Many clinical trials are followed-up throughout multiple phases, or multiple periods of time. For example, CYD-14 and CYD-15 are 2 randomized, observer-blind, placebo-controlled multicenter dengue trials funded by Sanofi Pasteur. Participants in CYD-14 or CYD-15 were followed-up for 25 months under an active phase (AP) in which an active surveillance system was used to identify symptomatic, virologically-confirmed dengue (VCD) cases and hospitalized, viologically-confirmed dengue (HVCD) cases. After 25 months, most participants were followed-up for at least 12 months of a hospital phase (HP) in which HVCD cases were identified if a participant had a febrile illness and required hospitalization. Due to the difference in the dengue case surveillance, the HVCD rate (naively estimated based on the data from passive surveillance) in the HP is likely to be lower than that in the AP. The cases from the AP could represent the gold standard for ascertainment against the passive surveillance in the HP. Passive surveillance in the HP would likely lead to under-ascertainment of cases. That is, some cases may be misclassified as controls (non-cases) due to passive surveillance. In this paper, we develop methodology to analyzing the follow-up data from the AP and HP. Our methodology is based on the use of serum antibody neutralization titers as predictors of the dengue outcome to adjust for under-ascertainment of the HVCD in the HP. We apply our methods to CYD14 and CYD15 trials, and we show that if under-ascertainment in the HP is a possible scenario, then statistical inference using antibody titers can adjust for the bias in estimating the vaccine efficacy against HVCD. We demonstrate that our proposed methods are valid for a combined analysis of the data for vaccine efficacy against HVCD from the 2 follow-up phases.