One of the main goals of safety management in clinical trials is to detect suspected unexpected serious adverse reactions (SUSARs). The unexpectedness concerns the nature, frequency or severity of an adverse reaction. Drug safety signals could thus be retrieved, and a study was performed to investigate whether SUSARs allow signal detection in pharmacovigilance. Data from six academic safety units were collected from 2005 to 2016. Characteristics of SUSARs were analysed and signals were identified i) by evaluating the presence of other causes, ii) by assessing the summary of product characteristics (SPC), iii) by searching for specific safety information in Pubmed and health agencies, and iv) by investigating the narrative of each case. Pharmacological plausibility was evaluated by compatible mechanism of reaction and time-to-onset. During the study period, 211 SUSARs were collected. They mostly concerned general disorders (26.1%) and protein kinase inhibitors (24.6%). After eliminating SUSARs with other causes or those considered as expected, 50 SUSARs (23.7%), involving a total of 115 drug-reaction pairs, concerned potential safety signals. Among these pairs, 12 (10.4%) were considered as pharmacologically plausible. This study indicates that one quarter of SUSARs collected in academic clinical trials refers to potential safety signals, especially for oncologic drugs. One tenth of drug-reaction pairs was considered to have a pharmacological plausibility and could merit further evaluation. This is the first study suggesting that SUSARs could be a source of safety signals and that their routine analysis should be complementary to spontaneous reporting.< Réduire