4 Reasons to Question Percutaneous LAAC After AF Ablation

If electrophysiologists were as skilled in critical appraisal of evidence as they are with procedures, the results of the OPTION trial comparing two strategies of stroke prevention after atrial fibrillation (AF) ablation would not lead to more percutaneous left atrial appendage closure (LAAC) procedures. 
Sadly, though, the positive results probably will lead to many more of these unproven devices being implanted. As I wrote earlier this week, OPTION was designed to deliver positive results. And so it did. 
Brief Recap of OPTION 
The trial enrolled 1600 patients after AF ablation and randomly assigned them to continue with oral anticoagulation (OAC) or have a Watchman LAAC device inserted and discontinue anticoagulation after 90 days. 
The idea was to show that LAAC could decrease the rate of bleeding while maintaining a low risk for stroke after ablation. 
Patients were 70 years old on average, 35% were female, and the mean CHA2DS2–VASc score was 3.5. Just over 40% of patients had LAAC at the same time as their ablation and the rest had the device implanted weeks later. OPTION Results 
The primary safety endpoint of nonprocedural major bleeding or clinically relevant nonmajor bleeding occurred in 8.5% of patients in the device arm vs 18.1% in the anticoagulation group. The hazard ratio (HR) was 0.44 (95% CI, 0.33-0.59; P <.001 for superiority). Two thirds of the bleeding events were nonmajor, including bruising, epistaxis, lacerations, and oral bleeding. 
The primary efficacy endpoint of stroke, systemic embolism, and death occurred in 5.3% of patients in the device arm and 5.8% of patients in the anticoagulation arm and met noninferiority. 
The secondary safety endpoint of major bleeding, which includes procedure-related bleeding, occurred in 3.9% of patients in the device arm and 5.0% of patients in the anticoagulation arm (HR, 0.77; 95% CI, 0.48-1.24). This met noninferiority but not superiority.
Death rates were similar in both arms (3.8% vs 4.5% for anticoagulation). Ischemic stroke rates were very low and also similar in both arms (1.2% vs 1.3%). 
In a noninferiority trial, you aim to establish noninferiority of an efficacy endpoint and superiority in safety. The topline results of OPTION confirm this for LAAC vs anticoagulation, but a closer look at endpoints and trial procedures refute this contention. 
1. Wrong Primary Safety Endpoint
Safety cannot be established as superior with a nonprocedural bleeding endpoint. The main safety concern with LAAC is procedure-related bleeding. 
To assess safety properly, we should use the secondary safety endpoint of major bleeding including procedural bleeding. These rates were 3.9% for LAAC and 5.0% for anticoagulation, which did not establish superiority for the device. 
Excluding procedural bleeding from the primary safety endpoint isn’t my only issue with its choice. The inclusion of nonmajor clinically relevant bleeding creates a possible bias in an unblinded trial. Patients on anticoagulation may be more likely to report nonmajor bleeding events. Indeed, bruising and oral bleeding were higher in the OAC arm. 
Evidence for this bias is that the Kaplan-Meier curves for the primary safety endpoint diverge almost immediately in favor of the device arm — despite the fact that the LAAC patients were initially taking an anticoagulant and aspirin. This more intense antithrombotic regimen should cause more, not less, bleeding than anticoagulation alone. 2. Wrong Efficacy Endpoint
The indication for LAAC or anticoagulation is to prevent thrombotic events, not death. Including death with stroke and systemic embolism in the composite endpoint simply inflated the numbers of primary efficacy endpoints. As expected, death rates were similar in both arms. 
Yet even with these inflated rates, the actual event rates of 5.3% and 5.8% (LAAC vs OAC) were far lower than the expected 10% vs 15%. This is important because the fixed noninferiority margin of 5% is based on the expected event rates and translates to a relative risk of 1.5%. In other words, LAAC could have been 50% worse in terms of efficacy and still have met the threshold for noninferiority. 
The actual event rates came in much lower than expected at 5.3% vs 5.8%. The risk difference of -0.5% has an upper bound of 1.8%, which is well below the noninferiority margin of 5%.
Proponents might argue that even with the lower event rates, noninferiority is still met using the relative risk ratio. This is true. The HR of 0.91 had a 95% CI of 0.59-1.39, and the upper bound of 1.39 is indeed less than the margin of 1.5. But this is deeply misleading because the actual stroke/systemic embolism rates were 1.2% and 1.3%, respectively. It would have taken nearly five times more patients to properly establish noninferiority in efficacy. OPTION, therefore, cannot tell us whether LAAC is noninferior to OAC for the reduction of stroke and systemic embolism. The fact that stroke rates are this low 3 years after ablation leads me to believe that an arm of no device and no OAC would have performed as well. In fact, one of the discoveries in OPTION was that stroke rates were at least three times lower than predicted based on the mean CHA2DS2–VASc score of the patients. This observation counters the view of many LAAC proponents who argue that the device could be an alternative option for patients who do not want to take OAC. A better alternative in low-stroke risk patients could be no device and no OAC. This comparison has never been tested in a trial; OPTION argues for such a trial. 
3. Missing Data
The design of the trial is not the only reason I worry. The authors report missing data for 144 patients (8.8% overall), with most of those patients in the anticoagulation arm (10.5% vs 7.4% missing in the LAAC arm). 
One way to assess whether the lost data might affect the outcome is to compare it with the event rates. Here, the rate of missing data is higher than both, what I would deem the correct safety endpoint (as outlined above) and the inflated efficacy endpoint. 
While there is no perfect solution for missing data, the authors made no statistical adjustments for the issue. In a trial with low event rates, this much missing data adds even more uncertainty. 
4. Incomplete LAAC and Possibility of Harm 
The authors report that a complete seal of the opening to the left atrial appendage with the device was observed in 80% of patients at 12 months. The translation is that 1 in 5 patients had a leak. Device-related thrombus occurred in 1.9% of patients. 
I see two potential issues relating to harm. The OPTION trial was done at expert centers. In the United States, large numbers of AF ablation and LAAC will be done by low-volume operators. The incomplete-seal numbers in OPTION probably represent a best-case scenario. If the results of this trial lead to a large increase in the implantation of these devices, there is a risk for serious harm on a population level. 
The second problem with incomplete seals is the short duration of the trial. An implanted device that doesn’t completely seal the opening to the left atrial appendage may be forever. OAC can be stopped. A device in the heart cannot be easily removed. Conclusion 
Despite enrolling 1600 patients in a randomized trial, we don’t know whether percutaneous LAAC post-ablation is as safe or effective as OAC (or even if it’s better than doing nothing). 
Despite the declaration of positive results, the OPTION trial was uninformative. It worries me greatly that this was easily predicted before knowing the results. 
I hope my field of electrophysiology sees through the positive topline results and rejects the coming spin. Putting foreign bodies into the heart should pass a much higher bar than the one set by OPTION. There is potential for a serious amount of harm if this leads to more LAAC procedures.  
John Mandrola practices cardiac electrophysiology in Louisville, Kentucky, and is a writer and podcaster for Medscape. He espouses a conservative approach to medical practice. He participates in clinical research and writes often about the state of medical evidence.