Outflow Graft Twist Occlusion in the HeartMate 3 Left Ventricular Assist System in 7 Cases: Analysis of Potential Mechanisms Using Contrast-Enhanced Multislice Computed Tomography

Objectives: Twist of the outflow graft (OG) of the HeartMate 3 left ventricular (LV) assist device (HM3) with subsequent occlusion is a rare (incidence app. 0.7%), but life-threatening complication. We analyzed implantation technique dependent mechanisms which may potentially cause the OG twist, using morphological characteristics obtained with multislice computed tomography (MSCT) and echocardiography.

Methods: We retrospectively analyzed clinical, echocardiographic and MSCT data of 7 patients with angiographically proven and surgically corrected OG twist and of 11 consecutive patients without any type of pump obstruction, in whom MSCT was performed for other reasons at least 3 months after implantation. MSCT parameters were (1) position of inflow cannula related to anatomical LV apex, (2) angulation of axis of inflow cannula related to LV axis, (3) orientation of outflow channel of the pump housing related to LV axis, (4) OG course. Echocardiographic parameters were LV end-diastolic diameter (LVEDD) and aortic valve opening.

Results: Baseline characteristics of both groups, including age, sex, initial diagnosis and INTERMAX level were similar. The pump was implanted in the same technique with OG anastomosis to the ascending aorta in all patients of both groups. Percentage of patients with concomitant surgery was 43 and 36%, respectively. There was one case of redo surgery in the twist group. Mean time from implantation to twist was 563 ± 161 days. Preoperative LVEDD (64.6 ± 9.8 mm vs. 67.4 ± 10.2 mm, p = 0.563) was similar. LVEDD reduction after implantation (20 vs. 28%) and prevalence of aortic valve opening was higher in the twist group (71 vs. 45%). The MSCT measurements showed a high degree of heterogeneity in both groups: position of inflow cannula related to anatomical LV apex differed from superior to inferolateral; angulation of the axis of inflow cannula related to LV axis varied from cranial (14 vs. 32 degrees) to caudal (37 vs. 36 degrees), lateral (28 vs. 0 degrees) and medial (20 vs. 37 degrees); orientation of the outflow channel of the pump housing related to LV axis also varied from antero-septal to inferior. The OG course was variable, but similar in both groups with incidence of nonobstructive kinking of 29% and 36%, respectively.

Conclusions: Marked heterogeneity of the measured MSCT parameters was observed in both groups. No specific pattern or geometric relation of the apex, pump, outflow graft or chest could be attributed to the OG twist phenomenon.