Outbreak of Salmonella enterica Goldcoast infection associated with whelk consumption , England , June to October 2013

T Inns (thomas.inns@phe.gov.uk)1,2,3, G Beasley1, C Lane4, V Hopps5, T Peters6, K Pathak6, R Perez-Moreno7, G K Adak4, A G Shankar1, on behalf of the Outbreak Control Team 1. Anglia and Essex Centre, Public Health England, United Kingdom 2. Field Epidemiology Training Programme, Public Health England, United Kingdom 3. European Programme for Intervention Epidemiology Training (EPIET), European Centre for Disease Prevention and Control (ECDC), Stockholm, Sweden 4. Centre for Infectious Disease Surveillance and Control, Public Health England, United Kingdom 5. Kings Lynn & West Norfolk Borough Council, United Kingdom 6. Microbiology Services, Public Health England, United Kingdom 7. Incidents Unit, Food Standards Agency, United Kingdom


Rapid communications
Outbreak of Salmonella enterica Goldcoast infection associated with whelk consumption, England, June to October 2013

Identification of outbreak
During September 2013, 17 laboratory-confirmed cases of Salmonella enterica serotype Goldcoast infection with gastroenteritis were reported in England. This number was greater than the annually expected number: in 2012, eight cases were reported in England, in 2011, five were reported and in 2010, 13 were reported. Public Health England initiated an investigation on 12 September 2013 in order to identify the source of the outbreak and enable suitable control measures to be put in place to prevent further cases. The last case, with symptom onset in October, was reported on 12 November 2013, bringing the number of reported cases to 38. The cases' ages ranged between six months and 83 years (median: 64 years); two were aged under 16 years. Of the 38 cases, 25 were male. All cases were resident in England, predominantly in the east of the country ( Figure 1).

Epidemiological investigation
Outbreak cases were defined as persons resident in England diagnosed with Salmonella Goldcoast infection by the Salmonella Reference Service (Public Health England, London) after 1 June 2013. All laboratories in England and Wales send all Salmonella samples to this reference service.
To generate a hypothesis as to the source of this outbreak, we undertook detailed telephone interviews of the first 10 cases, with a trawling questionnaire. The questionnaire included demographics, clinical details, information on travel and contact with symptomatic persons, events attended and a detailed food history (including a general seafood question) for the week before symptom onset, including venues eaten at, and types and origin of foods eaten at home. Of the three cases with the earliest symptom onset dates, two reported having eaten whelks. Whelk consumption was reported by five of the 10 cases interviewed; this was much higher than the expected level of whelk consumption (for gastroenteritis questionnaires routinely completed in the east of England, eating whelks is reported by less than 1% of cases). Therefore, our primary hypothesis was that illness was associated with eating whelks (Buccinum undatum).
We then undertook an unmatched case-control study to test the hypothesis that whelk consumption was associated with Salmonella Goldcoast infection in England in 2013. Cases were excluded from the study if they were aged under 16 years (for logistic reasons), had travelled outside the United Kingdom (UK) in the five days before onset of symptoms, had had close contact with other individuals with gastroenteritis in the five days before onset, were asymptomatic or had already been interviewed with the trawling questionnaire. Controls were recruited through a systematic digit dialling process, using the home telephone number of cases to generate telephone numbers of people who were then contacted. We initially aimed to recruit two controls per case, but subsequently reduced this due to the strength of association observed. Information was collected from participants using a pre-tested questionnaire that was administered over the telephone by trained investigators. The questionnaire included questions on the following: demographics, clinical details, travel history, contact with persons with diarrhoea, events attended and foods eaten, including location of purchase, in the week before symptom onset. Foods in addition to whelks that were eaten by at least eight cases in the trawling questionnaire were included in the case-control study.
Data from the questionnaire were entered into a database using EpiData 3.1. Data were checked and analysed using Stata v12.1. The association between illness and each variable was estimated using odds ratios (ORs) and 95% confidence intervals (CIs). Data were subjected to univariable analysis and stratification to test for effect modification and confounding together with multivariable analysis using logistic regression.

Figure 2
Cases of Salmonella enterica Goldcoast infection, by calendar week a of symptom onset, England, weeks 24-42 a 2013 (n=35) b a Week 24 started on 10 June 2013. b Onset dates were unobtainable for three cases.

Food chain investigation
Foods identified in the trawling questionnaire and case-control study were investigated by environmental health officers and the Food Standards Agency. From each point of sale, distributers and suppliers were traced. Links between suppliers were mapped to produce a food chain diagram.
Food and environmental samples were taken using appropriate media. Samples were tested for presence of Escherichia coli and Salmonella. Presumptive Salmonella isolates were screened using a real-time polymerase chain reaction (PCR) assay for identification of the most common subspecies of S. enterica [1,2]. Serum agglutination tests, using Kauffmann-White classification [3], confirmed the presence of S. enterica serovar Goldcoast (6,8:r:l,w).

Analytical epidemiology
In total, 22 cases were eligible for the case-control study: of these, 20 cases were included (two cases declined to participate). A total of 27 controls were included. In a univariable analysis, cases were significantly more likely than controls to have consumed whelks, cockles, lettuce, fish and peppers (Table). No effect modification was detected. In the final multivariable model, when adjusted for sex, cases were significantly more likely to have consumed whelks (OR: 109; 95% CI: 7.7-1,539).

Food and environmental microbiology
Whelk consumption was reported by 24 of the 38 cases and one of the 27 controls; venues that these cases reported purchasing whelks from were investigated by environmental health officers. A summary of the food chain inferred from these investigations is shown in Figure 3. The supplier could be traced for whelks eaten by 20 cases: all were traced back to whelks processed by the same factory (Factory X). The whelks eaten by the control were not supplied by Factory X.
Factory X is a seafood factory in England; in 2012 it processed 639,049 kg of whelk meat. Whelks are processed by cooking in a pressure cooker. Their shells are then crushed and removed before the meat is cooled in a water bath. A small proportion is then sold fresh, with the majority of cooked whelks being flash frozen before sale. Over 90% of cooked whelks from Factory X are shipped to a single non-European Union country for further processing and consumption; the remainder are sold in the UK.
A total of 11 samples of whelks that had been processed at Factory X were taken at the point of sale from four outlets; two of the samples were positive for Salmonella Goldcoast. Seven processed and two raw whelk samples were taken from Factory X: none were positive for Salmonella Goldcoast. We tested 33 swabs or water samples from whelk-processing machinery at Factory X: six tested positive for Salmonella Goldcoast.

Control measures
Following initial descriptive epidemiology and food tracing investigations, Factory X was visited by environmental health officers on 20 September 2013. Cooking temperatures could not be verified and the factory agreed not to produce ready-to-eat whelks until further notice. During a subsequent inspection on 23 September, inadequate product temperatures were recorded immediately following cooking and therefore the whelks processed on that day and stored frozen on the premises were kept on the site. Due to the problems identified in the processing of whelks, the factory   instigated a product recall on 23 September 2013. The effectiveness of this recall was monitored by the Food Standards Agency and associated local authorities. The Foods Standards Agency informed the competent authorities in the country that this product is exported to: to date, no cases have been reported.

Discussion
We present epidemiological, environmental, microbiological and food chain evidence, which all support the conclusion that this outbreak of Salmonella Goldcoast infection was associated with consumption of whelks processed by Factory X. Salmonella Goldcoast outbreaks have previously been associated with pork products (salami [4], pork cheese (cooked pig organs stuffed in a pig stomach) [5], French paté [6] and raw fermented sausage [7]), watercress [8] and hard cheese [9]. Whelks have previously been associated with toxin-based food poisoning [10] but to our knowledge, this outbreak is the first known report of bacterial food poisoning associated with whelk consumption.
Regarding the mechanism of Salmonella Goldcoast contamination, microbiological evidence suggests that production equipment in contact with cooked whelks was contaminated with Salmonella Goldcoast for a number of weeks, despite the use of a sanitiser and the cooked whelks passing through a highly saline bath. Salmonella Senftenberg has previously been observed to survive in high salinity environments [11]: it may be that Salmonella Goldcoast shares this characteristic.
There was limited evidence that could indicate the original source of the Salmonella Goldcoast contamination. One of the environmental samples that tested positive for Salmonella Goldcoast was a swab of the conveyer belt used to transport raw whelks to the cooker, indicating that it was present on at least some whelks before entering the factory. Whelks are not filter feeders, and it is unclear whether they ingested the bacteria or the pathogens were in water that contaminated the shells. On the basis of the epidemiological evidence, we hypothesise that contaminated whelks may have been produced over at least a three-month period, but this contamination may have been intermittent, at a consistently low level. This would account for the relatively small number of cases seen.
We hope to be able to undertake whole genome sequencing of all outbreak cases and environmental isolates in future. Providing suitable background isolates are sequenced, this should allow a higher level of discrimination within the cases to ascertain which may be an artefact of sporadic incidence. Cases who did not report whelk consumption may have been sporadic or secondary cases. Regarding the case who was six months-old, one possible explanation is that the infant may have been infected by a family member who had eaten contaminated whelks, but who was asymptomatic.
Of the 38 cases, 10 were hospitalised, of whom four were admitted to intensive care. This level of severity has not previously been reported for Salmonella Goldcoast. Salmonella incidence is usually highest among the youngest (aged 0 to 4 years) [12]. As these cases had a median age of 65 years, they may have been more likely to have co-morbidities that increase the risk of hospitalisation.
One limitation of this study was that memory recall may have been different in case and control groups. To minimise this, cases and controls were interviewed in as timely a fashion as possible. Another limitation was that by using sequential digit dialling, a preponderance of females were recruited (20/27), whereas 25/38 of the cases were male. Explicit or frequency matching of controls was not possible due to the resources available for the investigation. To address this, sex was adjusted for in the analysis. It should also be noted that due to the small number of cases and controls, the CIs around the estimates of effect were wide.
We consider that the measures put in place to control this outbreak were effective in preventing further cases in England. Whelks are a novel vehicle of Salmonella infection and should be considered in the investigation of future outbreaks. It is known that processed whelks are sold internationally, and so if contaminated, there is the potential for cases to occur in countries outside the UK.
TI designed the case-control study, analysed the data, drafted the manuscript and coordinated the outbreak response. GB organised data collection, food chain and environmental investigations. CL interpreted epidemiological and microbiological results. VH contributed to the environmental investigation. TP and KP were responsible for the microbiological results. RM coordinated the food chain investigation. GKA provided advice for the case-control study and the manuscript. AGS led and chaired the Outbreak Control Team, coordinated the outbreak response and advised on the manuscript. We report a brown hare (Lepus europaeus) infected with Francisella tularensis, the bacterium causing tularaemia, in the Netherlands in May 2013. This is the first case of tularaemia in Dutch wildlife since 1953. The finding results from the intensified surveillance of the disease in brown hares, which started in July 2011 after infected hares were detected in neighbouring countries.

Rapid communications
Tularaemia, caused by the bacterium F. tularensis, is a zoonotic disease that was reported in more than 800 humans in the European Union in 2010 [1]. The bacterium has a very wide host range that includes mammals, birds, amphibians, fish, and invertebrates [2]. It can remain infectious in water and mud for months [3]. It can be transmitted by inhalation of infective aerosols, contact with or ingestion of infected hosts or water, and arthropod bites [2].
Four subspecies can be distinguished: tularensis, holarctica, mediasiatica and novicida. The first two subspecies are important causes of tularaemia in humans and animals [2]. The bacterium has been detected in wildlife in various European countries, such as Denmark, Finland, France, Germany, Spain, Sweden and Switzerland in 2012; Belgium, Italy and Norway in 2011; and Austria in 2009 ( Figure 1) [4].
Lagomorphs and rodents are most susceptible to infection and disease by the bacterium [5]. In a number of European countries, brown hares are considered to be an important host of F. tularensis and transmission to humans is known to result from direct contact with hares [6,7].
In the Netherlands, the agent was last reported in 1953 when seven members of an eight-person family became ill after consuming a brown hare [8,9]. In contrast, in Lower Saxony, a German federal state that shares a common border with the Netherlands, F. tularensis has been detected in 2.9% of hares found dead as recently as in the period between 2006 and 2009 [10]. In addition, in the autumn of 2011 tularaemia was found in hares in Düren, a municipality in North Rhine-Westphalia, Germany, about 50 km from the eastern Dutch border [11], as well as in Anthisnes, a municipality in the Province of Liege, Belgium, approximately at the same distance from the southern Dutch border [12].
These recent reports suggested that F. tularensis may also be present in the Netherlands without being detected. Therefore the Dutch Wildlife Health Centre (DWHC) and the Central Veterinary Institute (CVI) in collaboration with the National Institute for Public Health and the Environment (RIVM) decided to intensify surveillance for tularaemia in brown hares in the Netherlands.

Finding of a brown hare testing positive for Francisella tularensis
Brown hares that are found either dead or terminally ill and then euthanised by hunters or game wardens can be submitted to the DWHC for post-mortem examination in the context of non-targeted surveillance. Since July 2011, these hares have been routinely tested for the presence of F. tularensis DNA by polymerase chain reaction (PCR) at CVI. DNA was extracted from lung and/or spleen using a DNA tissue kit (DNeasy Blood and Tissue Kit; Qiagen, Hilden, Germany). The extracted DNA was tested by TaqMan real-time PCR using the FTT0376 primers and probe published by Mitchell et al. [13] (forward primer 5'-CCATATCACTGGCTTTGCTAGACTAGT-3', reverse primer 5'-TGTTGGCAAAAGCTAAAGAGTCTAAA-3', probe 5'-FAM-A A AT TATA A A ACC A A ACCC AGACC T TC A A ACC AC A -BHQ1-3'). This assay is specific for the pathogenic subspecies of F. tularensis (subspecies tularensis, holarctica and mediasiatica). The positive samples were also sent to the Swedish Veterinary Institute (SVA), Uppsala, Sweden for confirmation.
By May 2013 a total of 49 animals from nine of the 12 Dutch provinces had been examined ( Figure 2). Of the 49 specimens, 26 had one or more macroscopic or microscopic lesions consistent with tularaemia in this species [14]. The first 48 hares tested negative for the presence of F. tularensis DNA by PCR. The 49th hare examined in May 2013 was an adult male hare from the province of Limburg. Prior to death, the animal had been seen with an unsteady gait, had been reluctant to move and was easy to catch.
The slightly autolytic carcass of this animal had an enlarged spleen at necropsy and histopathology revealed multiple foci of hepatocellular necrosis, consistent with F. tularensis infection [14]. Real-time PCR analyses of spleen and lung samples of this specimen were positive for F. tularensis. Culture of the samples from this animal on chocolate agar medium with cysteine and sodium sulphite provided negative results. Infection by F. tularensis was confirmed at the SVA, both by PCR (spleen) and immunohistochemistry (lung). The subspecies of F. tularensis was subsequently typed by CVI as holarctica based on the concatenated partial sequences of five metabolic housekeeping genes as described by Nübel et al. [15].

Discussion and conclusion
F. tularensis ranks among the top twenty emerging zoonotic pathogens considered to be relevant for the Netherlands [16]. The emergence or re-emergence of the disease in other countries has been associated with factors such as climate change, human-mediated movement of infected animals, as well as with conditions of war with subsequent increase in rodent populations. In some cases, detection due to enhanced surveillance revealed the presence of the disease [2,17]. Enhanced surveillance also likely contributed to the apparent re-emergence of tularaemia in Dutch wildlife after 60 years, as reported here.
The subspecies F. tularensis holarctica detected in this study is consistent with the subspecies detected in wildlife in the neighbouring countries [10]. The infected hare was found only 6 km away from the home of the family in the 1953 report [8,9] (Figure 2). It is unclear as to how widespread the occurrence of the bacterium is in wildlife in the Netherlands and therefore whether the proximity of both events indicates a hot spot or a coincidence. Heightened surveillance is needed in order to answer this question.
Given the proximity of these cases to the border, emergence due to import of the disease from neighbouring countries should also be considered. Indeed, in 2012 four cases of tularaemia in hares were identified in the area of Heinsberg, Hückelhoven and Erkelenz in Germany, 10 km from the Dutch border and 30 km from the case reported here [18]. It is unlikely that hares are deliberately introduced from abroad into the Netherlands, since release of hares is illegal. It is also unlikely that the specific hare found infected in this study came from abroad on its own, as hares do not usually cover such distances and in addition, a large river separates the two locations [19,20]. However, the infection may have moved more gradually into the Dutch area through other hosts or vectors or both without having been detected.
Only two human tularaemia cases likely to be autochthonous have been recorded in the Netherlands since 1953, though human infection was notifiable from 1976 to 1999 [9,21,22]. One case occurred in 2011 and one in 2013, and neither had a history of contact with dead hares or other animals [21,22]. However, it is presumed that the 2013 case may have contracted the disease in Limburg through insect bites [22]. Both, these human cases and the hare case, highlight the importance of raising the awareness of physicians and veterinarians with regards to the disease.

Figure 2
Location and year of sampling of hares with polymerase chain reaction test results for Francisella tularensis (n=49) relative to where an infected hare was found in 1953, Netherlands, 2011-2013 In the legend, the terms 'negative' and 'positive' are used to describe the sampled hares' polymerase chain reaction test results for Francisella tularensis. These findings support the continuation of non-targeted disease surveillance in hares and other wildlife species, and heightened targeted surveillance for tularaemia, with focus on the affected regions. Hunters, game wardens, and other groups that are likely to come into close contact with wildlife will be informed and included in these activities.

Rapid communications
Case of vaccine-associated measles five weeks postimmunisation, British Columbia, Canada, October 2013  We describe a case of vaccine-associated measles in a two-year-old patient from British Columbia, Canada, in October 2013, who received her first dose of measlescontaining vaccine 37 days prior to onset of prodromal symptoms. Identification of this delayed vaccineassociated case occurred in the context of an outbreak investigation of a measles cluster.
In this report we describe a case of measles-mumpsrubella (MMR) vaccine-associated measles illness that was positive by both PCR and IgM, five weeks after administration of the MMR vaccine. Based on our literature review, we believe this is the first such case report which has implications for both public health follow-up of measles cases and vaccine safety surveillance.
Between 29 August and 2 September 2013, three unlinked persons from across the Fraser Valley, British Columbia, Canada, presented with rash illness consistent with clinical measles [1]. Based on the outbreak investigation by the local health authority, none of the three cases had an identified exposure to a measles case or travel history outside of Canada during the incubation period, and a source case was never identified. All three cases had the same measles genotype B3 sequence type (MVs/British Columbia. CAN/34.13, MeaNS id 39928, GenBank accession numbers KF704002 and KF704001). Measles genotype B3 is endemic in the World Health Organization's African and Eastern Mediterranean regions [2]. Two additional cases of measles due to secondary transmission from one of the above cases were identified in British Columbia in the third week of September.

Case report
In early October 2013, a two-year-old child living in the Fraser Valley presented to the family physician with fever, rash, conjunctivitis and coryza. Symptoms had begun two days before, with a runny nose, followed by fever on the day hereafter. A macular rash appeared on the day of visiting the physician, starting on the face and progressing to the rest of the body; fever measured by the parents was at 39 °C. Clinical examination of the child by the family physician found a fever of 39.5 °C, marked bilateral conjunctivitis, and macular rash over the body. Three days later, fever had dissipated, rash was fading and symptoms resolved without complications.
Public health alerts had been issued to community physicians regarding the recent cluster of measles in September, which may have raised suspicion for measles in this case. Additionally, the child's family was aware of measles cases in the community from a relative who attended the same church as one of the original cases, but no direct link was identified and they had no travel history outside of Canada. Contact investigation revealed no ill household members or preschool contacts. The child's past medical history indicated anaphylaxis to peanuts and eggs. Primary series of immunisations were not up-to-date, as she had just received her first dose of MMR vaccine 37 days prior to the onset of illness. At the same visit, the child had received meningococcal C and pneumococcal conjugate vaccines.

Laboratory investigations
Laboratory testing for measles was performed on specimens collected on the day of rash onset. Measles RNA was detected in the nasopharyngeal swab by the RT-PCR assay [3]. Acute and convalescent measles specific IgM and IgG antibodies were detected in the blood by ELISA ( [4]. Other virology testing found no detectable Parvovirus B19 specific IgG or IgM antibody, and detectable human herpesvirus (HHV)-6 specific IgG antibody but no detectable HHV-6 DNA.

Public health measures
While genotyping results were pending, case management proceeded as for a wild-type measles infection. Public health follow-up lead to the identification of 87 contacts. As per guidelines, post-exposure prophylaxis was provided within six days of exposure to 45 susceptible contacts (41 contacts with a history of one dose of MMR vaccine received an additional MMR dose, and four contacts with no history of MMR vaccine or with contraindications to MMR vaccination, received immunoglobulin) [1]. All contacts received education on signs and symptoms of measles, and those who received immunoglobulin were recommended to subsequently receive MMR vaccine, if this was not contraindicated.

Discussion
The incubation period of measles is typically eight to 12 days from exposure to rash onset, with a range from seven to 21 days. Public health interventions are based on this established incubation period for determining the epidemiological links between cases and for estimating periods of exclusion for contacts in high risk settings [5,6]. Based on our review of the literature, this report documents the first case of MMR vaccineassociated measles, 37 days post-immunisation, well beyond 21 days and the routine 30 days post-MMR immunisation period used by the Canadian adverse event following immunization (AEFI) surveillance system.
Measles-containing vaccines are used globally, have been part of the British Columbia immunisation schedule since 1969, and have an impressive record of safety validated by careful, ongoing AEFI surveillance. Rash and/or mild clinical illness following MMR vaccine are not uncommon [7]. Clinically significant vaccine-associated illness is rare, but when it occurs it is indistinguishable from wild-type measles, except by genotyping [8]. Detection of vaccine virus has been documented up to 14 days post-immunisation by RT-PCR, and up to 16 days by immunofluorescence microscopy of urine sediment [9][10][11][12]. Complications from vaccine-associated measles have been documented in both immune-competent and compromised individuals [13,14]. Of note, only one case report of transmission from vaccineassociated measles has been identified [15,16].
Possible explanations for this prolonged shedding of measles vaccine virus include interference with the immune response by host or vaccine factors. Immunoglobulin administration early in the incubation period has been reported to extend the time to onset of symptoms, but in this child there was no such history and no known immunosuppressive illness [5]. The two-fold rise between acute and convalescent measles-specific IgG suggests the vaccine-mediated immune response had been underway prior to the onset of symptoms. Investigations clarified that there were no shipping, handling or cold-chain deviations for the specific vaccine used, and that it was administered by a public health nurse trained in immunisations. The potential immunological impact of the older age of the child at the time of receiving the first dose of MMR vaccine, 33 months versus the typical 12-15 months of age, and the co-administration of meningococcal C and pneumococcal conjugate vaccines are areas for future investigation.
It is possible that the case's symptoms were not measles-vaccine-related but an inter-current illness confounding the presentation. However, symptoms of marked conjunctivitis, continued fever with rash, and progression of macular rash from face to the whole body, are all more suggestive of measles versus other exanthems caused by viral diseases. Parvovirus and HHV-6 results were negative, and the absence of intake of medications excludes a drug reaction. Rubella serology was not done as it was expected to be positive given the recent MMR vaccine administration. Therefore, the combination of classic measles symptoms, detection of measles vaccine virus and reactive measles IgM, and lack of evidence of an alternative illness explanation, were highly suggestive of measles vaccine-associated illness.
Heightened surveillance and awareness of measles because of the ongoing outbreak likely contributed to the identification of this case. Although this is the first such reported case, it likely represents the existence of additional, but unidentified, exceptions to the typical timeframe for measles vaccine virus shedding and illness. Such cases have important public health implications for the investigation of measles clusters because while there is uncertainty about case classification (wild-type vs vaccine-type), case and contact management should proceed as if for wild-type to prevent secondary transmission. In this case, uncertainty from the presence of a measles outbreak, symptom onset on day 37 after MMR vaccine administration, and a two-week period between the RT-PCR findings and genotype determination, resulted in the initially reasonable presumption that this was a wild-type measles case and subsequent resource-intense follow-up of contacts. Awareness of the frequency of such exceptions to the typical measles timeframe and improving the timeliness of measles vaccine virus genotyping could help focus public health resources on cases of wild-type measles. Further investigation is needed on the upper limit of measles vaccine virus shedding based on increased sensitivity of the RT-PCR-based detection technologies and the immunological factors associated with vaccine-associated measles illness and virus shedding.

Surveillance and outbreak reports
A Shigella sonnei outbreak traced to imported basil -the importance of good typing tools and produce traceability systems, Norway, 2011 Introduction Shigellosis is endemic throughout the world. Symptoms are usually mild but range from watery, selflimiting diarrhoea to life threatening dysentery [1]. Of four Shigella species, S. sonnei is the most frequently isolated in industrialised countries [2]. Symptoms of S. sonnei infection are usually milder than those caused by S. dysenteriae or flexneri [3]. The bacteria are transmitted by ingestion of contaminated food or water, or through person-to-person contact. The incubation period ranges from 12 hours to one week [4]. The infective dose is very low: ingestion of 100 to 200 microorganisms can lead to disease [3].
In the European Union (EU) shigellosis infections are relatively uncommon. With a rate of 1.64 cases per 100,000 population in 2010, they are far less frequent than Campylobacter and Salmonella infections, which have respective incidences of 56.95 and 21.31 per 100,000 population [5]. In Norway, between 120 and 190 cases of shigellosis have been reported annually in the last ten years, corresponding to an incidence between 2.5 to 4.0 per 100,000 population. Only 10 to 20% of the cases are domestically acquired [6]. Imported fresh vegetables have been identified as the vehicle of several outbreaks over the last years [7,8].
In order to control imported feed and food of non-animal origin, the European Commission Regulation (EC) No 669/2009 specifies a list of risk products subjected to increased level of official controls upon entry into the European Economic Area, which includes Norway [9].

The alert
On 8 October 2011, clinicians at the University Hospital of North Norway in Tromsø, northern Norway (2011 population: 68,200 inhabitants) [10], attended six patients with bloody diarrhoea. On 9 October, the hospital's Department of Microbiology and Infection Control confirmed three patients with S. sonnei infection. None of them had a travel history outside or within Norway in the previous week. The microbiologist on call reported the cluster to the Municipal Medical Officer in Tromsø and to the Department of Infectious Diseases Epidemiology at the Norwegian Institute of Public Health (NIPH), and isolates were forwarded to the National Reference Laboratory for Enteropathogenic Bacteria (NRL). The NRL verified the isolates as being S. sonnei with an identical multilocus variable-number tandem repeat analysis (MLVA) profile that had not been identified in Norway before. Concurrently, the Local Food Safety Authority in Tromsø interviewed the patients on food consumption, who reported having eaten at delicatessen X in downtown Tromsø or having participated in social events with food provided by delicatessen X during the week before becoming sick. In addition, the owners of delicatessen X, who were also interviewed, had received complaints from customers who had fallen ill.
On 14 October, the Municipal Medical Officer in Sarpsborg (2011 population: 52,800) [10], 1,700 km south of Tromsø, notified a second cluster of shigellosis, whereby none of the patients had a travel history to Tromsø. S. sonnei isolates had the same MLVA profile as those from Tromsø.
Since more than one county was affected, the further coordination of the investigation was transferred to the national level. NIPH, in collaboration with the Norwegian Veterinary Institute, the Norwegian and Local Food Safety Authorities, the Municipal Medical Officers of the municipalities involved, and the Department of Microbiology and Infection Control at the University Hospital of North Norway, investigated the outbreak to identify the source, implement control measures and prevent further cases.

Epidemiological investigation
A case was defined as (i) a person in Norway with laboratory-confirmed S. sonnei infection after 1 October 2011 with the MLVA profile identified in the outbreak with absence of travel history abroad, or (ii) a person who had an isolate with one-locus difference from the MLVA outbreak profile and an epidemiological link to (i). The Local Food Safety Authority interviewed cases in Tromsø by telephone using a standard food-borne disease trawling questionnaire to generate hypotheses about common exposures among cases. Once the suspicion was focused towards Delicatessen X, their menu was used as basis for the interviews.
In order to gather more information on the second cluster, NIPH interviewed all cases in Sarpsborg by telephone using the same food-borne disease trawling questionnaire looking for common exposures among them and to those in Tromsø.

Cohort study in Tromsø
Delicatessen X provided a list of social events they catered for from 30 September to 8 October, including a banquet with 50 guests in Tromsø on 1 and 2 October. Since the organiser of the banquet had reported to the delicatessen that some of them had fallen ill, and this event included participation of a greater number of participants, the NIPH studied a cohort among the banquet attendees to identify risk factors for disease. For the cohort study we defined a 'probable case' as a person who developed diarrhoea (more than three loose stools in 24 hours) and fever (self-reported) up to seven days after the banquet.
On 14 October, the NIPH sent a link to a web-based questionnaire via e-mail to the attendees. It contained questions on demographic information, symptoms and food eaten. NIPH attempted to interview persons who had not replied within five days by phone. Attack rates and relative risks with 95% confidence intervals (CI) were calculated. Variables with p<0.1 in the univariate analysis were included in a multivariable logistic regression model, using. STATA 11.0 (Stata Corporation, College Station, TX, USA).

Microbiological investigation
The Department of Microbiology and Infection Control at University Hospital of North Norway identified the initial isolates as S. sonnei by fermentation tests, agglutination and Vitek 2 automated identification. They sent them to the NRL, where all Shigella spp. isolates identified in Norway are received for identification to species-level, O-serogrouping, antimicrobial resistance testing and MLVA-typing [11]. The Norwegian Veterinary Institute analysed, using a polymerase chain reaction (PCR)-based method elaborated by the Nordic Committee on Food Analysis, the food items served by Delicatessen X in Tromsø and those served in Sarpsborg. Specimens were analysed for Shigella spp., Enterobacteriaceae and thermotolerant coliforms [12][13][14].

Trace-back investigation
The Norwegian Food Safety Authority performed trace back of products and inspection of the premises where the food likely associated to this outbreak was distributed, prepared and served. The supplier of the relevant food items was contacted to document and provide an overview of the supply chain process.

Epidemiological investigation
Forty-six cases with identical MLVA profile were reported: 42 cases linked to Tromsø and four to Sarpsborg ( Table 1). None of them reported travel outside Norway during the week prior to the onset of symptoms.
Of the Tromsø cases, all were diagnosed in Tromsø, with the exception of one who was diagnosed in Oslo, but reported travel to Tromsø in the previous week ( Figure 1). The cases in Tromsø had isolates collected and tested for gastrointestinal pathogenic bacteria between 5 and 21 October and those in Sarpsborg between 11 and 25 October ( Figure 2).

Cluster in Tromsø
The median age of all 42 cases in Tromsø was 41 years (range: 19-84 years); twenty-four of the cases were female ( Table 1). The first case of the outbreak sought medical attention and was tested on 5 October and the last one on 21 October (Figure 2). In total, four patients were hospitalised. All of them were admitted during the first days of the outbreak and had bloody diarrhoea; fever and abdominal pain, with a mean C-reactive protein of 234 mg/L (range: 120-364 mg/L; norm <10mg/L). The mean length stay in hospital was 2.8 days (range: 1-4 days). Three of the hospitalised patients received antibiotic treatment and all the admitted patients recovered well.
The Local Food Safety Authority interviewed 38 of the 42 cases: 37 had eaten food containing pesto sauce made with fresh basil from the Delicatessen X.

Cohort study in Tromsø
Forty-two of the fifty banquet attendees answered the web-based questionnaire. Eleven met the probable case definition (attack rate: 26%). All of them had diarrhoea. Frequent symptoms were also abdominal pain (8 persons) and fever and nausea (7 persons). The highest attack rate occurred among those aged 20 to 29 years (4/11; 36%) although there were persons from all ages affected. Both sexes were equally affected.
Six probable cases sought medical attention and three of them had a stool sample taken and were laboratory confirmed with the S. sonnei outbreak strain.
Ten banquet food items were significantly associated with disease in the univariate analysis ( Table  2). Attendees exposed to basil pesto sauce had the The case diagnosed with Shigella sonnei infection in Oslo, reported travel to Tromsø in the week before being diagnosed.
highest attack rate (78%). Among the three food items with the highest relative risks (RR), basil pesto sauce had the smallest confidence interval. The lower limit of the confidence interval was higher than for any other item (RR: 5.4; 95% CI: 2.1-14.4).

Cluster in Sarpsborg
Four cases were reported from Sarpsborg ( Figure 1). The first case sought medical attention and was tested for gastrointestinal pathogenic bacteria on 11 October and the last one on 25 October (Figure 2

Microbiological investigation
The NRL received 48 S. sonnei isolates from stool samples during October 2011. Of these, 46 were confirmed by microbiological characteristics to be part of the outbreak. The MLVA-profiles of these isolates were identical (44 isolates) or with one locus difference (2 isolates). This MLVA-profile differed to a great extent from earlier profiles available in the NRL database including approximately 600 isolates and 405 distinct profiles.
The Norwegian Veterinary Institute analysed 20 food specimens from Tromsø and Sarpsborg consisting of diverse vegetables, fresh herb spices, fruits, nuts, herb dressings (including basil for pesto) and spiced butter. All food specimens analysed were negative for Shigella spp., but harboured high Enterobacteriaceae counts with relatively low levels of thermotolerant coliforms. One basil pesto product in particular originating from Delicatessen X had high levels of both Enterobacteriaceae and thermotolerant coliform counts.

Trace-back investigation
The Norwegian Food Safety Authority identified that the same supplier had provided fresh basil both to Delicatessen X in Tromsø and to restaurant Y in  11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

Discussion
The results of the trawling interviews with the laboratory-confirmed cases in Tromsø made us hypothesise that an ingredient used in pesto served in Delicatessen X could be the source of the outbreak. The results of the cohort study among the banquet attendees reinforced this hypothesis: eating basil pesto was an independent risk factor for disease and had the highest attack rate among exposed. In addition, basil pesto was the only common food item that had been consumed by the other laboratory-confirmed cases occurring in Tromsø and Sarpsborg. The findings from the cluster investigation in Sarpsborg strongly supported fresh basil as the vehicle ingredient of S. sonnei. The role of the other food item highlighted in the cohort study, fish soup, remains unclear. We considered whether an ingredient of the basil pesto could also be part of the fish soup.
The hypothesis was rejected as since the soup was not made by Delicatessen X, no common ingredients were used. None of the ingredients used in the soup had been eaten by the other laboratory-confirmed cases in Tromsø. The role of the food handlers in a potential cross contamination of the two food items remains unclear.
This outbreak, with 46 laboratory-confirmed cases, is the second largest shigellosis outbreak reported by 2013 in Norway [15]. A larger S. sonnei infection outbreak occurred during 1994 and affected several countries in Europe, including Norway, Sweden and the United Kingdom. In the 1994 outbreak, there were 110 laboratory-confirmed cases within Norway and investigations traced it to imported iceberg lettuce [8].

Table 2
Univariate analysis of foods to which probable cases of Shigella sonnei infection (n=11) were exposed at a banquet in Tromsø, Norway, 1-2 October 2011 Food items Exposed For each food exposure, there were between one and 10 attendees missing a response because they did not recall having consumed or not a given food item. Several shigellosis outbreaks reported in Scandinavian countries have been associated with imported fruits or vegetables consumed raw or minimally-processed [7,16]. These food items might become contaminated during preparation by infected food handlers or during production by irrigation water contaminated with sewage [2]. The low infective dose and the considerable amount of fresh basil as an ingredient in pesto may have contributed to the large number of people becoming sick after eating basil pesto from delicatessen X in Tromsø, despite the growth-inhibitory effect of fresh herbs like basil or thyme on S. sonnei reported by some studies [17]. In this outbreak few affected individuals were admitted to hospital and no patients reported serious extra-intestinal symptoms. A noteworthy high C-reactive protein in affected patients has also been reported in previous studies of shigellosis [18].
Despite the epidemiological evidence which seemed to conclusively identify basil as the likely source of the outbreak, none of the specimens were positive for Shigella. Detection of Shigella spp. in food items is difficult and no reliable method is available. High levels of both Enterobacteriaceae and thermotolerant coliform counts were obtained from a suspected pesto product. This indicates faecal contamination and makes contamination also by Shigella more likely.
The Norwegian importer decided to temporarily stop importing basil from the exporter upon the identification of the batch. The exporter went bankrupt, so no decision on when to resume importation was necessary. It is unclear at which point in the process of cultivation, production and importation of the basil the S. sonnei contamination may have occurred. Currently, basil from certain third countries outside EU/European Free Trade Association (EFTA), as Israel, are not included in the European Commission regulation (EC) No 669/2009 list of certain feed and food of non-animal origin subjected to increased level of official controls on imports. As a result of this investigation, the Norwegian Food Safety Authority, together with NIPH, planned to develop recommendations for food providers on how to handle fresh plant produce prior to consumption.
The multidisplinary collaboration during this investigation helped to identify and find the source of this outbreak of S. sonnei infection in Norway: The routine genotyping of all isolates of enteropathogenic bacteria in Norway was crucial to determine that the two clusters happening in two regions of the country were part of the same outbreak. The epidemiological and product trace-back investigations pointed to imported fresh basil as likely causing the outbreak.