Dengue incidence has increased globally, but empirical burden estimates are scarce. Prospective methods are best-able to capture all severities of disease. CYD14 was an observer-blinded dengue vaccine study conducted in children 2-14 years of age in Indonesia, Malaysia, Thailand, the Philippines, and Vietnam. The control group received no vaccine and resembled a prospective, observational study. We calculated the rates of dengue according to different laboratory or clinical criteria to make inferences about dengue burden, and compared with rates reported in the passive surveillance systems to calculate expansion factors which describe under-reporting. Over 6,933 person-years of observation in the control group there were 319 virologically confirmed dengue cases, a crude attack rate of 4.6%/year. Of these, 92 cases (28.8%) were clinically diagnosed as dengue fever or dengue hemorrhagic fever by investigators and 227 were not, indicating that most symptomatic disease fails to satisfy existing case definitions. When examining different case definitions, there was an inverse relationship between clinical severity and observed incidence rates. CYD14's active surveillance system captured a greater proportion of symptomatic dengue than national passive surveillance systems, giving rise to expansion factors ranging from 0.5 to 31.7. This analysis showed substantial, unpredictable and variable under-reporting of symptomatic dengue, even within a controlled clinical trial environment, and emphasizes that burden estimates are highly sensitive to case definitions. These data will assist in generating disease burden estimates and have important policy implications when considering the introduction and health economics of dengue prevention and control interventions.