|Outcomes||Risk of bias||Inconsistency||Indirectness||Imprecision||Publication bias||No of participants (studies)||Relative effect (95% CI)||Quality of evidence (GRADE)|
|Myocardial infarction||Not likely†||No serious inconsistency‡||No serious indirectness§||No serious imprecision¶||Not likely**||1 082 977 (10)||1.23 (1.15 to 1.31)||Moderate††⊕⊕⊕О|
|Coronary events||Not likely†||Inconsistency‡‡||No serious indirectness§||No serious imprecision¶||Not likely**||1 530 070 (28)||1.24 (1.10 to 1.39)||Low††⊕⊕ ОО|
|Ischaemic stroke||Not likely†||No serious inconsistency‡||No serious indirectness§||No serious imprecision¶||Undetected§§||80 787 (2)||1.05 (1.01 to 1.09)||Moderate†† ⊕⊕⊕О|
†Median Downs and Black score for included studies was 60% (interquartile range 34-86%).
§Population, outcome, and intervention were consistent with question of interest, although individual studies varied.
¶No of events and participants studied in review is large, and confidence interval does not include null value.
**Estimates adjusted for publication bias did not differ from observed estimates.
††Dilution effect of single time point exposure ascertainment allows upgrading of evidence.
§§Publication bias could not be tested for two studies.