum test captures differences in the frequency of consumption. Overall, 55% of case patients ate turkey deli meat more than 1–2 times in the 4-week period, compared with 29% of control patients (OR, 4.5; 95% CI, 1.3–17.1). Other turkey deli meats, such as turkey ham and turkey pastrami, were eaten by fewer patients and primarily by persons who also ate turkey breast products. Therefore, it was difficult to exclude risk from these deli meats as well. No other single food item was significantly associated with outbreak strain infection, except for lettuce, which was protective.
Traceback. Interviews of the first 29 case patients who had eaten turkey deli meat identified 80 purchase sites (mean, 3 sites per person [range, 1–8 sites]). None of the patients recalled the brand names of all turkey deli meat purchased. Sixty-eight of the 80 purchase sites could be located; 57 (84%) of these were investigated, with identification of products from >50 turkey processing plants. FSIS evaluated the 15 most frequently identified plants, using a standardized survey to address such factors as previous food safety infractions, history of recent construction, and records of in-plant Listeria species sampling. On the basis of these survey results, plants A, B, C, and D were found to warrant in-plant investigation.
Processing plant and marketplace investigations. Multiple L. monocytogenes strains were identified in the environment of plant A ( table 3). Two samples, both collected from floor drains, yielded the outbreak strain. Another isolate yielded a PFGE pattern differing from the outbreak pattern by 1 band, which was indistinguishable from that of 1 patient' isolate during the outbreak. Records of the plant' in-house environmental sampling in the room where cooked turkey was handled revealed an increase in the number of samples that yielded Listeria species during July and August 2002, compared with previous months and with the same months in 2001. This increase in positive test results coincided with a large construction project performed in the same room. Testing of previously unopened turkey products from plant A did not yield the outbreak strain ( table 3). However, it did yield 2 L. monocytogenes strains that were also found in the plant' environment and that were isolated from 2 control patients. The outbreak strain of L. monocytogenes was not recovered from any environmental samples from plant B. However, the outbreak strain was isolated from 2 of 18 previously unopened turkey samples produced by plant B. Plant B is located ∼ 30 miles from plant A.
Sampling of previously opened packages of turkey yielded the outbreak strain on 4 occasions (table 4). At the time of interview, only 1 case patient still had a sample of the turkey eaten before illness. A sandwich made with the same turkey deli meat consumed on 2 previous days was left in a cooler when an otherwise healthy young man was admitted to a hospital with L. monocytogenes bacteremia. The sandwich was retrieved from the cooler 2 weeks later, and L. monocytogenes with the outbreak PFGE pattern was isolated from the turkey. Invoice records revealed that only turkey products from plant A and plant B were sold at the market where the patient purchased the turkey.
Regulatory outcomes. In response to these findings, plants A and B voluntarily suspended operations and began intensive cleanup, and together they recalled >13.5 million kg (>30 million lbs) of ready-to-eat poultry products. The first recall occurred on 9 October 2002. In December 2002, immediately following the outbreak, FSIS issued a policy directive outlining an intensified microbiological testing program for ready-to-eat meat and poultry plants . Over the following year, FSIS completed a risk assessment for L. monocytogenes in deli meats  and finalized and issued a new regulation aimed at further reducing L. monocytogenes contamination of ready-to-eat meat and poultry products .
We investigated a multistate outbreak of listeriosis that affected 54 persons and caused 8 deaths and an additional 3 fetal deaths. An epidemiologic investigation linked outbreak strain infection with consumption of precooked, ready-to-eat turkey deli meat, making this the third multistate listeriosis outbreak related to ready-to-eat meat and poultry products in recent years [11, 11]. The outbreak strain of L. monocytogenes was found in the environment of plant A and in turkey breast products from plant B. Both plants suspended production and together recalled >13.5 million kg (>30 million lbs) of products, resulting in one of the largest meat recalls in US history. In the year following the outbreak, federal regulators issued new, more-stringent rules designed to prevent further L. monocytogenes contamination of ready-to-eat meat and poultry products [8, 10].
L. monocytogenes has 2 unique characteristics that influence its transmission to humans through ready-to-eat foods. First, it is a tenacious colonizer that favors moist, cool environments, such as food processing plants; eradication is difficult . Second, although it is easily killed by cooking, L. monocytogenes multiplies readily at refrigeration temperatures, whereas most other competing microflora do not . A small amount of L. monocytogenes contamination at a processing plant that occurs after cooking but before packaging may lead to a large infectious dose being delivered to a susceptible consumer, because of multiplication of bacteria during storage . Ready-to-eat poultry products, such as turkey deli meat, provide a particularly favorable environment for growth of L. monocytogenes during refrigeration . Recent risk assessment models have estimated that, among ready-to-eat foods, deli meats and nonreheated hot dogs have the highest risk of listeriosis per serving .
Outbreak investigations can highlight high-risk food items and target regulatory, industry, and public health action. Our investigation led us to 2 plants. Plant B was likely linked to illnesses, because the outbreak strain was found in its turkey products. It is also possible that plant A was linked to illnesses, because the outbreak strain and several other L. monocytogenes strains were found throughout the main room where unpackaged turkey products were handled. In addition, 2 nonoutbreak strains were found in both the plant environment and in packaged turkey products, demonstrating the likelihood of ongoing contamination of products within the plant. The geographic proximity of plants A and B could allow introduction of a shared strain by receiving raw turkey products from the same slaughterhouse, purchasing equipment from the same source, or having employees in common . A better understanding of the ecology of L. monocytogenes strains in ready-to-eat food plants would help guide efforts to prevent L. monocytogenes contamination of high-risk products.
Findings from this investigation helped guide FSIS in developing its new regulatory policy [15, 10]. Under these regulations, plants producing high-risk, ready-to-eat meat and poultry products must develop scientifically validated L. monocytogenes—control programs, which are stratified according to the number of control measures taken. One such measure is the institution of pathogen-elimination treatments after the product has been packaged, such as postpackaging heat treatments or “pasteurization” [18, 19]. Irradiation has been shown to be effective as a postpackaging treatment in studies, yet it has not been approved by the US Food and Drug Administration for use with ready-to-eat meat and poultry . Additives that suppress the growth of L. monocytogenes in products are another strategy . Plants with less rigorous control programs, particularly those relying solely on sanitation, are placed under an intensified microbiological testing program by FSIS. The new regulations also clarify, for the first time, that recalls can be based on identification of L. monocytogenes on equipment, not just in products, and that plant-generated microbiological testing data must be shared with FSIS officials.
A 2003–2004 FSIS survey revealed that the percentage of plants using new technologies to kill or suppress the growth of L. monocytogenes inside packaging has increased dramatically since release of the first FSIS policy changes, which were made in December 2002, as has the number of plants testing the processing environment for L. monocytogenes . More than 87% of the >2900 establishments surveyed had made at least 1 change in their food safety process . One year after the outbreak, FSIS reported a 25% decrease in the number of L. monocytogenes—positive samples detected by means of its regulatory testing program between January and September 2003, compared with 2002 . Most importantly, preliminary 2004 national surveillance data revealed a 40% decrease in the incidence of human listeriosis, compared with 1996–1998, with 2.7 listeriosis cases per million persons . This rate approaches the 2005 national objective of cutting listeriosis incidence to 2.5 cases per million persons .
In addition to regulatory and industry changes, another important aspect of listeriosis control is educating high-risk populations and their food preparers about high-risk foods, such as deli meats and hot dogs, soft cheeses made from unpasteurized milk, and smoked seafood . Current recommendations to prevent foodborne listeriosis can be found at the CDC' Web site . Prevention messages are straightforward for pregnant women and immunocompromised patients, who have a markedly increased risk of listeriosis . Older age has long been considered a risk factor for listeriosis ; however, most patients with listeriosis aged ⩾ 65 years that we identified (93%) had other immunocompromising conditions, a finding in other outbreak investigations [29, 30]. Interestingly, patients infected with the outbreak strain were significantly more likely to be either young and healthy or pregnant than were patients with sporadic listeriosis. High doses of L. monocytogenes have been linked to illness in healthier people, yet primarily in the context of febrile gastroenteritis . The role of organism pathogenicity in determining populations affected by invasive listeriosis is poorly understood [32, 33].
Although the precise risk to older persons may be difficult to define, institutions providing food to any at-risk population, including elderly persons, should implement appropriate policies regarding deli meats and other high-risk foods to reduce the risk of listeriosis. Several hospitalized or institutionalized patients identified during this outbreak probably acquired foodborne L. monocytogenes infection nosocomially, which is a scenario that has been documented previously [34, 34]. These findings indicate an urgent need for changes in these health care settings.
Finally, this investigation demonstrates the importance of PFGE subtyping through PulseNet in the detection and investigation of listeriosis outbreaks, as was recently discussed by Olsen et al. . In 2003, the Council of State and Territorial Epidemiologists issued a position statement calling for prompt interviewing of all patients with listeriosis with a nationally standardized form and expedited referral of L. monocytogenes isolates from clinical laboratories to state public health laboratories for real-time PFGE subtyping . We encourage clinicians and laboratories to promptly report all cases of listeriosis to health departments, and we encourage states to adopt the Council of State and Territorial Epidemiologists' proposal.
In conclusion, our investigation linked a large outbreak of listeriosis to turkey deli meat and reinforced findings of previous outbreaks [11, 12] and risk models  that ready-to-eat meat and poultry products continue to be a concern for at-risk populations. This investigation helped stimulate and guide new regulatory policies designed to prevent contamination at processing plants that produce these products. Industry surveys and national surveillance data already suggest a positive impact. Additional prevention measures implemented by institutions that serve high-risk populations could complement regulatory and industry efforts to reduce L. monocytogenes infections nationwide.
Balasubramanian Swaminathan, CDC (Atlanta); Caroline Johnson, Philadelphia Department of Public Health (Philadelphia); André Weltman, Pennsylvania Department of Health (Harrisburg); Bridget Anderson, Al Bugenhagen, Curtis Vincent, Ray Woron, and Shelley Zansky, New York State Department of Health (Albany); Sharon Balter and Faina Stavinsky, New York City Department of Health and Mental Hygiene (New York); Michelle Malavet and Corey Robertson, New Jersey Department of health and Senior Services (Trenton); Jessica Totaro, Maryland Department of Health and Mental Hygiene (Baltimore); Sally Bidol and John Tilden, Michigan Department of Community Health (Lansing); Ruthanne Marcus, Connecticut Emerging Infections Program (New Haven); Quyen Phan, Connecticut Department of Public Health and Social Services (Hartford); Marci Drees and Marjorie Postell, Delaware Health and Social Services (Dover); Patrick Piercy, Illinois Department of Public Health (Springfield); James Borda and Raza Haq, US Department of Agriculture—Food Safety and Inspected Service (Philadelphia); and David Goldman, Gerri M. Ransom, and Richard Roloff, US Department of Agriculture—Food Safety and Inspected Service (Washington, DC).
We thank Lauren Doctoroff, Taha Kass-Hout, Michael F. Lynch, Paul Mead, John Painter, and Andi L. Shane (CDC); Arthur Bagelman, Wendy Bamberg, Michael Eberhart, Robert Levenson, and George Zameska (Philadelphia, PA); Wayne Chmielecki and Maria Moll (Pennsylvania); Barbara Gerzonich, Christina Hidalgo, Dina Hoeffer, Stephanie Noviello, Glenda Smith, Nancy Spina, and Christine Waters (New York); Carrie Bogert, Eddy Bresnitz, Vaidehi Dave, Layla Fakhrzadeh, Barbara Gibson, Lisa McHugh, Maria Orsini-Lidicker, Marc Paladini, Nancy Sullivan, and Bijal Tamakuwala (New Jersey); John Fontana and Ellen Gould (Massachusetts); Naomi Barker and Jason Spangler (Maryland); James Averill (Michigan); Frank Green, Stacey Kinney, and Tracey Weeks (Connecticut); Gregory Huhn (Illinois); and Jan Behney, Walter Hill, Diane Loulis, Michael Ronczka, Zygmunt Sala, Susan Scarcia, and Armia Tawadrous and field, headquarters, and laboratory personnel (FSIS).
Potential conflicts of interest. All authors: no conflicts.