Abstract
Importance
Dual antiplatelet therapy after percutaneous coronary intervention (PCI) reduces ischemia at the expense of increased bleeding.
Objective
To develop a clinical decision tool to identify patients expected to derive benefit vs. harm from continuing thienopyridine beyond one year after PCI.
Design/Setting
From DAPT Study data, a prediction rule was derived stratifying patients into groups to distinguish ischemic and bleeding risk 12–30 months after PCI. Validation was internal via bootstrap resampling and external within the Patient Related OuTcomes with Endeavor versus Cypher stenting (PROTECT) trial.
Participants
Derivation group: 11,648 randomized DAPT Study patients from 11 countries (August 2009–May 2014). Validation group: 8,709 randomized PROTECT trial patients from 36 countries (June 2007–July 2014).
Exposure
12 months open-label thienopyridine plus aspirin; 18 months randomized continued thienopyridine plus aspirin vs. placebo plus aspirin.
Main Outcome Measure
Ischemia (stent thrombosis or myocardial infarction [MI]) and bleeding (GUSTO moderate/severe) 12–30 months after PCI.
Results
Among the derivation population (mean age 61.3 years, 25.2% female), ischemia occurred in 348 (3.0%) and bleeding in 215 (1.8%). Derivation cohort models predicting ischemia and bleeding had c-statistics of 0.70 and 0.68, respectively. The prediction rule assigned 1 point each for MI at presentation, prior MI or PCI, diabetes, stent diameter <3 mm, smoking, and paclitaxel-eluting stent; 2 points each for history of congestive heart failure/low ejection fraction and vein graft intervention; −1 point for age 65–<75; and −2 points for age ≥ 75. For patients with high scores (≥ 2, n=5917), continued thienopyridine (vs. placebo) was associated with reduced ischemic events (2.7% vs. 5.7%, risk difference [RD] −3.0%, 95% CI −4.1% to −2.0%, p<0.001), compared to those with low scores (<2, n=5731, 1.7% vs. 2.3%, RD −0.7%, 95% CI −1.4% to 0.09%, p=0.07, interaction p< 0.001). Conversely, continued thienopyridine was associated with smaller increases in bleeding among scores ≥2 (1.8% vs. 1.4%, RD 0.4%, 95% CI −0.3% to 1.0%, p=0.26) compared with low scores <2 (3.0% vs. 1.4%, RD 1.5%, 95% CI 0.8% to 2.3%, p<0.001; interaction p=0.02). Mortality rates were 2.1% for continued thienopyridine vs. 2.1% for placebo (p=0.99) for scores ≥2, compared to 1.7% vs. 0.9%, respectively (p=0.02, interaction p=0.14 – non-significant) for scores <2. Among the validation cohort (mean age 62 years; 2,061 (23.7%) women), ischemia occurred in 365 (4.2%) and bleeding in 171 (2.0%), with c statistic 0.64 for ischemia model and 0.64 for bleeding. In this cohort, high score patients had increased ischemic events (1.5% vs. 0.7%, RD 0.73%, 95% CI 0.23% to 1.23%, p=0.002), and no significant difference in bleeding (0.4% vs. 0.5%, RD −0.16%, 95% CI −0.46% to 0.13%, p=0.31).
Conclusion and Relevance
Among patients not sustaining major bleeding or ischemic events one year after PCI, a prediction rule assessing ischemic and bleeding risks showed modest accuracy in derivation and validation cohorts. This rule requires further prospective evaluation to assess potential effects on patient care, as well as validation in other cohorts.
Trial Registration
ClinicalTrials.gov number NCT00977938.
INTRODUCTION
The optimal duration of dual antiplatelet therapy with aspirin and thienopyridine after percutaneous coronary intervention (PCI) with stents is the subject of debate. Among patients who complete one year of dual antiplatelet therapy after PCI without an ischemic or bleeding event, continuing therapy decreases stent thrombosis and myocardial infarction (MI), but increases bleeding.1,2 Continuing dual antiplatelet therapy thus involves a careful assessment of the trade-offs between reduced ischemia and increased bleeding for individual patients.
However, assessing the balance between ischemia and bleeding risks can be challenging for clinicians and patients. Factors related to recurrent ischemic events and bleeding in patients undergoing PCI overlap substantially, making it difficult to determine optimal treatment.3 While subgroup analyses have been helpful in determining groups with larger absolute benefits from continuing therapy (e.g. patients presenting with MI) 4,5, there remain patients within these broad categories who may also experience serious bleeding events. Most data estimating ischemia and bleeding risk following PCI have focused on early risks, including periprocedural events. 6,7 It remains unclear which patients are at high risk for late ischemic events and may thus benefit most from longer term dual antiplatelet therapy versus those who are high risk for late bleeding events and may thus be harmed.
The goal of this study was to identify factors predicting whether the expected benefit of reduced ischemia would outweigh the expected increase in bleeding associated with continued dual antiplatelet therapy beyond one year for individual patients, using data from the Dual Antiplatelet Therapy (DAPT) Study. These factors were used to develop a decision tool to help select the duration of therapy for individual patients being evaluated one year after stenting.
METHODS
This secondary analysis of the DAPT Study was approved by the Institutional Review Board of Partners Healthcare. The Patient Related OuTcomes with Endeavor versus Cypher stenting (PROTECT) protocol was approved by ethical boards in accordance with local regulations. All patients in both studies provided written, informed consent. The DAPT Study, conducted from August 2009 to May 2014 in 11 countries, enrolled patients after PCI with either drug-eluting stents (DES) or bare metal stents (BMS), and treated them with open-label thienopyridine plus aspirin for 12 months; at 12 months, eligible patients who were free from major bleeding and ischemic events and compliant with therapy remained on aspirin and were randomized to continued thienopyridine vs. placebo for 18 months. 8 The full enrollment and randomization criteria are listed in the Appendix. Patients on long-term anticoagulation therapy, those with planned surgeries necessitating discontinuation of antiplatelet therapy for > 14 days, and those with life expectancy < 3 years were excluded from enrollment. At 12 months, only those patients who were compliant with thienopyridine therapy and free from MI, stroke, repeat coronary revascularization, stent thrombosis, and moderate or severe bleeding by the GUSTO (Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Arteries) criteria9 during the first 12 months after enrollment were randomized.
As permitted by regulatory authorities, race and ethnicity data were collected via patient self-report. Race categories were prespecified as American Indian or Alaska Native, Asian, Black or African American, Native Hawaiian or other Pacific Islander, White, and other. Ethnicity was collected as Hispanic or Latino and not Hispanic or Latino. This information was collected in order to assess potential heterogeneous treatment effects among different subgroups.
The primary follow-up period of the study was 12–30 months after the index procedure (or 18 months after randomization). Details of the study design and results have been described previously. 1,2,8 As the results of the study were consistent across DES- and BMS-treated cohorts2, all randomized patients were included in this analysis.
Study Goals
The goal of this study was to distinguish patients within the DAPT Study who derived the greatest benefit from those who experienced the most harm from continuation of dual antiplatelet therapy more than 1 year after PCI, considering individual patient characteristics and their independent effects on ischemic and bleeding events. This study sought to stratify outcomes based on a single multivariable risk score. 10 This entailed, first, identifying factors associated with ischemic and bleeding risks, second, choosing those that selectively predicted either ischemic or bleeding risk to generate a simplified risk score, and, third, assessing the randomized treatment results observed in the trial, stratified by the new risk score. An ideal score would identify patients with simultaneous high ischemic risk (and corresponding high benefit with continued thienopyridine) and low bleeding risk (and corresponding low risk of harm with continued therapy), and vice versa. In addition, the ability of the score to stratify ischemic and bleeding risk within an external sample was assessed.
Ischemic and Bleeding Endpoints
The primary ischemic endpoint was a composite of MI or Academic Research Consortium definite or probable stent thrombosis11 and the primary bleeding endpoint was moderate or severe bleeding, as defined by the GUSTO criteria9.
Predictors
A total of 37 candidate variables potentially associated with ischemic or bleeding events based on a comprehensive literature review and clinical plausibility were identified. Variables included sociodemographic variables, cardiovascular history, non-cardiovascular medical comorbidities, anatomical and procedural factors, and concomitant medical therapy. (see Appendix “Candidate Variables for Model Building”).
Statistical Analysis
Development of Ischemic and Bleeding Event Models
Clinical and procedural characteristics were compared between patients experiencing events between 12–30 months and those without events, using chi-square or t-tests as appropriate. Cox regression was used to develop two separate models within the DAPT randomized study population (derivation cohort), the first to predict ischemic events and the second to predict bleeding events after randomization. Data were censored at the time of a stent thrombosis or MI for the ischemia model, a moderate or severe bleed for the bleeding model, or at the time of death, last known contact, or 30 months, whichever was earliest. Candidate variables that differed in univariable comparisons at a significance level of <0.30 were incorporated. Stepwise selection was then performed, using the 0.05 significance level. To identify possible heterogeneous treatment effects, simple Cox regression models were developed for each outcome including treatment arm, variable of interest, and their interaction term. Interactions terms significant at p<0.15 were entered into the stepwise selection process with other candidate variables.
Proportionality was evaluated for all variables in the models. Model discrimination was assessed using the c-statistic. Calibration was assessed through the examination of calibration plots and using the corrected Nam and D’Agostino goodness of fit test.12,13 The primary models were internally validated using bootstrap resampling for 200 iterations.14 For each resampling, the stepwise selection process was re-run, and the discrimination of the bootstrap model was assessed in the bootstrap sample and the full dataset. The mean difference between these bootstrap model values was defined as the “optimism”, and was subtracted from the final reported discrimination of the models.15
Development of a Simplified Clinical Prediction Score
For each patient, the predicted risk (cumulative incidence) of an ischemic event between 12–30 months was estimated, assuming treatment with continued thienopyridine plus aspirin beyond 12 months and separately assuming treatment with aspirin alone beyond 12 months; similarly, bleeding event risks were predicted under these two assumptions. The difference between these two predicted values represented the predicted absolute risk reduction in combined stent thrombosis or MI anticipated with continued thienopyridine from the ischemic model, and the predicted absolute risk increase in moderate or severe bleeding anticipated with continued thienopyridine from the bleeding model. The absolute difference between the predicted ischemic reduction and bleeding increase was defined as the “benefit-risk difference”, and estimated for each patient.
A linear regression model was created, using benefit-risk difference as the outcome, and all predictors that were selected in the ischemia and bleeding models. Variables that explained more than 1% of the observed variation in estimated benefit-risk difference were included in a simplified clinical prediction score. To facilitate ease of use, continuous variables such as age and stent diameter were categorized based on visual inspection of results, and all variables were assigned an integer score of 1 or 2 (or −1 to −2) based on the beta coefficient. (Appendix, Development of a Predictive Score) The range of potential scores was between −2 and 10.
Evaluation of Randomized Treatment Effect Stratified by Clinical Prediction Score
The DAPT Study population was divided into approximate quartiles based on the score, and 12–30 month Kaplan-Meier event rates were compared within each score quartile by randomized treatment group. Event rates were additionally examined among patients receiving only everolimus-eluting stents (EES). Based on these results, clinically relevant score groupings were created, defining patients more likely to benefit from thienopyridine continuation (high score) versus those more likely to be harmed (low score). The absolute risk differences in ischemic and bleeding event rates associated with continued thienopyridine versus placebo across high versus low clinical prediction score groups were compared using a Z test for interaction.
External Validation
The risk models and the clinical prediction score were externally validated within the PROTECT Trial, conducted from June 2007 to July 2014 in 36 countries, in which patients undergoing PCI were randomized to receive sirolimus-eluting (SES) vs. zotarolimus-eluting stents (ZES) and were followed for 5 years.17 This trial was selected for validation due to its large inclusive population of stent-treated patients, with similar definitions and adjudicated outcomes as those used in the DAPT Study. Those patients not sustaining MI, stent thrombosis, or a moderate/severe bleeding event within the first 12 months in the PROTECT Trial served as the validation cohort (n = 8136). Two forms of validation were conducted: 1) evaluation of the DAPT Study-derived ischemic and bleeding models, and 2) evaluation of prediction score performance in stratifying risks of ischemic and bleeding events. First, for the validation of the models, because PROTECT Trial patients were not randomized to different durations of dual antiplatelet therapy, dual antiplatelet therapy duration was likely confounded by treatment indication and was therefore not included in the validation. The anticipated statistical effect of omitting this variable in the validation would be to yield a conservative estimate of each model’s performance, given that randomized treatment group is strongly associated with both bleeding and ischemic events. Models were validated via the estimation of c-statistics and goodness-of-fit tests by applying the function derived in the DAPT Study to PROTECT patients between 12–30 months after PCI, limited to patients not sustaining MI, stent thrombosis, or a moderate/severe bleeding event within the first 12 months. Because PROTECT had lower overall ischemic and bleeding event rates than the DAPT Study, the calibration of the models was assessed after accounting for this difference in baseline hazard18, and then the goodness-of-fit of the recalibrated model was assessed.
Second, the ability of the clinical prediction score to stratify ischemic and bleeding risk was evaluated by comparing overall rates of stent thrombosis, MI and GUSTO moderate or severe bleeding among patients with high vs. low score in the PROTECT Trial.
A two-tailed alpha of 0.05 was used to define the significance threshold for all comparisons. All analyses were performed at the Harvard Clinical Research Institute (HCRI), using SAS version 9.4.
Results
Study Population
A total of 11,648 patients undergoing PCI with coronary stents were randomized in the DAPT Study and included in this analysis (Figure 1). Of these, 40.3% received EES, 22.9% paclitaxel-eluting stents, 10.9% ZES, 9.6% SES, 14.4% BMS, and 1.8% received more than one stent type. Between 12–30 months after their index procedure, 348 (3.0%) patients developed stent thrombosis or MI (97 with stent thrombosis, 251 with MI without stent thrombosis), and 215 (1.8%) patients developed moderate/severe bleeding (142 moderate, 72 severe bleeding, and 1 patient with two different events adjudicated as moderate, and severe). Thirty-three patients had both an ischemic and bleeding event in follow-up. Patients who had an ischemic event in follow-up had higher rates of cardiovascular risk factors, including diabetes, hypertension, and smoking, and had higher rates of cardiovascular disease, including history of congestive heart failure (CHF), low ejection fraction, prior MI and prior PCI, and were more likely to have been randomized to placebo compared with patients without an ischemic event (Table 1). Patients with a bleeding event were older and had a higher prevalence of smoking, hypertension, prior CHF, renal insufficiency/failure, atrial fibrillation, prior stroke/transient ischemic attack (TIA), prior PCI, or history of cancer, and were more likely to have been randomized to continued thienopyridine, compared with patients without a bleeding event.
Figure 1. DAPT Study patient flow diagram.
A total of 11,648 randomized patients comprised the cohort used to derive a clinical prediction score to stratify individual risk of benefit and harm with continuation of dual antiplatelet therapy beyond 1 year after PCI.
Table 1. Characteristics of patients with vs. without ischemic or bleeding events.
Baseline characteristics of patients with myocardial infarction and/or definite or probable stent thrombosis; and with vs. without a GUSTO moderate or severe bleeding event from 12–30 months.
| MI and/or Definite or Probable Stent Thrombosis Events | GUSTO Severe or Moderate Bleeding Events | |||||
|---|---|---|---|---|---|---|
|
| ||||||
| Measure* | Event(N=348 Patients) | No Event(N=11300 Patients) | P Value | Event(N=215 Patients) | No Event(N=11433 Patients) | P Value |
| Demographics | ||||||
| Age (years) | 61.7±10.8 | 61.3±10.3 | 0.47 | 66.4±10.3 | 61.2±10.3 | <.001 |
| Female | 92 (26.4%) | 2833 (25.1%) | 0.57 | 63 (29.3% | 2862 (25.0%) | 0.15 |
| Hispanic or Latino ethnic group | 17 (4.9%) | 389 (3.5%) | 0.18 | 8 (3.8%) | 398 (3.6%) | 0.85 |
| Non-White race* | 35 (10.3%) | 950 (8.6%) | 0.28 | 17 (8.0%) | 968 (8.6%) | 0.90 |
| BMI (Kg/m2) | 30.1±5.6 | 30.4±5.7 | 0.28 | 29.5±5.1 | 30.4±5.8 | 0.01 |
| Medical History | ||||||
| Diabetes mellitus | 138 (39.9%) | 3253 (28.9%) | <.001 | 67 (31.3%) | 3324 (29.2%) | 0.50 |
| Hypertension | 282 (81.0%) | 8240 (73.1%) | <.001 | 181 (84.2%) | 8341 (73.2%) | <.001 |
| Cigarette smoker | 113 (33.0%) | 3029 (27.2%) | 0.02 | 39 (18.2%) | 3103 (27.6%) | 0.002 |
| Stroke or TIA | 20 (5.8%) | 381 (3.4%) | 0.02 | 16 (7.6%) | 385 (3.4%) | 0.003 |
| Congestive heart failure | 36 (10.4%) | 488 (4.3%) | <.001 | 17 (8.0%) | 507 (4.5%) | 0.02 |
| LVEF < 30% | 15 (4.6%) | 192 (1.9%) | 0.002 | 6 (3.1%) | 201 (1.9%) | 0.28 |
| Renal insufficiency/failure | 27 (7.9%) | 441 (3.9%) | 0.001 | 20 (9.4%) | 448 (3.9%) | <.001 |
| Peripheral arterial disease | 37 (10.9%) | 612 (5.5%) | <.001 | 30 (14.3%) | 619 (5.5%) | <.001 |
| Prior PCI | 147 (42.4%) | 3221 (28.6%) | <.001 | 81 (37.7% | 3287 (28.9%) | 0.01 |
| Prior CABG | 61 (17.5%) | 1188 (10.5%) | <.001 | 31 (14.4% | 1218 (10.7%) | 0.09 |
| Atrial fibrillation | 13 (3.8%) | 327 (2.9%) | 0.33 | 12 (5.6%) | 328 (2.9%) | 0.04 |
| Prior myocardial infarction | 112 (32.7%) | 2344 (21.1%) | <.001 | 47 (22.2%) | 2409 (21.4%) | 0.80 |
| History of cancer | 36 (10.5%) | 1034 (9.2%) | 0.39 | 34 (16.0%) | 1036 (9.1%) | 0.002 |
| Cancer reported prior to randomization (0–12M) | 2 (0.6%) | 48 (0.4%) | 0.66 | 3 (1.4%) | 47 (0.4%) | 0.07 |
| Indication for index procedure | ||||||
| STEMI | 50 (14.4%) | 1630 (14.4%) | 1.00 | 22 (10.2%) | 1658 (14.5%) | 0.08 |
| NSTEMI | 77 (22.1%) | 1819 (16.1%) | 0.004 | 26 (12.1%) | 1870 (16.4%) | 0.11 |
| Stable angina | 110 (31.6%) | 4039 (35.7%) | 0.13 | 74 (34.4%) | 4075 (35.6%) | 0.77 |
| Unstable angina | 57 (16.4%) | 1764 (15.6%) | 0.71 | 37 (17.2%) | 1784 (15.6%) | 0.51 |
| Other | 54 (15.5%) | 2048 (18.1%) | 0.23 | 56 (26.1%) | 2046 (17.9%) | 0.003 |
| Lesion and Procedure Characteristics | ||||||
| In-stent restenosis | 30 (8.6%) | 513 (4.5%) | 0.001 | 13 (6.1%) | 530 (4.6%) | 0.33 |
| No. treated vessels (per patient) | 1.1±0.3 | 1.1±0.3 | 0.84 | 1.1±0.3 | 1.1±0.3 | 0.87 |
| No. stents (per patient) | 1.5±0.8 | 1.4±0.7 | 0.11 | 1.4±0.7 | 1.4±0.7 | 0.58 |
| >2 vessels stented | 0 (0.0%) | 49 (0.43%) | 0.41 | 0 (0.0%) | 49 (0.4%) | 1.00 |
| Reference Vessel Diameter (mm) | 2.9±0.5 | 3.0±0.5 | <.001 | 3.1±0.6 | 3.0±0.5 | 0.09 |
| Modified ACC lesion class B2 or C1 | 168 (50.8%1) | 5128 (47.1%) | 0.20 | 97 (45.8%) | 5199 (47.3%) | 0.68 |
| Vein bypass graft stented | 22 (6.3%) | 300 (2.7%) | <.001 | 8 (3.7%) | 314 (2.8%1) | 0.40 |
| Thrombus-containing lesion | 50 (15.3%) | 1482 (14.2%) | 0.57 | 19 (9.6%) | 1513 (14.3%) | 0.06 |
| Stent Type | <0.001 | 0.16 | ||||
| Drug-Eluting | 301 (86.5%) | 9960 (85.5%) | 192 (89.3%) | 9769 (85.4%) | ||
| Everolimus-eluting | 28 (8.1%) | 1090 (9.7%) | 28 (13.0%) | 1090 (9.5%) | ||
| Paclitaxel-eluting | 27 (7.8%) | 1237 (11.0%) | 25 (11.6%) | 1239 (10.8%) | ||
| Zotarolimus-eluting | 114 (32.8%) | 2552 (22.6%) | 45 (20.9%) | 2621 (22.9%) | ||
| Sirolimus-eluting | 122 (35.1%) | 4581 (40.5%) | 87 (40.5%) | 4616 (40.4%) | ||
| >1 type | 10 (2.9%) | 200 (1.8%) | 7 (3.3%) | 203 (1.8%) | ||
| Bare Metal | 47 (13.5%) | 1640 (14.5%) | 23 (10.7%) | 1664 (14.6%) | ||
| Minimum stent diameter (mm) | <.001 | 0.78 | ||||
| <3mm | 193 (55.5%) | 4848 (42.9%) | 95 (44.2%) | 4946 (43.3%) | ||
| ≥3mm | 155 (44.5%) | 6452 (57.1%) | 120 (55.8%) | 6487 (56.7%) | ||
| Total stent length (mm) | 28.1±16.8 | 27.0±16.43 | 0.21 | 26.1±15.0 | 27.1±16.5 | 0.39 |
| Thienopyridine at randomization | 0.51 | |||||
| Prasugrel | 138 (39.7%) | 3548 (31.4%) | 0.002 | 63 (29.3%) | 3623 (31.7%) | |
| Clopidogrel | 210 (60.3%) | 7752 (68.6%) | 152 (70.7%) | 7810 (68.3%) | ||
| Aspirin at randomization | 0.41 | 0.46 | ||||
| >100mg | 127 (41.2%) | 4424 (43.7%) | 78 (40.8%) | 4473 (43.7%) | ||
| ≤100mg | 181 (58.8%) | 5698 (56.3%) | 113 (59.2%) | 5766 (56.3%) | ||
| Statin use at randomization | 300 (86.2%) | 10098 (89.4%) | 0.06 | 185 (86.1%) | 10213 (89.4%) | 0.12 |
| Randomization arm | <.001 | <.001 | ||||
| Placebo | 225 (64.7%) | 5561 (49.2%) | 80 (37.2%) | 5706 (49.9%) | ||
| Continued thienopyridine | 123 (35.3%) | 5739 (50.8%) | 135 (62.8%) | 5727 (50.1%) | ||
Plus–minus values are means ±SD; all other numbers are N (%). Zero to 2.3% of patients had missing values, except for the following variables, for which up to 11.5% of the patients had missing values: LVEF <30%, Modified ACC lesion class B2 or C1, Thrombus-containing lesion, and Aspirin at randomization.
Race was self-reported.
Abbreviations: BMI, body mass index; BMS, bare metal stent; CABG, coronary bypass artery graft; DES, drug-eluting stent; LVEF, left ventricular ejection fraction; MI, myocardial infarction; No., number; NSTEMI, non-ST-elevation myocardial infarction; PCI, percutaneous coronary intervention; TIA, transient ischemic attack; SD, standard deviation; STEMI, ST-elevation myocardial infarction.
Risk Prediction Models
In multivariable Cox regression, significant predictors of both ischemic and bleeding events included randomized treatment arm, peripheral arterial disease (PAD), hypertension, and renal insufficiency/failure. Variables that predicted the risk of ischemic events only included history of PCI or MI prior to the index procedure, stent diameter < 3 mm, MI at presentation, history of CHF or left ventricular ejection fraction (LVEF) < 30%, paclitaxel-eluting stent, vein graft stent, cigarette smoking, and diabetes (Table 2). No tested interactions between covariates and randomized treatment for ischemic events were retained in the model. The ischemic model had moderate discrimination (c-statistic=0.70, 95% CI 0.68–0.73) and was well-calibrated (goodness of fit p=0.81).
Table 2. Stent thrombosis or myocardial infarction prediction model and GUSTO moderate or severe bleeding prediction model.
Predictors of events between 12 and 30 months after coronary stenting.
| Predictors of Events | Predictors of Stent Thrombosis or MI | Predictors of Moderate/Severe Bleeding | ||
|---|---|---|---|---|
| HR (95% CI) | P Value | HR (95% CI) | P Value | |
|
| ||||
| Continued Thienopyridine vs. Placebo | 0.52 (0.42 – 0.65) | <0.001 | 1.66 (1.26, 2.19) | <.001 |
| Myocardial Infarction at Presentation | 1.65 (1.31 – 2.07) | <0.001 | - | - |
| Prior PCI or Prior MI | 1.79 (1.43 – 2.23) | <0.001 | - | - |
| History of CHF or LVEF < 30% | 1.88 (1.35 – 2.62) | <0.001 | - | - |
| Vein Graft PCI | 1.75 (1.13 – 2.73) | 0.01 | - | - |
| Stent Diameter < 3 mm | 1.61 (1.30 – 1.99) | <0.001 | - | - |
| Paclitaxel-Eluting Stent | 1.57 (1.26 – 1.97) | <0.001 | - | - |
| Cigarette Smoker | 1.40 (1.11 – 1.76) | 0.01 | - | - |
| Diabetes | 1.38 (1.10 – 1.72) | 0.01 | - | - |
| Age (per 10 years) | - | - | 1.54 (1.34, 1.78) | <0.001 |
| Peripheral Arterial Disease | 1.49 (1.05 – 2.13) | 0.03 | 2.16 (1.46, 3.20) | <0.001 |
| Hypertension | 1.37 (1.03 – 1.82) | 0.03 | 1.45 (1.00, 2.11) | 0.05 |
| Renal Insufficiency | 1.55 (1.03 – 2.32) | 0.04 | 1.66 (1.04, 2.66) | 0.03 |
The ischemia model had a c-statistic of 0.70 within the DAPT Study randomized population, and goodness of fit p = 0.81.
The bleeding model had a c-statistic of 0.68 within the DAPT Study randomized population, and a goodness of fit p = 0.34.
Abbreviations: CHF, congestive heart failure; PCI, percutaneous coronary intervention; LVEF, left ventricular ejection fraction; MI, myocardial infarction.
Increasing age was a significant independent predictor of bleeding, but not of ischemic events (Table 2). No tested interactions between covariates and randomized treatment for bleeding were retained in the model. The bleeding model showed similar discrimination to the ischemia model (c-statistic=0.68, 95% CI 0.65−0.72), and was well-calibrated (goodness of fit p=0.34). After bootstrap internal validation, optimism-corrected c-statistics for both the ischemia (c-statistic=0.68, 95% CI 0.65–0.70) and bleeding models (c-statistic=0.66, 95% CI 0.62–0.70) were similar.
Clinical Prediction Score
A simplified risk score was generated to predict the difference between the anticipated reduction in ischemic events and the anticipated increase in bleeding events with continued thienopyridine (i.e. the benefit-risk difference). (Supplementary Appendix) The score, ranging from −2 to 10, assigned points as follows: 0 for age <65, −1 for age 65–<75, −2 for age≥ 75, 2 for vein graft PCI, 1 for current cigarette smoker or within past year, 1 for diabetes mellitus, 1 for MI at presentation, 1 for stent diameter: <3mm, 2 for history of CHF or LVEF<30%, 1 for prior PCI or prior MI, and 1 for paclitaxel-eluting stent. (Figure 2). Among the randomized DAPT Study population, higher score quartile was associated with higher rates of stent thrombosis or MI (p<0.001), whereas lower score quartiles were associated with higher rates of moderate or severe bleeding (p=0.006). In addition, higher score quartiles were associated with larger observed risk reductions in stent thrombosis and MI with randomization to continued thienopyridine (p=0.001), and lower score quartiles were associated with greater observed risk increases in bleeding (p=0.04, Table 3).
Figure 2. Clinical prediction ccore to stratify individual risk of benefit vs. harm with continuation of dual antiplatelet therapy beyond 1 year after PCI.
A. Summation of points for factors included in score. B. Distribution of prediction scores among all patients randomized in the DAPT Study (n = 11,648).
Abbreviations: CHF, congestive heart failure; LVEF, left ventricular ejection fraction; MI, myocardial infarction; PCI, percutaneous coronary intervention.
Table 3.
Observed outcomes by treatment group between 12–30 months after index procedure stratified by prediction score quartile.
| Event | N (Continued Thieno-pyridine) | N (Placebo) | All Patients(N = 11648) | Continued ThienopyridineN = 5862 | PlaceboN = 5786 | Risk DifferenceContinued Thienopyridine - Placebo [95% CI] | Interaction P Value* |
|---|---|---|---|---|---|---|---|
| Stent Thrombosis | |||||||
| Quartile 1: Score in [−2, 0] | 1373 | 1356 | 3 (0.12%) | 1 (0.08%) | 2 (0.2%) | −0.07% [−0.33%, 0.19%] | <0.001 |
| Quartile 2: Score = 1 | 1501 | 1501 | 11 (0.4%) | 5 (0.3%) | 6 (0.4%) | −0.06% [−0.51%, 0.39%] | |
| Quartile 3: Score = 2 | 1525 | 1486 | 29 (1.0%) | 5 (0.3%) | 24 (1.7%) | −1.34% [−2.08%, −0.59%] | |
| Quartile 4: Score in [3, 10] | 1463 | 1443 | 54 (1.9%) | 12 (0.9%) | 42 (3.0%) | −2.18% [−3.23%, −1.12%] | |
| MI | |||||||
| Quartile 1: Score in [−2, 0] | 1373 | 1356 | 40 (1.5%) | 15 (1.2%) | 25 (1.9%) | −0.73% [−1.68%, 0.21%] | 0.001 |
| Quartile 2: Score = 1 | 1501 | 1501 | 71 (2.4%) | 31 (2.1%) | 40 (2.7%) | −0.59% [−1.72%, 0.55%] | |
| Quartile 3: Score = 2 | 1525 | 1486 | 82 (2.8%) | 23 (1.6%) | 59 (4.1%) | −2.56% [−3.80%, −1.33%] | |
| Quartile 4: Score in [3, 10] | 1463 | 1443 | 151 (5.4%) | 52 (3.7%) | 99 (7.2%) | −3.48% [−5.20%, −1.76%] | |
| Stent Thrombosis or MI | |||||||
| Quartile 1: Score in [−2, 0] | 1373 | 1356 | 40 (1.5%) | 15 (1.2%) | 25 (1.9%) | −0.73% [−1.68%, 0.21%] | 0.001 |
| Quartile 2: Score = 1 | 1501 | 1501 | 71 (2.4%) | 31 (2.1%) | 40 (2.7%) | −0.59% [−1.72%, 0.55%] | |
| Quartile 3: Score = 2 | 1525 | 1486 | 85 (2.9%) | 24 (1.6%) | 61 (4.3%) | −2.63% [−3.88%, −1.38%] | |
| Quartile 4: Score in [3, 10] | 1463 | 1443 | 152 (5.4%) | 53 (3.8%) | 99 (7.2%) | −3.41% [−5.13%, −1.68%] | |
| MACCE | |||||||
| Quartile 1: Score in [−2, 0] | 1373 | 1356 | 99 (3.7%) | 52 (3.9%) | 47 (3.5%) | 0.40% [−1.06%, 1.86%] | 0.02 |
| Quartile 2: Score = 1 | 1501 | 1501 | 110 (3.8%) | 50 (3.4%) | 60 (4.1%) | −0.65% [−2.04%, 0.75%] | |
| Quartile 3: Score = 2 | 1525 | 1486 | 137 (4.7%) | 51 (3.4%) | 86 (6.0%) | −2.54% [−4.10%, −0.98%] | |
| Quartile 4: Score in [3, 910 | 1463 | 1443 | 221 (7.9%) | 91 (6.4%) | 130 (9.3%) | −2.95% [−4.97%, −0.92%] | |
| Death | |||||||
| Quartile 1: Score in [−2, 0] | 1373 | 1356 | 43 (1.6%) | 28 (2.1%) | 15 (1.1%) | 0.99% [0.02%, 1.96%] | 0.33 |
| Quartile 2: Score = 1 | 1501 | 1501 | 29 (1.0%) | 18 (1.2%) | 11 (0.7%) | 0.49% [−0.24%, 1.22%] | |
| Quartile 3: Score = 2 | 1525 | 1486 | 48 (1.6%) | 25 (1.7%) | 23 (1.6%) | 0.09% [−0.85%, 1.02%] | |
| Quartile 4: Score in [3, 10] | 1463 | 1443 | 70 (2.5%) | 35 (2.5%) | 35 (2.5%) | −0.06% [−1.24%, 1.11%] | |
| GUSTO Moderate or Severe Bleed | |||||||
| Quartile 1: Score in [−2, 0] | 1373 | 1356 | 72 (2.7%) | 49 (3.7%) | 23 (1.7%) | 1.97% [0.71%, 3.23%] | 0.04 |
| Quartile 2: Score = 1 | 1501 | 1501 | 51 (1.8%) | 34 (2.3%) | 17 (1.2%) | 1.17% [0.20%, 2.14%] | |
| Quartile 3: Score = 2 | 1525 | 1486 | 45 (1.5%) | 28 (1.9%) | 17 (1.2%) | 0.69% [−0.22%, 1.60%] | |
| Quartile 4: Score in [3, 10] | 1463 | 1443 | 47 (1.7%) | 24 (1.7%) | 23 (1.7%) | 0.03% [−0.95%, 1.01%] | |
| GUSTO Moderate Bleed | |||||||
| Quartile 1: Score in [−2, 0] | 1373 | 1356 | 45 (1.7%) | 28 (2.1%) | 17 (1.3%) | 0.83% [−0.17%, 1.84%] | 0.33 |
| Quartile 2: Score = 1 | 1501 | 1501 | 37 (1.3%) | 26 (1.8%) | 11 (0.8%) | 1.03% [0.21%, 1.86%] | |
| Quartile 3: Score = 2 | 1525 | 1486 | 26 (0.9%) | 18 (1.2%) | 8 (0.6%) | 0.66% [−0.04%, 1.35%] | |
| Quartile 4: Score in [3, 10] | 1463 | 1443 | 35 (1.3%) | 19 (1.3%) | 16 (1.2%) | 0.18% [−0.66%, 1.03%] | |
| GUSTO Severe Bleed | |||||||
| Quartile 1: Score in [−2, 0] | 1373 | 1356 | 28 (1.1%) | 21 (1.6%) | 7 (0.5%) | 1.07% [0.27%, 1.86%] | 0.08 |
| Quartile 2: Score = 1 | 1501 | 1501 | 14 (0.5%) | 8 (0.6%) | 6 (0.4%) | 0.14% [−0.37%, 0.65%] | |
| Quartile 3: Score = 2 | 1525 | 1486 | 19 (0.7%) | 10 (0.7%) | 9 (0.6%) | 0.04% [−0.56%, 0.63%] | |
| Quartile 4: Score in [3, 10] | 1463 | 1443 | 12 (0.4%) | 5 (0.4%) | 7 (0.5%) | −0.15% [−0.66%, 0.35%] | |
Abbreviations: GUSTO, Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Arteries; MACCE, major adverse cardiovascular and cerebrovascular events; MI, myocardial infarction.
P-value for interaction assesses whether the absolute risk reduction observed between randomized treatment groups differs across quartiles of the clinical prediction score, as assessed by the Q-statistic for heterogeneity.
When separated into groups (≥ 2, the median value vs. < 2), among patients with predictive scores ≥ 2 (n=5917), randomization to continued thienopyridine was associated with larger reductions in stent thrombosis or MI (2.7% vs. 5.7%, RD −3.0%, 95% CI −4.1% to −2.0%, p<0.001) compared with those with scores < 2, (n=5731; MI 1.7% vs. 2.3%, RD −0.7%, 95% CI −1.4% to 0.09%, p=0.07; interaction p < 0.001). Conversely, randomization to continued thienopyridine was associated with smaller increases in bleeding among high score patients (1.8% vs. 1.4%, RD 0.4%, 95% CI −0.3% to 1.0%, p=0.26) compared with low score patients (3.0% vs. 1.4%, RD 1.5%, 95% CI 0.8% to 2.3%, p<0.001; interaction p=0.02, Figure 3A and B; eTable 3).
Figure 3. Observed rates of MI or stent thrombosis, MACCE, and moderate or severe bleeding, stratified by clinical prediction score group.
Kaplan-Meier curves for continued thienopyridine vs. placebo for MI or ST, MACCE, GUSTO moderate or severe bleeding, and death at 12–30 months after PCI in randomized patients (Panels A, B), and those treated with everolimus-eluting stents only (Panels C, D), stratified by clinical prediction score. The number at risk was defined as the number of patients who had not had the event of interest and who were available for subsequent follow-up. The P values are based on the subgroup log-rank test.
Abbreviations: MACCE, major adverse cardiovascular and cerebrovascular events; GUSTO, Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Arteries.
The risk reduction in MACCE with continued thienopyridine was significantly greater among the high score group (4.9% vs. 7.6%, P<0.001) vs. the low score group (3.7% vs. 3.8%, p=0.73, interaction p=0.001). The all-cause mortality rate was 2.1% for continued thienopyridine vs. 2.1% for placebo (p=0.99) for high score patients, compared with 1.7% for continued thienopyridine vs. 0.9% for placebo (RD 0.73%, 95% CI0.13%–1.33%, p=0.02, interaction p=0.14 – non-significant) for low score patients.
Outcomes in Patients Treated with EES
After restricting the population to those treated with EES (n=4,703), the rates of stent thrombosis or MI were 2.9% for continued thienopyridine vs. 4.7% for placebo (RD −1.89%, 95% CI −3.70% to −0.08%, p=0.04) among high score patients (n=1869), and were 1.7% for continued thienopyridine vs. 2.2% for placebo (RD −0.50%, 95% CI −1.55% to 0.56%, p=0.33, interaction p=0.18 – non-significant) among low score patients (n=2834). The corresponding rates of bleeding were 1.8% for continued thienopyridine vs. 1.2% for placebo (RD 0.52%, 95% CI −0.63% to 1.67%, p=0.38) for high score patients, and 3.0% vs. 1.4% (RD 1.67%, 95% CI 0.55% to 2.78%, p=0.003; interaction p=0.15 - non-significant). (Figures 3C and D, eTable 4). All-cause mortality occurred in 2.5% for continued thienopyridine vs. 1.8% for placebo (p=0.31) among the high score group, and 1.9% for continued thienopyridine vs. 0.7% for placebo (p=0.01, interaction p=0.54 - non-significant) among the low score group.
External Validation
Among 8,136 patients who did not have a MI, stent thrombosis or moderate/severe bleeding within the first 12 months after PCI in PROTECT, the models used to derive the predictive score (excluding the variable reflecting randomization to continued thienopyridine vs. placebo) showed modestly reduced discrimination (ischemic model c-statistic 0.64, 95% CI 0.58–0.70, bleeding model c-statistic 0.64, 95% CI 0.55–0.73). These results were overall similar within the ZES and SES populations of PROTECT [ischemic model C-statistic: ZES group 0.62 (95% CI 0.52–0.72); SES 0.64 (95% CI 0.57–0.72); Bleeding model C-statistic: ZES group 0.63 (95% CI 0.51–0.76).; SES 0.65 (95% CI 0.53–0.76]. Because PROTECT enrolled a lower risk population than the DAPT Study, both ischemic and bleeding event rates were overestimated. After recalibration to the baseline event rates observed in PROTECT, the models were well fit (P=0.91 for ischemia model, P=0.81 for bleeding model, Appendix).
Within the PROTECT validation cohort, the rate of stent thrombosis or MI between 12–30 months after PCI was greater among high score patients compared with low score patients (1.5% vs. 0.7%, respectively, HR = 2.01, 95% CI 1.29 to 3.13, p=0.002). Rates of moderate or severe bleeding were not significantly different by score group (0.36% among high score patients and 0.52% among low score patients, HR 0.69, 95% CI 0.33 to 1.42, p=0.31).
Discussion
Among patients in the DAPT Study, a clinical prediction score based on ischemic and bleeding risk factors was developed to help identify those patients with greater expected benefit vs. greater expected harm from continuation of dual antiplatelet therapy, from among those who had completed 1 year of dual antiplatelet therapy after coronary stent treatment without a major ischemic or bleeding event. For patients randomized in the DAPT Study with predictive scores ≥ 2 (50.8% of patients), randomization to continued thienopyridine was associated with an absolute risk reduction in MI or stent thrombosis that was 8.2 times greater than the absolute risk increase in moderate or severe bleeding. Conversely, among patients with scores < 2 (49.2% of patients), randomization to continued thienopyridine was associated with an absolute increase in bleeding that was 2.4 times the absolute reduction in MI or stent thrombosis. Within the PROTECT Trial, (validation cohort), patients with high scores were observed to have significantly greater ischemic risk and no significant difference in bleeding risk, compared with patients with low scores. Despite prior evidence suggesting that ischemic and bleeding risk are strongly correlated, 3,19 these results suggest that that it may be possible to identify individual patients with discordant ischemic risks and bleeding risks.
Numerous randomized trials evaluating duration of dual antiplatelet therapy after coronary stenting have demonstrated a tradeoff between reductions in ischemia and increases in bleeding associated with longer durations of treatment.20–24 While clinical trial results are expected to be applied to the population represented by enrolment criteria, in the setting of discordant risks and benefits of treatment, tailoring therapies to individual patient profiles in order to maximize benefits and minimize harms affords an opportunity to further optimize outcomes. However, the results of this study should be interpreted with the understanding that patients enrolled in clinical trials may not represent those cared for in routine practice on the basis of the exclusion and inclusion criteria of the trial, as well as other unmeasured differences between study participants and non-participants. Patients on oral anticoagulation were not enrolled in the DAPT Study, and make up 4–7% of all PCI patients. 25–27 Patients who interrupted therapy for more than 14 days or sustained a major bleeding or ischemic event were also not randomized in the DAPT Study, and represented 22.7% of enrolled subjects. Similarly, in a recent large registry of patients undergoing coronary stenting, discontinuation of antiplatelet therapy for more than 14 days occurred in 11.5% of patients, cessation due to a clinical event or non-compliance in 9.7%, and major bleeding in 1.4%, whereas MI occurred in 2.2%, and target-vessel revascularization in 5.1%, altogether representing ~30% of all PCI patients.27 While there remains a sizeable proportion of patients undergoing PCI who do not have events that would have disqualified them from randomization in the DAPT Study, nevertheless the patients used to derive the clinical risk score make up a group of patients that may not be representative of those seen in clinical practice.
Variables in the predictive score included patient and procedural characteristics that have demonstrated an association with either ischemic or bleeding events after PCI in prior studies. For instance, prior PCI, presentation with MI, current smoking, and diabetes have each been predictive of stent thrombosis occurring within the first year after PCI. 28 Similarly, advanced age, renal disease and history of PAD have correlated with both in-hospital and 30-day bleeding post-PCI.29,30 In this study, PAD, renal insufficiency and hypertension were predictive of both ischemic and bleeding events. Because these factors did not help identify discordant bleeding and ischemic risk, they were not included in the predictive score.
On the other hand, certain variables uniquely predicted either bleeding risk or anti-ischemic benefit: advanced age was predictive of increased bleeding only, while presentation with MI, history of CHF and prior PCI were predictive of MI or stent thrombosis but not bleeding. Deaths due to heart failure or arrhythmia not preceded by myocardial infarction or stent thrombosis were not considered in the creation of the prediction model because such deaths may not be directly modified by dual antiplatelet therapy. This may explain why age was not a significant predictor of the composite ischemia endpoint.
The median predictive score was 2, and patients with a score ≥ 2 had a clinically meaningful reduction in ischemic events (number needed to treat to benefit, or NNTB, of 34) with a smaller impact on bleeding events when randomized to continued thienopyridine (number needed to treat to harm, or NNTH, of 272), whereas those with scores < 2 had a larger increase in bleeding events (NNTH of 64) and a smaller reduction in ischemic events (NNTB of 153). Nonetheless, scores ranging from −2 to 10 likely define a continuous gradient of risk and benefit. The model used to derive the point values for variables required an assumption that bleeding and ischemic events were of equal weight. However, examination of the results stratified by score quartile allows assessment of different score cut-offs with varied weighting of the relative importance of bleeding and ischemic events, as well as to consider the influence of the score on other relevant endpoints, including bleeding events not classified as moderate or severe. The ischemic and bleeding events as defined in this analysis may not have an equivalent effect on patient outcomes, including mortality, and the results may have been different had more or less sensitive criteria for ischemic and bleeding endpoints been chosen.
Although the statistical test for interaction did not show a difference in the effect of continuation of long-term dual antiplatelet therapy on mortality in high vs. low prediction score groups, it is of interest that the numerical difference in all-cause mortality was concentrated in patients with low scores. After analyzing the results of 12 randomized trials enrolling 56,799 patients, the United States Food and Drug Administration recently concluded that there was no evidence of an increase in either cancer or mortality with extended thienopyridine treatment. 31 Whether different subgroups of patients may, in fact, have greater mortality with continuation of long term dual antiplatelet therapy has been suggested 32 and may be a topic of future inquiry.
Paclitaxel-eluting stents were found to be associated with higher risk of stent thrombosis or MI. While these results are consistent with other studies16, stent type was not randomized in the DAPT Study. As these stents are rarely used, the use of this predictive score going forward is unlikely to utilize this variable. In addition, among the stents used in the DAPT Study, only EES are widely used today. Among the EES subgroup (n=4,703), tests for interaction comparing treatment effect among high vs. low score groups were not significant. However, interaction testing is generally underpowered in clinical trials and more underpowered when performed within a subset of patients..Given that approximately half of the risk reduction in risk of MI attributed to continued thienopyridine therapy in the DAPT Study was not attributable to stent thrombosis1, and bleeding risk should not influenced by stent type, the impact of stratification by the prediction rule on stent-independent events would be expected to be generalizable to similarly characterized patients treated with various types of coronary stents other than those included in the study.
The incorporation of more variables into the individual bleeding and ischemia models may have improved discrimination, at the expense of parsimony. In addition, the estimation of risks based on the use of the separate ischemic and bleeding model coefficients rather than use of the simplified score could improve the ability to predict such events, and provide the opportunity for clinicians to identify patients with concordantly high ischemic and bleeding risks, in addition to those with discordant risks. (See Appendix, “Estimation of ischemic and bleeding risk”). A parsimonious score was also constructed in order to reduce the potential barriers to clinical use should the score be further validated, at the expense of some predictive accuracy.
While the development of the score was prespecified, the analysis should be considered exploratory. Thus, use of this prediction score should be cautious until further validation is performed, and optimal clinical and procedural care to reduce overall bleeding and ischemic risks should be practiced independent of score. Preexisting anemia, prior bleeding, and granular measures of atherosclerosis extent and severity were not available, and may, in part, explain the modest discrimination of the ischemia and bleeding prediction. In addition, patients receiving ticagrelor or other antiplatelet combinations could have different risk-benefit profiles. The score is relevant to patients with similar characteristics to those enrolled in the DAPT Study and its generalizability to other patient populations not studied in the trial may be limited. Although BMS-treated patients were included, the score is not applicable to patients for whom BMS are selected due to high risk of bleeding or non-compliance. The endpoints considered in developing the score, although well defined and adjudicated, are heterogeneous in severity. Next, while the PROTECT Trial served as an external population for validation, it was not a randomized trial of dual antiplatelet therapy duration, and the observed duration of therapy was likely influenced by patient risk factors. Therefore, these data could only be used to evaluate whether the score stratified patient ischemic or bleeding risk, and not actual benefit or harm with long-term dual antiplatelet therapy. Finally, these results would ideally be replicated in a similarly designed, large prospective randomized trial of different durations of dual antiplatelet therapy among PCI patients, and use of the clinical score has not been demonstrated to improve patient outcomes.
Conclusions
Among patients not sustaining major bleeding or ischemic events one year after PCI, a prediction rule assessing ischemic and bleeding risks showed modest accuracy in derivation and validation cohorts. This rule requires further prospective evaluation to assess potential effects on patient care, as well as validation in other cohorts.
Supplementary Material
Acknowledgments
Drs. Yeh and Mauri had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Sponsored by Harvard Clinical Research Institute. Dr. Yeh was funded by the National Heart, Lung and Blood Institute (K23 HL 118138). Additional funding was provided by a grant from the US Department of Health and Human Services (1RO1FD003870-01). The eight stent and pharmaceutical manufacturers who contributed to the funding of the DAPT Study included Abbott Vascular, Xience everolimus-eluting stent, Santa Clara, CA; Boston Scientific Corporation, TAXUS paclitaxel-eluting and PROMUS everolimus-eluting stents, Marlborough, MA; Cordis Corporation, Cypher sirolimus-eluting stent, Bridgewater, NJ; Medtronic, Inc., Endeavor zotarolimus-eluting stent, Minneapolis, MN; Bristol-Myers Squibb Company [New York, NY]/Sanofi Pharmaceuticals [Bridgewater, NJ] Partnership; Eli Lilly and Company, Indianapolis, IN; and Daiichi Sankyo Company Limited, Parsippany, NJ. The authors would like to thank Joanna Suomi, MSC for her assistance with the preparation of this manuscript; she is employed by Harvard Clinical Research Institute and was compensated for her contribution.
Footnotes
Author Contributions:
Study concept and design: Yeh, Mauri.
Acquisition, analysis, or interpretation of data: Yeh, Secemsky, Apruzzese, Song, Massaro, Mauri.
Drafting of the manuscript: Yeh, Secemsky, Mauri.
Critical revision of the manuscript for important intellectual content: Yeh, Secemsky, Kereiakes, Normand, Gershlick, Cohen, Spertus, Steg, Cutlip, Rinaldi, Camenzind, Wijns, Apruzzese, Song, Massaro, Mauri.
Statistical analysis: Yeh, Apruzzese, Song, Massaro, Mauri.
Role of the Funding Source
The stent manufacturers who funded the DAPT Study were participating members in the design of the trial, the conduct of the study, and in the collection of the data. Specifically, the DAPT Study included patients contributed from each of four industry-designed and conducted post-market studies. The funders had no role in the management, analysis, and interpretation of the data. HCRI was responsible for the scientific conduct of the DAPT Study and independent analysis of the data. For this substudy, funding sources had no role in study design or the decision to submit the manuscript for publication, but were given the opportunity to review and approve the manuscript prior to submission.
Declaration of Interests
Dr. Yeh reports personal fees from Abbott Vascular, Boston Scientific, and Merck, and research salary from Harvard Clinical Research Institute, outside the submitted work.
Dr. Secemsky has nothing to disclose.
Dr. Kereiakes has nothing to disclose.
Dr. Normand has nothing to disclose
Dr. Gershlick has nothing to disclose.
Dr. Cohen reports grants and personal fees from Abbott Vascular, Medtronic, Astra-Zeneca, and Eli Lilly, and grants from Daichii-Sankyo, outside the submitted work.
Dr. Spertus reports ownership of equity interest in Health Outcomes Sciences.
Dr. Steg reports personal fees from Amarin, AstraZeneca, Bayer, Boehringer-Ingelheim, Bristol-Myers-Squibb, Daiichi-Sankyo, GlaxoSmithKline, Lilly, Merck-Sharpe-Dohme, Novartis, CSL-Behring, Regeneron, Janssen, Pfizer, Regado, and Roche; grants and personal fees from Sanofi and Servier; and personal fees and non-financial support from The Medicines Company; outside the submitted work.
Dr. Cutlip reports grants from Medtronic and Boston Scientific, outside the submitted work.
Dr. Rinaldi reports serving on the advisory boards of Abbott Vascular and Boston Scientific.
Dr. Camenzind has nothing to disclose.
Dr. Wijns reports grants from Medtronic, during the conduct of the study; grants from Medtronic, Boston Scientific, Terumo, MiCell, Stentys, and Abbott Vascular, outside the submitted work; and non-executive Board member and shareholder of Argonauts Partners, Celyad and Genae Inc.
Ms. Apruzzese has nothing to disclose.
Mr. Song has nothing to disclose.
Dr. Massaro reports personal fees from Harvard Clinical Research Institute during the conduct of the study.
Dr. Mauri reports grants from Abbott, Boston Scientific, Cordis, Bristol-Myers Squibb, and Daiichi Sankyo; grants and personal fees from Medtronic, Eli Lilly and Company, Sanofi, Boehringer Ingelheim, Biotronik; personal fees from Amgen, Recor, Astra Zeneca, and St. Jude, outside the submitted work
References
- 1.Mauri L, Kereiakes DJ, Yeh RW, et al. Twelve or 30 months of dual antiplatelet therapy after drug-eluting stents. The New England journal of medicine. 2014 Dec 4;371(23):2155–2166. doi: 10.1056/NEJMoa1409312. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Kereiakes DJ, Yeh RW, Massaro JM, et al. Antiplatelet therapy duration following bare metal or drug-eluting coronary stents: the Dual Antiplatelet Therapy Randomized Clinical Trial. Jama. 2015 doi: 10.1001/jama.2015.1671. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Matteau A, Yeh RW, Camenzind E, et al. Balancing Long-Term Risks of Ischemic and Bleeding Complications After Percutaneous Coronary Intervention With Drug-Eluting Stents. The American journal of cardiology. 2015 Sep 1;116(5):686–693. doi: 10.1016/j.amjcard.2015.05.036. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Yeh RW, Kereiakes DJ, Steg PG, et al. Benefits and Risks of Extended Duration Dual Antiplatelet Therapy After PCI in Patients With and Without Acute Myocardial Infarction. Journal of the American College of Cardiology. 2015 May 26;65(20):2211–2221. doi: 10.1016/j.jacc.2015.03.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Udell JA, Bonaca MP, Collet JP, et al. Long-term dual antiplatelet therapy for secondary prevention of cardiovascular events in the subgroup of patients with previous myocardial infarction: a collaborative meta-analysis of randomized trials. European heart journal. 2015 Aug 31; doi: 10.1093/eurheartj/ehv443. [DOI] [PubMed] [Google Scholar]
- 6.Dangas GD, Caixeta A, Mehran R, et al. Frequency and predictors of stent thrombosis after percutaneous coronary intervention in acute myocardial infarction. Circulation. 2011 Apr 26;123(16):1745–1756. doi: 10.1161/CIRCULATIONAHA.110.981688. [DOI] [PubMed] [Google Scholar]
- 7.Rao SV, McCoy LA, Spertus JA, et al. An updated bleeding model to predict the risk of post-procedure bleeding among patients undergoing percutaneous coronary intervention: a report using an expanded bleeding definition from the National Cardiovascular Data Registry CathPCI Registry. JACC. Cardiovascular interventions. 2013 Sep;6(9):897–904. doi: 10.1016/j.jcin.2013.04.016. [DOI] [PubMed] [Google Scholar]
- 8.Mauri L, Kereiakes DJ, Normand SL, et al. Rationale and design of the dual antiplatelet therapy study, a prospective, multicenter, randomized, double-blind trial to assess the effectiveness and safety of 12 versus 30 months of dual antiplatelet therapy in subjects undergoing percutaneous coronary intervention with either drug-eluting stent or bare metal stent placement for the treatment of coronary artery lesions. American heart journal. 2010 Dec;160(6):1035–1041. 1041 e1031. doi: 10.1016/j.ahj.2010.07.038. [DOI] [PubMed] [Google Scholar]
- 9.An international randomized trial comparing four thrombolytic strategies for acute myocardial infarction. The GUSTO investigators. The New England journal of medicine. 1993 Sep 2;329(10):673–682. doi: 10.1056/NEJM199309023291001. [DOI] [PubMed] [Google Scholar]
- 10.Hayward RA, Kent DM, Vijan S, Hofer TP. Multivariable risk prediction can greatly enhance the statistical power of clinical trial subgroup analysis. BMC medical research methodology. 2006;6:18. doi: 10.1186/1471-2288-6-18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Cutlip DE, Windecker S, Mehran R, et al. Clinical end points in coronary stent trials: a case for standardized definitions. Circulation. 2007 May 1;115(17):2344–2351. doi: 10.1161/CIRCULATIONAHA.106.685313. [DOI] [PubMed] [Google Scholar]
- 12.Nam BH, D’Agostino RB., Sr . In: Goodness-of-Fit Tests and Model Validity. Huber-Carol C, editor. Boston: Birkhauser; 2002. pp. 267–279. [Google Scholar]
- 13.Demler OV, Paynter NP, Cook NR. Tests of calibration and goodness-of-fit in the survival setting. Statistics in medicine. 2015 May 10;34(10):1659–1680. doi: 10.1002/sim.6428. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Steyerberg EW, Harrell FE, Jr, Borsboom GJ, Eijkemans MJ, Vergouwe Y, Habbema JD. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. Journal of clinical epidemiology. 2001 Aug;54(8):774–781. doi: 10.1016/s0895-4356(01)00341-9. [DOI] [PubMed] [Google Scholar]
- 15.Harrell FE., Jr . Regression modeling strategies: with applications to linear models, logistic regression, and survival analysis. New York: Springer; 2001. [Google Scholar]
- 16.Palmerini T, Biondi-Zoccai G, Della Riva D, et al. Stent thrombosis with drug-eluting and bare-metal stents: evidence from a comprehensive network meta-analysis. Lancet. 2012 Apr 14;379(9824):1393–1402. doi: 10.1016/S0140-6736(12)60324-9. [DOI] [PubMed] [Google Scholar]
- 17.Camenzind E, Wijns W, Mauri L, et al. Stent thrombosis and major clinical events at 3 years after zotarolimus-eluting or sirolimus-eluting coronary stent implantation: a randomised, multicentre, open-label, controlled trial. Lancet. 2012 Oct 20;380(9851):1396–1405. doi: 10.1016/S0140-6736(12)61336-1. [DOI] [PubMed] [Google Scholar]
- 18.D’Agostino RB, Sr, Grundy S, Sullivan LM, Wilson P Group CHDRP. Validation of the Framingham coronary heart disease prediction scores: results of a multiple ethnic groups investigation. Jama. 2001 Jul 11;286(2):180–187. doi: 10.1001/jama.286.2.180. [DOI] [PubMed] [Google Scholar]
- 19.Joyner CD, Peters RJ, Afzal R, et al. Fondaparinux compared to enoxaparin in patients with acute coronary syndromes without ST-segment elevation: outcomes and treatment effect across different levels of risk. American heart journal. 2009 Mar;157(3):502–508. doi: 10.1016/j.ahj.2008.10.028. [DOI] [PubMed] [Google Scholar]
- 20.Valgimigli M, Campo G, Monti M, et al. Short- versus long-term duration of dual-antiplatelet therapy after coronary stenting: a randomized multicenter trial. Circulation. 2012 Apr 24;125(16):2015–2026. doi: 10.1161/CIRCULATIONAHA.111.071589. [DOI] [PubMed] [Google Scholar]
- 21.Feres F, Costa RA, Abizaid A, et al. Three vs twelve months of dual antiplatelet therapy after zotarolimus-eluting stents: the OPTIMIZE randomized trial. Jama. 2013 Dec 18;310(23):2510–2522. doi: 10.1001/jama.2013.282183. [DOI] [PubMed] [Google Scholar]
- 22.Collet JP, Silvain J, Barthelemy O, et al. Dual-antiplatelet treatment beyond 1 year after drug-eluting stent implantation (ARCTIC-Interruption): a randomised trial. Lancet. 2014 Nov 1;384(9954):1577–1585. doi: 10.1016/S0140-6736(14)60612-7. [DOI] [PubMed] [Google Scholar]
- 23.Colombo A, Chieffo A, Frasheri A, et al. Second-generation drug-eluting stent implantation followed by 6- versus 12-month dual antiplatelet therapy: the SECURITY randomized clinical trial. Journal of the American College of Cardiology. 2014 Nov 18–25;64(20):2086–2097. doi: 10.1016/j.jacc.2014.09.008. [DOI] [PubMed] [Google Scholar]
- 24.Gilard M, Barragan P, Noryani AA, et al. Six-month versus 24-month dual antiplatelet therapy after implantation of drug eluting stents in patients non-resistant to aspirin: ITALIC, a randomized multicenter trial. Journal of the American College of Cardiology. 2014 Nov 16; doi: 10.1016/j.jacc.2014.11.008. [DOI] [PubMed] [Google Scholar]
- 25.Oudot A, Steg PG, Danchin N, et al. Impact of chronic oral anticoagulation on management and outcomes of patients with acute myocardial infarction: data from the RICO survey. Heart. 2006 Aug;92(8):1077–1083. doi: 10.1136/hrt.2005.074070. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Wang TY, Robinson LA, Ou FS, et al. Discharge antithrombotic strategies among patients with acute coronary syndrome previously on warfarin anticoagulation: physician practice in the CRUSADE registry. American heart journal. 2008 Feb;155(2):361–368. doi: 10.1016/j.ahj.2007.09.003. [DOI] [PubMed] [Google Scholar]
- 27.Mehran R, Baber U, Steg PG, et al. Cessation of dual antiplatelet treatment and cardiac events after percutaneous coronary intervention (PARIS): 2 year results from a prospective observational study. Lancet. 2013 Nov 23;382(9906):1714–1722. doi: 10.1016/S0140-6736(13)61720-1. [DOI] [PubMed] [Google Scholar]
- 28.Dangas GD, Claessen BE, Mehran R, et al. Development and validation of a stent thrombosis risk score in patients with acute coronary syndromes. JACC Cardiovascular interventions. 2012 Nov;5(11):1097–1105. doi: 10.1016/j.jcin.2012.07.012. [DOI] [PubMed] [Google Scholar]
- 29.Mehta SK, Frutkin AD, Lindsey JB, et al. Bleeding in patients undergoing percutaneous coronary intervention: the development of a clinical risk algorithm from the National Cardiovascular Data Registry. Circulation. Cardiovascular interventions. 2009 Jun;2(3):222–229. doi: 10.1161/CIRCINTERVENTIONS.108.846741. [DOI] [PubMed] [Google Scholar]
- 30.Mehran R, Pocock SJ, Nikolsky E, et al. A risk score to predict bleeding in patients with acute coronary syndromes. Journal of the American College of Cardiology. 2010 Jun 8;55(23):2556–2566. doi: 10.1016/j.jacc.2009.09.076. [DOI] [PubMed] [Google Scholar]
- 31.United States Food and Drug Administration. FDA Drug Safety Communication. FDA review finds long-term treatment with blood-thinning medicine Plavix (clopidogrel) does not change risk of death. 2015 http://www.fda.gov/Drugs/DrugSafety/ucm471286.htm.
- 32.Palmerini T, Benedetto U, Bacchi-Reggiani L, et al. Mortality in patients treated with extended duration dual antiplatelet therapy after drug-eluting stent implantation: a pairwise and Bayesian network meta-analysis of randomised trials. Lancet. 2015 Mar 13; doi: 10.1016/S0140-6736(15)60263-X. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.