Several previous epidemiological studies have focused on school-based surveillance aiming at improving dengue-vector control measures 
, studying the dynamics of patterns of dengue transmission 
or describing a model that takes into account the role of human movement in the transmission dynamics of vector-borne pathogens 
. Earlier cluster investigation methods were designed as an alternative approach to the commonly used prospective cohort study method for investigating the natural history of dengue virus infection in South-East Asia and Latin America 
. Although different study designs have demonstrated the feasibility of identification of inapparent dengue cases, it remains difficult to recruit these subjects. We designed our study to include family household investigation in order to identify a group of inapparent dengue-infected subjects and to compare them with symptomatic dengue-infected and non-dengue-infected subjects living in the same family household. The study design was based on family household recruitment specifically in order to collect data and biological samples, and to study secondarily the host susceptibility to dengue infection and disease. Unlike studies based on cohorts from hospital referrals, this multi-country study captured dengue cases ranging from inapparent infections, through mild disease to severe dengue fever, using definitions of clinical cases and diagnostic methodology standardized across the four sites. The period of inclusion, from July 2006 to June 2007, spanned the dengue season at each site, although incidence of dengue was low that year in French Guiana.
The main objective of this study was to identify dengue infections and particularly inapparent infections among dengue patients' household family members in South-East Asia and Latin America. Based on our data, we estimated the proportion to be about 45% among those participating in the household study. Most of the dengue cases studied had symptomatic infections, covering the spectrum of disease from dengue fever to severe dengue cases. We also identified inapparent infections in the population. We observed dengue-infected subjects classified as DIC and some of their HHM without acute dengue infection but with a positive IgM detection, suggesting an early convalescent phase after dengue infection with no clinical symptoms. In this study we identified 29 inapparent dengue infections but we believe this number underestimates the proportion of inapparent dengue cases because we were not able to take blood samples from non-symptomatic subjects at Home Visit 2.
We postulated that dengue is transmitted to members of the DIC's family household during the period of the index subject's infection, and thus designed our study to detect inapparent dengue infections with a home visit organized shortly after identification of DIC. Obviously, we cannot confirm whether the index subject's DIC was always the source of infection in other family members, but we can postulate that a non-hospitalized DIC who remains at home during acute illness represents a potential source of DENV transmission to Aedes
. According to our study design, clustering of cases within a household could be the result of a single or very few infected mosquitoes biting different household members during a short period of time, perhaps within a single gonotrophic cycle as previously suggested 
. This is also consistent with a previous observation that over periods from 1 to 3 days, dengue cases were clustered within short distances, i.e., within a household 
. No mosquito captures were, however, conducted in our study to identify DENV-positive Aedes
mosquitoes. DENV sequencing would help resolve the extent of localized transmission.
We characterized subjects with acute dengue infection using virus isolation and detection of the genome. We also used NS1 antigen detection, a more recently recognised diagnostic tool. As for many tropical infectious diseases, there is an urgent need for validated diagnostic tools for dengue. In parallel with the virological techniques, we evaluated detection of the NS1 antigen with the Platelia Dengue NS1 Ag test. In this study, this test was found to have good sensitivity (83.6%; 95% CI: 78.5–88.6) and specificity (98.9%; 95% CI: 96.6–99.9) in both Asia and Latin America, as reported in previous studies 
. A recent multi-country study observed unequal sensitivity between geographical regions that remains unexplained, suggesting further assessments are needed 
. The use of viral detection antigen is particularly useful during the first five days of illness with NS1 assays that are significantly more sensitive for primary than secondary dengue 
. However, NS1 antigen could be detected in only 20% of inapparent DENV-infection. This finding suggests that NS1 antigen may have a role in dengue disease pathogenesis and also indicates that this test cannot be relied upon for detection of inapparent dengue infection.
By comparing HHM not infected with dengue with those presenting with inapparent dengue infection, we showed that neutrophil and monocyte counts were early indirect biological markers of dengue infection, whereas platelet counts and the frequency of IgG detection at the first visit did not differ between the two groups (). A comparison of inapparent dengue-infected HHM with symptomatic dengue-infected subjects showed that lymphocyte counts and detection of the NS1 antigen differed significantly between these two groups (). Moreover, the NS1 antigen was detected during the acute phase in most of the dengue cases tested, and the sensitivity of this test was even higher in severe dengue cases (26/28, Table S1
), possibly reflecting higher viral loads. These findings may indirectly reflect the progression of the immune response to DENV, leading in some cases to severe acute lymphopenia and a lack of virological control, with high rates of NS1 antigen circulation in the blood that may be correlated with high-level or prolonged viremia 
. Severe dengue cases were also more likely to be male, to have lower monocyte counts or normal liver enzyme levels, and to be infected with DENV-2, although quantitative RT-PCR did no permit study of the magnitude of the viremia. We showed that half of the severe dengue cases had not previously been infected with DENV, as confirmed by the occurrence of DENV IgG seroconversion during convalescent phase 
. In all dengue-infected subjects, including inapparent, we observed a decrease in neutrophil and monocyte counts. On one hand, it may suggest a direct effect of dengue illness on hematopoiesis, although such an effect is in conflict with data reported elsewhere in the literature 
. On the other hand, DENV is detected in peripheral monocytes during acute disease, and the infection of monocytes leads to cytokine production, suggesting that virus-monocyte interactions are relevant to pathogenesis 
. Moreover, DENV can induce apoptosis in monocytes, and this may lead to decreases in the number of these cells in severe dengue cases 
In this study we only observed severe dengue cases in South-East Asia. Disease severity and pathogenesis remain largely unexplained and certainly related to complex interactions of several factors, including virus strain, immune response to previous dengue infection and host genetic background. The introduction of the Asian 1 DENV-2 genotype into the Americas in the 1980s led to the emergence of severe dengue cases on this continent. Following this introduction a new genotype emerged, named Asian/American DENV-2 genotype 
. During the study period, this Asian/American genotype was circulating in French Guiana (Philippe Dussart, personal data) and probably in the north of Brazil, however DENV-2 did not cause an outbreak and we did not report any severe dengue case among Brazilian subjects.
Two constraints of the study design deserve mention. All methods (biological markers, virological testing, NS1 antigen detection and IgM serology) were standardized across the four reference laboratories, with the exception of the IgG ELISA. As a consequence, we were unable to calculate the IgM/IgG ratio 
. However, as the intention was to include dengue cases during the acute phase of infection, this ratio was not a crucial endpoint for the study. Another constraint of this study was that we did not include infants and children below 24 months of age in the DENFRAME project. However, several previous reports already provide insight into the epidemiology of dengue in this specific population 
These findings confirm the complexity of dengue disease in humans and the need to strengthen multidisciplinary research efforts to improve our understanding not only of virus transmission but also host responses to DENV in various human populations. It will therefore be interesting, based on clinical data and biological samples collected in this study, to further evaluate the host susceptibility to dengue infection and disease using family-based association analyses. Moreover, we think that technological transfer of standardized diagnostic methods in laboratories based in tropical countries is essential if we are to estimate disease burden and to optimize vector control interventions. Together with improvements in clinical care for dengue patients and better understanding of dengue pathogenesis, the development of a preventive vaccine and antiviral drugs would complete the arsenal of weapons for combating dengue worldwide.