Supplemental Tables and Figures

Supplemental table 1: Codes used to identify any bleeding events in routine data Supplemental table 2: Algorithm used to categorise major bleeding events in routine data Supplemental table 3: Codes used to identify any serious vascular events in routine data Supplemental table 4: Agreement of routine data versus adjudicated direct follow-up for any major bleeding, by subgroups Supplemental table 5: Comparison of routine data event date versus adjudicated direct follow-up Supplemental table 6: Agreement of routine data versus adjudicated direct follow-up, by routine data algorithm Supplemental table 7: Agreement of routine data versus adjudicated direct follow-up, by subcomponents of the major bleeding outcome Supplemental table 8: Major bleeding events identified in adjudicated direct follow-up alone or routine data only Supplemental table 9: Comparison of estimated treatment effects between adjudicated direct followup and routine data follow-up for the aspirin randomised comparison Supplemental table 10: Effect of allocation to aspirin versus placebo on any major bleeding, by source of outcome data Supplemental table 11: Effect of allocation to aspirin versus placebo on serious gastrointestinal and other major bleeding, by routine data algorithm Supplemental table 12: Comparison of bleeding categories between pre and post adjudicated direct follow-up Supplemental table 13: Comparison of bleeding categories between pre and post adjudicated direct follow-up, by components of the serious gastrointestinal bleeding outcome Supplemental table 14: Comparison of bleeding categories between pre and post adjudicated direct follow-up, by components of the other major bleeding outcome Supplemental figure 1: Flow diagram of post hoc analyses in the ASCEND trial Supplemental figure 2: Bleeding severity criteria for ASCEND adjudicated direct follow-up Supplemental figure 3: Sources of major bleeding events in ASCEND Supplemental figure 4: Effect of allocation to aspirin versus placebo on any major bleeding, by routine data algorithm BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Heart

Percentages in parentheses are % of total number of ASCEND participants. *Heterogeneity test compares kappa statistics between subgroups. **We categorised the predicted 5-year risk of serious vascular event without the use of aspirin or fatty acids as follows: low risk as less than 5%, moderate risk as 5% to less than 10%, and high risk as 10% or more. CI = Confidence interval.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Other major bleeding 29 (56.9%) 10 (19.6%) Percentages in parentheses are % of total number of ASCEND participants. *All algorithms presented were considered, selection of the main algorithm for major bleeding in routine data was conducted by clinicians when blinded to the results of the randomised comparison. Selection were based on agreement statistics, simplicity of algorithm, and a definition of major bleeding broadly similar to ASCEND (see online supplemental figure 2). 1st/2nd = Bleeding code in 1 st or 2 nd position. CI = Confidence interval. Emergency = Emergency admission. Nights = Nights in hospital. Primary = Bleeding code as primary diagnosis.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Percentages in parentheses are % of total number of ASCEND participants with an outcome in either adjudicated follow-up alone or routine data only. *Only events occurring within 90 days of the reported major bleeding event were used for these analyses.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Log-rank methods were used to calculate the rate ratio and 95% confidence intervals. *Bootstrap methods were used to calculate the 95% confidence intervals for the difference in rate ratios between adjudicated follow-up and routine data. These methods used 1000 resamplings, with replacement, where the difference in point estimate was recalculated in each bootstrap sample.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Log-rank methods were used to calculate the rate ratio and 95% confidence intervals. *Events identified in routine data only did not undergo adjudication and were excluded if within 90 days of an adjudicator refuted major bleeding event. Observed absolute effects of aspirin versus placebo on major bleeding for adjudicated direct follow-up and routine data combined were +8.4 per 5000 person-years (mean standard error ± 2.6). CI = Confidence interval.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Log-rank methods were used to calculate the rate ratio and 95% confidence intervals. *All algorithms presented were considered, selection of the main algorithm for major bleeding in routine data was conducted by clinicians when blinded to the results of the randomised comparison. Selection were based on agreement statistics, simplicity of algorithm, and a definition of major bleeding broadly similar to ASCEND (see online supplemental figure 2). 1st/2nd = Bleeding code in 1 st or 2 nd position. CI = Confidence interval. Emergency = Emergency admission. Nights = Nights in hospital. Primary = Bleeding code as primary diagnosis.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)

Supplemental figure 4: Effect of allocation to aspirin versus placebo on any major bleeding, by routine data algorithm
Log−rank methods were used to calculate the rate ratio and 95% confidence intervals. Absolute effects expressed as numbers of events per 5000 person−years, plus−minus values are means ± standard error. *All algorithms presented were considered, selection of algorithm for major bleeding in routine data was conducted by clinicians when blinded to results of the randomised comparison. Selection were based on agreement statistics, simplicity of algorithm, and a definition of major bleeding broadly similar to ASCEND (see online supplemental figure 2). 1st/2nd = Bleeding code in 1 st or 2 nd position. CI = Confidence interval. Emergency = Emergency admission. Nights = Nights in hospital. Primary = Bleeding code as primary diagnosis.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)