Virtual reality (VR) head-mounted displays (HMDs) can be used in a variety of body postures, including standing, sitting, and lying down (supine). If VRHMD can be used effectively even from the supine position, it will not only allow bedridden people to comfortably access VR content, but also evolve VRHMDs into familiar devices that can be used in any body position, like smartphones. Though existing interaction methods for VR, such as moving the controller in the air, are assumed to be used while standing or sitting, the gaze input interface is a promising method that is available while supine. However, there are some issues unique to gaze input. It is unsuitable for fine pointing because the estimated gaze position movers continually, even if the user tries to gazes at a particular target. We propose a hybrid method that uses both gaze and the controller to overcome this issue. The proposed method combines controller rotation, a single button, and gaze input from an eye tracker. An experiment is conducted to compare three methods: the proposed method, a method using only gaze, and a method using only the controller. The experiment included the task of clicking tiny buttons that required high pointing resolution. The results show that when the eye tracker works as intended, the proposed method has an input speed almost equal to or faster than the method that uses only the controller, and it is superior to the method that uses only the eye tracker.