Asymptomatic Dengue Virus Infections, Cambodia, 2012–2013

We investigated dengue virus (DENV) and asymptomatic DENV infections in rural villages of Kampong Cham Province, Cambodia, during 2012 and 2013. We conducted perifocal investigations in and around households for 149 DENV index cases identified through hospital and village surveillance. We tested participants 0.5–30 years of age by using nonstructural 1 rapid tests and confirmed DENV infections using quantitative reverse transcription PCR or nonstructural 1–capture ELISA. We used multivariable Poisson regressions to explore links between participants’ DENV infection status and household characteristics. Of 7,960 study participants, 346 (4.4%) were infected with DENV, among whom 302 (87.3%) were <15 years of age and 225 (65.0%) were <9 years of age. We identified 26 (7.5%) participants with strictly asymptomatic DENV infection at diagnosis and during follow-up. We linked symptomatic DENV infection status to familial relationships with index cases. During the 2-year study, we saw fewer asymptomatic DENV infections than expected based on the literature.

We investigated dengue virus (DENV) and asymptomatic DENV infections in rural villages of Kampong Cham Province, Cambodia, during 2012 and 2013. We conducted perifocal investigations in and around households for 149 DENV index cases identified through hospital and village surveillance. We tested participants 0.5-30 years of age by using nonstructural 1 rapid tests and confirmed DENV infections using quantitative reverse transcription PCR or nonstructural 1-capture ELISA. We used multivariable Poisson regressions to explore links between participants' DENV infection status and household characteristics. Of 7,960 study participants, 346 (4.4%) were infected with DENV, among whom 302 (87.3%) were <15 years of age and 225 (65.0%) were <9 years of age. We identified 26 (7.5%) participants with strictly asymptomatic DENV infection at diagnosis and during follow-up. We linked symptomatic DENV infection status to familial relationships with index cases. During the 2-year study, we saw fewer asymptomatic DENV infections than expected based on the literature.
In Cambodia, the national dengue surveillance system reported 60,000 cases and 135 deaths attributed to DENV in 2012 and 2013 (3). Syndromic surveillance and random testing of dengue-like cases in referral pediatric hospitals in Cambodia likely underestimate the true disease burden (4). By definition, syndromic surveillance does not detect asymptomatic DENV infections, which increase vector transmission potential (5). Mammen et al. used both dengue-positive and dengue-negative index cases of febrile children to initiate perifocal investigations and found no cases in proximity to dengue-negative index cases (6). To maximize the number of recruited cases, we investigated homes around preidentified, dengue-positive index cases, as per a previous study (7). Our objectives were to document the proportion of strictly asymptomatic infections in this region of Cambodia; characterize human, sociodemographic, household-level, and mosquito control-related factors associated with DENV infection; and identify factors associated with asymptomatic DENV infection.

Ethics Considerations
The study protocol was approved by the Cambodian National Ethics Committee on Health Research. We obtained informed consent from participants or their guardians documented during hospital or village surveillance or perifocal investigations.

Study Site
We conducted a study in rural villages of Kampong Cham Province, 120 km northeast of Cambodia's capital, Phnom Penh. The study area included 368 villages with ≈60,000 households and 3 hospitals within a 30-km radius. Dengue is endemic in the region and mainly affects children <15 years of age during the annual rainy season (June-October). under active surveillance for febrile illness (5). We targeted persons 0.5-30 years of age. In the 3 hospitals, blood samples were drawn at admission and discharge for all patients suspected of having DENV infection on the basis of clinical assessment and platelet count. In the 26 villages, volunteers monitored eligible residents weekly, measuring axillary body temperature using a digital thermometer to identify persons with temperatures >38°C. Blood samples were drawn 1-2 days after fever onset, as described elsewhere (4,8,9). All samples were screened for DENV infection by using a nonstructural (NS) 1 IgM/IgG combination rapid test. We confirmed DENV by using quantitative reverse transcription PCR (qRT-PCR) or NS1-capture ELISA and included case-patients with confirmed DENV infection as index cases in the study.

Perifocal Investigations
Within 1-2 days of identifying an index case, whether from village or hospital surveillance, we began a perifocal investigation of the index case-patient's village of origin (7). For each perifocal investigation, we used a rapid dengue test kit to screen eligible residents in the index case-patient's household for DENV and completed a baseline questionnaire on individual symptoms, socioeconomic status, and household characteristics. We did the same in 20 households in a 100-meter radius of the index case-patient's household. We included persons 0.5-30 years of age who consented or whose guarantor consented. We tested adults >20 years of age during the first year of the study but found no DENV-positive cases and did not test this age group during the second year. All consecutive cases were eligible for inclusion. To avoid bias through overlapping investigations of a potentially common source of infection, we did not conduct a perifocal investigation within 1 week of a previous investigation for >2 index cases consecutively detected from the same village.

DENV Testing and Case Definitions
To screen for DENV infection during surveillance and perifocal investigations, investigators tested all blood samples on-site using SD BIOLINE Dengue Duo kit (Standard Diagnostics, https://www.alere.com), according to the manufacturer's instructions. Investigators interpreted results within 15-20 minutes and ruled out possible cases if the control band was negative. Blood samples from DENV-positive participants were sent to Institut Pasteur du Cambodge (Phnom Penh, Cambodia) for qRT-PCR testing, as described previously (10,11), or confirmation using an NS1-capture ELISA (11,12) with positive controls diluted to the limit of detection, negative, and nontemplate controls used during extraction and PCR steps to reduce inaccuracies (10). We considered cases confirmed when a blood sample tested positive by NS1 rapid test and was confirmed by NS1-capture ELISA or qRT-PCR. During the first year, we also tested participants for Japanese encephalitis virus (JEV) and chikungunya virus (CHIKV) IgM antibodies by ELISA and confirmed IgM-positive results using specific RT-PCR to ensure that symptoms were not related to CHIKV, JEV, or co-infections (11)(12)(13).
Symptomatic DENV-confirmed case-patients had fever, muscle or joint pain, rash, bleeding, prolonged headaches, or digestive signs. We asked participants whether they had taken antipyretics in the previous 24 hours. We termed afebrile all symptomatic DENV-positive participants without a fever and no antipyretic use. We considered participants asymptomatic when they had confirmed DENV infection, no antipyretic use, and no signs or symptoms, including fever. Participants who were symptomatic at initial diagnosis on day 0 received follow-up monitoring on days 2 and 7. We monitored asymptomatic participants daily on days 0-7 and again on day 10 using a questionnaire to document signs and symptoms of DENV. In our analyses, we recategorized participants who were asymptomatic at baseline to symptomatic if they reported any symptoms during the follow-up period.

Statistical Analysis
We described DENV infection attack rates for perifocal investigations and the proportion of asymptomatic cases among all DENV infections and circulating serotypes. To explore participant-and household-level factors associated with DENV infection, we conducted a multivariable Poisson regression estimating attack rate ratios (ARRs) (14), excluding index cases. We built explanatory models around each participant-level and household-level factor, with and without adjusting for covariates. Participant-level factors included age, sex, occupation or schooling, and relationship to an index case-patient. Because we found collinearity between age and occupation, we adjusted only for age. We placed participants 0.5-1 year of age into a specific category to account for differences in immunity and exposure to vectors due to reduced mobility. Household-level factors included the main source of income, source of water, measures against mosquitoes, and environmental factors favorable to mosquito development. We further divided the source of water into 2 categories: piped water (from indoor or outdoor taps with a tube well and pump) or nonpiped water (from a pond, river, lake, or a well without pump).
Considering the limited flight range of a female Aedes mosquito, we assumed that the probability of DENV transmission would be higher within a household than across households. To account for this factor and measure potential clustering, we developed a random-effects multilevel model. We computed the intraclass correlation coefficient as the proportion of the variability in the probability of infection attributable to differences between households versus differences within households (15). We excluded 19 participants with missing covariates or predictors from the regression analyses. We explored associations between asymptomatic DENV infection and DENV serotype, participant-level factors, and the main source of income as socioeconomic indicators. We used the Fisher exact test for comparing proportions, the Student t test for means, and an empty multilevel model to search for a cluster effect. We conducted analyses using Stata version 13 (StataCorp, https://www.stata.com).
Besides headache and fever, symptomatic casepatients mainly experienced muscle, retro-orbital, and joint pain. Although fever is considered a typical symptom of DENV infection, careful interview, rigorous clinical assessment, and follow-up interviews showed that participants remained afebrile in 110 (31.8%) of the 320 symptomatic DENV infections, even without antipyretics. Only 6 (1.7%) of the DENV-positive case-patients required hospitalization, 2 with bleeding.
The 2 annual outbreaks were dominated by DENV-1. However, DENV-2 and DENV-4 emerged in 2013, and we detected DENV-3 sporadically (Table 1). During the first year of the study, samples from all symptomatic and asymptomatic DENV cases were negative for CHIKV by MAC-ELISA. Because we diagnosed no CHIKV in year 1, and our national surveillance system also did not detect any CHIKV cases (data not shown), we did not perform CHIKV testing during year 2. Of 26 asymptomatic cases confirmed by qRT-PCR or NS1-capture ELISA, 6 had positive JEV serology and also were positive for DENV IgM. We could not conclude whether JEV-positive results were indicative of a recent or acute JEV co-infection or the result of cross-reaction among flaviviruses. Among hospitalized patients, 2 had positive JEV results without detectable DENV IgM, even though qRT-PCR or NS1-capture ELISA was positive. These results could suggest a recent or acute JEV co-infection. During perifocal investigations, 42 participants tested positive for JEV by MAC-ELISA with negative DENV results, NS1, and qRT-PCR, supporting evidence of JEV co-circulation in the country (16).
Among DENV-positive cases, boys and girls were equally affected at a mean (+ SD) age of 8 . Households reporting mosquito control-related parameters did not have a lower risk for DENV infection ( Table 2).
The main source of income was similarly distributed between households with ≥1 case and households with no cases (p = 0.272). Our multilevel model showed a notable clustering effect at the household level after adjustment (intraclass correlation coefficient 40.8%).
We found 26 (7.5%) case-patients, 17 (65.4%) male and 9 (34.6%) female, who were positive for DENV infection but remained asymptomatic. We found serotypes DENV-1, DENV-2, and DENV-4 in our study group (Table 3). We used a multilevel approach to explore the role of specific serotypes and participant-level factors, such as age, gender, and relationship to the index case-patient, a proxy for common genetic background, with being DENVpositive and asymptomatic. We found that only family relationship to the index case-patient was associated with asymptomatic infection. We did not identify a cluster effect or associated factors.

Discussion
We screened 7,960 participants in communities in Cambodia during 2012 and 2013 and found 346 (4.3%) participants  (17), multiple factors could explain this observation and deserve further research.
Our study documented cases of DENV infection in transmission clusters located around index case-patients. We found that 26.6% of DENV-confirmed case-patients reported clinical symptoms, including headache and muscle pain, but no fever even in the absence of antipyretics, comparable to data from Thailand, where 40.4% of the DENV cases remained afebrile (17). The appearance of afebrile DENV-infected patients raises potential concerns for case definitions for detection, especially of imported cases in at-risk countries.
An additional 7.5% of DENV-confirmed case-patients had no symptoms during the 10-day course of clinical monitoring, a considerably lower rate than estimates from other prospective studies (5,(18)(19)(20)(21). Published sources refer to inapparent infections, often defined as afebrile clinical complaints with biologic evidence of DENV infection, ranging from 20% to 80% of cases (19,22,23). Previous studies used different definitions of asymptomatic infection than ours, but the major difference lies in follow-up monitoring. Other retrospective studies used school or work absenteeism as a basis for follow-up (19). Strictly asymptomatic patients, such as those we describe, escape detection by surveillance or control measures, infect mosquitoes, and might disproportionately contribute to DENV transmission (5).
The DENV burden documented through hospitalbased surveillance of febrile case-patients in Thailand and Vietnam showed a shift to older age groups (24,25). Our active, systematic case-finding system to identify DENV in villages in Cambodia found the attack rate was highest in children <10 years of age, which is what we expected in a dengue-endemic country with frequent outbreaks demonstrated in other careful studies (26). This finding raises concerns because recommendations for the only licensed dengue vaccine are for use in persons 9-45 years of age with demonstrated evidence of past DENV infection (27). Our study demonstrates that children in rural Cambodia might have undergone >1 DENV infection before 9 years of age, reducing the potential cost-effectiveness of vaccination.
Few studies have explored the role of socioeconomic status, which might be a proxy for peridomestic environmental management, on DENV infection in Southeast Asia. Often, the direction of the association is unclear and socioeconomic status has entirely different associations depending on the setting (28). Our study shows that the adjusted risk for DENV infection was highest in households in which the main source of income was from fishing, farming, or animal husbandry, activities associated with lower average household income in Cambodia.
We found temephos provided no additional protection against DENV infection after adjusting for other factors. Although temephos is effective in reducing Aedes spp. larval populations in water storage jars, its use did not correlate with lower DENV transmission in Cambodia or elsewhere (8), due to incorrect distribution coverage, dosage, and placement (29) or multiple vector breeding sites. In addition, increases in temephos-resistant A. aegypti mosquito larvae have been documented in Cambodia (30).
Unexpectedly, we found a higher risk for DENV with piped water as a main water source after adjusting for other factors, contrary to a study in Thailand (6). However, piped water with suboptimal sanitation in Cambodia might contribute to collection of water in or around households that could become breeding sites for DENV-transmitting mosquitoes.
We found that 40.8% of the variability in probability of being DENV infected was explained by differences between households. Those living in the same household as an index case-patient were 2.11 times more likely to be infected, consistent with other published sources. A study in Mexico found that the risk for infection for those living with an index case-patient was twice that of someone living in a 50-meter radius of an index case-patient (31). This relationship was further described in a cluster study in Thailand that showed decreased risk for infection with increasing distance from the index case household (31). This clustering effect around an index case, however, seems to occur only on a short temporal scale, at least in urban settings (32).
Rates and severity of illness after infection with the different DENV serotypes differ widely (33,34). The only notable epidemiologic factor associated with asymptomatic DENV infections in our study was being family-related to the index case-patient. Another study showed that adaptive immune responses against DENV differ between persons with symptomatic and asymptomatic DENV infection (35), which might explain our observations. We found no other associated epidemiologic factor, including age or cluster effects. Although the ratio of male to female participants was twice as high among asymptomatic than symptomatic participants, this finding was not statistically significant, likely due to sample size.
Although the size and duration of our study confer strength to our data, it might suffer from biases and limitations, especially due to the small number of strictly asymptomatic DENV-positive participants after stratifying by DENV serotype. We found dengue incidence rates highest in young children. These data might be biased because we focused on investigating clusters around an index case, perhaps overestimating the incidence in the general population. DENV circulation, however, is intense in children in Cambodia, and these figures remain comparable to those found in dengue studies that use different methods, ranging from 20 to 80 per 1,000 person-seasons (1). Furthermore, we did not capture cases referred to the private sector, lowering our estimates somewhat. Healthy male workers often were away at the time of the investigations, possibly leading to an overestimation of DENV incidence. These workers, however, are >18 years of age, but DENV infections occur mainly in persons <15 years of age in Cambodia (4).
Documentation bias might also have pulled our risk factor estimates toward the null. We did not document solid waste disposal in our study, but comparatively high Breteau index values have been reported in Cambodia (29). In addition, we could have missed details or misrepresented implementation of mosquito-control measures. Despite the potential misclassification, mosquito-control measures remain nondifferential and likely had no major effect on our risk estimates. Further, 7.5% of our DENV-infected participants remained strictly asymptomatic. Aside from case definition issues we discuss, our method of screening for DENV around symptomatic cases might have underestimated the number of asymptomatic DENV infections. In addition, we did not enroll persons who tested negative for DENV IgM, NS1-capture ELISA, and qRT-PCR. Some of these persons might have been infected but not yet mounted an IgM response, so that NS1 and viral RNA titers had already receded to undetectable levels when we tested them. This strict case definition might have underestimated the incidence of asymptomatic cases, but a precise retrospective documentation of such cases would be extremely difficult. Similarly, we retrospectively conducted MAC-ELISA on samples collected during perifocal investigations and identified 11 cases of IgM seroconversion in the absence of PCR-or NS1-positive tests. Even in the context of JEV cocirculation, some of these cases could have been true DENV infections, but including them would not have changed the overall estimated attack rate. Previous studies suggested virus serotype might affect severity and types of symptoms and observed that DENV-1 infections more frequently were associated with clinically apparent illness (36,37). Virus molecular analysis studies are ongoing to determine whether specific strains cause more asymptomatic infection than others. Furthermore, DENV infection in Cambodia occurs mainly in children who might be more likely to answer positively to daily-repeated questions on dengue symptoms, somewhat underestimating asymptomatic cases. Having implemented careful and thorough 10day clinical assessment of objective symptoms in each asymptomatic DENV-positive participant, we believe these figures reflect the true proportion of strictly asymptomatic DENV infection in our setting. However, we collected our findings mainly in children with DENV-1 infection in Cambodia. Whether these findings are directly applicable to other epidemiologic settings, populations, or virus serotypes or genotypes remains to be determined (33).
Finally, vaccination against JEV might have led to cross-protection against symptomatic dengue. Data on JEV vaccination were not collected during perifocal investigations. According to local health centers, however, JEV vaccine has been provided only recently and only for children 9-24 months of age. In our study, only 3 children were DENV-positive in that age category.
Our study demonstrates that systematically relying on fever for DENV case definition can underestimate cases and hinder control efforts in areas with potential vectors and at risk for DENV introduction. We found 7.5% of DENV-infected participants remained strictly asymptomatic, which has wide-ranging epidemiologic consequences. Undetected sources can increase transmission (5), a factor that must be taken into account in future vaccine coverage and vaccine effectiveness studies. The attack rate differences observed around febrile index case-patients detected in village surveillance and index case-patients detected in hospital surveillance deserve further study. In-depth virus (36) and human genetic studies could contribute useful insights (33,35). Our strict definition of asymptomatic DENV infections should be considered when designing studies that aim to elucidate the pathophysiological mechanisms of dengue disease.