BackgroundIt is infeasible to occlude a doubly committed juxtaarterial ventricular septal defect (DCVSD) percutaneously. The previous perventricular device closure technique was performed through an inferior median sternotomy approach. The purpose of this study is to evaluate the feasibility, safety and efficacy of perventricular device closure of DCVSDs through a left parasternal approach.MethodsSixty-two patients, with the DCVSD of less than 6 mm in diameter, were enrolled in this study. The pericardial space was approached through a left parasternal mini-incision without entering into the pleural space. Two parallel pursestring sutures were placed on the right ventricular outflow tract for puncture. Under transesophageal echocardiographic guidance, a new delivery sheath loaded with the device was inserted into the right ventricle and advanced through the defect into the left ventricle. The device, connected with a device stay suture, was deployed subsequently.ResultsSuccessful device closure of the defects was achieved in 58/62 patients (94 %). The DCVSD failed to close in 4 (6 %) patients due to device-related aortic regurgitation and device migration. The mean DCVSD diameter was 3.4 ± 1.0 mm (range, 2.0 to 6.0 mm). The implanted device size was 5.2 ± 1.3 mm (range, 4 to 8 mm). Forty-four out of 58 patients (76 %) was implanted with an eccentric occluder. The mean intracardiac manipulation time was 14 ± 13 min (range, 2 to 60 min). The procedure time was 66 ± 15 min (range, 42 to 98 min). During the follow-up period of 180 to 1860 (median 880) days, new mild pulmonary regurgitation occurred in 2 patients. No other device-related complications were found. The complete closure rate was 95 % at discharge, 98 % at 1-, 6- and 12-month, 96 % at 2-year, and 100 % at 3-year follow-up.ConclusionsPerventricular device closure of a DCVSD through a left parasternal approach is feasible, safe, and efficacious in selected patients. This minimally invasive technique permits easy defect crossing and accurate device positioning.