Background: The problem of ghostwriting in corporate-sponsored clinical trials is of concern to medicine, bioethics and government agencies. We present a study of the ghostwritten archival report of an industry-sponsored trial comparing antidepressant treatments for bipolar depression: GlaxoSmithKline (GSK) paroxetine study 352. This analysis is based upon publicly available evidence presented in a complaint of research misconduct filed with the Office of Research Integrity of the US Department of Health and Human Services. Objectives: We performed a deconstruction of the published study to show how primary and secondary outcome analyses were conflated, turning a 'negative' clinical trial into a 'positive' study - with conclusions and recommendations that could adversely affect patient health. Methods: The paroxetine 352 study was a randomised, double-blind, placebo-controlled, 19-site trial comparing paroxetine and imipramine in 117 patients with bipolar type I major depressive episode which was unresponsive to prior lithium carbonate therapy. Results: Analysis of the primary outcome measures found no statistically significant difference between paroxetine or imipramine versus placebo. However, the published article concluded that both drugs were efficacious versus placebo for a post hoc subgroup of patients. Conclusions: Few industry-sponsored studies gain public scrutiny. It is important to make these articles transparent to the scientific and medical community.