The epidemiology of dengue in southern Vietnam: distribution of disease and determinants of risk in an endemic setting

Over the past half-century dengue has re-emerged as a major public health challenge throughout the tropical and sub-tropical world, associated with global dispersal of both the mosquito vector, Aedes aegytpi, and the four dengue virus serotypes. No vaccine nor therapeutic drug is currently licensed for dengue, with vector control and careful clinical management the only tools currently available for prevention and control. This thesis explores the distribution of dengue in southern Vietnam, where dengue is endemic, and the factors that shape individual and population-level infection risk and disease outcome. An improved understanding of the distribution of dengue risk in endemic settings is essential for optimising clinical management, maximising the effectiveness of current vector control interventions, identifying populations most likely to benefit from novel therapeutics and vaccines, and informing the design of trials to evaluate those novel interventions. Analysis of a large retrospective dataset of more than 100,000 dengue patients admitted to three hospitals in Ho Chi Minh City (HCMC) over a 14-year period demonstrated a significantly increased risk of mortality among girls and very young children. This raises questions about gender- and age-specific dengue pathogenesis and health-seeking behaviour, and emphasises the need for vaccines and antiviral drugs to be trialled in and delivered to pre-school-aged children in order to achieve the greatest impact on disease burden and mortality. The dynamics of dengue in endemic areas are characterised by strong seasonality and multiannual epidemic peaks. Spatiotemporal analysis of a ten-year time series of dengue surveillance data from southern Vietnam demonstrated that the timing and magnitude of annual dengue epidemics were significantly spatially correlated, suggesting local biological and ecological drivers operating at a scale of 50-100km. Interestingly, seasonal peaks in dengue occurred later in HCMC each year and with lower per capita incidence than in most rural provinces. Dengue incidence during the dry inter-epidemic months accounted for 63% of the variability in the magnitude of the subsequent seasonal epidemic peak. These findings provide insights into the processes underlying dengue transmission and can aid the targeting of prevention and control efforts. Spatial heterogeneity in dengue transmission was explored further through a prospective community-based study of focal dengue transmission around households of confirmed dengue cases in HCMC. An assumption that dengue virus (DENV) infections cluster within and around households underlies the commonly employed strategy of perifocal insecticide spraying around the homes of reported dengue cases and their neighbours, which is a central feature of the national dengue control program in Vietnam, despite a lack of evidence for its effectiveness. This study demonstrated limited evidence for household clustering of DENV infections in highly urban HCMC, and importantly no evidence for clustering of prospectively detected infections within two weeks of a dengue index case, suggesting that targeted insecticide spraying is likely to have limited impact on dengue transmission in this setting. This study used dried blood spot (DBS) samples for dengue serological testing of community participants, based on the results of a hospital-based evaluation of less invasive alternatives to venous blood samples for dengue diagnostics, which demonstrated high sensitivity and specificity of DBS compared to plasma. This research provides new insights into the epidemiology of dengue in southern Vietnam, highlighting patient groups at increased risk of poor clinical outcomes and elucidating spatiotemporal patterns in dengue transmission at both broad and fine scales. These findings emphasise that the assumptions underlying current reactive vector control, specifically household clustering of DENV transmission, may not hold true in all settings. Rather spatial and temporal variation in DENV transmission seems to be setting dependent, and analysis of local epidemiological data is needed in order to optimise the targeting of current and novel interventions.