Background: Several risk factors for coronary artery disease (CAD) have been described, some of which are genetically determined. The use of a polygenic risk score (PRS) could improve CAD risk assessment, but predictive accuracy according to age and sex is not well established. Methods: A PRS _CAD including the weighted effects of >1.14 million single nucleotide polymorphisms associated with CAD was calculated in UK Biobank (n=408 422), using LDpred. Cox regressions were performed, stratified by age quartiles and sex, for incident myocardial infarction (MI) and mortality, with a median follow-up of 11.0 years. Improvement in risk prediction of MI was assessed by comparing PRS _CAD to the pooled cohort equation with categorical net reclassification index using a 2% threshold (NRI ^0.02 ) and continuous NRI (NRI ^>0 ). Results: From 7746 incident MI cases and 393 725 controls, hazard ratio for MI reached 1.53 (95% CI, 1.49–1.56; P =2.69×10 ⁻²⁹⁶ ) per SD increase of PRS _CAD . PRS _CAD was significantly associated with MI in both sexes, with a stronger association in men (interaction P =0.002), particularly in those aged between 40 and 51 years (hazard ratio, 2.00 [95% CI, 1.86–2.16], P =1.93×10 ⁻⁷² ). This group showed the highest reclassification improvement, mainly driven by the up-classification of cases (NRI ^0.02 , 0.199 [95% CI, 0.157–0.248] and NRI ^>0 , 0.602 [95% CI, 0.525–0.683]). From 23 982 deaths, hazard ratio for mortality was 1.08 (95% CI, 1.06–1.09; P =5.46×10 ⁻³⁰ ) per SD increase of PRS _CAD , with a stronger association in men (interaction P =1.60×10 ⁻⁶ ). Conclusions: Our PRS _CAD predicts MI incidence and all-cause mortality, especially in men aged between 40 and 51 years. PRS could optimize the identification and management of individuals at risk for CAD.