Trial samples. For each product in our sample, we create three trial samples: (1) ‘all trials,’ (2) ‘patient trials’ and (3) ‘FDAAA applicable trials.’ The ‘all trials’ sample contains all trials submitted to the FDA for initial approval of each reviewed product (i.e., all trials in an approved new drug application (NDA)). The ‘patient trials’ sample contains only trials in the targeted patient population for the approved indication (excluding, e.g., trials conducted in healthy volunteers). ‘FDAAA applicable trials’ are those highly likely to be subject to FDAAA or Final Rule trial registration and results reporting requirements, generally phase II and III controlled trials begun after 27 September 2007 or ongoing as of 26 December 2007 that (1) have at least one US site; (2) were conducted under an FDA investigational new drug application; or (3) involved a drug, biologic or device manufactured in the USA and exported for research.
Transparency measures, product level. We examine three outcome measures for the trials supporting each product’s approval. The first pertains to trial registration: we determine whether trials in the ‘all trials’ and ‘patient trials’ samples for each product are registered within 6 months of initial FDA approval of each product. Second, for trials completed by a product’s FDA approval, we determine whether results are reported in a public registry or published in a journal indexed by PubMed, Google Scholar or EMBASE within 6 months of initial FDA approval. Third, we examine FDAAA implementation—that is, whether applicable trials are registered within 21 days of their start date and results reported within 30 days of initial FDA approval of each product.
Data sharing measures, company or sponsor level. We examine companies’ data sharing practices using five measures: (1) whether they have a public policy committing to sharing analysis-ready datasets and clinical study reports (CSRs) for applicable studies; (2) whether their policy explains how such data can be requested; (3) whether the policy commits to making data available by 6 months after approval by the FDA or European Medicines Agency or 18 months after a trial’s completion date, whichever was later; (4) whether the company reports the number of data requests received and how each was handled (granted or denied); and (5) the proportion of ‘data sharing applicable’ trials registered in a public registry. The overall data sharing score for each company or institution is the average of the five component scores.
Scoring companies and sponsors on their overall transparency. Lastly, we determine an overall company or sponsor transparency score. For companies and sponsors with only one product approved by the FDA in our target year of focus, we averaged their scores on their (1) patient trials analysis, (2) FDAAA compliance and (3) data sharing analysis. Each component is weighted equally for consistency with past GPS analyses, and because each component is essential to achieving the full benefits of transparency. For companies or sponsors with multiple products approved, we pool the trials from all their products into our three trial samples and then apply our outcome measures to the pooled samples. We then calculate an overall score by averaging the pooled components. All sponsors, public and private, are held to the same standards.
A complete description of the methodology for each reporting year is available below.