Travel-related Dengue Virus Infection, the Netherlands, 2006–2007

To assess the incidence of and risk factors for clinical and subclinical dengue virus (DENV) infection, we prospectively studied 1,207 adult short-term travelers from the Netherlands to dengue-endemic areas. Participants donated blood samples for serologic testing before and after travel. Blood samples were tested for antibodies against DENV. Seroconversion occurred in 14 (1.2%) travelers at risk. The incidence rate was 14.6 per 1,000 person-months. The incidence rate was significantly higher for travel during the rainy months. Dengue-like illness occurred in 5 of the 14 travelers who seroconverted. Seroconversion was significantly related to fever, retro-orbital pain, myalgia, arthralgia, and skin rash. The risk for DENV infection for short-term travelers to dengue-endemic areas is substantial. The incidence rate for this study is comparable with that in 2 other serology-based prospective studies conducted in the 1990s.

D engue virus (DENV) is a fl avivirus transmitted through the bite of an infected Aedes spp. mosquito. The disease is endemic to tropical and subtropical regions and each year affects ≈100 million persons worldwide. Most cases are reported from Southeast Asia and Central and South America (1,2).
The incubation period for DENV infection is 3-14 days. Manifestations are nonspecifi c and range from asymptomatic to severe (3,4). Most infections are benign, with few deaths. In 2%-3% of patients, dengue hemorrhagic fever develops (5).
DENV has 4 distinct serotypes. Primary infection with 1 serotype confers lifelong immunity to that serotype but increases the risk for severe dengue illness after secondary infection with another serotype (6). No specifi c therapy exists for dengue.
During the past few decades, dengue has emerged in tropical and subtropical countries worldwide (1,7). In addition, the number of reported symptomatic DENV infections among international travelers has increased (5,8,9). This increase may refl ect increased incidence of dengue among travelers, increased number of travelers to areas in which dengue is endemic, increased awareness of the disease among physicians, or a combination of these factors.
However, valid research on travelers' risk for DENV infection is scarce. To our knowledge, only 2 prospective studies, performed in the 1990s (10,11), have been conducted. Most other epidemiologic studies are retrospective and focus on symptomatic patients who seek care at a clinical site, thus disregarding the nonspecifi c nature of the infection and the increased number of travelers to dengue-endemic areas (5,8,9,12,13).
We prospectively estimated the prevalence and incidence of DENV infection and its risk factors. Our fi ndings are based on serologic testing in a cohort of shortterm travelers from the Netherlands to dengue-endemic areas.

Study Population
We prospectively studied persons attending the travel clinic of the Public Health Service Amsterdam during Dengue endemicity was based on information from The Global Infectious Diseases and Epidemiology Online Network (14). All participants were seen by a medical doctor or nurse who specialized in travel medicine. They received vaccinations, a prescription for antimalarial chemoprophylaxis if required, and oral and written information about how to avoid acquiring travel-related diseases, including mosquito-borne infections, according to Dutch National Guidelines on Traveler's Health Advice (15).

Survey Methods
Before departure (just before they received the required vaccinations) and 2-6 weeks after return, participants donated venous blood samples for serologic testing. Before departure,they completed a standard questionnaire to collect data on sociodemographic factors, including travel history, vaccination status, and the purpose of travel (tourism, work/education, or visits to friends and/or relatives).
Participants were given a digital thermometer and asked to take their temperature (axillary or rectally) when feeling feverish. They kept a structured travel and posttravel diary to record symptoms and signs of dengue illness (fever, myalgia, arthralgia, headache, retro-orbital pain, skin rash, signs and symptoms of gastrointestinal disease), use of insect repellent containing N,N-diethylmeta-toluamide (DEET), and use of mosquito netting. Participants were asked to report both the presence and absence of symptoms and preventive measures on a daily basis; after travel, the diary was checked by a registered nurse in the participant's presence.
To encompass the incubation period of DENV infection, we used data from the third day after arrival at the travel destination until 2 weeks after return from travel. With respect to symptomatic outcome, we used the term travel-related to refer to this period.
Destinations were grouped into regions by using the classifi cation of the United Nations Development Agency (16). Travel duration was recorded as the total days spent in areas meeting the inclusion criteria.
The Medical Ethics Committee of the Academic Medical Centre Amsterdam approved the study protocol. Participants were included only if they provided informed and written consent.

Laboratory Methods
Blood samples were immediately stored at 6°C, then centrifuged and frozen at -80°C within 24 h after collection. After all study participants had returned, all posttravel serum samples were thawed and tested for immunoglobulin (Ig) M antibodies to DENV antigen by using an IgMcapture ELISA and for IgG antibodies to DENV antigen serotypes 1, 2, 3, and 4 by using an indirect ELISA (Panbio Diagnostics, Brisbane, Queensland,Australia) according to the manufacturer's instructions. For participants whose posttravel sample yielded positive test results for anti-DENV IgM or IgG, pretravel samples were also tested for anti-DENV IgM or IgG, respectively.
Serology suggestive for previous DENV infection was defi ned as an anti-DENV IgG >11 Panbio units in both the pretravel and the posttravel sample. The posttravel-topretravel anti-DENV IgG ratio was <4:1.

Data Analysis
To calculate risk factors for previous DENV infection, we used SPSS for Windows version 17.0 (SPSS Inc., Chicago, IL, USA) to obtain prevalence rates, univariate and multivariate prevalence rate ratios, and 95% confi dence intervals (CIs) by using logistic regression modeling. All variables with an overall p value <0.05 in univariate analysis were included in multivariate analysis. Statistical interactions between variables were checked for all variables. If signifi cant (p<0.05), they were included in multivariate analysis.
We calculated attack rates of recent DENV infection per 100 travelers by dividing the number of seroconversions by the total number of participants. Incidence rates (IR) per 1,000 person-months were calculated by dividing the number of seroconversions by the total number of months in which participants were at risk for DENV infection. To calculate risk factors for recent DENV infection, incidence rate ratios (IRR) were obtained as conditional maximumlikelihood estimates of the rate ratio, with mid-P exact test, 95% CIs, and p values. We quantifi ed the application of a protective measure against mosquito bites by dividing the number of days the measure was applied by the number of travel days; it was dichotomized by using the mean of this proportion in the total study population as the cutoff.
Travel during the rainy months was assessed by using the midtravel dates. The rainy months per destination were defi ned as months with an average total precipitation of >100 mm, based on World Weather Information Service data from the United Nations' World Meteorological Organisation (22).
We used χ 2 tests to evaluate the predictive value of travel-related signs and symptoms for recent DENV infection. Fever was defi ned as a body temperature >38°C. Dengue-like illness was defi ned as any episode of fever accompanied by >1 of the following: myalgia, arthralgia, headache, retro-orbital pain, or skin rash. A p value <0.05 was considered signifi cant.

Study Population
The study began with 1,276 participants who intended to travel to dengue-endemic areas. Of these, 69 (5.4%) were excluded: 33 were lost to follow-up; 23 had their travel arrangements cancelled; 7 did not supply the posttravel blood donation; 3 changed travel plans such that inclusion criteria were not met; 2 did not complete the structured diary; and for 1, the posttravel sample was lost.
Most of the remaining 1,207 participants were persons of Dutch origin who were traveling for tourism and previously had traveled to dengue-endemic areas (Tables  1, 2). The median travel duration was 21 days, with an interquartile range (IQR) of 16-28. The median interval between return from travel and posttravel blood donation was 25 days (IQR 21-29).
Twenty-six percent of travelers had received yellow fever vaccination for a previous journey, and 23% received a yellow fever vaccination for the current trip, just after blood donation (not shown in Table 1). Only 4 (0.3%) participants had ever received Japanese encephalitis vaccination, all for a previous journey. No one had ever received vaccination against tick-borne encephalitis.

Serologic Results Suggestive for Previous DENV Infection
Serologic results suggestive of previous DENV infection were found for 78 (6.5%; 95% CI 5.2%-8.0%) participants (Table 1). Previous DENV infection was positively correlated with increasing age. The rate was signifi cantly higher for travelers born in a dengue-endemic country, for those who traveled frequently to dengueendemic countries (all 78 reported >1 previous trips to such a country), and for those whose purpose was to visit friends or relatives. In multivariate analysis, previous DENV infection was not related to yellow fever vaccination for previous travel because it was confounded by previous trips to a developing country. We found no interactions indicating effect modifi cation.
All 14 persons with recent infection were born in non-dengue-endemic countries; 9 were male, and 11 had traveled previously to a dengue-endemic country ( Table  3). The median age was 36 years (IQR 27-56 years). The primary purpose of travel was tourism (12 persons) or work/education (2 persons). The median travel duration was 24 days (IQR 19-30).
The incidence rate was signifi cantly higher for travel during the rainy months (23.6 vs. 7.5/1,000 person-months; IRR 3.2, 95% CI 1.01-11.6; p = 0.048). Travel during the rainy months was not related to use of insect repellent containing DEET or use of a mosquito net (odds ratio 1.0; p>0.05).
The incidence rate appeared to be higher for male travelers (  previous travel to a dengue-endemic country, yellow fever vaccination for the current trip, and use of a mosquito net, the IRRs equaled 1.0 (p>0.05). Recent DENV infection was positively correlated with fever, retro-orbital pain, myalgia, arthralgia, and skin rash ( Table 4). Dengue-like illness was reported by 5 (36%; 95% CI 14%-62%) of 14 participants. Evidence for recent DENV infection was 7.3× more prevalent among those with than without dengue-like illness. No traveler reported symptoms signs or suggestive of dengue hemorrhagic fever.

Discussion
In this prospective study, the serology-based attack rate and incidence rate for recent DENV infection in short-term travelers to areas in which dengue is endemic were substantial: 1.2% and 14.6 per 1,000 person-months, respectively. Although most seroconversions occurred in travelers to Southeast and southern Asia, 3 occurred in travelers to sub-Saharan Africa.
The risk for recent infection was signifi cantly related to travel during the rainy months, as others have reported (10). The use of insect repellent containing DEET appeared to be protective, although not signifi cantly so. The presence of fever, retro-orbital pain, myalgia, arthralgia, and skin rash all had predictive value for recent infection. Denguelike illness was the strongest predictor. Nevertheless, 64% of infections were asymptomatic.
For as many as 6.5% of travelers, serology results suggested previous DENV infection. This rate was strongly related to birth in a dengue-endemic country and a history of frequent travel to such countries. Increasing age and the travel purpose of visiting friends or relatives were also predictive for previous DENV infection, as described (5).
Studies based on febrile travelers returning from the tropics have reported dengue fever in ≈2% of those who sought medical help (12,23). Some studies have suggested that this proportion has increased in the past 2 decades (3,8,9,24); others showed no signifi cant increase (25).
All these studies were retrospective. They were based on persons who sought medical attention after travel and thereby missed persons who never had symptoms, whose symptoms resolved during travel, or who never consulted a doctor. These studies were infl uenced by referral bias and relied only on posttravel serologic confi rmation. They could not compare characteristics between patients and persons who remained well, and the studies lacked valid denominator data to determine absolute risk. Interpreting their fi ndings is therefore diffi cult.
The best methodologic approach for estimating incidence rates of clinical and subclinical travel-related DENV infections is to follow a cohort of travelers prospectively (26). We found only 2 such studies. In 1 study, 6.7% of 104 travelers from Israel who spent on average 6.1 months (range 3-16 months) in Southeast Asia, South America, or Africa seroconverted (11), yielding an incidence rate of 11.0 per 1,000 person-months. This incidence rate is comparable to the rate found in our study, although our study differed in travelers' destination and duration of exposure and in test characteristics.
In the other prospective study, performed during 1991-1992, the seroconversion rate for 447 travelers from the Netherlands who spent a median of 4 weeks in Asia was 3.6%, resulting in an incidence rate of 36.9 per 1,000 person-months (10). In our study, the seroconversion and incidence rates in travelers to Southeast Asia were lower: 1.6% and 18.5 per 1,000 person-months, respectively. Our different fi ndings may refl ect different test characteristics and differences in exposure and risk behavior related to factors such as travel destination, year of exposure, and measures taken to prevent mosquito bites.
The incidence rate for persons in our study who traveled to sub-Saharan Africa appeared to be substantial (13.4/1,000 person-months), although the 95% CI was wide (3.4-36.5), and cross-reactions with other fl aviviruses or previous yellow fever vaccination cannot be excluded. However, cross-reactions with yellow fever vaccination are unlikely because the travelers to Africa who seroconverted for anti-DENV either did not receive a yellow fever vaccination during the study period (2 travelers) or had dengue-like illness (1) ( Table 3). These travelers visited Senegal and the Gambia, Mali and Burkina Faso, and Madagascar. All these countries reported dengue cases during the study period (14,27,28). According to data from the GeoSentinel Network (12) and the TropNetEurope Network (5), only a small proportion (5%-8%) of clinical dengue cases in travelers are acquired in Africa. However, distributions yielded by these data refl ect not only global dengue activity but also the popularity of tourist destinations and physicians' expectations about dengue endemicity and their inclination to test for the disease. Dengue endemicity in Africa may have changed, and the risk for travelers may thus be higher than expected (27). More studies are needed. The ratio of apparent to unapparent infection in our study was 1:1.8. The 2 other prospective studies reported ratios of 1:0.75 (11) and 1:3 (10). Variations in these ratios may refl ect variation in strain virulence, the infl uence of initial viral load, or host factors such as susceptibility.
Our study has some limitations. First, our sample size was too small to yield a precise estimate of disease incidence per region or per destination. For example, travelers to Southeast Asia are considered to be at higher risk than travelers to other regions (5,12). This fi nding could not be confi rmed in our study. Second, selection bias may have occurred. Although participants in our study are comparable to the average traveler (29), they were all seeking pretravel health advice. Thus, they perhaps had a higher than average health awareness, particularly after receiving oral and written travel advice, learning about the study, and agreeing to participate. Nevertheless, their average use of insect repellent containing DEET was only 45% of travel days, and 17% of participants used no repellent at all. Third, the risk for travel-related DENV infection also can depend on endemicity and on outbreaks in a particular country during a particular time of travel. Also, among travelers, DENV transmission has annual oscillations (12,30). Furthermore, sometimes epidemics occur in tourist areas, other times in areas never reached by tourists. Although 2006-2007 were not years with high endemicity for most dengue-endemic countries, the exact contribution of these factors on our fi ndings is unknown.
A fi nal limitation is that, although the serologic tests used are highly sensitive and specifi c, false-negative and false-positive test results may have occurred. For example, 7 travelers seroconverted only for anti-DENV IgG and not for anti-DENV IgM. Although this may be because of false-positive IgG or false-negative IgM test results, these results may also suggest secondary infection. Unfortunately, no test characteristics are available that apply exactly to our study design; the best available test characteristics are probably those from studies that used paired serum samples (6,17,18,20). With test specifi city of <96%, >4% of positive samples actually give false-positive results. Subclinical infections with fl aviviruses other than DENV may account for some of the seroconversions (31). Also, cross-reactions can affect the interpretation of test results in travelers who have received a fl avivirus vaccine. In 1 study, 2 (15%) of 13 persons vaccinated against yellow fever had positive test results for DENV-IgG 1-2 months after the vaccination (31). In our study, however, recent and previous DENV infections were not related to yellow fever vaccination. False-positive test results can also occur just by chance. Because the serologic tests used were validated in a clinical environment among patients suspected to have dengue, their positive predictive value will be lower when applied as a screening tool in a more general sample of travelers in which disease occurrence is lower. Unfortunately, no true standard exists for confi rming or ruling out DENV infections. Even a plaque-reduction neutralization test can cross-react (32), and PCR is useful only in the very early stage of dengue disease (33).
In conclusion, although valid data regarding the actual frequency of DENV infection in international travelers is elusive, our prospective study confi rms that the incidence rate for recent DENV infection in short-term travelers to areas in which dengue is endemic is substantial. We found no differences in incidence rate compared with other serology-based prospective cohort studies performed in the 1990s. Thus, the increase in annually reported dengue cases from network studies is possibly related to the increase in international travel and the expansion of DENV and its vector to new areas rather than to an increase in DENV transmission. Until an effective and safe vaccine that provides effective, long-term immunity against all 4 serotypes of DENV has been developed, the only useful preventive measure for travelers to areas where dengue is endemic is to avoid mosquito bites, particularly during the rainy season.