The investigation focused on all UK passengers who had been on the cruise at any point from January through March 2008. Contact addresses were provided by the cruise ship company, and 2,850 passengers were sent letters inviting them to participate in the investigation and explaining why. On the basis of when they were most likely to have been exposed (ascertained from the first 4 cases), participants were asked to go to their own doctors to give blood samples within 2 weeks (the time frame for detection of immunoglobulin [Ig] M). HPA provided sample kits with prepaid return packaging. Blood samples were tested for HEV antibodies (IgG and IgM) by using the Fortress Diagnostics ELISAs (Fortress Diagnostics Limited, Antrim, Northern Ireland) at the Virus Reference Department at the HPA Centre for Infections. Assays were run in accordance with the manufacturer’s instructions. The Fortress assays were chosen for this investigation because our validation exercises (data not shown) had demonstrated these assays to be more sensitive and specific than some other commercially available assays. Samples were screened for IgG, and those that were positive were then tested for IgM. The IgM-seropositive samples were further analyzed for HEV RNA, and those that were RNA positive were genotyped as previously described (7). Briefly, phylogenetic analysis of a 300-bp region of open reading frame 2 was conducted. The generated sequences were compared with genotype 3 sequences from the United Kingdom, Europe, and the United States and with genotype 1, 2, and 4 sequences retrieved from GenBank.

Participants returned self-completed questionnaires by mail. Detailed information was collected about demographic characteristics, potential risk factors and cofactors for disease (medical conditions, food and drink consumption, excursions, water exposure such as water activities in pools on the cruise ship and swimming in the sea while off the ship) and any relevant signs and symptoms. All HPA documents and files containing patient identifiable information were handled and stored in compliance with Caldicott guidance (8).

Participants were classified according to their serologic results as having had recent acute infection (serologically confirmed by HEV IgM and IgG), past infection (serologically confirmed by HEV IgG only and therefore unlikely to have been acquired during the cruise), or no infection (serologically negative for HEV IgG). Those with recent acute infection also provided further blood samples for liver function testing and confirmatory HEV antibody testing. Patients with acute cases were followed up by the local Health Protection Unit of the HPA. Liver function tests were performed by local services, and results were reported back to the HPA on a specific form. Blood samples for HEV testing were returned by mail as described above. A symptomatic hepatitis case was defined as recent acute infection in a patient with signs and symptoms compatible with HEV infection, e.g., jaundice and/or dark urine and pale feces. To ensure no false-positive results, additional testing was conducted on samples taken at least 1 month later from these IgM-seropositive participants.

We compared risk factors and exposure for those with recent acute infection with those for seronegative controls. Persons who had evidence of past infection were excluded because they had probably been immune during the study period. Single and multivariable logistic regression, using Stata statistical software, release 10.1 (StataCorp, College Station, TX, USA), was performed to identify the most likely cause of the outbreak and to estimate time and place of exposure. Specific exposures with estimated odds ratios (ORs) >1, p<0.2, and at least 50% of cases of recent acute HEV infection, were included in a multivariable model. The least significant factor was dropped from the model in a stepwise fashion until all remaining exposures exhibited a significant association: p<0.05 and OR >1.

After the multivariable model was finalized, we added each port visited, 1 at a time, to identify where participants may have been exposed to HEV. The interaction between food item and location was also considered by using a stricter selection criterion of significance level p<0.01.

Of the 2,850 passengers, >1,100 volunteered and 851 (30%) participated in the investigation (Figure). Blood samples and questionnaires were available from 659 participants. Of these 659, age range was 22–92 years old (mean age 68 years); >90% participants were 55–79 years of age. Of 789 participants who gave blood samples, 426 (54%) were female and 363 (46%) were male.

This hepatitis E outbreak among passengers during a 3-month world cruise was reported at the end of the cruise, after passengers had already disembarked and returned home. Nevertheless, the fact that approximately one third of eligible passengers were able to give blood samples within a short time enabled detection of an acute antibody response. The testing algorithm was adopted on the basis of the onset dates of the first 4 cases and the expected time delay between contacting the passengers and actually receiving samples in the laboratory for testing. It was therefore thought to be unlikely that screening for HEV RNA would provide a useful marker of recent acute infection.

One study limitation was that participants were self-selected and may not therefore represent a random sample of passengers on the cruise. Also, the time between the cruise and completion of questionnaires may have made recalling foods consumed during the cruise difficult. However, most returned questionnaires were completed comprehensively, leaving no reason to suspect differential recall between case-patients and others.

Evidence of recent acute hepatitis E infection was found for 33 participants. The evidence of past infections for 162 (21%) is consistent with hepatitis E seropositive rates of ≈25% of UK residents >55 years of age (9). Only 11 participants with acute infection reported illness compatible with hepatitis; 22 (two thirds) were asymptomatic or had unrelated symptoms. This investigation provided a unique opportunity to diagnose asymptomatic infection in an exposed group and found a much higher asymptomatic rate than previously reported (≈3%–4%) (3,10). Elevated liver function test results appeared to be associated with symptomatic cases; however, because those without apparent clinical signs were tested a longer time after exposure, their liver function could have returned to within reference limits.

HEV RNA was detected in only 3 case-patients, suggesting that RNA had cleared by the time most samples were tested. Virus RNA sequences were identical and belonged to genotype 3, suggesting a common-source outbreak. Genotype 3 is the main type of HEV found in industrialized counties, including the United Kingdom. Although it has also been reported in North America, Southeast Asia, Australia, and New Zealand (4), the genotype 3 virus found in this outbreak had close sequence homology with genotype 3 strains reported in Europe.

The evidence implicates the second leg of the cruise (January 26–February 11). Although infection could have been acquired while ashore, the epidemiologic investigation suggests that exposure occurred while on board the ship. The estimated incubation period for this outbreak, 25–40 days, is shorter than but within the range of the reported incubation period for hepatitis E (15–60 days) (3).

The 3 associated risk factors (gender, age, shellfish consumption) remained significant in multivariate analysis. First, male passengers were more likely to have been infected than female passengers; 76% of recent acute infections were in men. A marked excess of indigenously acquired HEV cases in middle-age and elderly men has been reported in England, Wales (5), and other European countries (10–12). Our study suggests that this observed excess is not caused by ascertainment bias because men are at higher risk for disease, but rather it appears to be a genuine difference in exposure. Second, alcohol consumption was associated with recent infection. Although excess alcohol consumption could compromise hepatic function and predispose to symptomatic hepatitis E infection, as suggested by our study, alcohol consumption is probably not causally linked to exposure. The association may indicate a propensity for risk behavior that could not be controlled for in the analysis. Third, consumption of shellfish on board the ship was strongly associated with HEV infection. Further analysis did not implicate a particular type of shellfish, and cross-contamination of shellfish from a single vehicle is possible, but because the virus is waterborne and has been shown to contaminate shellfish (13–15), this association is biologically plausible. Other studies have shown that HEV can be foodborne, and illness has been linked to consumption of undercooked or raw meat (16–18).

Many of the 33 recent acute HEV infections, predominantly in men who drank alcohol, were asymptomatic and would otherwise have gone undiagnosed. We found no evidence of continuing transmission, other than 1 potential secondary case, or of any breaches of public health standards on board the ship. However, the analytical study and supporting genotype findings challenged the initial assumption that the outbreak was due to infection acquired while ashore. This investigation suggests that shellfish, which are known to be a common source of other viral infections, are a potential source of HEV infection in Europe.