Unblinded non-inferiority RCT with high loss-to-follow-up/analysis with differences between groups (24% vs 13%) that was stopped early & analyzed using modified intention-to-treat population.
Non-inferiority trial details:
- Non-inferiority margin: 1.0% PER YEAR absolute risk increase in 1o outcome (composite of thromboembolism/bleed) with intervention vs control
- No minimum clinically important difference provided
- Final sample size calculation of 1000 patient-years of follow-up/group would have provided 80% power for non-inferiority with 1-sided alpha of 0.05 (actual follow-up ~755 and ~675 pt-y/group)
Patients (n=374)
- 33 centres in Canada + US
- Enrolled from June 2006 to October 2009, followed to March 1, 2013
- Inclusion criteria:
- Adult (18+ y/o)
- Isolated mechanical AVR
- 1+ risk factor for thromboembolism, including:
- Hx of neurologic events
- Chronic AF
- LV ejection fraction <30%
- Left or right ventricular aneurysm
- Enlarged left atrium (diameter >50 mm)
- Spontaneous echo contrast in left atrium
- Hypercoagulability (as determined by testing for factor V Leiden, prothrombin mutation, antithrombin III, protein C, S & factor VIII activity, LDL)
- Vascular pathologic features
- Lack of platelet response to ASA or clopidogrel
- Women receiving estrogen replacement therapy
- Received warfarin with target INR 2-3 + ASA 81 mg/d for the first 3 months post-op
- No thromboembolism during first 3 months post-AVR
- Key exclusion criteria:
- Note: Did not exclude if concomitant cardiac surgery such as CABG or mitral/tricuspid repair, ascending aortic replacement, maze procedure
- Double left-sided valve replacement (i.e. also had MVR)
- Right-sided valve replacement
- Active endocarditis at implantation
- 425 screened -> 375 randomized -> 374 analyzed as part of intention-to-treat analysis
- Average patient:
- 55 y/o
- 20% female
- Concomitant procedures during AVR
- CABG 27%
- Aortic aneurysm repair 14%
- Other 25%
- Valve characteristics
- Etiology: Calcified ~2/3, bicuspid ~1/3
- Lesion: Stenosis (~50%), regurgitation (~25%), both (~25%)
- Risk factors
- Hx neurologic event 3-5%
- AF <5%
- LVEF <30% 5%
- Enlarged left atrium ~10%
- Estrogen replacement therapy 1-2%
- Abnormal labs
- Antithrombin III activity 15%
- Other hypercoagulable states (<5% each)
- P2Y12 inhibition ~25%
- Urine thromboxane ~35-45%
Intervention and Control
- Co-interventions common to both groups:
- 1st 3 months post-AVR: Both groups treated with warfarin to target INR 2.0-3.0 + ASA 81 mg/d
- Patients in both groups received a home "point-of-care" INR monitor & instructed to check INR q1week
- >80% of patients checked their INR at least 3x/month
- Warfarin dose adjustment made by study sites (rather than by family doctor or other local anticoagulation service)
- Intervention: Lower INR target
- Starting 3 months post-AVR: Warfarin to target INR 1.5-2.0 + ASA 81 mg/d
- Time in therapeutic range (INR 1.5-2.0): 66%
- Mean INR: 1.9 +/- 0.5
- Patients who experienced a thromboembolic event had their INR target increased to 2.0-3.0
- Control: Standard INR target
- 3 months post-AVR: Continue warfarin to target INR 2.0 to 3.0 + ASA 81 mg/d
- Mean INR 2.5 +/- 0.6
- Time in therapeutic range (INR 2.0-3.0): 55%
Outcomes
- Median follow-up: Low-target 5.1 years, standard target 5.7 years
- Major bleeding: Low target 1.6%/year, standard target 3.9%/year (NNT 44/year)
- Thromboembolism: Non-inferiority NOT demonstrated
- Study did not report any formal non-inferiority statistical testing; reported as "similar/no different" rates of thromboembolic events between groups
- Stroke/TIA: Low target 2%, standard target 1.65% (rate ratio 1.22, 95% CI 0.64-2.32)
- 95% CI crosses the 1.0%/y non-inferiority margin, & cannot rule out 2% absolute increase PER YEAR
- Peripheral thromboembolism: Low target 0.4%, standard target <0.1% (rate ratio 4.61, 95% CI 0.52-41.28)
- 12% of patients in the low-target group crossed over to the standard target due to thromboembolism or valve thrombosis
Key Considerations
The study did not demonstrate non-inferiority of the lower INR target
- Based on the 95% CI, the absolute risk of stroke/TIA could be as much as 2.2% higher per year, which is far greater than the 1.0%/y non-inferiority margin.
- Notably, theis non-inferiority margin was not based on any statistically appropriate criteria or an agreed-upon minimum clinically important difference (MCID)
- Additional problems with this non-inferiority analysis:
- Primary outcome included competing events (thrombosis & bleed). Inclusion of bleeding events in this composite outcome attenuates any difference between groups in favor of the lower target INR group, biasing the results towards non-inferiority
- The initial 2014 publication was based on an unplanned interim analysis
- The non-inferiority margin changed between publications (1.5%/year in the 2014 publication and 1.0%/year in the 2018 publication), and was based on absolute instead of relative risk differences
- Done on ITT population only (non-inferiority trials should report both ITT & per-protocol analyses)
- In summary, the results of this trial suggest that loosening the INR target from 2.0-3.0 to 1.5-2.0 may increase the risk of the primary outcome by up to 8% over ~4 years.
Generalizability: Most patients in this trial did not have typical high-risk features for thromboembolism, These risk factors (including AF, enlarged LA, reduced LVEF, hypercoagulability, vascular disease or history of embolism) were each present in <10% of the study population. Therefore, these results only truly apply to patients with On-X AVR at relatively low risk of thromboembolism.
Internal validity: Unclear/high risk of bias in multiple domains likely underestimate differences between groups in terms of thromboembolic events, & overestimate differences in favor of the lower-target group for bleeding events.
Summary updated: July 30, 2018
