What is the DAPT Score? Robert W. Yeh, MD MSc MBA Smith Center for Outcomes Research in Cardiology Beth Israel Deaconess Medical Center Harvard Clinical Research Institute Harvard Medical School

Disclosures Funding The DAPT Study was sponsored by Harvard Clinical Research Institute, and funded by Abbott, Boston Scientific Corporation, Cordis Corporation, Medtronic, Inc., Bristol-Myers Squibb Company/Sanofi Pharmaceuticals Partnership, Eli Lilly and Company, and Daiichi Sankyo Company Limited and the US Department of Health and Human Services (1RO1FD003870-01). This study was supported by the National Heart, Lung and Blood Institute (K23HL118138) and Harvard Clinical Research Institute. Disclosures Proctor and consult for Abbott Vascular, Boston Scientific Expert witness testimony for Merck The DAPT Study was funded through a collaboration between the US Food and Drug Administration, industry and the Harvard Clinical Research Institute. This analysis was funded through a grant from the National Heart Lung and Blood Institute and by HCRI.

Risk Difference (Continued Thienopyridine – Placebo), 12-30M Background In the DAPT Study, continuation of dual antiplatelet therapy beyond 12 months reduced ischemic complications after coronary stenting compared with aspirin alone, yet increased moderate or severe bleeding. Risk Difference (Continued Thienopyridine – Placebo), 12-30M Stent Thrombosis Death, MI, Or Stroke (MACCE) Myocardial Infarction GUSTO Mod/Severe Bleed Death HR 0.47 (0.37–0.61) P<0.001 HR 0.71 (0.59–0.85) P<0.001 In the DAPT Study, continuation of dual antiplatelet therapy beyond 12 months reduced stent thrombosis and myocardial infarction after coronary stent treatment compared with aspirin alone. This benefit was counterbalanced by an increase in moderate or severe bleeding, as well as a nominal increase in all-cause mortality among drug-eluting stent-treated patients. However, while these results reflect average treatment effects in the study population, there may be some individuals who are more likely to derive greater overall benefit from continuation of dual antiplatelet therapy, and others who are more likely to be harmed. Mauri, Kereiakes, Yeh et al. NEJM. 2014 Dec 4:371:2155-66.

The Goal of the DAPT Score We need a decision tool to identify whether an individual patient is more likely to derive benefit or harm from continuation of dual antiplatelet therapy beyond 1 year. Simultaneously accounting for risks of ischemia AND bleeding with continued therapy. The objective of this analysis, therefore, was to develop a clinical decision tool that could be used to identify whether an individual patient is more likely to derive benefit or harm from continuation of dual antiplatelet therapy beyond 1 year. While prior risks models have separately identified predictors of ischemic risk and bleeding risk, our goal was to develop a single tool that simultaneously accounted for both of these risks.

Design Inclusion: FDA-approved DES or BMS, candidates for thienopyridine Excluded: Oral anticoagulant therapy; life expectancy < 3y Randomized: Free from MI, stroke, repeat revascularization, moderate/severe bleeding, and adherent with therapy at 12 months In the DAPT Study, eligible patients who had completed a year of dual antiplatelet therapy after PCI were randomized to continuation of thienopyridine plus aspirin, vs. aspirin alone for an additional 18 months. Notable exclusions included patients who were on oral anticoagulant therapy or had life expectancy< 3years, and only those patients free from MI, stroke, repeat revasculrization, or bleeding, and were adherent with antiplatelet therapy at 12 months were randomized. Mauri, Kereiakes et al. AHJ. 2010;160(6): 1035-41. ClinicalTrials.gov number NCT00977938 5 5 5

Methods – Models to Predict Ischemic and Bleeding Events Development of 2 Prediction Models within the randomized DAPT Study population (N=11648). Ischemic Model: Myocardial infarction or stent thrombosis between 12-30 months after index PCI. Includes fatal events. Bleeding Model: GUSTO moderate or severe bleeding between 12-30 months after index PCI. Includes fatal events. Cox regression, stepwise selection among 37 candidate variables, including randomized treatment arm. In addition, several interaction terms with treatment arm evaluated. P value of 0.05 for retention. We started the analysis with the creation of 2 prediction models, one for ischemia and one for bleeding. The primary endpoint of the ischemia model was the composite of myocardial infarction or stent thrombosis including fatal events. The primary endpoint of the bleeding model was GUSTO defined moderate or severe bleeding, including fatal bleeds. To develop these models, we used standard methods of Cox regression, with stepwise selection from a wide range of clinical and procedural variables, including randomized treatment arm. These models were validated internally and then externally within the PROTECT randomized trial.

The Critical Component of the DAPT Score – Predicting Net Treatment Effect Predicted Ischemic Event Rate with Placebo Predicted Ischemic Event Rate with Rx Predicted Bleeding Event Rate with Rx Predicted Bleeding Event Rate with Placebo Predicted Risk Reduction in Ischemic Events (Beneficial Effect) Predicted Risk Increase in Bleeding Events (Harmful Effect) Predicted Net Treatment Effect (Range from Negative to Positive) Predictors of net treatment effect with continued thienopyridine determined from linear regression and simplified to an integer point score (DAPT Score) From these two models, for each individual subject randomized in the DAPT study, 4 values could be predicted: The predicted rate of ischemic events first assuming treatment with placebo and then assuming treatment with thienopyridine, and then the predicted bleeding rate assuming treatment with placebo and then assuming treatment with thienopyridine. These values could be subtracted from one another to estimate 2 values (CLICK): the predicted absolute reduction in ischemic events and the predicted increase in bleeding events with continued thienopyridine treatment compared with placebo. The difference (CLICK) between the expected ischemic benefit and the expected harm from bleeding represents the predicted net treatment effect for each patient, with more positive numbers reflecting greater overall benefit and more negative numbers reflecting greater overall harm from continuation of dual antiplatelet therapy. (CLICK) A linear regression model was then used to identify the factors associated with greater or lesser net treatment effect, the coefficients of which were used to create a simplified risk score called the DAPT Score.   We then examined the actual observed outcomes among patients randomized to continued therapy and compared them to the outcomes of those randomized to placebo for different levels of DAPT score. We further examined these results limited to the population of patients not receiving paclitaxel-eluting stents.

Multivariable Prediction Models Predictors of Myocardial Infarction or Stent Thrombosis Predictors of Moderate/Severe Bleeding Predictors of Events HR (95% CI) P Continued Thienopyridine vs. Placebo 0.52 (0.42 – 0.65) <0.001 1.66 (1.26 - 2.19) MI at Presentation 1.65 (1.31 – 2.07) - Prior PCI or Prior MI 1.79 (1.43 – 2.23) CHF or LVEF < 30% 1.88 (1.35 – 2.62) Vein Graft PCI 1.75 (1.13 – 2.73) 0.01 Stent Diameter < 3 mm 1.61 (1.30 – 1.99) Paclitaxel-Eluting Stent 1.57 (1.26 – 1.97) Cigarette Smoker 1.40 (1.11 – 1.76) Diabetes 1.38 (1.10 – 1.72) Peripheral Arterial Disease 1.49 (1.05 – 2.13) 0.03 2.16 (1.46, 3.20) Hypertension 1.37 (1.03 – 1.82) 1.45 (1.00, 2.11) 0.05 Renal Insufficiency 1.55 (1.03 – 2.32) 0.04 1.66 (1.04, 2.66) Age (per 10 years) 1.54 (1.34, 1.78) After multivariable modeling, randomization to continued thienopyridine was associated with a significant decrease in myocardial infarction or stent thrombosis, and to significant increases in bleeding events. There were several factors associated exclusively with an increased risk of ischemic events, related to prior cardiovasculr history and risk factors as well as procedural characteristics. Age was significantly associated with increased bleeding risk but not with an increase in stent thrombosis or MI. And PAD, hypertension and renal insuffiicency were found to be associated with an increased risk of both ischemic and bleeding events. The c-statistics of the models ranged from 0.68 to 0.70 in the DAPT Study, and 0.64 in the PROTECT study. *The ischemia model C-statistic: 0.70 in DAPT Study **The bleeding model C-statistic: 0.68 in DAPT Study

Predictors of Net Treatment Effect Characteristics Impact on Net Treatment Effect % of Variation Explained Age ≥ 75 Age 65 - < 75 Age < 65 (reference) -1.2% -0.5% - 6.0% 2.1% Prior PCI or MI 1.1% 14.6% Stent Diameter < 3 mm 0.9% 10.1% CHF or LVEF < 30% 1.9% 9.9% MI at Presentation 1.0% 9.6% Paclitaxel-Eluting Stent 8.8% Cigarette Smoker 0.7% 4.3% Diabetes 0.6% Vein Graft PCI 1.6% 3.7% Hypertension 0.2% 0.4% Renal Insufficiency 0.3% PAD -0.1% 0.04% Bleeding Predictors Ischemia Predictors This tables shows how these factors were associated with greater or lesser overall net treatment effect taking into account both bleeding and ischemic risk. As age was seen to increase bleeding but not ischemia, it had an overall negative impact on net treatment effect. Predictors of ischemia had an overall beneficial or positive impact on the net treatment effect. As expected, factors associated with both ischemia (CLICK) and bleeding had virtually no impact on net treatment effect and explained very little of the variation between individuals. As a result (CLICK), these variables were not included in the clinical tool. Bleeding and Ischemia Predictors

The DAPT Score Variable Points Patient Characteristic Age ≥ 75 -2 65 - <75 -1 < 65 Diabetes Mellitus 1 Current Cigarette Smoker Prior PCI or Prior MI CHF or LVEF < 30% 2 Index Procedure Characteristic MI at Presentation Vein Graft PCI Stent Diameter < 3mm Distribution of DAPT Scores among all randomized subjects in the DAPT Study The remaining factors were converted to a simplified integer risk score, the DAPT score, which I present here. The DAPT score ranges from -2 to 9, with an increasing score reflecting an increasing expected benefit from continuation of dual antiplatelet therapy. Roughly ¼ of patients had a score if 0 or less, ¼ with a score of 1, another ¼ with a score of 2, and the final quarter having a score of 3 or more. s

Risk Difference (Continued Thienopyridine – Placebo), 12-30M Continued Thienopyridine vs. Placebo Treatment Effect by DAPT Score Quartile (N = 11,648) Q1 = DAPT Score -2 to 0 Q3 = DAPT Score 2 Q2 = DAPT Score 1 Q4 = DAPT Score > 2 Risk Difference (Continued Thienopyridine – Placebo), 12-30M Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Here are the risk differences in events rates that were observed between randomized treatment groups in the DAPT Study, stratified by quartiles of the DAPT Score. There was a clear gradient in the effect of thienopyridine continuation by DAPT Score, in which increasing DAPT score was associated with progressively larger observed reductions in stent thrombosis and myocardial infarction. Conversely, increasing DAPT score was associated with a progressively smaller impact of continued therapy on bleeding.

Risk Difference (Continued Thienopyridine – Placebo), 12-30M Continued Thienopyridine vs. Placebo Treatment Effect by DAPT Score Quartile (N = 11,648) Risk Difference (Continued Thienopyridine – Placebo), 12-30M Mortality Net Adverse Events Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Notably, this gradient of effect was also observed for overall mortality, with the increase in mortality associated with continued thienopyridine concentrated in the lower DAPT score quartiles. Overall net adverse events, that is, the combination of MI, stent thrombosis and moderate or severe bleeding, were increased with continued dual antiplatelet therapy in the lower quartiles but decreased in the higher quartiles of DAPT score. We therefore dichotomized the DAPT score at this cutoff, with scores of < 2 representing low DAPT scores and scores of 2 or more representing high dapt scores. DAPT Score < 2 DAPT Score ≥ 2 DAPT Score < 2 DAPT Score ≥ 2 12 12

Continued Thienopyridine vs. Placebo DAPT Score <2 (Low); N=5731 Myocardial Infarction or Stent Thrombosis Death, MI, or Stroke (MACCE) 10% Continued Thienopyridine Placebo 10% Continued Thienopyridine Placebo 8% 8% 6% 6% Cumulative Incidence of ST/MI Cumulative Incidence of MACCE 4% 1.7% vs. 2.3% P=0.07 4% 3.7% vs. 3.8% P=0.73 2% 2% 0% 0% 12 15 18 21 24 27 30 12 15 18 21 24 27 30 Months After Enrollment Months After Enrollment 10% Continued Thienopyridine Placebo These are the observed kaplan meier event rates among low DAPT score patients, patients with scores of less than 2. In this group, there were no significant differences in MI or stent thrombosis, or the combination of death MI or stroke between randomized treatment groups, but a significant increase in gusto moderate or severe bleeding with randomization to continued thienopyridine. 8% GUSTO Moderate/ Severe Bleeding 6% Cumulative Incidence of GUSTO Moderate/ Severe Bleed 4% 3.0% vs. 1.4% P<0.001 2% 0% 12 15 18 21 24 27 30 Months After Enrollment

Continued Thienopyridine vs. Placebo DAPT Score ≥ 2 (High); N=5917 Myocardial Infarction or Stent Thrombosis Death, MI or Stroke (MACCE) 10% Continued Thienopyridine Placebo 10% Continued Thienopyridine Placebo 8% 8% 6% 6% 4.9% vs. 7.6% P<0.001 Cumulative Incidence of ST/MI 2.7% vs. 5.7% P<0.001 Cumulative Incidence of MACCE 4% 4% 2% 2% 0% 0% 12 15 18 21 24 27 30 12 15 18 21 24 27 30 Months After Enrollment Months After Enrollment 10% Continued Thienopyridine Placebo On the other hand, among high DAPT Score patients, those with scores of 2 or more, we observed the reverse. Sizable reductions ischemic events in the continued thienopyridine group, but no significant differences in bleeding. 8% GUSTO Moderate/ Severe Bleeding 6% Cumulative Incidence of GUSTO Moderate/ Severe Bleed 1.8% vs. 1.4% P=0.26 4% 2% 0% 12 15 18 21 24 27 30 Months After Enrollment

FAQs re: the DAPT Score Question: Paclitaxel-eluting stents perform poorly, is the DAPT score relevant to today’s stents. Answer: Yes.

Continued Thienopyridine vs. Placebo, by DAPT Score, Excluding PES Myocardial Infarction or Stent Thrombosis GUSTO Moderate or Severe Bleed Net Adverse Events Mortality (Continued Thienopyridine – Placebo), 12-30M Risk Difference P=0.06 P=0.07 P=0.003 P=0.17 After excluding paclitaxel-eluting stents from the analysis, the effects were consistent with the primary analysis, with the continued observation of opposite directions of effect of continued dual antiplatelet therapy on net adverse events for low vs. high DAPT score patients. P values are for comparison of risk differences across DAPT Score category (interaction).

Question: Has the DAPT Score been validated in an external population. FAQs re: the DAPT Score Question: Has the DAPT Score been validated in an external population. Answer: Yes.

Observed Event Rates in High vs. Low DAPT Score Groups in PROTECT Add details of population Discuss C-statistics or add them here of the individual models. ?back up, or reference in a slide on valdation and sensitiviy analysis? LM: back up, can put it as a line in the model limitations that it was good for predictive models, but did not have a large enough randomized duration dataset to validate DAPT score stratification.

FAQs re: the DAPT Score Question: Does use of the DAPT Score assume that bleeding and ischemia are weighted the same? Answer: No.

Risk Difference (Continued Thienopyridine – Placebo), 12-30M Continued Thienopyridine vs. Placebo Treatment Effect by DAPT Score Quartile (N = 11,648) Q1 = DAPT Score -2 to 0 Q3 = DAPT Score 2 Q2 = DAPT Score 1 Q4 = DAPT Score > 2 Risk Difference (Continued Thienopyridine – Placebo), 12-30M Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Here are the risk differences in events rates that were observed between randomized treatment groups in the DAPT Study, stratified by quartiles of the DAPT Score. There was a clear gradient in the effect of thienopyridine continuation by DAPT Score, in which increasing DAPT score was associated with progressively larger observed reductions in stent thrombosis and myocardial infarction. Conversely, increasing DAPT score was associated with a progressively smaller impact of continued therapy on bleeding.

Conclusions DAPT Score may help clinicians decide who should, Among patients who have not had a major ischemic or bleeding event within the first year after PCI: The DAPT Score identified patients for whom ischemic benefits outweighed bleeding risks, and patients for whom bleeding risks outweighed ischemic benefits. Low DAPT Score (< 2) NNT to prevent ischemia = 153 NNH to cause bleeding = 64 High DAPT Score ≥ 2 NNT to prevent ischemia = 34 NNH to cause bleeding = 272 In Conclusion, among patients who have not had a major ischemic or bleeding event within the first year after PCI, the DAPT Score identified those patients for whom the ischemic benefits of continued thienopyridine therapy outweighed the bleeding risks, as well those patients for whom the bleeding risks outweighed the ischemic benefits. The number needed to treat to prevent ischemic events were more than 4 x higher for Low DAPT score patients compared with high, while the number needed to harm to cause bleeding were 4 x lower. Although we split patients into these two groups, it is important to acknowledge that risks and benefits exist on a continuum which the score can help to quantify. We believe that the DAPT score is a useful tool to help clinicians decide who should and who should not be treated with extended duration dual antiplatelet therapy after coronary stent treatment. -2 10 DAPT Score may help clinicians decide who should, and who should not be treated with extended DAPT

DAPT Score Calculator DAPT Score calculator www.daptstudy.org Thank you! ryeh@bidmc.harvard.edu

DAPT Score Calculator DAPT Score calculator www.daptstudy.org Thank you! ryeh@bidmc.harvard.edu