A step forward for data protection and biomedical research.

1374 www.thelancet.com Vol 387 April 2, 2016 be “the difference in home systolic blood pressure between each active drug (spironolactone, doxazosin, bisoprolol) and placebo”. The function and obligation of the SAP was to specify how the primary hypothesis would be addressed, using the prespecified primary outcome measures, capturing the maximum amount of information that has been discovered by the trial. Hierarchical co-primary endpoints permit each endpoint in the hierarchy to be tested only if all previous tests are signifi cant, so rendering it progressively harder for each hypothesis after the fi rst to yield a positive outcome. Moreover, because the SAP prespecified use of all data without selection, averaging blood pressure responses for both the low-dose and high-dose periods, our primary analyses under-represent the true superiority of spironolactone at the end of each drug cycle. This fi nding is apparent in the fi gure, showing that the 50 mg dose of spironolactone reduced blood pressure almost twice as much as 25 mg, whereas there was little advantage from increasing the dose of bisoprolol or doxazosin. The obstacles to performing all clinical trials these days are immense, with the power to control what is done for patients inversely related to evaluating patients’ real benefi ts and risks.