Although initially considered an unlikely vehicle for enteric pathogens because of its high acid content, unpasteurized apple cider has been associated with two outbreaks of E. coli O157:H7 infection (7, 8) and with one cluster of the hemolytic uremic syndrome (9). In all three outbreaks, the implicated cider was produced locally on a small scale; in the two outbreaks for which information on apple processing was available, the apples were unwashed and had been picked up from the ground. This finding led the Centers for Disease Control and Prevention (CDC) to recommend that processors wash and brush apples before pressing them (7).

On 30 October 1996, the Seattle-King County Department of Public Health in Washington State announced that a cluster of E. coli O157:H7 infections was associated with exposure to brand A unpasteurized apple juice (10). In the case–control study done by the Department of Public Health, all 10 patients with E. coli O157:H7 infection but none of 9 matched controls reported exposure to brand A unpasteurized apple juice (Alexander ER. Personal communication). Brand A juice is made by the largest commercial producer of fresh unpasteurized juice in the United States and is distributed widely throughout the western United States and British Columbia, Canada. The case–control study and subsequent announcement prompted the company to recall voluntarily all of its apple and apple-containing juices from the market. Our investigation sought to define the extent of the outbreak, confirm its source, and determine how the product became contaminated.

Methods

Top Methods Results Discussion Author & Article Info References

Our outbreak investigation had three concurrent areas of inquiry: 1) case finding to define the scope of the outbreak; 2) a traceback investigation of the juice-production plant, apple-packing houses, and apple farms to determine how the juice may have become contaminated; and 3) laboratory investigation of recalled juices, clinical specimens, and specimens from the juice-production plant and from apple farms.

Case Finding

The CDC sent notification of the outbreak to epidemiologists from all U.S. states and territories and from British Columbia, Canada. Methods of case finding varied by jurisdiction and included public health alerts, press releases, and passive and active surveillance. Widespread media attention enhanced case reporting. We defined a primary case as 1) evidence of infection with E. coli O157:H7 as defined by a stool culture positive for E. coli O157:H7, presence of the hemolytic uremic syndrome, or bloody diarrhea with serum antibody to O157 lipopolysaccharide; 2) illness onset on or after 1 October 1996; and 3) exposure to brand A apple-containing juice in the 10 days before illness onset. We excluded cases if the pulsed-field gel electrophoresis pattern of the stool isolate did not match that of the outbreak strain. We classified a case as secondary if the criteria for evidence of infection and onset date were met but exposure was to a primary case rather than to an apple-containing juice. We classified a case as having an indeterminate source if the exposure was to juice and to a primary case. For each case, we collected demographic, clinical, and product-exposure information.

Traceback Investigation

Our traceback investigation began with an inspection of the company's sole juice-production plant. Between 30 October and 12 November 1996, investigators from the Food and Drug Administration (FDA) and from the California Department of Health Services (CDHS) reviewed all records concerning production of apple juice and of juice blends that contained apple juice, microbiological testing, sanitation, and employee health and absentee records for the period from 20 September to 24 October 1996.

We inspected all packing houses and farms that shipped apples to the juice-production plant between 14 September and 15 October 1996. These dates bracketed shipment of all apples that could have caused confirmed cases.

To support the findings from the initial case–control study and to confirm that contaminated brand A apple juice could have caused the entire outbreak of E. coli O157:H7 infection, we compared the distribution of a batch of contaminated apple juice with the distribution of reported outbreak-associated cases of E. coli O157:H7 infection. The plant shipped juice to distribution nodes, and the nodes distributed juice to retailers; we were able to classify cases by distribution node if only one retailer was identified by a case-patient. We estimated the distribution of apple juice that was produced on a particular date to distribution nodes using the daily warehouse inventory log, daily shipping records, and shipping conventions as described by management.

Laboratory Investigation

All stools were cultured on sorbitol MacConkey agar. State laboratories forwarded serum specimens from some case-patients who had bloody diarrhea or the hemolytic uremic syndrome but negative stool cultures to the CDC for testing for antibody to O157 lipopolysaccharide (specimens were positive when the IgG titer 1:160 and the IgM titer 1:340) (11).

The CDHS Microbial Diseases Laboratory enriched the recalled juices (both pure apple and apple-containing blends) in modified trypticase soy broth with novobiocin (sodium salt) to recover E. coli O157:H7 (12), screened the enrichment broths with a commercial enzyme immunosorbent assay (EIA) (Assurance EIA, E. coli O157:H7, BioControl Systems, Inc., Bothell, Washington), and plated dilutions of the presumptive positive broths on sorbitol MacConkey agar and cefixime-tellurite sorbitol MacConkey agar (13). A subset of juice samples was analyzed at the CDC laboratory by using the immunomagnetic bead capture and culture method (Dynabeads anti- E. coli O157, Kynal, Lake Success, New York) and polymerase chain reaction (PCR). Targets for PCR were the uidA allele that is specific for E. coli O157:H7 (14) and the genes that encode Shiga toxins (15). Polymerase chain reaction was done directly on juice (5 µL), diluted juice (5 µL of a 1:500 dilution), and juice that had been enriched in gram-negative broth. Controls included juice samples spiked with boiled E. coli EDL 933 to ensure that negative results on PCR were not caused by inhibitors present in the juice samples.

On 31 October 1996, various points along the juice press line were swabbed. The swabs were transported in Cary-Blair transport media and enriched in modified trypticase soy broth with novobiocin, as previously described. The enriched broths were screened by EIA and were dilution plated on sorbitol MacConkey and cefixime-tellurite sorbitol MacConkey agars. Enrichment broths showing any growth were also plated on modified Endo, LES agar to detect growth of E. coli.

The CDHS microbial diseases laboratory used a defined substrate test system (Colilert, IDEXX Laboratories, Inc., Westbrook, Maine) to test water samples from farms and packing houses for E. coli. Researchers enriched specimens of animal feces in modified trypticase soy broth with novobiocin, screened the broths with EIA, and dilution plated positive broths on sorbitol MacConkey and cefixime-tellurite sorbitol MacConkey agars.

All available isolates of E. coli O157:H7 from humans, animals, and juice were subtyped by pulsed-field gel electrophoresis by using Xba I for restriction of genomic DNA according to a CDC protocol (16) or a protocol recommended by Bio-Rad Laboratories (Bio-Rad, Hercules, California). The predominant pulsed-field gel electrophoresis pattern among case isolates was designated the outbreak pattern.

Statistical Analysis

We used EpiInfo (version 6) (17) for data storage and management. We used Spearman rank correlation and the Wilcoxon rank-sum test to analyze nonparametric data.

Results

Top Methods Results Discussion Author & Article Info References

Description of Outbreak

We identified 70 cases of E. coli O157:H7 infection from California (n = 26), Colorado (n = 5), Washington (n = 29), and British Columbia (n = 10). Sixty-five cases were primary, 2 were secondary, and 3 were indeterminate. Dates of diarrhea onset for the primary cases ranged from 7 October 1996 to 3 November 1996 (Figure 1). On 7 October, one primary case-patient developed nonbloody diarrhea that resolved but then became more severe; after continued consumption of brand A juices as rehydration therapy for the initial illness, the case-patient developed bloody diarrhea on 14 October.

View larger version (18K): [in this window] [in a new window]   Figure 1. Outbreak-associated cases ofEscherichia coliO157:H7 by date of diarrhea onset (n= 70) for British Columbia, California, Colorado, and Washington (fall 1996). Each box represents one outbreak-associated case.

Of the 65 primary case-patients, 48 (74%) drank pure apple juice in the 10 days before illness onset, 7 (11%) drank a juice blend that contained apple juice, and 10 (15%) drank both types of juice. Thirty-one of the primary case-patients reported drinking only one type of juice and reported only one date of exposure (Figure 2). For these case-patients, we could calculate the incubation period between exposure to brand A juice and onset of diarrhea.

View larger version (15K): [in this window] [in a new window]   Figure 2. Dates of consumption of brand A apple juice for case-patients exposed to only one juice on one date (n= 31) for British Columbia, California, Colorado, and Washington (fall 1996). Apple juice and blended juices containing apple juice that were produced on 7 October 1996 could have been consumed at any time during their 17-day shelf life or after the expiration date. At least 2 days usually elapse between production of juice and retail sale.

Case-patients were young (56% were 5 years of age [range, 1 to 46 years]), and sex distribution was even (51% of case-patients were male). Sixty-four case-patients (91%) reported bloody diarrhea, and 25 case-patients (36%) were hospitalized. The ways in which cases met criteria for evidence of infection in the case definition are summarized in Table 1. The hemolytic uremic syndrome developed in 14 case-patients (20%) and was fatal in 1 case-patient (1%). Thirteen of the 14 case-patients who developed the hemolytic uremic syndrome were 3 years of age or younger (median age, 2 years). Patients who developed the hemolytic uremic syndrome had a longer incubation period (median, 5 days [range, 3 to 9 days]) than patients who did not develop hemolytic uremic syndrome (median, 4 days [range, 1 to 6 days]) (P = 0.03).

Plant Inspection and Juice Production

The plant seemed to be clean and well run. No verbal or written reports had documented mechanical problems with the press line, sewage problems, or gastrointestinal illness among employees during the period of interest. No sources of contamination from inside the plant were discovered.

We reviewed apple-shipping and apple-processing procedures. The plant received apple shipments in 900-pound bins from apple-packing houses or directly from apple growers. The company provided a written statement advising suppliers that it would accept only handpicked apples, but it had no mechanism to ensure compliance. Each shipment (13 to 48 bins) was assigned a lot number and was to be unloaded only if a sampling procedure estimated the level of decay by weight to be less than 10%; however, this procedure was not strictly followed.

The company produced vegetable juices and two kinds of apple juice on the same press line. The pure apple juice was made in large quantities on a standing schedule, whereas the apple juice for blends was made on an as-needed basis. Although production times were distinct for the two juices, one lot of apples (making up many 900-pound bins) could be used in both products. After a bin of apples was selected, the apples were hand-sorted on a conveyer belt, mechanically scrubbed and washed with a diluted phosphoric-acid solution, rinsed with water, ground, and pressed to extract juice. The juice was immediately filtered, cooled, and stored.

The plant used a chlorine-based wash solution until summer 1996 and then changed to a phosphoric-acid-based solution. However, neither of the two brands of phosphoric acid wash used by the plant was used correctly. According to their respective labels, one brand was not intended for use on produce and the other was not intended for use on waxed produce; in addition, the latter brand was sometimes used at concentrations below the recommended level.

On 4 November 1996, we learned that the FDA laboratory had isolated E. coli O157:H7 from one unopened container of brand A pure apple juice that was produced on 7 October 1996. According to plant records, apple juice for blends and for pure apple juice was produced on that day. We assumed that apples from only one farm were contaminated and reasoned that the contaminated apples should account both for case-patients who reported consuming only pure juice and for those who reported consuming only blended juice. We also reasoned that the production dates on which the possibly contaminated apples were used should be consistent with the consumption dates given by case-patients who had only one exposure date (at least 2 days usually elapse between production of juice and retail sale). Only three of the seven lots of apples used on 7 October (lots 1, 2, and 3) met these criteria. Lots 1 and 2 were used only to produce apple juice for blends. On 7 October, however, the apple juice press line ran continuously during the transition from juice for blends to pure apple juice, potentially allowing a pathogen in lot 1 or 2 to contaminate the pure juice. Only one of the five lots used to make pure apple juice on 7 October could also explain the infection of case-patients who consumed only blended juice. This lot, lot 3, was not used in juice for blends but originated from the same farm as lot 1; a source from that farm could explain contamination of lots 1 and 3 and consequently contamination of the pure apple juice and of the juice for blends.

Using the available data from case-patients and records from the juice-production plant, we could not determine whether contaminated juice was produced on 7 October only or on 7 October and additional days. The juice produced on 7 October could account for all cases except two that resulted from a single juice exposure (Figure 2) and could account for at least one exposure for all case-patients who had multiple exposures. However, contaminated juice could also have been produced on additional dates (probably 4, 8, 9, or 11 October).

Investigation of Apple Farms and Packers

We inspected the 6 packing houses and 31 farms that shipped apples to the juice-production plant between 14 September and 15 October 1996; however, only the 2 packing houses and 6 farms that supplied the 3 suspect lots (lots 1, 2, and 3) used on 7 October are reviewed here. We considered that the apples could have been contaminated at any time before their arrival at the plant: on the farm, at the packing plant, or during transport.

Lot 2 consisted of 48 bins of apples that were shipped from a packing house to the juice-production plant on 2 October 1996. Investigation of the packing house and the five farms that supplied apples to the packing house provided several plausible but no confirmed sources of contamination. Plant managers could not locate documentation describing the condition of lot 2 on receipt; however, lot 2 was described as "bad" in notes made by the press operator on 4 and 7 October. On those days, three employees were assigned to the hand-sorting station on the press line to remove defective fruit (usually only one employee is assigned to this task). The packing house waxed all apples before sorting them by grade, potentially sealing in any pathogens that were present on the surfaces of the apples. At the packing house, the wash water was changed only once per day; water that we collected from the reservoir drain of the drencher on 5 November 1996 was estimated to have 11.1 E. coli organisms/100 mL, but culture did not show E. coli O157:H7. Reservoir and well water obtained on 13 November from one of the five farms also showed low levels of E. coli on screening (27.1 E. coli organisms/100 mL and 2.0 E. coli organisms/100 mL), but culture did not show E. coli O157:H7. Owners of two other farms that were located near property with cattle admitted to shipping some dropped apples.

Lots 1 (13 bins) and 3 (11 bins) contained different apple varieties but were harvested from the same multivarietal orchard on the same farm. Documentation describing the condition of lots 1 and 3 on arrival at the plant could not be found by plant managers, as had been the case for lot 2. The source orchard for lots 1 and 3 hosted a large deer population. Fresh deer feces collected from an adjacent wildlife refuge area approximately 0.25 miles from the orchard on 18 December 1996 grew E. coli O157:H7, but the pulsed-field gel electrophoresis pattern of the isolate did not match the outbreak pattern. The owner of the apple farm stated that seasonal workers, who are paid by number of bins harvested, were instructed not to harvest apples from the ground, but no mechanisms were in place to enforce this policy.

Juice Distribution

The company distributed juices from its sole production plant in central California to British Columbia, California, Colorado, New Mexico, Oregon, Texas, and Washington through 20 distribution nodes. No outbreak-associated cases of E. coli O157:H7 infection were reported in New Mexico, Oregon, or Texas. The estimated number of gallons of apple juice produced on 7 October that was shipped to each distribution node was positively correlated with the number of cases of infection associated with apple juice that corresponded to each distribution node (r = 0.80; P < 0.001) (Table 2).

Laboratory Analysis

The FDA, CDC, and CDHS laboratories tested a total of 184 bottles of unopened, recalled juices representing 26 different production dates between 23 September and 29 October 1996, and local public health laboratories tested an unknown number of additional samples forwarded by families of case-patients. Only one juice sample yielded E. coli O157:H7. Cultures, immunomagnetic bead tests, and PCR of all other products, including 10 other samples of apple juice from the same production date, yielded negative results for E. coli O157:H7 and other enteric pathogens. Direct PCR of undiluted juice samples was not valid because none of the corresponding spiked controls were positive, indicating that some juice components were inhibiting the PCR amplification.

Cultures of swabs taken from the juice-production plant yielded no coliform bacteria. The fresh deer feces that grew E. coli O157:H7 were from 1 of 9 samples collected from an area near the orchard that supplied lots 1 and 3; the other 8 samples tested negative. The orchard was fertilized with composted turkey manure; samples of fresh turkey feces, turkey manure, and cattle feces collected from the source turkey farm did not yield E. coli O157:H7.

The pulsed-field gel electrophoresis pattern of the apple juice isolate was indistinguishable from that of the outbreak strain. Both produced Shiga toxins 1 and 2. In contrast, the pattern of the isolate from deer feces differed from the outbreak pulsed-field gel electrophoresis pattern by four bands; the former produced only Shiga toxin 2.

Discussion

Top Methods Results Discussion Author & Article Info References

We have described a widespread outbreak of E. coli O157:H7 infection from contaminated unpasteurized apple juice made by the largest commercial producer of fresh juice in the United States. An epidemiologic association identified by a small case–control study prompted the voluntary recall of all potentially contaminated products. This association was supported by the isolation of the outbreak strain from an unopened, recalled juice sample and by a positive correlation between the distribution of contaminated juice and the distribution of reported cases. The outbreak occurred despite sorting, washing, and brushing procedures at a sophisticated, state-of-the-art facility and despite a policy of accepting only hand-picked apples that would have been considered more than adequate in the wake of the previous apple juice-associated E. coli O157:H7 outbreak, which occurred in 1991 (7). We did not find a plausible source of contamination inside the juice-production plant, and we documented cases associated only with apple juice rather than with vegetable or citrus juices. We therefore suspect that the pathogen arrived on incoming apples.

The existing procedures at the juice-production plant were inadequate to eliminate contamination. The company's written policy of accepting only hand-picked fruit could not realistically be enforced; it would be impossible to know whether some fruit had been harvested from the ground. Three oversights at the plant might have allowed pathogens on incoming apples to persist and to contaminate the final juice product. First, although the company had procedures to assess fruit for decay after arrival at the plant, the application of these procedures was inconsistent. Lot 2, which might have caused the outbreak, clearly surpassed the acceptable cut-off point for decay, yet all 48 bins of this lot were processed. Second, the plant managers, who were apparently unaware that one packer was shipping waxed apples, were using a wash solution whose label specifically stated that wax removal was required before washing. Third, the acid wash was inadvertently run at concentrations lower than the recommended concentration and probably did not kill all of the pathogens on the fruit.

Several studies have demonstrated the acid tolerance of E. coli O157:H7 (18-20). In one study, after 21 days in unpreserved cider at a pH of 3.7 to 3.9 at 4 °C, 91% to 98% of the organisms survived (21). The acid tolerance of the organism may be related to rumen physiology: The organism may primarily inhabit the rumen, an acidic environment, and be shed intermittently in the feces (2). Once shed, the organism can grow in manure slurries for several weeks (22). The outbreak described here demonstrates, once again, that fresh apple juice is an adequate vehicle for E. coli O157:H7.

However, E. coli O157:H7 is not the only enteric pathogen that can survive the acidic conditions in apple juice, nor is apple juice the only acidic fruit juice demonstrated to harbor enteric pathogens. Apple juice has been implicated in outbreaks of cryptosporidiosis (8, 23) and salmonellosis (24) in addition to outbreaks of E. coli O157:H7. Orange juice has been implicated in several outbreaks of viral gastroenteritis (25) and salmonellosis, even at a pH of 4.0 (26).

Our outbreak investigation had two main limitations. First, we could not estimate the infectious dose of E. coli O157:H7 or the number of contaminated apples that might have been necessary to cause the outbreak. The recalled juice samples were tested after the shelf life of the product had expired, and most were fermented. Although the FDA laboratory did recover the organism from juice, fermentation and acidity prevented us from isolating the pathogen from more than one sample. Second, although evidence supported an apple farm as the source of contamination, we were unable to pinpoint which lot of apples was contaminated or how it became contaminated, a frequent limitation in outbreaks associated with fresh produce. The E. coli O157:H7 isolate obtained from the deer did not match the outbreak strain; however, it is plausible that deer, like cattle (27), may carry and excrete more than one strain. We believe that dropped apples were probably shipped to the juice-production plant despite the company's stated policy of accepting only hand-picked fruit and that these apples were most plausibly contaminated by cattle or deer feces.

These limitations, however, in no way diminish the basis from which we make our recommendations. The 14 cases of the hemolytic uremic syndrome with one death underscore the seriousness of E. coli O157:H7 infection and the reality that unpasteurized apple juice, even when made at a state-of-the-art plant, carries a risk for infection. Previous outbreaks of E. coli O157:H7 associated with fresh juice have involved fewer cases, have been confined to smaller geographic areas, and have not prompted any federal regulatory action. The magnitude and severity of this outbreak, however, prompted a series of meetings with farmers, fresh juice processors, consumers, and representatives from public health and regulatory agencies to discuss the safety of fresh fruit juices. Ultimately, the meetings led the FDA to propose two new regulations. The first, implemented on 5 November 1998, requires unpasteurized fruit and vegetable juices to carry a label stating, "Warning: This product has not been pasteurized and therefore may contain harmful bacteria that can cause serious illness in children, the elderly, and persons with weakened immune systems" (28). The second regulation would mandate application of Hazard Analysis and Critical Control Point (HACCP) principles to fruit and vegetable juice processing (29). A HACCP program is a systematic method of identifying key production steps in which contamination may occur and instituting specific monitoring and interventions, such as pasteurization, at those steps. These regulations will greatly improve the safety of fruit and vegetable juices.

The public and regulatory responses to this outbreak reflect broad concerns about the safety of the food supply. The change of the U.S. diet to a heart-healthy diet rich in fruits and vegetables and the globalization of the food supply have changed the epidemiology of foodborne disease in the United States (30). Although consumers demand that their foods be completely safe, they also desire less processed foods, more choice, and fresh produce year-round. The increasingly international food supply brings more choice and more risk, as demonstrated by the widespread outbreaks of cyclosporiasis from imported raspberries (31) and hepatitis A from imported strawberries (32). Food distribution systems are increasingly complex; a nationally distributed food product with a low level of contamination can cause widely dispersed foodborne outbreaks that are difficult to detect (33). After an outbreak is detected, it is frequently difficult to trace the pathogen back to a single source if the food item is composed of products from multiple sources, as was the case with tracing contaminated apple juice back to many apple farms.

Continued vigilance on the part of clinicians who diagnose enteric infections and collaboration among public health agencies that detect and investigate outbreaks of foodborne illness will be necessary to identify emerging pathogens, such as E. coli O157:H7, and uncommon vehicles for contamination, such as unpasteurized apple juice. Until the safety of fresh fruit juices can be assured, persons who want to reduce their risk for E. coli O157:H7 infection and other enteric infections should consume only pasteurized juices. This is especially important for children and for other persons who are at increased risk for severe sequelae of E. coli O157:H7 infection.