In the Netherlands, HAV infection is a reportable disease. Physicians and medical laboratories report cases to a municipal health service (MHS) according to national notification criteria: presence of a predefined set of clinical signs of hepatitis combined with HAV IgM in serum. MHS Consultants for Communicable Disease Control contact the patient and administer a questionnaire that collects routine demographic and epidemiologic data consisting of age, sex, country of birth, time of disease onset, related cases, travel history, homosexual contacts, and other possible modes of transmission (full questionnaire available on request to M.P.). MHS enters the suspected transmission route and other anonymized information into a national electronic registration system hosted by the National Institute for Public Health and the Environment (RIVM). All cases reported during July 1, 2008, through June 30, 2010, were included in the study.

During the 2-year period, an enhanced surveillance system, which included systematic typing of viruses from patients, was deployed. All medical microbiological laboratories and MHSs in the Netherlands were asked to send serum samples from all reported patients to the Laboratory for Infectious Diseases Research, Diagnostics and Screening at RIVM. RNA was extracted from the serum and tested for HAV by reverse transcription PCR selective for the viral protein (VP) 1–2A region of the genome (3,4). HAV genotyping was conducted by sequencing of a 460-nt fragment of the VP1–2A region. Sequence data were stored in a Bionumerics database (Sint-Martens-Latem, Belgium). Sequencing results were merged with the national registration data, according to laboratory name and serum sample number. For cases lacking a unique serum sample number, notification data and sequences were linked by using combinations of variables to match records (birth year, 4-digit postal code, date of illness onset, date of diagnosis).

The reporting MHSs were contacted twice by telephone for interviews. We asked for the MHS conclusion as to the most probable modes of transmission immediately after the notification and then after results from sequencing were available. This approach was taken because public health measures for different transmission categories might differ (Table 1) and interventions could be adjusted accordingly. The initial interviews were also used to inform MHS about the study and to emphasize the need for collection of serum samples. The conclusions as to possible mode of transmission before and after inclusion of typing information were logged separately.

Because we used serum already available for diagnostic purposes, ethics approval was not needed. Patients were approached according to existing guidelines, and analyses maintained patient anonymity.

The Bionumerics database already contained patient data and strain sequences from previous studies conducted in the Netherlands (3–6) and all available sequences from GenBank (19). These data were used for background comparison if sequences covered a minimum of 300 nt of the VP1–2a region and if information was available on the most probable country of infection (for travelers) or other risk activities (20,21). The geographic fingerprints and other risk-group associations (e.g., Dutch MSM strains) from the background data were used to classify strain sequences from patients with unknown exposure to a probable source. This association was reported to the MHS only if the association was considered robust; robust clusters consisted of at least 3 identical sequences from independent patients with the same country of infection or MSM association, branching separately in a maximum-parsimony tree with >75% reproducibility of bootstraps. Clusters were defined when the following were found: at least 2 identical sequences branching separately in a maximum-parsimony tree with >75% reproducibility of bootstraps. Maximum-parsimony trees (phylogenetic trees based on finding the simplest or minimal evolutionary change between strains) were built by using Bionumerics, and reproducibility was tested by performing 1,000 bootstraps. Cases with strains meeting this cluster definition and sharing the same suspected mode of transmission were considered confirmed clusters within the assigned transmission category.

Data analyses were performed by using SAS software version 9.2/9.3 (SAS Institute, Cary, NC, USA). We described the study population by age and sex, disease incidence, and the number and percentage of patients for whom the virus could be typed. We analyzed the representativeness of age distribution for patients for whom sequencing was performed. If date of onset of disease was unknown, we used the date of diagnosis as a proxy. We compared age distribution and, when available, lag time between onset of disease and PCR diagnosis of positive and negative cases to weigh a negative result. We described the number and percentage of most probable modes of transmission in 5 categories (Table 1) before and after inclusion of typing results.

The 421 cases reported over the 2-year period resulted in incidence rates of 1.2 cases/100,000 population during the study year 2008–09 and 1.3 cases during 2009–10 among a total population of 16.4 million at the start of the study period and 16.5 million at the start of the second year. Most reported patients were in age groups from 0–9 through 40–49 years (range 13.3%–18.8% per group; Table 2). For patients in the youngest age group (0–9 years), sequenced cases were underrepresented, although distributions for patients for whom sequencing was performed did not differ significantly from reported patients (data not shown).

MHS determined the most probable modes of transmission for 268 of the 421 reported cases before typing (64%). Travel-associated transmission dominated (141 cases), followed by person-to-person transmission (76), male-to-male sexual contact (33), and foodborne transmission (18) (Table 3).

Of 292 samples received (69% of reported cases) PCR results for HAV were negative for 39 and positive for 253 (5 of which could not be typed and were excluded). The remaining 248 (59% of reported cases) were included in the final analysis. For 21 strains, sequencing was limited to 402–458 nt instead of the goal of 460 nt; for 1 strain, sequencing was limited to 100 nt.

Logistic regression showed that a longer lag time between onset of disease and diagnosis and belonging to the youngest or oldest age groups correlated with negative PCR results for HAV. This finding was expected because of unclear date of disease onset (data not shown).

Typing results confirmed all clusters of suspected person-to-person transmission, nearly all reported cases of male-to-male sexual transmission, and a large proportion of travel-associated infections (Table 3). One third of patients with travel-associated infections had traveled to countries with insufficient HAV sequence information in the public databases for reference. Therefore, the strain sequences for the virus in these patients could not be definitively assigned (category unresolved, Table 3).

In the category of suspected foodborne infections, nearly half of the cases for which sequencing had been performed could be confirmed. Only 1 case was misclassified; this infection was assigned to male-to-male sexual contact because the strain from this patient matched the dominant strain for MSM and the patient’s sexual orientation was concordant with this finding. The remaining cases were considered unresolved because the virus sequences did not cluster with known sequence clusters in the database.

For almost half of the 87 patients with unknown mode of transmission for whom sequencing was performed, the mode of transmission was resolved according to interpretation of the typing results. A remarkably high proportion (52%) of these infections were foodborne (Table 4).

Cluster 1 began with 2 cases linked to the same restaurant according to notification alone. A cook working in the restaurant had been infected by the dominant strain usually identified in MSM. He had continued working during his illness and was the probable source of infection. After genotyping and additional questioning, 2 more cases were added to this cluster.

Cluster 2 consisted of 2 cases clustered in time. Each patient had a unique genotype IA strain not previously detected, and both patients had eaten mussels.

Clusters 3 and 4 were associated with 2 consecutive outbreaks related to semidried tomatoes (12 and 5 primary cases, respectively). Cluster 3 turned out to be part of the largest foodborne outbreak thus far reported in the Netherlands, reaching 17 cases (including primary and secondary cases). The cases were clustered in time (reported in February and March) but were geographically dispersed, and the national notification rate was at an expected low level for this time of year, according to the 5 previous years. The strain sequences clustered with those from a large outbreak (at least 144 cases) in Australia and an outbreak (59 cases) in France, both of which were associated with consumption of semidried tomatoes (22–24). Cluster 4 was caused by a genotype IB strain closely resembling the strain involved in cluster 3.

Cluster 5 consisted of 1 case in a food handler of a dinner and 5 secondary cases. Cluster 6 consisted of 5 cases that were clustered strongly in time and for which virus strain sequences were identical, but the cases were geographically dispersed. Although the strain sequences were similar to those of strains typically detected in travelers returning from Morocco, the patients reported no travel history and no contact with patients with HAV infection imported from Morocco. Moreover, they clustered in April, a time of year when secondary or tertiary infections following travel-related imported cases are rare (8). Therefore, this cluster was considered a point-source—and very probably foodborne—cluster, although a source could not be determined.

Of 29 foodborne cases confirmed by a combination of epidemiologic and typing information (7 previously suspected foodborne and 22 previously unknown source), 20 additional reports were made. These cases were reported to the national food safety authority and international alerts through the Rapid Alert System for Food and Feed and Early Warning and Response System of the European Commission and the European Centre for Disease Prevention and Control.

For 45 (52%) cases initially reported as having no known source of infection (Table 3), conclusive evidence for a source was not found despite molecular typing. Nevertheless, some clustering occurred among these unresolved cases. The dominant MSM strain was found in 11 patients; however, these patients were not epidemiologically linked (time, place, food consumption), and among them were women and children, indicating spillover from the MSM risk group to the general population. Several other strains matched background strains previously imported from or known to circulate in Morocco and Egypt and even an outbreak strain from the Czech Republic (25). None of these patients had a history of travel. This finding could indicate unnoticed endemic transmission from persons with imported cases, although transmission through food or food handlers could not be excluded.