In pivotal clinical effectiveness trials, the primary endpoint needs to be precisely defined and quantified because any misclassification could introduce noise and possible bias, potentially leading to incorrect trial conclusions. The underlying premise is that adjudication will be more accurate, reduce noise, and limit any misclassification by the site investigator, with any residual errors being consistently distributed across the treatment groups. Some observers have been critical of the absence of endpoint adjudication in trials using these data and point to inaccuracies of hospital coding and statistics.25 In systems where payment to the health care service is reliant on coding data, external factors such as reimbursement incentives or local practice variations can cause bias.26 In large international trials, a further concern is the heterogeneity and reporting biases of different health care systems. ...subsequent follow up for over 20 years has now been done entirely through routinely collected data.34,35 Similar findings have also been found in the more contemporary aspirin for primary prevention in people with diabetes mellitus (ASCEND) trial,36,37 where hospital episode statistics in England, UK were compared with adjudicated clinical endpoints and effect size estimates for the primary outcome were again very similar (personal communication, Jane Armitage, University of Oxford).