Subgroup analyses
Subgroup analyses were performed using the stent length (≤10 cm versus >10 cm), the stent type (Thoraflex®Hybrid, E-vita®, Frozenix®, and Cronus®), male predominance (≤74% versus >74% male), and the chronicity of the operation (elective versus non-elective). The categorization by the stent length led to pooled estimates associated with a lower heterogeneity regarding reoperation for bleeding among those implanted longer device (6% [95% CI 2 to 10]; I2 = 79.36% in ≤10 cm versus 8% [95% CI 5 to 11]; I2 = 65.74% in >10 cm); however, the test of group differences for reoperation for bleeding was non-significant (p=0.45; Figure 4). The subgroup by the stent type showed substantial decrease in the heterogeneity (I2 = 0.01% for Thoraflex Hybrid, I2 = 53.95% for Evita, I2 = 0.01% for Frozenix, and I2 = 54.41% for Cronus; Figure 5) and a significant difference between groups (p = 0.001). However, no effect on heterogeneity was observed in subgroups by the male predominance (Figure 6). The subgroup by the chronicity of operation also resulted in less heterogeneity among patients undergoing elective compared to non-elective operations (8% [95% CI 3 to 13]; I2 = 29.22% in elective versus 6% [95% CI 3 to 9]; I2 = 80.56% in non-elective; Figure 7). The effect of subgroup analyses on the pooled rate of in-hospital mortality was more pronounced. The heterogeneities improved among patients implanted stent >10 cm compared to those implanted stent ≤10 cm (I2 = 0.06% versus I2 = 81.90%; Supplementary Figure 2-A). The categorization by the stent type led to substantial improvement of heterogeneity (I2 = 0% for Thoraflex Hybrid, I2 = 0.06% for Frozenix, I2 = 0.01% for Evita and I2 = 42.48% for Cronus®, Supplementary Figure 2-B). However, the subgroups by the male predominance did not significantly decrease the heterogeneity (Supplementary Figure 2-C). Moreover, it disappeared among patients undergoing elective surgeries compared to those undergoing non-elective surgeries (I2 = 0% versus I2 = 66.18%; Supplementary Figure 2-D). When excluding two of five studies with a wide 95% CI for the number of patients with postoperative bleeding,31,44the heterogeneity was significantly improved 11% (95% CI 8 to 15; I2 = 12.97%).