Fourteen years after a drug trial had been published, investigators —using multiple sources, including regulatory and legal documents—restored a version of the same trial and came to completely opposite conclusions.

The original study of 275 adolescents with depression—Study 329—concluded “paroxetine is generally well tolerated and effective for major depression in adolescents.” Yet the 2015 reanalysis in The BMJ, under the restoring invisible and abandoned trials (RIAT) initiative, is utterly at odds with this finding: “antidepressant trial finds popular drug ineffective and unsafe for adolescents.”

We know trial 329 widely influenced prescribing of paroxetine. Problems, however, emerged straightaway: the trial was largely ghostwritten, leading to the manipulation of data favouring the medication. Outcome definitions were changed—making an insignificant result significant—and children with suicidal thoughts or ideation were wrongly reported as emotionally labile or as hospitalised—downplaying the suicidal risks.

The problems of unreported and all too often misreported trials led to the RIAT initiative. This initiative calls for sponsors and investigators of abandoned studies to publish unreported trials or to republish those trials with concerns over their findings.

Using publicly available clinical study reports, other publicly available documents, and individual patient data, along with 77 000 pages of de-identified individual case report forms, the RIAT authors reconstructed the 329 trial. The original 329 trial publication reported 265 adverse events with paroxetine compared to the restored trial’s 481 events. Also, the original publication only reported 5 suicidal and self-injurious behaviours in the 93 participants on paroxetine; the RIAT paper—after detailed examination of the case report forms—reported 11 such events.

This reanalysis provides some insight into why GSK, in 2012, paid a fine of $3bn, part of which was for the unlawful promotion of paroxetine for patients under the age of 18. But it also questions the fundamental nature and basis of clinical research, the ethical principles underpinning research—including the impact conflicts of interest have on distorting the truth—and a journal publication process that, at times, allows inaccuracies to persist.

What is promptly needed to restore public trust is an overhaul of the current trial publication system. Journals should openly allow corrections and comments to be addressed: science should be self-correcting. Otherwise, we end up with a situation where an incorrect study, cited 576 times (oops, and 577 times if you are reading this), remains in the public domain, further reinforcing public suspicion over clinical trial results; making it more difficult to discern beneficial effects from harmful ones.

Carl Heneghan is professor of EBM at the University of Oxford, director of CEBM, and a GP.

His research interests span chronic diseases, diagnostics, use of new technologies, and investigative work with The BMJ on drugs and devices that you might stumble across in the media. He is also a founder of the alltrials.net campaign.

I declare that I have read and understood BMJ Policy on declaration of interests and I hereby declare the following interests: Carl Heneghan is a co founder of the AllTrials initiative and has an active interest in all studies being reported and published in full. CH has also been a co-recipient of a WHO grant on barriers to effective publication in public health emergencies and received a UK NIHR research grant for the update and amalgamation of two Cochrane reviews using unpublished data. He has received expenses and payments for his media work and for his teaching.