It's been a backburner project of mine to data-mine the publicly-available clinical study information...while not all results are reported, there are useful meta-analyses that can be done (i.e. which companies are least likely to report, under what circumstances, such as study size and disease, are studies likely to omit info?).

It's not immediately evident, but ClinicalTrials.gov does provide their database in a structured format...probably the most convenient endpoint is this giant sitemap, which they have generated for scraping uses: http://www.clinicaltrials.gov/ct2/about-site/crawling