Larger effect sizes in nonrandomized studies are associated with higher rates of EMA licensing approval

Benjamin Djulbegovic, Paul Glasziou, Farina A Klocksieben, Tea Reljic, Magali VanDenBergh, Rahul Mhaskar, John P A Ioannidis, Iain Chalmers

Research output: Contribution to journalArticleResearchpeer-review

9 Citations (Scopus)
128 Downloads (Pure)


OBJECTIVES: Evaluate how often the European Medicines Agency (EMA) has authorized drugs based on non-randomized studies and whether there is an association between treatment effects and EMA preference for further testing in RCTs.

STUDY DESIGN AND SETTING: We reviewed all initial marketing authorizations in the EMA database on human medicines between 1995 and 2015 and included authorizations granted without randomized data. We extracted data on treatment effects and EMA preference for further testing in RCTs.

RESULTS: Of 723 drugs, 51 were authorized based on non-randomized data. These 51 drugs were licensed for 71 indications. In the 51 drug-indication pairs with no preference for further RCT testing, effect estimates were large [OR 12.0 (95% CI: 8.1 to 17.9)] compared to effect estimates in the 20 drug-indication pairs for which future RCTs were preferred [OR 4.3 (95%CI 2.8 to 6.6)], with a significant difference between effects (p=0.0005).

CONCLUSIONS: Non-randomized data were used for 7% of EMA drug approvals. Larger effect sizes were associated with greater likelihood of approval based on non-randomized data alone. We did not find a clear treatment effect threshold for drug approval without RCT evidence.

Original languageEnglish
Pages (from-to)24-32
Number of pages9
JournalJournal of Clinical Epidemiology
Early online date9 Feb 2018
Publication statusPublished - Jun 2018


Cite this