Use of Traceback Methods To Confirm the Source of a Multistate Escherichia coli O157:H7 Outbreak Due to In-Shell Hazelnuts

Abstract

Traceback methods by state regulatory agencies were used to complement traditional epidemiological cluster investigation methods and confirmed hazelnuts (also referred to as filberts) as the vehicle in a multistate outbreak of Escherichia coli O157:H7 infections. Bulk in-shell hazelnut and mixed-nut purchase locations were identified during the initial epidemiological interviews. Based on purchase dates and case onset dates, regulators in Minnesota, Michigan, and Wisconsin traced product back through the supply chain. Six (86%) retail locations received the suspect hazelnut or mixed-nut shipments from a Minnesota distributor, with one retailer (14%) receiving products from a Wisconsin distributor. Both distributors received 100% of their bulk in-shell hazelnuts and mixed nuts from a distributor in California. The California distributor received 99% of their hazelnuts from a packing company in Oregon. The California distributor received the hazelnuts in 50-lb (22.7-kg) bags and either resold them without opening the bags or used the in-shell hazelnuts in the manufacture of their in-shell mixed nuts. Records at the packing company in Oregon were incomplete or lacked sufficient detail needed to identify a suspect farm or group of suspect farms. Laboratory samples collected from human cases and subsequently recalled product matched the outbreak pulsed-field gel electrophoresis subtype of E. coli O157:H7. Hazelnut harvesting practices create a plausible route of contamination from fecal matter from domestic ruminants or wild deer. This outbreak investigation demonstrates the use of product traceback data to rapidly test an epidemiological hypothesis.

In the United States, an estimated 63,000 cases of Escherichia coli O157:H7 infection occur every year, including approximately 3,700 laboratory-confirmed cases and 20 deaths (28, 30). E. coli O157:H7 outbreaks have been primarily associated with ground beef and leafy green vegetables, reflecting both the primary reservoir and environmental spread of the agent (28). The apparent complexity of E. coli O157:H7 reservoir systems results in unusual or new food vehicles, such as cookie dough and mechanically tenderized steaks, being periodically documented through outbreak investigations (3, 14, 25).

Food supply chains are integrated at the point of consumption. Complex foods may contain a combination of globally sourced and locally produced ingredients. Ingredients from a single supplier may be incorporated into hundreds of different products. These complex food systems present a considerable challenge for analytic epidemiologic investigation methods in which exposures have typically been analyzed at the level of a specific food item or commodity rather than by the source of the commodity or ingredient. This lack of detailed exposure information can limit the ability of an analytic study to identify and confirm the vehicle of outbreaks caused by commercially distributed food items.

Tracing the distribution pathway of suspect food items to their respective production sources has been a critical part of epidemiologic outbreak investigations, providing the food exposure specificity necessary to identify the outbreak vehicle (23, 31, 33).

In February 2011, a multistate cluster of E. coli O157:H7 cases with isolates of the same pulsed-field gel electrophoresis (PFGE) subtype (Centers for Disease Control and Prevention [CDC] XbaI designation EXHX01.1159, BlnI designation EXHA26.3665) was identified in Wisconsin (four cases), Minnesota (three cases), and Michigan (one case) (10).

Hypothesis-generating interviews conducted by public health agencies in each state, along with reinterviews of each case with additional specific questions about a number of food items, identified that in-shell hazelnuts were the only food item consumed by all cases. In some instances the hazelnuts were purchased as part of a mixed-nut product. However, brand names for the hazelnuts were not available, as in each instance they were purchased from bulk bins in grocery stores. Due to the unavailability of brand information and the higher-than-expected rate of reported hazelnut consumption among cases (19), investigators determined that tracing back the hazelnuts for all of the cases in an attempt to identify a common distribution source would be the fastest and most effective way to test the epidemiological hypothesis and facilitate an effective public health intervention.

We describe here the criteria and methods used to conduct these tracebacks and consequently confirm in-shell hazelnuts as the outbreak vehicle.

MATERIALS AND METHODS

Case definition and follow-up.

A case was defined as a person who had an E. coli O157:H7 isolate with the outbreak PFGE pattern (EXHX01.1159, EXHA26.3665) and who had illness onset on or after 1 December 2010. State-specific hypothesis-generating questionnaires were administered by each state, and patients were reinterviewed several times about consumption of various specific food items.

Traceback investigation.

The Wisconsin Division of Public Health (WDPH) and the Minnesota Department of Health (MDH) initially discussed the E. coli O157:H7 cluster on 4 February 2011 (Fig. 1). On 11 February 2011, the Minnesota Department of Agriculture (MDA), MDH, WDPH, the Wisconsin Department of Agriculture, Trade and Consumer Protection (WDATCP), the Michigan Department of Agriculture and Rural Development (MDARD), and the Michigan Department of Community Health (MDCH) conducted a conference call to share updated case exposure histories, discuss suspect food items, and plan further investigation approaches. During this call it was decided that the Minnesota and Wisconsin state regulatory agencies would initiate traceback investigations of in-shell mixed nuts (in all instances the mix consisted of in-shell hazelnuts, walnuts, almonds, and Brazil nuts) and in-shell hazelnuts consumed by cases to determine if product distribution data could confirm the epidemiologic hypothesis that hazelnuts were the outbreak vehicle.

FIGURE 1.

Timeline of the epidemiologic and traceback investigation starting on 3 February 2011 and concluding on 5 March 2011, when the outbreak PFGE pattern was identified from a hazelnut sample collected from a case patient’s home. Agency acronyms: WDPH, Wisconsin Division of Public Health; MDH, Minnesota Department of Health; MDA, Minnesota Department of Agriculture; CDPH, California Department of Public Health.

MDA contacted the California Department of Public Health (CDPH) to inform them that a distributor in California (distributor C) was a point of convergence in the traceback investigation. On 25 February, CDPH conducted an inspection of this distributor and collected invoices and other records to identify the source of this distributor’s in-shell mixed-nut and hazelnut products.

Record collection.

Table 1 lists the type of information that was collected by MDA, WDACTP, MDARD, and CDPH (34). Records obtained during this traceback investigation were collected in person by field investigation staff and remotely by phone, e-mail, or fax. The record collection time window was based on case exposure information, product shelf life, and product residence time in the supply chain.

TABLE 1.

Type of records and information that should be collected during a traceback investigation^a

Information type	Type of records or information
Record collection	Examples of records that typically need to be collected include, but are not limited to, invoices, shipping and receiving records, bills of lading, inventory record queries, label information, packaging type and size, lot codes, UPCs or GTINs, and production dates.
Product ordering and shipping	How and when is product ordered? How much of the product is used or shipped daily? What is done if the establishment runs out of product before another shipment is received (e.g., purchase from grocery store, request more from supplier, etc.)? How are deliveries and receipt dates recorded? Compare the shipping dates to the dates received; determine suppliers during the time period of interest, including cash transactions. What is the transportation time from supplier(s) to the establishment? Was the product repacked during distribution?
Product storage and handling	How is the product unloaded and added to existing inventory? Is the suspect food item used as an ingredient in preparation or manufacture of another food item? How is stock inventory recorded? How are partial cases and/or containers accounted for if carryover is recorded? What does each inventory number represent? Review the standard procedures for stock rotation and determine if the facility is capable of internal traceability or follows a FIFO model.

^aUPCs, universal product codes; GTINs, global trade item numbers; FIFO, first-in, first-out.

Each regulatory agency obtained invoices from each retailer from which a case had reported purchasing either in-shell mixed-nut or in-shell hazelnut products. A traceback target time frame from 1 November 2010 to 31 December 2010 was established, and invoices for all bulk in-shell nuts within this time frame were requested from each retailer.

The MDARD food inspection staff visited the Michigan retail store to obtain invoices. WDATCP and MDA contacted retail locations and distributors both in person and by telephone, and invoices were obtained in person, by fax, or by e-mail.

MDA investigators contacted a distributor in Minnesota by telephone and e-mail and requested invoice and purchase order (PO) records pertaining to this distributor’s source of in-shell nuts that would have shipped to six retail stores where case patients had purchased in-shell hazelnuts or in-shell mixed nuts. WDATCP contacted a Wisconsin distributor by telephone and e-mail and requested invoice and PO records and bills of lading pertaining to this distributor’s source of in-shell nuts that corresponded to product shipped to a retail store where a case patient had purchased in-shell mixed nuts.

Wisconsin and Minnesota investigators analyzed product distribution information to identify shipments most likely associated with illness, based on case-reported purchase and consumption dates.

An iterative approach was used in Minnesota to collect traceback information, similar to that used in epidemiological investigations (29, 32). This approach, as it applies to traceback investigations, involves confirming product distribution and receipt backwards through the supply chain to confirm that product that was shipped was actually received. This involved verifying that documents related to incoming shipments matched the quantities and descriptions of the outgoing shipments one step back in the supply chain. Where discrepancies were identified, investigators contacted both entities in the supply to seek clarification.

A traceback diagram that included case onset dates, case purchase dates, product description, quantities, shipment and receipt dates, invoice and PO numbers, and notes on case exposures or product handling practices was constructed (Fig. 2). Figure 2 has been simplified for publication to include only the shipments and product exposures most likely to have been associated with human illnesses.

FIGURE 2.

Traceback diagram for E. coli O157:H7 1102WIEXH-1 cluster investigation of seven cases in three states. The hazelnuts were traced back to two packing facilities and did not undergo further processing at the distributor or retail level. The mixed nuts were made by distributor C and included the same hazelnuts from the packing facilities. Distributor A assigned unique identifiers based on purchase order number (PO #), which created internal traceability and facilitated rapid traceback. This figure depicts shipments limited to 2 weeks prior to a case’s approximate purchase date and does not represent all of the hazelnuts or mixed nuts distributed during the outbreak investigation.

Laboratory investigation.

A sample of in-shell bulk hazelnuts was collected from one case household and tested by the MDA using the standard method, which involves a PCR screen followed by culture confirmation incorporating immunomagnetic separation isolation techniques (11). One 50-lb (22.7-kg) bag of hazelnuts and one 50-lb bag of walnuts were collected from the California distributor by the CDPH during one on-site inspection. Wisconsin DATCP collected in-shell hazelnuts from a case household as well as a 50-lb bag of in-shell mixed nuts from a Wisconsin distributor that was returned after the California distributor announced its recall (9).

RESULTS

Traceback investigation.

Tracebacks were completed for seven of the eight cases linked to the cluster (cases A through G; Fig. 2); the eighth case reported an exposure to the implicated hazelnuts but was detected in Wisconsin after the products had been recalled. Each case reported either an in-shell mixed-nut or in-shell hazelnut exposure, and no case reported consuming both products prior to illness onset. Four (57%) cases reported the purchase of bulk in-shell mixed nuts and three (43%) reported purchase of bulk in-shell hazelnuts.

The purchase dates reported by the cases ranged from 16 December 2010 to 7 January 2011, although five cases reported purchase dates around 25 December 2010. Each case reported purchasing nuts from a separate retailer (retailers A through G).

Six retailers (86%) reported purchasing nuts from a common distributor (distributor A). The one (14%) retailer who did not receive in-shell nuts from distributor A received in-shell mixed nuts from a distributor in Wisconsin (distributor B). Both distributor A and distributor B received 100% of their in-shell mixed nuts and hazelnuts from a common third distributor, distributor C, during the investigation time frame.

Among the six retailers that received in-shell mixed nuts or in-shell hazelnuts from distributor A in the 2 weeks prior to case purchase, one (14%) lot of in-shell mixed nuts was shipped to multiple retailers (A, B, and F). This common lot (Lot x20–04 in Fig. 2) was in three (75%) retail locations 2 weeks prior to patient purchase dates and was the only shipment to two (50%) of the retail locations associated with mixed-nut exposures. The source of the hazelnuts in Lot x20–04 of mixed nuts could not be directly traced to an incoming shipment of hazelnuts at the California distributor (distributor C) due to a lack of internal traceability. However, based on the documented first-in, first-out product handling practices at distributor C, it is likely that the hazelnuts received on 22 November 2010 from packer B or on 24 November 2010 from packer A were the likely source. Given the volume of hazelnuts received from packer A, it is more likely that packer A was the source of the contaminated hazelnuts.

There were no common lots of in-shell hazelnuts shipped to all retailers during the 2 weeks prior to case purchase dates. Based on shipments to retailers C, D, and E, it is likely that product received by distributor A on or after 10 December 2010 was most likely to be associated with illness, and these shipments also traced back to product received by distributor C on 22 November 2010 and 24 November 2010.

Initially, retailer C denied that distributor A supplied any hazelnuts to the store during the investigation time frame. After repeated phone interviews with the quality assurance manager for retailer C and the quality assurance manager for distributor A, a single shipment of hazelnuts to retailer C occurring on 14 December 2010 was identified.

Ninety-eight percent (124,000 lb [56,363.6 kg]) of in-shell hazelnuts that distributor C distributed during the target time frame were received from packer A while <2% (1,750 lb [795.5 kg]) came from packer B (a single shipment received on 22 November 2010). A specific hazelnut grower was not identified during the traceback investigation because packer A did not provide records to the regulatory agencies. However, during the course of the investigation packer A indicated that between 20 and 60 growers might have provided product that shipped to distributor C in the time frame of interest and packer A sourced only domestically harvested hazelnuts.

Some 50-lb bulk bags of hazelnuts that distributor C received from packers A and B were used to make the in-shell mixed nuts shipped to distributors A and B. Those hazelnuts not used to manufacture mixed nuts at distributor C followed an approximate first-in, first-out pattern of shipment. Records at distributor C were not of sufficient detail to link incoming shipments of hazelnuts from packers A and B to outgoing shipments of mixed nuts and hazelnuts to distributors A and B.

Recall announcement.

On 4 March 2011 distributor C issued a voluntary recall of all hazelnuts and mixed-nut products distributed from 2 November to 22 December 2010. Recalled product was distributed to stores in seven states: Minnesota, Iowa, Michigan, Montana, North Dakota, South Dakota, and Wisconsin.

Regulatory and health agencies in Minnesota and Wisconsin issued press releases on 4 March 2011 to inform the public. All persons who possessed recalled in-shell hazelnuts were encouraged to discard them or return them to the store (7, 8). MDA and WDATCP provided a list of stores where recalled product was sold based on distribution records obtained during the traceback investigation.

Laboratory investigation.

On 5 March 2011, the MDA laboratory reported isolation of E. coli O157:H7 from bulk in-shell hazelnuts collected from a case patient’s home; on 7 March, the isolate was determined to match the outbreak pattern by PFGE and multilocus variable number of tandem repeats analysis. The outbreak strain of E. coli O157:H7 was also isolated from an intact in-shell mixed-nut sample collected from a 50-lb bag collected from distributor B by WDATCP on 11 March 2011. WDATCP also isolated Shiga toxin–producing E. coli O64:H34 from an intact in-shell mixed-nut sample collected from a 50-lb bag from distributor B. CDPH isolated the outbreak PFGE subtype of E. coli O157:H7 from an intact mixed-nut sample collected from a 50-lb bag from distributor C. Because of inadequate record keeping at distributor C, investigators could not definitively link positive product samples to a particular incoming shipment from packer A or B. The first-in, first-out practices and quantity of product received suggested that the 40,000-lb (18,181.8-kg) shipment from packer A to distributor C on 24 November 2010 likely contained the contaminated bolus of hazelnuts.

DISCUSSION

This was an outbreak of E. coli O157:H7 infections associated with bulk in-shell hazelnuts sold at retail food locations in Minnesota, Wisconsin, and Michigan. This is the first recognized E. coli O157:H7 outbreak associated with nuts. However, previous Salmonella outbreaks or recalls have been associated with almonds, pistachios, and peanuts (4, 5, 13, 18, 24, 26).

Regulatory agencies have historically conducted tracebacks to determine the source of a product after laboratory confirmation of the etiologic agent in food or after the food item was epidemiologically associated with illness in an analytic study. This outbreak demonstrates the usefulness of starting a traceback investigation before the food can be definitively implicated, with the goal of confirming a suspected association by identifying a common source via a point of convergence in the food supply. By starting earlier in the course of an investigation than is traditional, this type of traceback can provide meaningful information that can shorten the course of the investigation and lead to an earlier public health intervention.

The following criteria were used to determine if a traceback investigation was warranted as part of the epidemiologic investigation: (i) a PFGE subtype cluster of cases likely represented a common source outbreak; (ii) cases occurred in multiple locations or jurisdictions (in this instance, multiple states); (iii) interviews of case patients revealed no obvious point source exposures in common (e.g., they did not eat at the same restaurant or attend the same event), suggesting that the outbreak vehicle was a commercially distributed food item; and (iv) a suspect food vehicle was identified, and the frequency of exposure among cases provided a strong hypothesis that could be directly tested by identifying a common production source for exposed cases.

The following criteria were used to determine which mixed-nut and hazelnut exposures should initially be traced: (i) the likelihood that the exposure was truly the exposure of interest for a case; (ii) the availability of clear, documented details on the exposure; and (iii) geographic and/or temporal dispersion of case exposures with the goal of identifying multiple food distribution chains during the traceback.

In this outbreak a common PFGE subtype cluster was identified, cases occurred in three states with unique retail exposures, bulk nuts were epidemiologically suspected, and identifying a common source of production was determined to be the fastest and most effective way to test the epidemiological hypothesis. Each case represented a unique retail exposure, and all were traced with the same priority.

Bulk in-shell nuts, like many produce items, were not labeled, and therefore, the consumer could not report brand information when interviewed. In order to accurately identify commonalities associated with the hazelnut exposures, a traceback investigation was required. Because identifying a common distribution source was the most direct way to test the epidemiological hypothesis and maximize the public health intervention, the tracebacks needed to be conducted rapidly. The outbreak strain of E. coli O157:H7 was ultimately isolated from in-shell hazelnuts and mixed nuts containing in-shell hazelnuts. However, laboratory confirmation occurred after the tracebacks confirmed the epidemiologic hypothesis and thereby enabled investigators to implicate hazelnuts and prompt the recall and public advisories.

Multistate outbreaks in which cases are not uniquely associated with a single retailer suggest that the source is a commercially distributed food and that the source is not primarily associated with on-site environmental or food worker contamination. Although food workers and environmental contamination need to be addressed, in these types of outbreaks priority should be given to rapid tracebacks through record collection related to product receipt and distribution in order to identify common suppliers throughout the supply chain.

Distributor A possessed good internal data systems and provided accurate summary reports of their complete distribution. Using the iterative investigation approach, a review of distributor A’s records identified a shipment of in-shell hazelnuts to retailer C that was originally and repeatedly denied by retailer C during the initial stage of the investigation. This shipment was significant because it was the only shipment of in-shell hazelnuts received by retailer C and represented the exposure associated with the related case’s illness as well as the positive home sample. This demonstrates the importance of reinterviewing companies and reanalyzing the distribution data when there is an apparent discrepancy.

A common source was identified in this investigation by defining implicated shipments corresponding to case purchase dates and tracing these shipments back from retailers through distributors to a repackaging and distribution operation (distributor C). While this operation (distributor C) was not likely the original source of adulteration, it did represent a common point of convergence for all cases in terms of product distribution. Distributor C received over 99% of their in-shell hazelnuts from packer A, located in Oregon. Ninety-nine percent of domestically harvested hazelnuts are grown in Oregon (27).

Product handling practices at distributor C complicated the traceback investigation. Distributor C did not maintain records that would allow investigators to link an incoming shipment of bulk nuts to an outgoing lot of finished product when manufacturing mixed nuts.

Packer A did not maintain adequate internal traceability records to allow investigators to adequately identify a subset of farms that provided hazelnuts during the time frame of interest. Without access to these records, investigators were unable to conduct environmental assessments of the farms that supplied packer A to identify possible sources of contamination or adulteration.

The lack of internal traceability within the food processing industry and among distributors is not uncommon (22). The passage of the federal Food Safety Modernization Act in January 2011 increases record-keeping requirements and may improve internal traceability in the future (6).

Distributor A maintained internal traceability and assigned a unique lot number to all incoming products based on the PO and item number of incoming product (Fig. 2). This internal traceability allowed investigators to trace product in retail stores directly back to specific incoming shipments from distributor C to distributor A. If this level of traceability were available throughout the entire supply chain, it would be possible to easily identify a farm or producer as the source of contamination. Specifically, the positive hazelnuts collected from the Minnesota case household came from a single identifiable shipment to retailer C on 14 December 2010. Because this was the only shipment to this location and the hazelnuts tested positive, complete supply chain traceability would have identified the farm where the product was grown and harvested.

This outbreak investigation illustrates the importance of collaboration between epidemiologists and regulatory officials within individual states and between states. Regular conference calls were held among MDA, MDH, WDATCP, WDPH, MDARD, MDCH, CDC, and the U.S. Food and Drug Administration to discuss common exposures among cases and share traceback information as the evidence was collected and developed. Given the small number of cases, the likelihood that a single state could have implicated in-shell hazelnuts was small.

Similarly, detailed communication among regulatory agencies increased the speed and accuracy of the investigation. MDA was in regular communication with CDPH once distributor C had been identified during the traceback investigation. CDPH’s inspection of distributor C and record collection prompted a recall of the adulterated product on 4 March 2011, 3 weeks after the initial conference call between MDA and WDPH (Fig. 1).

Hazelnuts are mechanically harvested from the ground. Processing practices may vary by facility, where some may be treated with an antimicrobial wash prior to the drying process. The risk of fecal or environmental contamination of the exterior of the hazelnut is highly plausible. Previous outbreaks of fresh produce have been associated with fecal contamination of fields by wild animals and domestic ruminants (1, 2). In this instance, it is plausible that feces from wild deer or domestic cattle grazing in the orchards contaminated the surface of the hazelnuts prior to harvesting. No published data on the survivability of E. coli O157:H7 on in-shell nuts exist, but the traceback evidence suggests that the organism remained viable for at least 3 months from the time of initial distribution to the time when the home sample tested positive on 5 March 2011. There is evidence that Salmonella can persist for days to weeks on nut shells and in orchard soils (17, 21, 35). In pecans and almonds, Salmonella can infiltrate a damaged shell and remain viable in the kernel for over a year after drying (16, 20).

This investigation was timely and resulted in the recall of adulterated product, but it was limited by inadequate record keeping by distributor C and packer A, which did not allow for the identification of the farms that were the ultimate source of contamination. Without access to this information, investigators were unable to physically investigate and assess potential sources of contamination.

Better record keeping and internal traceability within the food industry will improve the timeliness and accuracy of future traceback investigations. The hazelnut industry benefited from implicated product being traced to a single distributor and a limited number of packers, rather than hazelnuts in general. Several recent outbreaks associated with tomatoes, spinach, and sprouts have seen consumer advisories that target an entire commodity group (1, 12, 15). Importantly, this outbreak demonstrated how a collaborative multijurisdictional rapid traceback investigation significantly reduced the time required to identify the source of adulterated product and initiate a meaningful public health intervention. The isolation of pathogenic Shiga toxin–producing E. coli from recalled product suggests that additional human illnesses were prevented as a result of the investigation and subsequent recall.