Objective: To compare clinical trial information contained in internal company documents with publications, to examine validity and completeness. Methods: We examined clinical trials, sponsored by Parke-Davis/Pfizer, assessing off-label uses of gabapentin. For each trial, we compared internal company documents, including protocols, statistical analysis plans (SAPs), and research reports, with publications. One author extracted data on number of participants for each item on the CONSORT statement’s flow diagram as well as details of types of analyses for primary outcomes (e.g., ITT, per protocol). A second author verified abstracted data against original documents. Results: We identified 12 published trials for four off-label indications of gabapentin: migraine prophylaxis (two trials), bipolar disorders (three trials), and neuropathic pain (seven trials). The number of randomized participants reported in the publication disagreed with the number in the internal company research report for 3/12 trials. Seven different types of analyses for the primary outcome were described in trial protocols, SAPs and publications: ITT, modified ITT, per protocol, evaluable, efficacy evaluable, evaluable for efficacy, and ‘‘population to be analysed.’’ Seven different definitions for ITT analysis were described in the protocols and publications across 7/12 trials and each definition used different criteria to specify which participants were included in the ITT analysis (Table 1). 5/12 trials did not specify an ITT analysis. There was a disagreement in the definition of ITT between the protocol and publication for all four trials that defined ITT analysis in both the protocol and publication. Disagreements were also observed for other analyses. Conclusions: Reporting of key analysis information from selected industry-sponsored trials was inconsistent, incomplete, and may be incorrect. Trial protocols should be made publicly available, and both protocols and published reports must specify the primary type of analysis and criteria used to include participants in the analysis.