Results of search
According to the previous search strategy, 617 studies were obtained from the four databases and 2 studies were found by searching the reference lists and reviewed articles. After removing the duplicates, 416 publications remained in total and 401 records were excluded by browsing title and abstract. Among the remaining 15 records, 4 citations were removed according to our inclusion and exclusion criteria. Finally, 11 full-text studies (12 trials, a study consisted of 2 parallel randomized controlled trials) were suitable for this meta-analysis. (Fig. 1). The characteristics, type of haemorrhageand intervention methods of included studies were summarized in Table 1.
Risk of bias within studies
Bias risk of twelve trials was assessed in Fig.2. Two experiments did not provide a detailed method of random processes. The blinding process was at high risk of bias in two studies and unclear risk of bias in one study. No study had unclear or incomplete descriptions of their outcome data. Five trials did not provide a satisfactory description of reporting bias and nine studies did not indicate the other bias in the article.
Analysis of Primary outcomes
A total of 2,040 patients from nine studies were included in the analysis of 30-day mortality in both rFVIIa and control groups. The results showed that rFVIIa did not increase the incidence of death events from a meta-analysis [220/1217 vs 139/823, RR=1.00 (0.82-1.21), P for effect= 0.98]. No heterogeneity was revealed in the results.[P for heterogeneity =0.83, I2 = 0%] (Fig. 3a).
Total thromboembolic events
Total thromboembolic events occurred in 279 of 1786 patients in the experimental group and 143 of 1,083 in the control group. rFVIIa did not increase the incidence of total thromboembolic events overall from meta-analysis. [RR=1.13 (0.94-1.36), P for effect= 0.20]. No heterogeneity was revealed in the result. [P for heterogeneity =0.54, I2 = 0%] (Fig. 3b).
Six trials reported the volume of RBC transfusions, with a total of 907 participants. The results showed that there was no significant difference between the two groups [MD = -168.12 (-381.84–45.60), P for effect=0.12]. Heterogeneity was revealed in the result. [P for heterogeneity ＜0.01，I2 = 76%].
Considering the significant heterogeneity of the experimental results, we performed subgroup analysis to compare the volume of red cells transfusion between the low-dose (＜200ug/kg ) and high-dose (＞200ug/kg) groups. In the subgroup of patients with low-dose, rFVIIa did not significantly reduce RBC transfusion compared with the control group. [MD = -106.72(-448.33–234.89), P for effect =0.54, P for heterogeneity< 0.0001, I2 = 90%]. In the subgroup of patients with high-dose, rFVIIa significantly reduced RBC transfusion. [MD=-232.34(-410.31–-54.37), P for effect =0.01, P for heterogeneity=0.77, I2 = 0%]. (Fig.3c).
Analysis of Second outcomes
Arterial thrombotic events
Compared with the control group, rFVIIa significantly increased the incidence of arterial thrombotic events, and there was no significant heterogeneity. [RR = 1.38(1.08–1.77), P for effect =0.01, P for heterogeneity=0.87, I2 = 0%]. (Fig.4a)
In total, 8 studies with 2295 participants reported the incidence of myocardial infarctions after using rFVIIa and placebo. The overall analysis showed no increased risk of myocardial infarction [86/1508 vs 30/787, RR =1.37 (0.92–2.05), P for effect = 0.13, P for heterogeneity =0.43, I2 = 0%]. (Fig.4b)
Deep vein thrombosis
Seven studies reported the incidence of deep vein thrombosis in this meta-analysis. The RR for deep vein thrombosis (DVT) with the use of rFVIIa versus placebo was 0.83 (95% CI 0.51–1.33) which suggested that rFVIIa would not increase the incidence of DVT. No heterogeneity was revealed in the result. [P for heterogeneity =0.59, I2 = 0%] (Fig. 4c).
Number of patients transfused RBC
Number of patients transfused RBC in the rFVIIa group and the control group was 414/497 and 378/433, respectively. rFVIIa did not reduce the number of patients transfused RBC overall from meta-analysis. [RR =0.94 (0.83–1.08), P for effect = 0.39, P for heterogeneity =0.002, I2 = 77%]. (Fig.5a).
Four trails with a total of 627 participants reported the ICU-stay days. Overall, rFVIIa did not significantly reduce length of ICU-stay. [RR = 0.40(-1.28–2.07), P for effect=0.64, P for heterogeneity=0.77, I2 = 0%] (Fig. 5b).
Publication Bias and Sensitivity Analysis
Funnel plots showed that there was asymmetric distribution of included studies. Egger tests and trims and fill method demonstrated that there was no publication bias in all the studies. Sensitivity analyses have confirmed that all results are robust. All the results were shown in the Supplementary Figures.