Background: Dementia is one the major problems of aging societies, and, novel and effective non-drug therapies are required as interventions in the oldest-old to prevent cognitive decline.Objective: This study aims to examine the efficacy and safety of reminiscence using immersive virtual reality focusing on anxiety that often appears with cognitive decline. The secondary objective is to reveal the preference for VR image types for reminiscence: live-action or computer graphics.Methods: This was a pilot, open-label, and randomized crossover study which was conducted (...) on January 2020 at a single nursing home. The subjects were randomly divided into two groups in equal numbers, and they alternately viewed two types of VR images themed on the mid- to late Showa era in Japan. In group A, the CG images were viewed first, and then the LA images were viewed. In group B, the images were viewed in the opposite order. Before VR viewing, subjects responded to Mini-Mental State Examination Japanese version and State-Trait Anxiety Inventory Japanese version. After viewing the first and second VR, subjects responded to STAI and the numerical rating scale for satisfaction and side effects.Results: Ten subjects participated in this study. The values of analyses are presented in the mean. The age was 87.1 years, and the MMSE was 28.5. The total STAI score before VR viewing was 36.1, but it significantly decreased to 26.8 after the first VR viewing, and further decreased to 23.4 after the second VR viewing. The NRS score for satisfaction tended to be higher after viewing LA in group A vs. 8.6, P = 0.0993), while in group B, the score after CG was slightly lower than that after LA. There were no serious side effects.Conclusions: This study suggests that iVR reminiscence can reduce anxiety in the oldest-old without causing serious side effects. Furthermore, the impacts might be better with LA images. (shrink)