Introduction: The emergence of real-life health databases has opened the door to studies for signal detection in pharmacovigilance without the formulation of an a priori hypothesis, i.e., without defining a drug/adverse drug event (ADR) pair. Our objective was to perform a systematic review of this type of study and the statistical methods used in this context. Methods: Studies about drug signal detection without a priori hypotheses published in the MEDLINE database between 2012 and 2021 were included. Database name and type, statistical methods, ATC class for the studied drug(s) and SOC MedDRA classification for the studied ADR were extracted. Results: Ninety-two studies were included. Pharmacovigilance databases were the most used type of database. Most studies performed a disproportionality analysis using frequentist or Bayesian methods. The most studied drug classes were anti-infectives, nervous system drugs, and antineoplastics and immunomodulators. No common procedure was implemented to correct for multiple testing. Conclusions: There are very few statistical methods used for drug signal detection without a priori hypotheses, with no consensus-based method and no interest in multiple testing correction. This review argues for the establishment of guidelines to perform such studies.