National Outbreak of Salmonella Serotype Saintpaul Infections: Importance of Texas Restaurant Investigations in Implicating Jalapeño Peppers

Background In May 2008, PulseNet detected a multistate outbreak of Salmonella enterica serotype Saintpaul infections. Initial investigations identified an epidemiologic association between illness and consumption of raw tomatoes, yet cases continued. In mid-June, we investigated two clusters of outbreak strain infections in Texas among patrons of Restaurant A and two establishments of Restaurant Chain B to determine the outbreak's source. Methodology/Principal Findings We conducted independent case-control studies of Restaurant A and B patrons. Patients were matched to well controls by meal date. We conducted restaurant environmental investigations and traced the origin of implicated products. Forty-seven case-patients and 40 controls were enrolled in the Restaurant A study. Thirty case-patients and 31 controls were enrolled in the Restaurant Chain B study. In both studies, illness was independently associated with only one menu item, fresh salsa (Restaurant A: matched odds ratio [mOR], 37; 95% confidence interval [CI], 7.2–386; Restaurant B: mOR, 13; 95% CI 1.3–infinity). The only ingredient in common between the two salsas was raw jalapeño peppers. Cultures of jalapeño peppers collected from an importer that supplied Restaurant Chain B and serrano peppers and irrigation water from a Mexican farm that supplied that importer with jalapeño and serrano peppers grew the outbreak strain. Conclusions/Significance Jalapeño peppers, contaminated before arrival at the restaurants and served in uncooked fresh salsas, were the source of these infections. Our investigations, critical in understanding the broader multistate outbreak, exemplify an effective approach to investigating large foodborne outbreaks. Additional measures are needed to reduce produce contamination.


Introduction
Long-distance transportation keeps stores and restaurants stocked with a wide assortment of fresh fruits and vegetables year round. However, with this abundance has come increased risk of large produce-associated foodborne outbreaks [1][2][3]. Among produce-associated outbreaks with known etiologic agents, the most common pathogen is non-typhoidal Salmonella, a diverse group of bacteria that live in intestinal tracts of many animals and survive in environments from the farm to the table [3,4]. The overall incidence of Salmonella infection from all sources in 2009 was more than double the national target for 2010; little progress towards this goal has been observed since 1996 [5].
In May 2008, a multistate outbreak of Salmonella enterica serotype Saintpaul infections with indistinguishable pulsed-field gel electrophoresis (PFGE) patterns was detected which ultimately became the largest known foodborne outbreak in the United States in over 10 years. Initial investigations identified an association between illness and the consumption of raw tomatoes [6]. On June 7, the US Food and Drug Administration (FDA) issued a national alert to avoid ''raw red plum, red Roma, and red round tomatoes'' [7].
As cases continued to occur, a specific variety of tomato could not be epidemiologically implicated and the sources of suspected tomatoes were not converging to any specific growing region or to a common distributor. Furthermore, a possible association between infections and eating Mexican-style foods began to emerge. Because tomatoes are a staple ingredient in many cuisines, not just Mexican, this finding suggested there might be something else, besides tomatoes, more unique to Mexican food involved. Collectively, these observations raised concern that an unidentified vehicle of infection might exist.
In mid-June, a cluster of cases was identified among persons who ate at Restaurant A, a Mexican-style restaurant in Wichita Falls, Texas, providing an opportunity for evaluation of the discrete list of foods served. A second cluster of cases was subsequently identified among persons who ate at two locations of Mexican-style Restaurant Chain B, in northern Texas. Here we describe the investigation of cases associated with Restaurant A and Restaurant Chain B conducted to further characterize the source of infections in the broader multistate outbreak.

Ethics statement
The National Center for Emerging and Zoonotic Infectious Diseases within the Centers for Disease Control and Prevention determined that these investigations did not meet the definition of research as provided by 45 CFR4 6.102(d) and therefore IRB review was not required. The basis for this determination was that the primary purpose of this activity was to identify, characterize, and control disease in response to an immediate public health threat. All participants were explained the purpose of the investigation and participation was voluntary.

Case definitions
We defined confirmed cases as culture-confirmed Salmonella infections reported to the Texas Department of State Health Services (TXDSHS) or Wichita Falls-Wichita County Public Health District (WFWCPHD) during June 2008 in persons who ate food from Restaurant A or from either of two Restaurant Chain B establishments during the seven days before diarrhea began. We defined probable cases as $3 loose stools in a 24-hour period starting within 7 days after eating at Restaurant A or at either Restaurant Chain B location.

Case finding
We obtained lists of culture-confirmed Salmonella infections in each of the three counties where the three restaurants were located. Outbreak strain infections, defined as Salmonella enterica serotype Saintpaul with XbaI PFGE pattern JN6X01.0048, were detected through routine laboratory-based surveillance, whereby clinical laboratories forward Salmonella isolates to public health laboratories for serotyping, PFGE analysis, and reporting [6]. Additional culture-confirmed Salmonella infections for which subtyping analyses were not performed were detected through mandated Salmonella case-based reporting by physicians.
We then determined which infections occurred in persons who had consumed food from Restaurant A or from either Restaurant Chain B establishment during the 7 days before diarrhea began through telephone interviews. These confirmed case-patients were asked to provide names and telephone numbers of their meal companions. Probable case-patients were identified from this list of meal companions. Several probable case-patients with resolved gastroenteritis submitted rectal swabs for culture at the TXDSHS laboratory during our investigation to more accurately classify case-patients.

Case-control studies
Beginning June 21, we conducted sequential, independent, matched case-control studies of Restaurant A and Restaurant Chain B patrons. Well meal companions, defined as persons who ate meals at either Restaurant A or B with confirmed case-patients but who did not develop diarrhea, abdominal cramping, fever, or vomiting during the 7 days following the meal, were enrolled as controls. Trained interviewers administered standardized questionnaires developed to measure exposure to all menu items. Because patrons could modify menu items by asking that standard ingredients or condiments be excluded or by adding non-standard ingredients or condiments, respondents were systematically questioned about any modifications made. In addition to assessing exposures to menu items, we obtained recipes for each item, to assess exposures to specific ingredients.

Statistical analysis
Data from each case-control study were separately analyzed using SAS 9.2 (SAS Institute, Cary, NC). We used exact conditional logistic regression to obtain matched odds ratios and 95% confidence intervals for the association between illness and each menu item and ingredient. Because both gender and age are often associated with diet preferences, we also calculated sex-and age group-(,10 and $10 years) adjusted estimates. We used a pvalue threshold of #0.05 for significance and did not adjust for multiple comparisons. Not all case-patients had consumed their meals with controls. Therefore, to include as many case-patients as possible in the analyses, we matched case-patients to controls by meal date. We allowed a variable ratio of controls to cases in each matched set. Sex-and age group-adjusted multivariate models were constructed when two or more menu items or ingredients were associated with illness in univariate analysis. To maximize our sensitivity to detect an association between raw tomato consumption and illness, all menu items that regularly contained raw tomatoes were treated as containing them even if the restaurant reportedly stopped using raw tomatoes in response to initial outbreak alerts.

Restaurant environmental investigations
Environmental investigations of the restaurants included review of food storage and preparation, inquiries into worker absenteeism and illness, and collection of environmental and food samples for testing using the FDA's Bacteriological Analytical Manual (BAM) Salmonella culture method at TXDSHS and WFWCPHD laboratories [8].

Traceback investigation
The FDA and Centers for Disease Control and Prevention (CDC) formed a joint team in July 2008 to trace the path of implicated product from both Restaurant A and Restaurant Chain B back to the farm. We used dates of patient exposures to the restaurants as the starting point to identify suspect shipments. Distribution records, including invoices and bills of lading, were used to determine shipment and receipt dates at each point in the supply chain. FDA investigators collected product and environmental samples for testing using the BAM method at FDA laboratories [8].

Restaurant A epidemiological investigation
Case finding. Thirty-two culture-confirmed outbreak strain or non-subtyped Salmonella infections were identified during June 2008 in Wichita County. Of 30 patients who could be contacted, 29 (97%), met the confirmed case definition and consumed food from Restaurant A. In addition, 25 probable case-patients were identified; three submitted rectal swabs and were reclassified as confirmed case-patients following isolation of the outbreak strain. All patients dined at Restaurant A from May 30 through June 2.
Case-control study. We enrolled 47 case-patients (25 confirmed, 22 probable). Of the 25 confirmed case-patients, 19 had outbreak strain infections; Salmonella isolates from the other six were not sent to the TXDSHS laboratory for serotyping or subtyping. We enrolled 40 meal date-matched controls.
Among enrolled case-patients, dates of dining at Restaurant A and dates of diarrhea onset were similar for culture-confirmed and probable cases ( Figure 1). The median incubation period was 2 days for both culture-confirmed (range, ,1-7 days) and probable (range, 1-5 days) cases. Case-patients ranged in age from 2 to 63 years; 51 percent were female, four (9%) were hospitalized, and none died.

Restaurant chain B epidemiological investigation
Case finding. Thirty-three culture-confirmed outbreak strain or non-subtyped Salmonella infections were identified through routine surveillance during June 2008 in the two counties in which the Restaurant Chain B establishments were located. Of 32 patients who could be contacted, 23 (72%) met the confirmed case definition and consumed food from either of the two Restaurant Chain B establishments. In addition, nine probable case-patients were identified. All patients dined at Restaurant Chain B from May 23 through June 12, with only two exposures before June 2.
Case-control study. We enrolled 30 case-patients (21 confirmed, 9 probable). Of the 21 confirmed case-patients, 13 had outbreak strain infections; Salmonella isolates from the other eight were not sent to the TXDSHS laboratory for serotyping or subtyping. We enrolled 30 meal date-matched controls. Matched controls could not be found for nine (5 confirmed, 4 probable) case-patients, leaving only 21 case-patients included in the matched analyses ( Figure 2).
Among enrolled case-patients, dates of dining at Restaurant Chain B and dates of diarrhea onset were similar for cultureconfirmed and probable cases (Figure 2). The median incubation period was 3 days for both culture-confirmed (range, 1-7 days) and probable (range, 1-5 days) cases. Case-patients ranged in age from 5 to 76 years; 47 percent were female, nine (30%) were hospitalized, and none died. Of 101 menu items individually evaluated, no items were significantly associated with illness in the absence of adjustment for possible confounders. With adjustment for age and sex, only red salsa was statistically associated with illness, consumed by all 21 case-patients and 24 (77%) controls (adjusted-mOR, 13; 95% CI, 1.3-infinity) ( Table 1). No ingredients were significantly associated with illness in the absence of adjustment for possible confounders. With adjustment for age and sex, the only ingredient in Restaurant Chain B dishes associated with illness was raw jalapeñ o peppers, consumed by all 21 case-patients and 25 (81%) controls (adjusted-mOR, 13; 95% CI: 1.3-infinity) ( Table 1). All nine case-patients excluded for lack of meal date-matched controls consumed red salsa and raw jalapeñ o peppers at the restaurants.

Environmental findings
The Restaurant A salsa contained raw Roma or red round tomatoes, raw jalapeñ o peppers, salt, granulated garlic, and red pepper flakes. During the 4 day period in which case-patients ate at Restaurant A, approximately 25 25-pound (11.3 kg) boxes of tomatoes of various brands were used by the restaurant, whereas only one box of jalapeñ o peppers was likely used.
The recipe for Restaurant Chain B salsa was similar to Restaurant A's, except that it always used commercially canned tomatoes, not raw tomatoes. Restaurant A and Restaurant Chain B used different brands of salt, granulated garlic, and red pepper flakes. Therefore, the only ingredient that might have been in common between the two salsas was raw jalapeñ o peppers. The  dry spices were purchased in bulk, lasting longer than the observed exposure periods. The implicated salsas were made in large batches. Individual servings were placed on multiple trays stacked above a single layer of ice in the service lines. The restaurants did not routinely measure the temperature of the salsa on these trays.
The restaurant managers denied any food worker illness or absenteeism during the case-patients' exposure periods. Cultures of food and environmental samples, including tomatoes, jalapeño peppers, salsa, spices, and swabs of food processing and storage surfaces collected from Restaurant A and one Restaurant Chain B location did not yield serotype Saintpaul. However, no foods served during the dates of exposure were available for testing.

Traceback investigation
Jalapeñ o peppers used in Restaurant A during the period of case-patient exposure were traced to Importer A in southern Texas. Jalapeñ o peppers used in the Restaurant Chain B restaurants during the period of case-patient exposure were traced to Importer B, located near Importer A. This area of Texas is home to many produce import firms. During our record review we directly observed that these firms commonly trade goods amongst themselves to fill orders.
Culture of a jalapeñ o pepper sample collected from Importer B on July 11 yielded the outbreak strain. These peppers were traced to a packing facility in Nuevo Leon, Mexico. The traceback from the packing facility was complex with commingling of product and a network of interrelated distribution points. FDA investigations continued on two Mexican farms (Farm A and Farm B) that were major suppliers of peppers to the packing facility, though records indicate that other farms also supplied the packing facility during this time period. Farm A grew Roma tomatoes in addition to jalapeñ o and serrano peppers. Environmental sampling at Farm A identified Salmonella, but none that were serotype Saintpaul. Farm B, located approximately 100 miles from Farm A, was the packing facility's main pepper supplier. It grew jalapeñ o peppers and serrano peppers, but not tomatoes, and harvested produce from mid-April to mid-June ( Figure 3). The outbreak strain was isolated from two samples taken from Farm B-agricultural water and serrano peppers found in the field.

Discussion
In the midst of a massive multistate outbreak of Salmonella serotype Saintpaul infections, investigation of two clusters of cases among patrons of restaurants in Texas provided evidence implicating jalapeñ o peppers. First, the only ingredient in the restaurants independently associated with illness was raw jalapeño peppers. Second, the only ingredient in common between the two restaurant salsa recipes was raw jalapeñ o peppers. Third, the outbreak strain was isolated from serrano peppers and irrigation water on a farm that grew epidemiologically implicated jalapeño peppers. Harvesting of peppers from the farm began shortly before the first cases of the multistate outbreak and continued for a duration that could account for the entire outbreak. Although, to our knowledge, no reports of Salmonella infections acquired from consumption of jalapeñ o peppers existed before our investigations, it was known that Salmonella grows well in extracts of homogenized jalapeñ o peppers [9]. This outbreak is one in a series of increasingly recognized, large, and widely-dispersed outbreaks caused by contaminated produce [10][11][12][13][14][15][16][17]. The proportion of all reported foodborne outbreaks in the United States associated with contaminated produce increased from 0.7% in the 1970s to 6% in the 1990s; this trend appears to be ongoing [2,3]. Related forces likely driving this shift include increasing consumption of fresh produce, increasing centralization of the produce industry to maintain year-round availability, expansion of growing fields to areas adjacent to animal production areas, and improved detection and investigation of widely dispersed outbreaks [1,2,18]. Centralization might foster outbreaks by increasing the number of points at which contamination may occur and the amount of product that could be contaminated in a given event. Centralization could also lead to more widely dispersed outbreaks because contaminated produce may be widely distributed [19].
Routine laboratory-based surveillance, especially molecular subtyping by PulseNet USA, has greatly improved our ability to detect widely dispersed outbreaks [20]. However, investigation of these outbreaks can be challenging when there is little apparent epidemiologic clustering by person, place, or time, and because of time delays inherent with laboratory-based surveillance. When generating hypotheses, investigators must consider an enormous list of possible food, water, animal, and environmental exposures [2]. Furthermore, patients are often interviewed several weeks after becoming ill, when recall of basic foods consumed is limited, and even more diminished for specific details, such as tomato type or ingredients within prepared dishes.
Our investigations highlight how these challenges can be reduced by searching for and investigating clusters of cases associated with specific food establishments or events within a larger outbreak. Investigations of localized clusters simplify hypothesis generation because the suspected sources of infection are usually limited to a finite list of menu items [2]. Likewise, patient recall is enhanced by focusing on a specific, often memorable, meal. Furthermore, investigators can obtain recipes to assess associations between illness and specific ingredients; this is especially helpful in identifying stealthy vehicles, such as jalapeñ o peppers, that some people might have been unaware they consumed. Additionally, turnover rate of ingredients within restaurants can be evaluated to find those ingredients that best fit the exposure period of patients in the outbreak. For example, at Restaurant A we noted that approximately 25 boxes of tomatoes would have been used during the 4-day period when patients ate there compared with one box of jalapeno peppers; thus contamination present in just one box of peppers could account for the Restaurant A outbreak duration. Finally, food delivery invoices kept by restaurants and caterers contain specific information that can improve the accuracy of traceback investigations; this is particularly important in fresh produce-associated outbreaks because fruits and vegetables sold in stores typically come in many varieties and may have minimal labeling to identify their source. Collectively, these attributes of localized cluster investigations serve to generate more specific exposure information [21].
Detailed exposure information allowed us to document the stealthy nature of jalapeñ o peppers in this outbreak. Three casepatients in our study of Restaurant Chain B also participated in a multistate case-control study [6]. Our study, aided by restaurant recipes, documented that all three consumed foods containing raw jalapeño peppers at the restaurant, as all reported having consumed salsa; one patient also reported adding raw jalapeñ o peppers to an entrée. In the multistate study, which asked patients to recall consumption of specific ingredients in restaurants, the two patients, whose only exposure to raw jalapeñ o peppers at Restaurant Chain B was from salsa, denied exposure to this ingredient, presumably because they were unaware that it was present in the salsa.
Some restaurant-prepared fresh salsas might be prone to amplifying and spreading small amounts of bacterial contamination present on a few individual produce items to a large number of servings because they often are made in large batches containing pooled and diced raw produce ingredients [21]. Salmonella grows better on diced, as compared with intact, tomatoes and jalapeñ o peppers [22]. Diced tomatoes, and any foods containing them, unless acidified to a pH of ,4.2, are included in the 2009 FDA Food Code as a potentially hazardous food, and thus require storage at or below 41uF (5uC) [23]. High Salmonella growth rates have been observed in salsas prepared using the Restaurant A and Restaurant Chain B recipes at temperatures at or above 54uF (12uC) [22]. Growth might be reduced by replacing granulated garlic with fresh garlic and adding lime juice to the recipes [22].
Because so much produce is consumed raw and disinfection methods are not highly effective on fresh produce, the keys to preventing produce-associated outbreaks are preventing the initial contamination and minimizing handling practices that lead to amplification [2]. Water used for irrigation and pesticide application is one possible source of contamination in this outbreak. Keeping water used for these purposes protected from animals and waste run-off is important. Modifying salsa recipes to include growth-inhibitory ingredients might limit amplification of Salmonella [22]; all fresh salsas with a pH $4.2 should be stored in adherence with established time and temperature recommendations to minimize growth [24]. Additionally, because Salmonella grows rapidly in diced jalapeñ o peppers, regulatory consideration is warranted to define foods containing them as potentially hazardous.
During the summer of 2008, 1,500 Salmonella serotype Saintpaul infections with the outbreak PFGE pattern were reported from 43 states and the District of Columbia and Canada. Shortly after jalapeñ o peppers were identified as the probable source of infections in the two Texas clusters described in this report, an independent investigation of restaurant-acquired serotype Saintpaul infections in Minnesota implicated raw jalapeñ o peppers [6]. During the outbreak, 33 restaurant clusters were reported nationally; 31 served foods that contained jalapeñ o or serrano peppers [6].
The Farm B chili pepper harvest period (April 14-June 14) closely mirrors the range of illness onset dates in the multistate outbreak (April 16-August 26) (CDC and FDA, unpublished data). Chili peppers might begin to wrinkle and lose quality 3 weeks after harvest, but refrigeration may extend their shelf life well beyond this period (FDA, personal communication). Therefore, Farm B peppers were harvested shortly before and available during this multistate outbreak (Figure 3). On July 30, national alerts advised persons to avoid raw jalapeñ o and serrano peppers grown or packed in Mexico. The two restaurant cluster investigations in Texas reported here were critical in solving this complex multistate outbreak.