In nature, biological vision systems are renowned for their efficient perception of dynamic environments. Inspired by these, event cameras emulate the sparse and low-latency characteristics of biological vision, offering new possibilities for stable and real-time state estimation in challenging scenarios. Although event-based odometry has gained significant attention in recent years, research incorporating depth information remains limited. This paper introduces DAB-VIO, a Visual-Inertial Odometry method that leverages the complementary strengths of event cameras, RGB-D images, and IMU measurements. The method compensates for motion in the event stream using IMU data combined with optimized estimation, converting the stream into a Time Surface for stable feature tracking. The proposed method can fully capitalize on the high temporal resolution and low power consumption of event cameras. Performance comparisons on the open-source VECtor dataset, which includes high dynamic range and aggressive motion scenarios, demonstrate the superior performance of the proposed method. The results indicate that the algorithm maintains high accuracy while proving its robustness in the challenging scenarios.