Bottom line: The DAPT Score is an easy-to-use tool to help select patients for extended DAPT duration after PCI, but it hasn't yet been sufficiently validated in a real-world population to use in practice.
Particularly important issues that may crop up in future validation include poor discrimination in a real-world heterogeneous population (& already fairly low with c-statistic ~0.70 or less in RCT populations), poor calibration, & over-simplification from combining ischemic & bleeding risk into a single score.
Until publication of the DAPT trial in 2014, limited evidence to guide optimal dual antiplatelet therapy (DAPT) duration following acute coronary syndrome (ACS) and/or percutaneous coronary intervention (PCI) with drug-eluting stent (DES) placement, so guidelines routinely recommended 12 months for all patients.
Numerous subgroup analyses looked at patient characteristics that could be used to identify patients that could obtain a net benefit from prolonged DAPT, but these only look at 1 characteristic at a time;
Decisions on DAPT duration are complex and require integration of multiple characteristics.
The newest AHA guidelines on DAPT recommend using a risk prediction tool such as the DAPT Score to aid in deciding on DAPT duration.
- Available here
Score-based risk calculator ranging from -2 to 10, with lower scores representing an unfavorable benefit/risk ratio from extended DAPT and higher scores representing a greater net benefit from extended DAPT.
The investigators divided the score into 2 categories:
With a DAPT Score <2 (-2 to 1), extending DAPT >1 yearr:
NNT 167 for MI/stent thrombosis
NNH 72 for GUSTO moderate-severe bleed
With a DAPT Score >2 (2-10), extending DAPT >1 year resulted in:
NNT 53 for MI/stent thrombosis
NNH 250 for GUSTO moderate/severe bleed
Level of evidence
- Derivation, internal validation & external validation: 2a (level of evidence ranges from 1 [highest] to 4 [lowest])
- Hierarchy of evidence for clinical prediction rules: Level 3 (validated in only 1 narrow prospective sample)
- Data sources:
- Study setting: Outpatient cardiology follow-up
- Patient population: Individuals undergoing PCI with DES (elective or urgent)
- Adequate proportion of patients with each predictor variables?
- es, in DAPT ~300 patients (3%) with rarest included predictor variable (stent in vein graft); other characteristics present in >2000 (>20%) of patients
- 37 predictor variables initially considered
- Demographics: Age, sex, race
- CV history: HF or LVEF <30%; prior CABG, PCI or MI; PAD; stroke/TIA; AF
- Comorbidities: BMI; cancer at time of randomization; current smoking; diabetes; history of major bleed; hypertension; renal insufficiency (SCr >2 mg/dL)
- Procedural characteristics: >2 lesions/vessel; bifurcation stenting; coronary lesion class C; number of stents; number of treated vessels; presentation with MI; pre-procedural stenosis; prior brachytherapy; prior in-stent thrombosis; severe coronary calcification; stenting of vein graft; stent diameter; stent type; thrombus-containing lesion; TIMI grade flow post-procedure; total stent length; unprotected left main stenting
- Meds: Randomization arm (DAPT 12 vs >12 months); clopidogrel vs prasugrel; statin at time of randomization
- Checklist for what makes good predictor variables (all were met):
- Clear & reproducible predictor definition
- Available at time of decision
- Assessors were blind to the outcome at time of predictor variable determination (inherent in prospective design of these RCTs)
- Predictor variable definition consistent between derivation, internal validation & external validation cohorts
- Composites of:
- MI/stent thrombosis
- Moderate/severe bleed (GUSTO definitions)
- Checklist for what makes good outcomes (all were met):
- Clinically important
- Clear & reproducible definition
- Caveat: Adjudicators not blind to predictor variables at time of outcome assessment (though blind to the existence of the DAPT Score derivation & related hypotheses)
- Analysis appropriate & well-described: Hazards ratios derived using Cox multivariable regression
- Handling of missing data: Unclear
- Overfitting? Low risk (far more than rule-of-thumb 10 outcome events/predictor variable included in statistical model)
- Discrimination (tool's ability to distinguish between patients who do & those who don't experience the outcome, using the c-statistic, where a c-statistic of 0.50 = model is as good as chance, & a c-statistic of 1.00 = perfect discrimination): Good, not great
- Internal validation (DAPT):
- Ischemic outcome: 0.70 (moderate)
- Bleeding outcome: 0.68 (moderate)
- External validation (PROTECT): c-statistics = 0.64 for ischemic & bleeding outcomes
- Internal validation (DAPT):
- Calibration (tool's ability to correctly estimate the incidence of an outcome in a population)
- Internal validation (DAPT): Good
- External validation (PROTECT): Overestimated risks of outcomes (because PROTECT enrolled a lower-risk population than the DAPT trial). Good after re-calibration for this lower incidence in the overall study population.
Generalizability (also known as transportability)
- To RCT populations similar to the DAPT & PROTECT trials: Excellent
- To different geographical areas, clinical settings, providers (e.g. GPs, internists, pharmacists, nurses. etc): Unknown
- To different follow-up intervals (e.g. extended DAPT beyond 3 years): Unknown
- To patients with different spectra of coronary artery disease or comorbidities: Unknown
- To patients on different antiplatelet agents (e.g. ticagrelor) or on anticoagulant therapy (e.g. DOAC for AF): Unknown