Postembolization Persistence of Pulmonary Arteriovenous Malformations: A Retrospective Comparison of Coils and Amplatzer and Micro Vascular Plugs Using Propensity Score Weighting

BACKGROUND. Endovascular embolization of pulmonary arteriovenous malformations (PAVMs) was historically performed with embolic coils. The Amplatzer Vascular Plug device (AVP) was introduced for this purpose in 2007 and the Micro Vascular Plug device (MVP) in 2013. OBJECTIVE. The purpose of this study was to compare coils, AVPs, and MVPs in terms of risk of persistence after PAVM embolization by use of propensity score weighting to account for biases in device selection. METHODS. This retrospective study included 112 patients (78 women and girls, 34 men and boys; mean age, 45 years) who underwent embolization of 393 PAVMs with a single device type (coil, MVP, or AVP) from January 2003 to January 2020. Persistence was defined as less than 70% reduction in PAVM sac size or contrast enhancement of the sac on follow-up pulmonary CTA. A Cox proportional hazards regression model was used to assess associations between embolic device selection and PAVM persistence. Inverse propensity score weighting was used to account for differences in embolic device selection based on patient and PAVM characteristics. RESULTS. The median postembolization follow-up period was 1.5 years (IQR, 0.3–5.6 years). Persistence was found in 10% (41/393) of PAVMs, including 16% (34/207) of those treated with coils, 8% (7/88) of those treated with AVPs, and 0% (0/98) of those treated with MVPs. Variables associated with embolization device (p < .25) were age, sex, pediatric versus adult status, smoking status, PAVM complexity, PAVM laterality, number of feeding arteries, and feeding artery diameter. The Cox regression model incorporated inverse propensity score weighting to account for the differences between treatment groups in these variables and incorporated feeding artery diameter because of imbalance remaining after weighting. With coils as the referent, MVPs had a hazard ratio for persistence of less than 0.01 (95% CI, < 0.01 to < 0.01; p < .001), and AVPs had a hazard ratio of 0.37 (95% CI, 0.16–0.90; p = .03). CONCLUSION. The risk of persistence after PAVM embolization was significantly lower for MVPs alone than for coils or AVPs alone. In addition, the risk of persistence was lower for AVPs than for coils. CLINICAL IMPACT. The findings support the clinical use of MVPs as the preferred device for PAVM embolization over coils and polytetrafluoroethylene-covered plugs.