ObjectiveTo ascertain contemporaryapproaches to the collection, reporting and analysis of adverse events (AEs)inrandomised controlled trials(RCTs)with a primary efficacy outcome.DesignA reviewof clinical trials of drug interventions from four high impactmedical journals.Data sourcesElectronic contents table of the BMJ, the Journal of the American Medical Association, the Lancet,andthe New England Journal of Medicine were searchedfor reports of original RCTs published between September 2015 and September 2016.MethodsA pre-piloted checklist was used and single data extraction was performed by three reviewers with independent check of a randomly sampled subset to verify quality. We extracted data on collection methods, assessment of severity and causality, reporting criteria, analysis methods and presentation of AE data.ResultsWe identified 184 eligible reports (BMJ n=3; JAMA n=38, Lancet n=62; and NEJM n=81).Sixty-two percent reported some form of spontaneous AE collection but only 29% included details of specific prompts used to ascertain AE data. Numbers that withdrew from the trial were well reported (80%), however only 35% of these reported whether withdrawals were due toAEs.Results presented and analysis performed was predominantly on ‘patients with at least 1event’ with 84% of studies ignoring repeated events. Despite a lack of power to undertake formal hypothesis testing, 47% performed such tests for binary outcomes. ConclusionsThis review highlighted that the collection, reporting and analysis of AE data in clinical trials is inconsistent and RCTs as a source of safety data are underutilised. Areas to improve include reducing information loss when analysing at patient level and inappropriate practice of underpowered multiple hypothesis testing. Implementation of standard reporting practices could enable a more accurate synthesis of safety data and development of guidance for statistical methodology to assesscausality ofAEs could facilitate better statistical practice.