Abstract： Objective To determine whether the adjustment of the pusher of GTF was useful to decrease the degree of tilting of the femoral Ginther Tulip filter (GTF) in an in vitro caval model. Methods The caval model was constructed by placement of a 25 mm × 100 mm and two 10 mm ×200 mm Dacron graft inside a transparent bifurcate glass tube. The study consisted of two groups: left straight group (GLS) (n =100) and left curved group (GLC) (n = 100). In the GLC, a 10° to 20° angle was curved on the introducer.The distance (DCH) between the caval right wall and the hook was measured. The degree of tilting (DT) was classified into 5 grades and recorded. Before and after the GTF being released, the angle (ACM1, 2)between the axis of IVC and the metal mount, the distance (DCM1) between the caval right wall and the metal mount, the angle (ACF) between the axis of IVC and the axis of the filter and the diameter of IVC (DIVC) were measured. The data were analyzed with Chi-Square test, t test, rank sum test and Pearson correlation test. Results The degree of GTF tilting in each group revealed a divergent tendency. In group LC, the apex of the filter tended to be grade Ⅲ compared in group LS (x2 value 37. 491 ,P ＜0.01).The differences of most variables between GLS and GLC were considered as statistical significance(16. 60° vs.3.05°, 20.60° vs. 3.50°, -3.90°vs. -0.40°, 2.98 mm vs. 10.40 mm, - 10.95° vs. -0.485°,13. 17 mm vs. 10.06 mm, - 1.70° vs. 0.70°, t or Z values - 12. 187, - 12. 188, -8.545, -51.834,-11. 395,9. 562, -3. 596, P ＜ 0. 01). There exist significant positive correlation between ACM1 and ACF,ACM1 - ACM2 and DCH1 - DCH2 in each group, respectively(r values 0. 978,0. 344,0. 879,0. 627 ,P ＜0. 01),while significant negative associations are detected between DCH1 and ACF in each group , ACP and ACF in group LC(r values -0.974, -0.322, -0.702,P＜0.01). Conclusion The technique of adjusting the orientation of filter pusher had minimized the incidence and extent of GTF filter tilting in vitro.