Ph: pancreas head, Pb: pancreas body, Pt: pancreas tail, SMA: superior mesenteric artery, SPA: spleen artery, GDA: gastroduodenal artery, RHA: right hepatic artery, PV: portal vein, SMV: superior mesenteric vein, SPV: spleen vein
The assessed blood vessels were the portal vein (PV) in 11 sites, superior mesenteric vein (SMV) in 7, splenic vein (SPV) in 17, superior mesenteric artery (SMA) in 1, splenic artery (SPA) in 10, gastroduodenal artery (GDA) in 1, and right hepatic artery (RHA) in 1. Among them, pathological vascular infiltration was positive in 5 cases of PV, 10 cases of SMV, 7 cases of SPV, 0 cases of SMA, 2 cases of SPA, 1 case of GDA, 1 case of RHA (Table 1). Dynamic CT was performed in all 58 subjects. Vascular invasion was assessed by EUS-B mode and EUS-EG in 58 and 36 subjects, respectively. Of subjects in whom vascular invasion was not assessed in EUS-EG, 6 were diagnosed with no vascular invasion because the vessels were distant from the tumor and 16 did not undergo EUS-EG.
The interobserver agreements of EUS B-mode and EUS-EG findings for vascular invasion were moderated (κ = 0.596, P < 0.0001) and substantial (κ = 0.634, P < 0.0001).
The sensitivity, specificity, PPV, NPV, and accuracy (95% CI)) for vascular invasion in the 58 subjects (62 sites) were 0.733 ( 0.519–0.881), 0.787 (0.719–0.834), 0.524 (0.371–0.629), 0.902 (0.824–0.956), and 0.774 (0.671–0.845) on dynamic CT (62 sites); 0.733 (0.519–0.881), 0.787 (0.719–0.881), 0.524 (0.371–0.629), 0.902 (0.824–0.956), and 0.774 (0.671–0.845) in EUS B-mode (62 sites); and 0.917 (0.712–0.984), 0.889 (0.798–0.919), 0.786 (0.610–0.844), 0.960 (0.862–0.992), and 0.897 (0.771–0.939) in EUS-EG (39 sites), respectively (Table 2).
Table 2
Overall results (95% confidence interval).
| Dynamic CT (62 sites) | EUS B-mode (62 sites) | EUS-EG (39 sites) |
sensitivity | 0.733 (0.519–0.881) | 0.733 (0.519–0.881) | 0.917 (0.712–0.984) |
specificity | 0.787 (0.719–0.834) | 0.787 (0.719–0.881) | 0.889 (0.798–0.919) |
PPV | 0.524 (0.371–0.629) | 0.524 (0.371–0.629) | 0.786 (0.610–0.844) |
NPV | 0.902 (0.824–0.956) | 0.902 (0.824–0.956) | 0.960 (0.862–0.992) |
accuracy | 0.774 (0.671–0.845) | 0.774 (0.671–0.845) | 0.897 (0.771–0.939) |
CT: computed tomography, EUS: endoscopic ultrasonography, EUS-EG: endoscopic ultrasonography elastography, PPV: positive predictive value, NPV: negative predictive value |
In the 27 subjects (29 sites) in difficult diagnosis sites group, the sensitivity, specificity, PPV, NPV, and accuracy for vascular invasion were 0.556 (0.303–0.772), 0.750 (0.636–0.847), 0.500 (0.273–0.694), 0.789 (0.670–0.892), and 0.690 (0.533–0.824) on dynamic CT; 0.667 (0.400-0.863), 0.700 (0.580–0.788), 0.500 (0.300-0.647), 0.824 (0.682–0.927), and 0.690 (0.524–0.811) in EUS B-mode; and 0.889 (0.635–0.979), 0.850 (0.736–0.890), 0.727 (0.520–0.801), 0.944 (0.818–0.989), and 0.862 (0.705–0.918) in EUS-EG, respectively. These results show that EUS-EG had the best diagnostic performance (Table 3). The diagnoses of all three modalities matched in 12 sites and the diagnostic accuracy was 100% (12/12) in this situation. Meanwhile, the diagnoses of either two out of three modalities (CT and EUS B-mode, CT and EUS-EG, or EUS B-mode and EUS-EG) matched in 5 sites, 7 sites, and 5 sites, and the diagnostic accuracy for each situation was 40% (2/5), 85.7% (6/7), and 80% (4/5), respectively.
Table 3
Results in group of difficult diagnosis sites (29 sites) (95% confidence interval).
| dynamic CT | EUS B-mode | EUS-EG |
sensitivity | 0.556 (0.303–0.772) | 0.667 (0.400-0.863) | 0.889 (0.635–0.979) |
specificity | 0.750 (0.636–0.847) | 0.700 (0.580–0.788) | 0.850 (0.736–0.890) |
PPV | 0.500 (0.273–0.694) | 0.500 (0.300-0.647) | 0.727 (0.520–0.801) |
NPV | 0.789 (0.670–0.892) | 0.824 (0.682–0.927) | 0.944 (0.818–0.989) |
accuracy | 0.690 (0.533–0.824) | 0.690 (0.524–0.811) | 0.862 (0.705–0.918) |
CT: computed tomography, EUS: endoscopic ultrasonography, EUS-EG: endoscopic ultrasonography elastography, PPV: positive predictive value, NPV: negative predictive value |
The 29 sites were divided into arterial (8 sites) and portal groups (21 sites). In the arterial group, the sensitivity, specificity, PPV, NPV, and accuracy for vascular invasion were 1.000 (0.418-1.000), 0.500 (0.306-0.500), 0.400 (0.167-0.400), 1.000 (0.612-1.000), and 0.625 (0.334–0.625) on dynamic CT; 1.000 (0.413-1.000), 0.833 (0.638–0.833), 0.667 (0.275–0.667), 1.000 (0.765-1.000), and 0.875 (0.581–0.875) in EUS B-mode; and 1.000 (0.413-1.000), 0.833 (0.638–0.833), 0.667 (0.275–0.667), 1.000 (0.765-1.000), and 0.875 (0.581–0.875) in EUS-EG, respectively. In the portal group, these results were 0.429 (0185-0.619), 0.857 (0.735–0.952), 0.600 (0.258–0.866), 0.750 (0.643–0.833), and 0.714 (0.552–0.841) on dynamic CT; 0.571 (0.290–0.806), 0.714 (0.573–0.832), 0.500 (0.253–0.705), 0.769 (0.617–0.896), and 0.667 (0.479–0.823) in EUS B-mode; and 0.857 (0.567–0.971), 0.857 (0.712–0.914), 0.750 (0.496–0.850), 0.923 (0.767–0.985), and 0.857 (0.664–0.933) in EUS-EG, respectively (Table 4).
Table 4
Results in group of difficult diagnosis sites arterial system (8 sites) and portal vein (21 sites).
| | Dynamic CT | EUS B-mode | EUS-EG |
sensitivity | artery | 1.000 (0.418-1.000) | 1.000 (0.413-1.000) | 1.000 (0.413-1.000) |
portal vein | 0.429 (0.185–0.619) | 0.571 (0.290–0.806) | 0.857 (0.567–0.971) |
specificity | artery | 0.500 (0.306-0.500) | 0.833 (0.638–0.833) | 0.833 (0.638–0.833) |
portal vein | 0.857 (0.735–0.952) | 0.714 (0.573–0.832) | 0.857 (0.712–0.914) |
PPV | artery | 0.400 (0.167-0.400) | 0.667 (0.275–0.667) | 0.667 (0.275–0.667) |
portal vein | 0.600 (0.258–0.866) | 0.500 (0.253–0.705) | 0.750 (0.496–0.850) |
NPV | artery | 1.000 (0.612-1.000) | 1.000 (0.765-1.000) | 1.000 (0.765-1.000) |
portal vein | 0.750 (0.643–0.833) | 0.769 (0.617–0.896) | 0.923 (0.767–0.985) |
accuracy | artery | 0.625 (0.334–0.625) | 0.875 (0.581–0.875) | 0.875 (0.581–0.875) |
portal vein | 0.714 (0.552–0.841) | 0.667 (0.479–0.823) | 0.857 (0.664–0.933) |
CT: computed tomography, EUS: endoscopic ultrasonography, EUS-EG: endoscopic ultrasonography elastography, PPV: positive predictive value, NPV: negative predictive value |