A systematic review aggregated data and individual participant data meta-analysis of percutaneous endovascular arteriovenous fistula

Objective: To assess the short-term and mid-term safety and efficacy of percutaneous endovascular arteriovenous fistula (pAVF) creation.

Methods: A systematic search was implemented corresponding to the PRISMA 2020 and the PRISMA for individual participant data (IPD) systematic reviews 2015. Aggregated data from the included studies were obtained and meta-analyzed regarding both the overall pAVF efficacy and the comparison of pAVF with surgical AVF (sAVF). We performed a two-stage IPD meta-analysis for studies comparing pAVF and sAVF regarding primary and secondary patency. Primary end points included primary patency, secondary patency, and functional cannulation.

Results: Eighteen studies with 1863 patients were included. The overall pAVF, primary patency, secondary patency, functional cannulation and abandonment rates were 54.01% (95% confidence interval [CI], 40.69-66.79), 87.27% (95% CI, 81.53-91.42), 79.94% (95% CI, 65.94-89.13), and 15.58% (95% CI, 7.77-28.79), respectively. The overall pAVF, technical success, maturation, reintervention per person-years and mean time to maturation rates were 97.08% (95% CI, 95.66-98.04), 82.13% (95% CI, 71.64-89.32), 0.80 (95% CI, 0.34-1.47), and 58 days (95% CI, 36.64-92.82), respectively. Secondary patency and pAVF abandonment rates where the only end points were WavelinQ and Ellipsys displayed statistically significant differences of 81.36% (95% CI, 76.15-85.65) versus 92.12% (95% CI, 87.94-94.93) and 32.54% (95% CI, 22.23-44.87) versus 11.13% (95% CI, 4.82-23.65). An IPD meta-analysis of hazard ratios for primary and secondary patency between pAVF and sAVF were 1.27 (95% CI, 0.61-2.67) and 1.25 (95% CI, 0.87-1.80), favoring sAVF. Statistically significant difference between pAVF and sAVF were solely depicted for steal syndrome relative risk of 5.91 (95% CI, 1.12-31.12) and wound infections relative risk of 4.19 (95% CI, 1.04-16.88). Plotting of pAVF smoothed hazard estimate displayed an upsurge in the probability of primary patency failure at 1 month after the intervention.

Conclusions: Although we failed to identify statistically significant differences between pAVF and sAVF regarding any of the primary end points, pAVF displayed a decreased risk for steal syndrome and wound infection. Although both the Ellipsys and WavelinQ devices displayed satisfactory secondary patency rates, Ellipsys demonstrated a statistically significant improved rate compared with WavelinQ. Additionally, and despite the borderline statistically insignificant inferior reintervention rate displayed by WavelinQ, one in three WavelinQ pAVFs resulted in abandonment. The introduction of pAVF as a treatment modality calls for standardized definition adjustment and improvement.