Tool condition monitoring (TCM) in machining operations is crucial to maximise the useful tool life while reducing the risks associated with tool breakage. Unlike progressive tool wear, tool breakage occurs randomly, with more severe implications for workpiece quality, machining system stiffness, and even operator safety. Existing literature reviews on TCM focus on tool wear monitoring, including wear state recognition and remaining useful life prediction. However, a comprehensive review of tool breakage monitoring (TBM) techniques is lacking. Generic signal processing and intelligent decision-making methods cannot fully satisfy the practical requirements of the TBM. In addition, developing and evaluating TBM models using imbalanced data is more challenging. Herein, we present the first systematic review on TBM to bridge these limitations, and provide adequate guidance for avoiding catastrophic tool failures during cutting processes. Signal acquisition, feature extraction, and decision-making methodologies for the TBM are outlined and compared with related techniques for tool wear monitoring. The effects of data imbalance on TBM models are considered, and feasible solutions are provided at the data and algorithm levels. Finally, the challenges faced by the TBM are discussed, and potential research directions are suggested. The research and application of TBM techniques will certainly better empower various machining operations in response to intelligent manufacturing demands.