IN RESPONSE

We appreciate readers' interest in and insight on our recent report of the association between residual shunt after PFO closure and stroke recurrence. We agree with Dr. Kumar and colleagues' comment about the challenge of TIA diagnosis and the difficulty of precise PFO RLS quantification with current methods, especially in light of the connection between PFO and migraine (1). Transient ischemic attack has no footprint on diffusion-weighted magnetic resonance imaging; however, embolic TIA and migraine can be differentiated clinically in conjunction with collateral history, advanced perfusion imaging, and clinical expertise. We used a rigorous protocol of 2 vascular-trained neurologists independently reaching consensus for PFO-associated TIA.

We further agree that TCD offers high accuracy in initial screening, but it has limitations in differentiating between intracardiac and extracardiac shunting. Our protocol found TTE with bubble with the Valsalva maneuver to have the highest degree of accuracy (2). It is also safer than TEE and is our first choice. However, because of the complex 3-dimensional dynamic nature of PFO, complementary methods (including TEE to better visualize fossa ovale and TCD as a sensitive initial screen) or such new methods as cardiac computed tomography may be the best strategy for some patients–especially those with smaller shunts or complex anatomy. Because of challenges posed by all current methods, we are exploring circulatory biochemical markers for better PFO shunt quantification (3).

We appreciate that Dr. Aggarwal and associates highlighted the crucial interplay among PFO, AF, and such other causes as aortic atheroma for recurrent stroke. To explore this, we performed a new analysis that they advised by adding AF as a covariate. We found that the hazard ratio did not change significantly (3.05 [95% CI, 1.65 to 5.62] vs. 3.04 [CI, 1.65 to 5.61]), which suggests that AF was not an important contributor to recurrent stroke. In addition, most device-related AF in our patients was nonsustained after PFO closure.

We agree that extending the cardiac monitoring period beyond 30 days would be of interest. However, the data that Dr. Aggarwal and associates cited noting higher risk for postclosure AF were obtained using a STARFlex closure device (NMT Medical), which is no longer on the market because of design issues resulting in clotting (4). Advances in the development of better PFO closure devices will probably lessen the risk for postclosure AF.

Because all patients were diagnosed with PFO-attributable cryptogenic stroke after thorough stroke workup to rule out other known mechanisms, aortic atheroma did not contribute to recurrent risk. Thus, AF and other traditional causes of stroke do not seem to have confounded our findings; residual shunt remained the most significant risk for recurrent stroke after adjustment for all covariates (see Table 2 of our article). Because PFO is finally coming of age as a cause of stroke, such novel PFO-specific risks as hypercoagulable state that increase stroke recurrence will be the next frontier (3, 5).

We thank Dr. Karagianni and coworkers for independently replicating our finding from their own cohort, further strengthening the association between recurrent stroke and residual shunting after PFO closure. They also raise 2 separate and important concerns. The first is whether residual shunting can really decrease over time; the short answer is that it can, as shown in Appendix Table 4 of our article and our prior publications (6). The reason is gradual device epithelialization, especially within the first few years. We have observed this phenomenon when following individual patients clinically over time. However, we do not claim that stroke risk itself diminishes over time, which Dr. Karagianni and coworkers misinterpreted. On the contrary, we urge vigilant follow-up with multidisciplinary care and concluded, “[P]atients with residual shunt, particularly those with moderate or large shunt, face continued risk.” In particular, as we noted in the clinical recommendations mentioned in the Discussion section of our article, traditional risk factors for stroke may interact and gain more importance as patients age. As such, risk for PFO-mediated stroke for those with residual shunting may also increase as other comorbid conditions (such as diabetes mellitus) create more hypercoagulable conditions.

Dr. Karagianni and coworkers' second concern pertains to the rationale for our sensitivity analysis. Following the editors' advice, we found that follow-up was shorter for the residual shunt group than for the complete-closure group (mean follow-up, 3.2 years vs. 3.8 years; P = 0.002). As stated in the Discussion section of our article, the direction of this differential gave us confidence in our result because it probably underestimated recurrent stroke/TIA events in the shunt group, biasing toward null. We confirmed this by performing a sensitivity analysis (a recalculation of outcome under alternative follow-up scenarios), which showed that the difference in loss to follow-up did not affect the conclusion.