Author + information
- Holger Eggebrecht, MD∗ (, )
- Axel Schmermund, MD and
- Rajendra H. Mehta, MD, MS
- ↵∗Cardioangiological Center Bethanien, Im Prüfling 23, 60389 Frankfurt, Germany
We have read with great interest the meta-analysis by Giustino et al. (1) of 5 modest-size randomized trials (n = 30 to 363) suggesting favorable effects of cerebral embolic protection (EP) devices during transcatheter aortic valve implantation (TAVI).
We believe that some issues inherent to this meta-analysis need discussion. First, EP devices were developed to decrease cerebral embolism and resultant stroke while they are deployed. However, only one-half of neurologic events occur during or early after the TAVI procedure. After this, neurological events are related to new-onset atrial fibrillation or thrombus on valve leaflet that EP device is unlikely to prevent. The timing of stroke in reference to the index procedure was not reported in the 5 studies of the meta-analysis. However, given the follow-up period of up to 6 months it is likely that strokes were included that occurred days after the EP was actually removed.
Second, we concur that it is not inappropriate to use a composite of periprocedural stroke or mortality of unknown cause where a determination that a stroke had occurred before death or resulting in it had not been possible. However, this is hardly justified when the cause of death was known and unlikely to be affected by use of EP, as were the 2 deaths caused by annular rupture in the control group of the DEFLECT-III trial. In fact, most deaths in the high-risk fragile TAVI patient population are not related to neurological sequelae.
Finally, Giustino et al. (1) have used a nonhierarchical definition of the combined endpoint stroke or death for their analysis, but rather double counted events that occurred in the same patient. In the MISTRAL-C study, 3 of 33 control group patients died during 30-day follow-up, 2 of them from severe stroke. In the meta-analysis, 6 events were counted (2 strokes, 3 deaths during 30 days and 1 additional death beyond 30 days), which we believe is incorrect because the 2 patients who died of stroke were double counted. For the largest trial of the meta-analysis, the SENTINEL trial, hierarchical outcomes were not reported. However, severe strokes occurred infrequently (control, n = 1 of 109 [0.9%] vs. EP, n = 2 of 231 [0.9%]). Thus, we concede that in this case a potential bias from nonhierarchical endpoint definition is likely to be rather small and unlikely to affect the overall results of the meta-analysis.
In conclusion, the current meta-analysis would have served well if it was based on patient-level data on outcomes including their timing so that appropriate endpoints likely to be impacted by EP could have been analyzed. Otherwise, such meta-analyses are unlikely to provide insights into the effectiveness of these devices for stroke prevention in TAVI patients.
Please note: All authors have reported that they have no relationships relevant to the contents of this paper to disclose. Deepak L. Bhatt, MD, MPH, served as Guest Editor-in-Chief for this paper.
- 2017 American College of Cardiology Foundation