AIM: There is no consensus about what constitutes a dose-limiting toxicity (DLT) in phase I cancer clinical trials. We aimed to evaluate how DLTs are defined in phase I trials of molecularly targeted agents (MTA). METHODS: We retrieved all phase I trials testing monotherapy with an MTA published over the last decade. In each trial, all items used to define DLTs were recorded. RESULTS: Reports of 155 phase I trials evaluating 111 different MTAs were reviewed. The most frequent determinant of whether a toxicity was regarded as a DLT was severity, usually assessed using the NCI CTCAE classification. However, for any given toxicity, there was substantial variability in the degree of severity required for a toxicity to be considered a DLT. Specifications about minimum duration of toxicity, degree of reversibility, the need to delay treatment and to reduce dose-intensity because of toxicity were infrequently incorporated in the definition of DLT. The definition of DLT varied with administration schedule. Discrepancies between the initial and the final definition of DLT were reported in 25% of trials. CONCLUSIONS: While our results do not support a standardisation of the definition of DLT, the inclusion of following specifications in its definition when relevant would reduce the heterogeneity observed across trials: (1) DLT assessment period, (2) absolute severity according to NCI CTCAE classification as well as severity relative to baseline status, (3) minimum duration of toxicity, (4) reversibility of toxicity within a certain period of time, and (5) necessity to delay treatment or to reduce dose-intensity.