Study 329

The trial failed. On all eight pre-specified outcomes paroxetine was no better than placebo. The company acknowledged this internally in 1998; a position paper later filed as a court exhibit records the decision to manage the presentation of the data rather than the finding.

The published paper concluded the opposite: that paroxetine was generally well tolerated and effective. It reached that conclusion by shifting to outcome measures not specified in the protocol, and by reporting the negative results in a table while writing a conclusion that did not follow from them. Adverse events were minimised — suicidal thinking and behaviour, more frequent in the paroxetine arm, were coded under 'emotional lability'.

The article was ghostwritten. A public relations firm engaged by the company drafted it; the named academic authors signed a manuscript they had not written, and the underlying data were not available to them or to anyone else.

The regulators reached the opposite conclusion once they saw the full dataset. Britain's MHRA analysed study 329 alongside GSK's other paediatric paroxetine trials in 2003 and found no evidence of efficacy in children and adolescents and, in its words, robust evidence of a causal link with suicidal behaviour. The MHRA and the FDA advised against prescribing to under-18s the following month.

In 2004 the New York attorney general sued GSK for withholding trial data; the settlement required the company to post its results publicly. In 2012 the Department of Justice imposed a $3 billion resolution that named study 329 specifically — the misleading article and the off-label promotion of the drug for adolescents that it enabled.

In 2015 the BMJ published a reanalysis under the Restoring Invisible and Abandoned Trials initiative, working from the raw case report forms. Joanna Le Noury and colleagues found neither paroxetine nor imipramine better than placebo, more suicidal ideation and behaviour in the paroxetine group, and more cardiovascular events in the imipramine group.