Link to original article.

http://journals.plos.org/plosmedicine/article?id=10.1371/jou...

there is a section on 'data', of which here is one sample

The data submitted by GlaxoSmithKline on paroxetine for review by the Committee on Safety of Medicines Expert Working Group are described as a complete set of data from all placebo-controlled trials of this drug [5]. The use of this dataset thus involves no selection by the authors, and any selection bias there might have been on the part of the company seems unlikely to have increased the size of the problem. Data from placebo-controlled trials of sertraline in children are also presented, as these also offer a complete dataset, so minimising any selection bias.