Abstract OBJECTIVES To compare device success and paravalvular leak rates of 3 new-generation transcatheter aortic valve replacement devices in patients with bicuspid aortic valve stenosis and to test their biomechanical performance in a computer-based simulation model of aortic root with increasing ellipticity. METHODS This retrospective multicentre study included 56 bicuspid aortic valve patients undergoing transcatheter aortic valve replacement with new-generation devices: Lotus/Lotus Edge (N = 15; 27%), Evolut-R (N = 20; 36%) and ACURATE neo (N = 21; 37%). Three virtual simulation models of aortic root with increasing index of eccentricity (0–0.25–0.5) were implemented. Stress distribution, stent–root contact area and paravalvular orifice area were computed. RESULTS Device success was achieved in 43/56 patients (77%) with comparable rates among Lotus (87%), Evolut-R (60%) and ACURATE neo (86%; P = 0.085). Moderate paravalvular leak rate was significantly lower in the Lotus group as compared to Evolut-R group (0% vs 30%; P = 0.027) and comparable to the ACURATE neo group (0% vs 10%; P = 0.33). By index of eccentricity = 0.5, Lotus showed a uniform and symmetric pattern of stress distribution with absent paravalvular orifice area, ACURATE neo showed a mild asymmetry with small paravalvular orifice area (1.1 mm2), whereas a severely asymmetric pattern was evident with Evolut-R, resulting in a large paravalvular orifice area (12.0 mm2). CONCLUSIONS Transcatheter aortic valve replacement in bicuspid aortic valve patients with new-generation devices showed comparable device success rates. Lotus showed moderate paravalvular leak rate comparable to that of ACURATE neo and significantly lower than Evolut-R. On simulation, Lotus and ACURATE neo showed optimal adaptability to elliptic anatomies as compared to Evolut-R.