Abstract-Device-to-device (D2D) communications in cellular networks enable direct transmissions between user equipments (UEs). If the source UE (SUE) and the destination UE (DUE) are far away from each other or the channel between them is too weak for direct transmission, then multi-hop D2D communications, where relay UEs (RUEs) forward the SUE's data packets to the DUE, can be used. In this paper, we propose an energyefficient optimal adaptive forwarding strategy (OAFS) for multihop D2D communications. OAFS adaptively chooses between the best relay forwarding (BRF) mode and the cooperative relay beamforming (CRB) mode with the optimal number of RUEs, depending on which of them provides the higher energy efficiency (EE). To reduce the computational complexity for selecting the optimal RUEs for CRB mode, we propose a low-complexity suboptimal adaptive forwarding strategy (SAFS) that selects between the BRF and the CRB with two RUEs by comparing their EE. Furthermore, a distributed forwarding mode selection approach is proposed to reduce the overhead for forwarding mode selection. The analytical and simulation results show that OAFS and SAFS exhibit significantly higher EE and spectral efficiency (SE) than BRF, CRB, direct D2D communications and conventional cellular communications. SAFS is almost as energy-and spectral-efficient as OAFS.