AUGUSTUS - Antithrombotic regimens including apixaban vs warfarin, & aspirin vs placebo, in patients with AFib plus PCI &/or ACS

Reference: Lopes RD, et al. Antithrombotic therapy after acute coronary syndrome or PCI in atrial fibrillation. NEJM 2019.

Bottom line: In patients with atrial fibrillation who either undergo PCI and/or have ACS, in combination with a P2Y12 inhibitor (almost always clopidogrel):

  • Apixaban reduces the risk of major or clinically-relevant non-major bleeding (NNT=24), hospitalizations (NNT=27), & stroke (NNT=84) compared to warfarin at 6 months;

  • Aspirin (beyond the first week) increases the risk major or clinically-relevant non-major bleeding (NNH=15), without a clear effect on hospitalization/death or ischemic events compared to placebo at 6 months;

  • Therefore, an antithrombotic regimen of apixaban + clopidogrel (without aspirin) should be routinely considered in these patients. Warfarin should be limited to patients for whom a DOAC is contraindicated, intolerable or unaffordable; & aspirin beyond the first week should be limited to patients with very high risk of stent thrombosis/recurrent coronary events.

Patients (n=4614 from 33 countries)

  • Included if (all of the following):

    • Age 18+ years

    • Known AF (paroxysmal, persistent or permanent) with planned long-term oral anticoagulation

    • Recent (<14 days) ACS &/or PCI with plan for 6+ months of P2Y12 inhibitor

  • Key exclusion criteria:

    • Other indication for anticoagulation (prosthetic valve, VTE, mitral stenosis, etc)

    • History of intracranial hemorrhage, ongoing bleeding or coagulopathy

    • Recent/planned CABG

    • “Severe” renal insufficiency

  • Average baseline characteristics:

    • Age 71 years, male (71%), white (92%)

    • Qualifying event: ACS+PCI (37%), medically-managed ACS (24%), elective PCI (39%)

      • ~6.6 days from ACS/PCI to randomization

    • CHA2DS2-VASc ~4, HAS-BLED ~3

    • Prior stroke/TIA/thromboembolism (14%), HF (43%), HTN (88%), diabetes (36%)

    • SCr >133 (8%)

    • Previous oral anticoagulant (49%)

Interventions x6 months

  • 2x2 factorial design: Patients were simultaneously randomized to apixaban vs warfarin & aspirin vs placebo within 14 days of ACS &/or PCI, so total of 4 different intervention groups.

  • Management prior to randomization: At the discretion of treating physicians according to local standard of care (likely that all at least received DAPT +/- anticoagulation leading up to randomization, though not recorded/reported)

  • Anticoagulation: Apixaban vs warfarin

    • Apixaban arm: 5 mg PO BID

      • Reduced to 2.5 mg PO BID if 2 of the following: Age >80 years, wt <60 kg, SCr >133 umol/L

      • Discontinued study regimen prematurely: 13%

    • Warfarin to target INR 2.0-3.0

      • Median time in therapeutic range (TTR)=59%; INR<2.0 23% of the time, INR>3.0 3% of the time

      • Discontinued study regimen prematurely: 14%

  • Antiplatelet: Aspirin 81 mg PO daily vs matching placebo

    • Discontinued study drug prematurely: 15-17%

  • All: P2Y12 inhibitor left at the discretion of the treating clinicians (clopidogrel 93%, prasugrel 1%, ticagrelor 6%)

  • After 6 months, anticoagulation & antiplatelets were managed according to local standard of care (i.e. not standardized for the trial)

Results @ 6 months

  • Primary outcome: Major or clinically-relevant non-major bleeding, ISTH definition

  • Key secondary outcomes: Composite of death or hospitalization; composite of death or ischemic events (stroke, MI, definite/probable stent thrombosis, or urgent revascularization).

Outcomes at 6 months of apixaban versus warfarin in combination with P2Y12 inhibitor +/- aspirin

Outcomes at 6 months of aspirin versus placebo in combination with P2Y12 inhibitor + apixaban or warfarin

Risk of bias

  • Low risk of: Allocation bias (allocation concealed via interactive voice-response system), attrition bias (low [0.3%] loss to follow-up & analyzed by intention-to-treat), outcome reporting bias (all outcomes of interest defined & reported).

  • Variable risk of performance/detection bias:

    • Apixaban vs warfarin comparison was open-label (i.e. patients & clinicians aware of treatment assignment):

      • All outcomes were adjudicated by a blinded clinical endpoint committee, therefore providing some protection against (but not eliminating) detection bias.

    • Aspirin vs placebo comparison was blinded (patients, clinicians, outcome adjudicators unaware of treatment assignment): Low risk of performance & detection bias.

REDUCE-IT - Icosapent ethyl (EPA) to reduce CV events

Bhatt DL, et al. Cardiovascular risk reduction with icosapent ethyl for hypertriglyceridemia. NEJM

Bottom line: In patients with existing ASCVD or diabetes + other CV risk factors, isocapent ethyl (esterified EPA) reduced the risk of CV events (NNT 21) versus placebo over 4.9 years. Conversely, icosapent increased the risk of AF (NNH 72), peripheral edema (NNH 67) and possibly serious bleeding (NNH 167) versus placebo.

It remains unclear how icosapent ethyl works to reduce CV events, or whether it benefits only patients with elevated triglycerides.

Patients (n=8179)

  • Enrolled in 11 countries from November 2011 to August 2016

  • 19,212 screened (10,429 did not meet inclusion criteria) -> 8179 randomized

  • Included

    • Either

      • Secondary prevention: 45+ y/o + established ASCVD, or

      • Primary prevention: 50+ y/o with diabetes + 1 other CV risk factor (male age 55+ y/o or female 65+ y/o; smoker; HTN; HDL-C <1 for men or <1.2 for women; hsCRP >3 mg/L; CrCl 30-60; retinopathy; albuminuria; ABI <0.9 without intermittent claudication)

    • Fasting triglyceride 1.7-5.6 mmol/L (amended in 2013 to 2.3-5.6)

    • LDL-C 1.0-2.6 mmol/L on a stable statin dose for at least 4 weeks

  • Key exclusion criteria

    • HF NYHA functional class 4; life-threatening condition other than CVD with expected prognosis <2y

    • BP >200/100 mm Hg; HbA1c >10.0%; CrCl <30 or need for peritoneal/hemodialysis

    • Planned PCI/CABG

    • Prior pancreatitis; ETOH abuse in past 6 months

    • Meds:

      • Lipid-lowering drugs other than statin +/- ezetimibe (niacin (>200 mg/d), fibrate, omega-3 supplements, bile acid sequestrants, PCSK9 inhibitors

      • Drugs that affect triglycerides & other lipids (tamoxifen, estrogen, progestins, thyroid replacement, systemic steroids.

    • Allergy to fish or shellfish

  • Typical baseline characteristics

    • 64 y/o, male (71%), white (90%)

    • Secondary prevention (71%), primary prevention (29%)

    • Type 2 diabetes (58%)

    • Labs

      • LDL-C 1.9 mmol/L, HDL-C 1.0 mmol/L, trigs 2.4 mmol/L

      • hsCRP 2.2 mg/L

    • Meds

      • Statin (100%): High-intensity (32%), moderate (62%), low (6%)

      • Ezetimibe 6%

Intervention & control

  • I: Icosapent ethyl 2 g PO BID

    • Purified formulation of eicosapentanoic acid (EPA), one of the main omega-3 fatty acids in fish oil;

    • Far exceeds doses found in over-the-counter (OTC) fish oil products, which are typically labeled to contain ~200 mg of EPA/capsule.

  • C: Matching “placebo” containing mineral oil

Results @ median 4.9 years

  • Reduction in primary CV outcome (composite of CV death, MI, stroke, PCI/CABG, or hospitalization for unstable angina [UA]): Icosapent ethyl 17.2% vs placebo 22.0% (NNT 21) & every individual component

    • Hazard ratio (HR) 0.75, 95% confidence interval (CI) 0.68-0.83

    • Reduction in 3-point MACE (CV death, MI, stroke): 11.2% vs 14.8% (NNT 28), HR 0.74, 95% CI 0.65-0.83

      • CV death: 4.3% vs 5.2% (HR 0.80, 95%CI 0.66-0.98)

      • Non-fatal or fatal MI: 6.1% vs 8.7% (HR 0.69, 95%CI 0.58-0.81)

      • Stroke: 2.4% vs 3.3% (HR 0.72, 95%CI 0.55-0.93)

    • Urgent/emergent PCI/CABG: 5.3% vs 7.8% (HR 0.65, 95%CI 0.55-0.78)

    • UA hospitalization: 2.6% vs 3.8% (HR 0.68, 95%CI 0.53-0.87)

  • Death from any cause: 6.7% vs 7.6% (HR 0.87, 95%CI 0.74-1.02 - inconclusive)


  • Increased:

    • Atrial fibrillation: 5.3% vs 3.9% (NNH 72)

    • Hospitalization for afib/flutter: 3.1% vs 2.1% (NNH 100)

    • Peripheral edema: 6.5% vs 5.0% (NNH 67)

  • Possible increased risk of serious bleeding: 2.7% vs 2.1% (NNH 167; p=0.06)

    • No difference in GI bleeds (1.5% vs 1.1%) or CNS bleeds (0.3% vs 0.2%)

Effect on surrogate outcomes

  • Trigs at year 1: -0.4 vs +0.05 mmol/L (-20% [-0.5 mmol/L] from baseline vs placebo

  • LDL-C at year 1: +0.05 vs 0.18 mmol/L (0.13 mmol/L lower vs placebo)

  • hsCRP at year 2: -0.2 vs +0.5 mg/L (-38% [0.8 mg/L] lower vs placebo

Internal validity

  • Low risk of allocation, performance and detection bias

    • Computer-generated randomization sequence stratified by CV risk group (2o or 1o prevention), use of ezetimibe & geographic region

    • Allocation concealment maintained by central allocation via interactive voice response system

    • Blinding of patients, investigators, clinicians maintained by use of mineral oil “placebo” in control group, which is similar in appearance to the intervention

  • Unclear (potentially high) risk of attrition bias

    • Low loss to follow-up (0.2%) for death

    • High loss to follow-up for non-fatal outcomes, with similar frequency between groups (icosapent ethyl 9.3%, placebo 10.0%)

    • Intention-to-treat analysis.

Other considerations

  • This is not a study of fish oil/omega-3 fatty acid supplements

    • High-quality evidence is exceptionally clear that fish oil/omega-3 fatty acid supplements, such as those sold at pharmacies, health food stores or over the Internet, do NOT reduce the risk of CV events in patients with or without CVD. This has been shown in a meta-analysis of 20 RCTs including 68,680 patients, as well as 2 other recent RCTs (ASCEND & VITAL).

  • The mechanism of action for CV event reduction with icosapent ethyl is unclear

    • Unlikely to be explained by triglyceride reduction

      • Identical CV relative risk reduction (RRR) regardless of baseline triglyceride concentration (<1.7 vs >1.7 or <2.3 vs >2.3 mmol/L);

      • Identical CV RRR regardless of whether achieved trigs <1.7 or 1.7+ mmol/L.

    • Not fully explained by LDL-C increase caused by mineral oil within placebo in comparator group

      • Identical RRR vs placebo patients who had LDL-C increase, decrease, or no change;

      • LDL-C difference of 0.13 mmol/L would only explain a ~3% RR difference based on estimates from the Cholesterol Treatment Trialists’ Collaboration meta-analysis of statins (where 1 mmol/L reduction in LDL-C associated with a ~25% RRR in CV events)

    • Other possible mechanisms include

      • Anti-inflammatory effect (or pro-inflammatory effect of mineral oil in placebo)?

      • Antiarrhythmic effect, or stabilization of cellular membranes?

        • Reduced tertiary outcomes of cardiac arrest (HR 0.52, 95% CI 0.31-0.86) & sudden cardiac death (HR 0.69, 95% CI 0.50-0.96)

      • Antithrombotic effect?

        • Reduced MI, stroke, as well as sudden cardiac events & likely increased risk of bleeding

  • Ongoing trials with EPA +/- DHA: STRENGTH (International) , RESPECT-EPA (Japan)

GLOBAL LEADERS: Ticagrelor-based DAPT x1 month, then ticagrelor monotherapy vs 12 months of standard DAPT in PCI

GLOBAL LEADERS. Lancet 2018;392:940-9.

Bottom Line: In patients who undergo PCI for either ACS or stable CAD, a regimen of ticagrelor plus ASA for 1 month, followed by ticagrelor monotherapy for 23 months, did not reduce the risk of death or Q-wave MI versus standard DAPT at 2 years.

Although major bleeding rates were no different in the overall population, in patients with PCI for ACS, ticagrelor monotherapy after the first month may reduce the risk of bleeding versus ticagrelor plus ASA for 12 months followed by ASA monotherapy. Conversely, in patients with stable CAD, there was no benefit to the ticagrelor-based regimen studied here.

Patients (n=15,991)

  • Included: Scheduled for percutaneous coronary intervention (PCI) for acute coronary syndrome (ACS) or stable coronary artery disease (CAD)

  • Excluded:

    • Taking strong CYP3A4 inhibitor, oral anticoagulant;

    • Use of fibrinolytic <24h before PCI;

    • Planned for CABG within 12 months of randomization;

    • Hx of intracranial hemorrhage, known (current) major bleed, stroke/TIA in last 30 days.

  • Average patient at baseline:

    • Age 64.5 y

    • Female 23%

    • Presentation: ACS 47% (STEMI 13%), stable CAD 53%

    • Cardiac Hx: Prior myocardial infarction (MI) 23%, prior PCI 33%, prior CABG 6%

    • PMHx: Smoker 26%, HTN 74%, diabetes 25%, dyslipidemia 70%, eGFR<60 14%

    • Angiographic characteristics

      • Lesion treated: Left main 2%, LAD 41%, bypass graft 1%

      • Mean total stent length 24.8 mm, stent diameter 3.0 mm

Intervention & Control

  • Intervention (both ACS & stable CAD as indication for PCI)

    • First month: Dual antiplatelet therapy (DAPT) with ticagrelor (load, then 90 mg BID) + low-dose ASA (75-100 mg/d); then

    • Next 23 months: Ticagrelor 90 mg BID monotherapy

    • Adherent (among those assessed): @ 1 month (95%), @ 1 year (81%), @ 2 years (78%)

  • Control

    • ACS as indication for PCI: Ticagrelor (load then 90 BID) + low-dose ASA x12 months, then ASA x12 months

    • Stable CAD as indication for PCI: Clopidogrel (load then 75 mg/d) + low-dose ASA x12 months, then ASA x12 months

    • Adherent (among those assessed): @ 1 month (96%), @ 1 year (89%), @ 2 years (93%)

  • Co-interventions standardized to all patients:

    • PCI performed with biodegradable biolimus-eluting stents;

    • Bivalirudin was used as anticoagulant (not heparin/enoxaparin) during PCI.

Results (intervention vs control) @ 2 Years

  • Primary outcome (all-cause death or Q-wave MI): 3.8% vs 4.4% (hazard ratio (HR) 0.87, 95% confidence interval 0.75-1.01)

    • Death from any cause: 2.8% vs 3.2% (HR 0.95, 0.74-1.22)

    • New Q-wave MI: 1% vs 1.3% (HR 0.80, 0.60-1.07)

    • Subgroup by indication for PCI (no interaction; p=.93):

      • ACS: HR 0.86 (0.69-1.08)

      • Stable CAD: HR 0.87 (0.71-1.08)

    • Landmark analyses:

      • Up to day 30: HR 0.81 (0.52-1.27)

      • Up to year 1 (including first 30 days): HR 0.79 (0.64-0.98)

      • After year 1: HR 0.97 (0.77-1.22)

  • Any MI: 3.1% in both groups (HR 1.00, 0.84-1.19)

  • Stroke: 1% in both groups (HR 0.98, 0.72-1.33)

  • Major bleeding (BARC grade 3 or 5): 2.0% vs 2.1% (HR 0.97, 0.78-1.20)

    • Subgroup by indication for PCI (interaction p=.007):

      • ACS: HR 0.73 (0.54-0.98) - where control is ticagrelor+ASA

      • Stable CAD: HR 1.32 (0.97-1.81) - where control is clopidogrel+ASA

  • Dyspnea: 13.8% vs 6.5% (p<.0001)

Generalizability (External Validity)

GLOBAL LEADERS tested a very complex intervention in a heterogeneous population:

  • First, this trial evaluates two populations with distinct standards of care: Patients with ACS, and patients with stable CAD. In patients undergoing PCI for ACS, the standard-of-care antiplatelet regimen is ticagrelor plus low-dose ASA for 12 months based on the PLATO trial, followed by lifelong ASA. Conversely, there is no evidence that ticagrelor is superior to clopidogrel in patients undergoing PCI for stable angina; no trial has been done to address this question.

  • Second, the interventions in GLOBAL LEADERS differ at 3 timepoints:

    • Day 0-30: Antiplatelet intensity in ACS subgroup (intervention = control), stable CAD subgroup (intervention > control).

    • Day 31-365: Antiplatelet intensity in ACS subgroup (intervention < control), stable CAD subgroup (intervention ? control).

    • Day 366-730: Antiplatelet intensity: intervention ? (> or =) control.

    • Therefore, the neutral results at 2 years may be because of a true lack of difference, or they may be due to a mixture of benefit in some timepoints and harms in other timepoints for these 2 subpopulations.

  • Third, adherence to ticagrelor in this trial was similarly poor to what was seen in PLATO and in clinical practice, driven by a higher rate of dyspnea and nuisance bleeding.

Risk of Bias: Moderate

  • Low risk of allocation bias:

    • Computer-generated randomization by otherwise-uninvolved 3rd party;

    • Blocked randomization with permuted blocks of 2 or 4;

    • Central randomization using locked web-based system.

  • High risk of performance and detection bias:

    • Open-label (patients, caretakers & clinicians caring for patient aware of which intervention they were randomized to);

    • No adjudication of outcomes.

PRAGUE-18: Ticagrelor vs prasugrel in ACS


Bottom line: PRAGUE-18 was underpowered to identify clinically-important differences & was at high risk of bias. As a result, it could not rule out differences between prasugrel & ticagrelor.

Patients (n=1230)

  • Multicentre (14 tertiary-care cardiology centres in the Czech Republic)
  • Included MI (STEMI or NSTE-ACS with ST depression) requiring emergent (<120 min of admission to cardiac centre) angiography +/- PCI 
  • Key exclusion criteria:
    • Administration of clopidogrel loading dose or non-ASA antiplatelet before randomization (but could be on long-term clopidogrel-based DAPT before enrolment)
    • Indication for oral anticoagulant
    • Serious bleeding in past 6 months
    • Stroke, age >75 years, body weight <60 kg
  • Baseline characteristics
    • Age 62 y
    • Female 23-26%
    • At admission
      • EKG: STEMI 89%, LBBB 5%, NSTEMI 6%
      • Killip class: 1 (89%), 2 (6-7%), 3-4 (5-6%)
    • PMHx: Prior MI (7-9%), prior PCI (~7%), prior CABG (<2%), HF (1%), CKD (1%)
    • Procedural characteristics: PCI >99%, stent 96%, drug-eluting stent 68%
    • Meds @ discharge: ASA (97%), ACEI/ARB (83%), beta-blocker (82%), statin (94%), PPI 61%


  • I: Ticagrelor (180 mg PO load, then 90 mg BID)
  • C: Prasugrel (60 mg PO load, then 10 mg daily [or 5 mg daily if age >75 y or wt <60 kg])
  • Common for both groups
    • Load generally administered immediately on hospital arrival before angiography
    • Duration recommended for 12 months
    • ASA administration required, with dose 100 mg/d recommended
    • Switch to clopidogrel if cost of prasugrel/ticagrelor was prohibitive for the patient

Results (prasugrel vs ticagrelor)

@ day 7 (or at discharge if discharged early)

  • Primary outcome (all-cause death, re-MI, stroke, serious bleed requiring transfusion or prolonging hospitalization, or urgent target vessel revascularization): 4.0% vs 4.1% (odds ratio [OR] 0.98, 0.55-1.73)
    • All-cause death: 1.3% vs 2.0% (p=0.30)
    • Re-MI: 1.0% vs 0.7% (p=0.59)
    • Urgent revascularization: 1.4% vs 1.2% (p=0.71)
    • Stroke: 0.2% vs 0.2% (p=0.96)
    • Serious bleed: 1.3% vs 1.2% (p=0.90)

@ day 30

  • CV composite (CV death, MI, stroke): 2.7% vs 2.5% (p=0.86)
  • All-cause death: 2.2% vs 2.7% (p=0.59)
  • Definite stent thrombosis: 0.5% vs 0.9% (p=0.43)
  • TIMI major bleed: 0.6% vs 0.7% (OR 0.86, 0.17-4.27)
  • No difference in bleeds based on BARC definition

@ 1 year

  • CV composite (CV death, MI, stroke): 6.6% vs 5.7% (hazard ratio [HR] 1.17, 0.74-1.84)
    • CV death: 3.3% vs 3.0% (p=0.77)
    • Non-fatal MI: 3.0% vs 2.5% (p=0.61)
    • Stroke: 1.1% vs 0.7% (p=0.42)
  • All-cause death: 4.7% vs 4.2% (p=0.65)
  • Definite stent thrombosis: 1.1% vs 1.5% (p=0.53)
  • TIMI major bleed: 0.9% vs 0.7% (p=0.75)
  • BARC major bleed: 2.4% vs 1.5% (p=0.31)
  • Dyspnea: Not reported


  • Generalizability: Widely applicable to patients with STEMI requiring primary PCI.
  • High risk of several biases
    • High risk for allocation bias
      • Simple randomization, no use of permuted blocks or stratification by site;
      • Allocation concealment by sealed envelopes (prone to tampering).
    • High risk for performance & detection bias
      • No blinding of patients & clinicians to study drug;
      • Differential study drug discontinuation
        • Patients who could not afford the study drug could switch to clopidogrel
        • Study drug not free for participants, & there was differential funding for prasugrel & ticagrelor, as ticagrelor was not funded by the public insurance plan, whereas prasugrel was covered for patients with STEMI plus either left main disease, pLAD or multivessel disease
        • Switch to clopidogrel for financial reasons higher in ticagrelor group (44.4%) vs prasugrel (34.1%, p=0.003).
      • Blinded endpoint adjudication, however issues upstream as described above limit the value of blinded adjudication of potentially biased reports.
    • Low risk of attrition bias: Loss to follow-up: <1% at 30 days, 0% at 1 year.
  • Other issue: Study completely underpowered to detect a clinically-significant difference.
    • Initial power calculation with target sample size 2500 was based on unrealistic expectations
      • Done to detect a 2.5% absolute risk difference (or 39% relative risk difference) between ticagrelor & prasugrel (larger than difference between either drug compared to clopidogrel)
    • Confidence around point estimate of trial stopped early with sample size of 1230 cannot rule out a ~3% absolute risk difference in CV events between these drugs.

TACT: Heavy metal chelation for cardiovascular risk reduction

Lamas, et al. Effect of disodium EDTA chelation regimen on cardiovascular events in patients with previous myocardial infarction: the TACT randomized trial. JAMA 2013;309:1241-50.

Bottom line:

  • The role of chelation in cardiovascular disease is unproven & unclear;

  • The TACT trial, which has unclear risk of bias due to possible unblinding, demonstrated a reduction in cardiovascular events in patients with prior MI (mainly driven by coronary revascularization) with a complex regimen that includes a chelator, as well as pharmacologically-active agents such as heparin & procaine;

  • If a patient chooses to undergo chelation, clinicians should encourage ongoing adherence to evidence-based pharmacological therapies, & implement harm-minimization strategy (i.e. avoid use in those with contraindication to chelation or frequent IV infusions; monitor for fluid overload & hypocalcemia).


  • Chelation therapy (the act of administering a drug to bind metal ions) has been used in a non-evidence-based fashion to treat coronary artery disease (CAD) since the 1950s. The original rationale was to bind calcium and "decalcify the heart" following observations that autopsied hearts had heavily calcified coronary arteries.
    • Notably, we now know that calcification of arteries is a late consequence of atherosclerosis & not the direct cause of coronary events.
  • The rationale for chelation therapy in atherosclerotic CVD now centers around observations that increased exposure to heavy metals, particularly cadmium & lead, is associated with cardiovascular disease, possibly by causing endothelial dysfunction (cadmium), increased BP (lead), & inducing macrophage dysfunction & apoptosis (cadmium).
  • The only supporting evidence for chelation therapy in CAD prior to this trial came from uncontrolled case reports & case series.


  • 134 sites in US & Canada
  • 60% of sites already practiced chelation therapy (generally naturopathic & alternative medicine clinics)

Patients (n=1708)

  • Inclusion
    • Age >50 years
    • Prior MI (>6 weeks before enrollment)
  • Exclusion
    • Past intolerance to chelation or vitamin components
    • Inability to tolerate 500-mL infusions weekly
    • Chelation therapy within 5 years
    • Coronary/carotid revascularization planned or in prior 6 months
    • "Active" HF or HF hospitalization within 6 months
    • Current smoker or smoking in last 3 months
    • BP >160/100 mm Hg
    • SCr >170 umol/L
    • Plt <100
    • Abnormal LFTs
  • Baseline characteristics
    • 65 years
    • Female 18%
    • White 94%
    • CAD characteristics: Last MI 4.6 years ago, anterior MI 40%, angina 55%, prior CABG/PCI 83%
    • PMHx: Former smoker 56%, CHF 18%, diabetes 40%
    • BP 130/76
    • Meds: Antithrombotic 91% (ASA in 83%), ACEI/ARB 63%, beta-blocker 72%, statin 73%, multivitamin 44%, herbal products 34%
    • Labs: LDL 2.3 mmol/L, HDL 1.1 mmol/L, SCr 97 umol/L


  • Intervention: Chelation+ infusion
    • Each infusion consists of 10 components commonly combined by US chelation practitioners: Disodium EDTA (ethylene-diamine-tetracetic acid) 3 g (decreased based on eGFR), procaine hydrochloride 100 mg, heparin 2500 units, ascorbic acid (vitamin C) 7 g, magnesium chloride 2 g, potassium chloride 2 mEq, sodium bicarbonate 840 mg, pantothenic acid 250 mg, thiamine 100 mg, pyridoxine (vitamin B6) 100 mg, & sterile water to total 500 mL
    • Each infusion administered over minimum 3h (slower if HF or albumin-corrected serum Ca 2.0-2.12 mol/L)
    • 30 weekly infusions, followed by 10 infusions q2-8 weeks
  • Control: Placebo infusion (500 mL of NS & 1.2% dextrose)
  • Co-intervention: Daily vitamin containing vitamin B6 25 mg, chromium 50 ug, copper 2 mg, manganese 15 mg, zinc 25 mg

Results @ median 4.6 years


  • Chelation reduced the primary outcome (all-cause death, re-MI, stroke, coronary revascularization, angina hospitalization): Chelation 26% vs placebo 30% (hazard ratio [HR] 0.82, 0.69-0.99)
    • Driven by fewer coronary revascularizations: 15% vs 18% (HR 0.81, 0.64-1.02)
  • No significant difference in secondary outcome (CV death, re-MI, stroke): 11% vs 13% (HR 0.84, 0.64-1.02)
  • No significant difference in death (10% vs 11%), MI (6% vs 8%), stroke (1% vs 1%), or angina hospitalization (2% vs 2%)
  • No difference in quality of life scores at 2 years (in either cardiac functional status measured using Duke Activity Status Index, or SF-36 Mental Health Inventory-5) in subset of 911 patients


  • Serious adverse events: Chelation 12% vs placebo 15% (p=0.10)
  • Non-serious adverse events: 68% vs 67% (p=0.60)
  • Hypocalcemia (serum Ca <2.0 mmol/L): 6.2% vs 3.5% (p=0.008)

Subgroups: 9 subgroups evaluated, 2 had significant interaction p-value (diabetes, anterior MI) for primary outcome

  • iabetes status (p=0.02 for interaction)
    • Diabetes: HR 0.61, 0.45-0.83
    • No diabetes: HR 0.96, 0.77-1.20
  • MI distribution (p=0.03 for interaction)
    • Anterior MI: HR 0.63, 0.47-0.86
    • Non-anterior MI: HR 0.96 (0.77-1.20)
  • No difference between complementary/alternative sites vs others (p=0.28)

Generalizability & other considerations

  • Key characteristics of included patients
    • Several groups of patients with CAD were excluded, notably current smokers (rationale unclear), and those with "active" heart failure or recent hospitalization for HF, uncontrolled HTN or renal dysfunction (who may not tolerate the weekly fluid infusions)
    • Patients with chelation in the past 5 years were also excluded, likely to remove bias from including any patients with a potential preference for this therapy
    • Use of evidence-based therapies for CAD at baseline was low (particularly for statins; only used at any dose in 73%), & rates later in the trial are not reported. It is unclear if use of chelation, or practitioners of chelation themselves, may have influenced the use of proven, evidence-based therapies.
  • The chelation regimen includes a number of other pharmacologically-active components, including heparin (anticoagulant) and procaine (pro/antiarrhythmic, vasodilator). TACT used this combination as it is the most commonly used by chelationists in the US, but the actual therapeutic rationale is unclear. The apparent benefit in the primary outcome of TACT may be in part explained by these other components rather than chelation. Moreover, these extra components could be harmful.
  • The difference in the primary outcome is marginally statistically significant (p=0.035, just below the modified 0.036 threshold), & completely driven by coronary revascularization, the softest endpoint in the composite.
  • The findings from the subgroup analysis should be interpreted cautiously in context of the limitations: large # of subgroup analyses performed (9); the beneficial effect found in diabetes & anterior MI was not expected; no biologic rationale to explain the benefit in diabetics & those with anterior MI but not in others; no other RCT with which to compare.

Internal validity: Unclear risk of bias

  • Allocation: Web-based randomization with permuted blocks, allocation concealed.
  • Performance: Patients, clinicians blinded, matching placebo
    • Concern has been raised about possible unblinding of participants due to the higher withdrawal of consent in the placebo group, suggesting that patients were somehow told that they were receiving placebo and decided to leave the trial. This is not proven, but is one possible explanation for this finding, and would therefore introduce bias in the softer outcome of revascularization, which drove the difference in the primary outcome.
  • Detection: Blinded adjudication committee.
  • Attrition: Intention-to-treat analysis, but high loss-to-follow-up with differences between groups
    • 18% lost-to-follow-up, mainly due to withdrawal of consent, which was higher with placebo (20%) than chelation (13%). Usually, differences in loss-to-follow-up between groups are because of tolerability issues with 1 of the interventions, however, in this case those receiving placebo left the trial more often. As discussed above, this suggests unblinding. Other explanations would include chance (unlikely with such a large difference), people receiving chelation feeling better than placebo, thus being motivated to continue treatment (but this is not corroborated by the quality of life results).
  • Selective outcome reporting: Reporting of all clinical outcomes of interest for both efficacy & safety.
  • Others:
    • Trial initially planned to enroll ~2400 patients, but decreased to 1700 midway due to slow patient recruitment.
    • Data safety monitoring board performed 11 interim analyses prior to the final analysis.