Selection and summary of studies
We first searched in various databases and obtained 95 articles. Specifically, there were 15 articles in PubMed, 26 articles in Embase, 15 trials in Cochrane Library, 23 articles in Web of Science, 13 articles in China National Knowledge Infrastructure, 1 article in China Science and Technology Journal Database, and 2 articles in Wanfang Database. Simultaneously, we added 21 additional documents by using methods such as document tracing. Then we removed 78 duplicate documents, 18 documents whose content or document type did not meet the requirements, and 3 abstract articles without full-text details. Finally,17 studies (10, 14-29) were included in our meta analysis,including 8 RCTs and 9 cohort studies. The flow diagram is shown in Figure 1. A total of 51294 patients were enrolled in all studies, of which 25747 patients were enrolled in the experimental (Vonoprazan) group, and 25547 patients were enrolled in the control (PPIs) group. The literature and baseline information of the included patients are shown in Table 1.
Table 1
Characteristics of each included studies.
Author
|
Year
|
Country
|
Study design
|
Intervention
|
N(I)
|
Control
|
N(C)
|
follow up (W)
|
Hirai(14)
|
2018
|
Japan
|
RCT
|
VPZ 20mg
|
61
|
LPZ 30mg
|
66
|
4,8
|
Abe(15)
|
2021
|
Japan
|
Cohort Study
|
VPZ, standard dose and low dose
|
23989
|
PPIs,standard dose and low dose
|
24040
|
NA
|
Ban(16)
|
2020
|
Japan
|
RCT
|
VPZ 20mg
|
101
|
LPZ 30mg
|
95
|
1,2,4,8
|
Hamada(17)
|
2019
|
Japan
|
RCT
|
VPZ 20mg
|
69
|
LPZ 30mg
|
70
|
8
|
Horikawa(18)
|
2018
|
Japan
|
Cohort Study
|
VPZ 20mg
|
62
|
LPZ 30mg
|
53
|
2
|
Ichida(19)
|
2019
|
Japan
|
RCT
|
VPZ 20mg,
RBP 300mg
|
43
|
EPZ 20mg,
RBP 300mg
|
39
|
4,8
|
Ishida(20)
|
2021
|
Japan
|
Cohort Study
|
VPZ 20mg
|
205
|
PPIs,standard dose
|
205
|
6
|
Ishii(21)
|
2018
|
Japan
|
RCT
|
VPZ 20mg,
RBP 300mg
|
27
|
EPZ 20mg,
RBP 300mg
|
26
|
4,8
|
Kagawa(22)
|
2016
|
Japan
|
Cohort Study
|
VPZ 20mg
|
75
|
RPZ 20mg
|
150
|
5,8
|
Kakushima(23)
|
2019
|
Japan
|
Cohort Study
|
VPZ 20mg
|
59
|
EPZ 20mg
|
71
|
NA
|
Komori(24)
|
2019
|
Japan
|
RCT
|
VPZ 20mg
|
18
|
RPZ 10mg
|
15
|
4
|
Marouka(25)
|
2017
|
Japan
|
Cohort Study
|
VPZ 20mg,
RBP 300mg
|
31
|
EPZ 20mg,
RBP 300mg
|
31
|
4
|
Shiratori(26)
|
2021
|
Japan
|
Cohort Study
|
VPZ
|
627
|
PPIs
|
627
|
8
|
Takahashi(27)
|
2016
|
Japan
|
RCT
|
VPZ 20mg
|
14
|
LPZ 30mg
|
12
|
4
|
Tsuchiya(10)
|
2017
|
Japan
|
RCT
|
VPZ 20mg
|
39
|
EPZ 20mg
|
41
|
8
|
Yamamoto(28)
|
2020
|
Japan
|
Cohort Study
|
VPZ 20mg
|
50
|
EPZ 20mg
|
116
|
NA
|
Yamasaki(29)
|
2018
|
Japan
|
Cohort Study
|
VPZ 20mg
|
77
|
LPZ 30mg
|
90
|
4
|
RCT: randomized controlled experiment; N:number of patients; NA:unknown; VPZ:Vonoprazan; PPIs:proton pump inhibitors; LPZ:lansoprazole; RPZ:rabeprazole; EPZ:esomeprazole; RBP:rebamipide; the standard dose in Japan is 30, 10, 20 and 20 mg in lansoprazole, rabeprazole, esomeprazole,and omeprazole.
Study quality
The assessment of risk of quality showed that among the RCTs, one article reached level A in evidence quality (17), and the remaining 7 articles reached level B in evidence quality (10, 14, 16, 19, 21, 24, 27). The detailed description of the risk of bias in the studies was shown in Figure 2. Meanwhile, according to the NOS score, all 9 cohort studies were rated as high-quality (15, 18, 20, 22, 23, 25, 26, 28, 29), as shown in Figure 3.
Outcomes measurements
delayed bleeding
There are 14 literatures that reported delayed bleeding after gastric ESD (10, 14-17, 19, 20, 22-24, 26-29), while the other 3 literatures had a weight of 0 in this aspect (18, 21, 25). The heterogeneity of the various studies was relatively low (I2=21.1%, P=0.224), so the fixed effects model was used for analysis. As shown in the forest diagram in Figure 4A, there was a statistical difference in the overall delayed bleeding rate between the test group and the control group. The overall mean RR for delayed bleeding following the administration of Vonoprazan was 0.72 (P<0.05) with a 95% CI (0.67-0.77).
A subgroup analysis was carried out according to the type of study, as shown in Figure 4B. There was a certain degree of heterogeneity (I2=52.5%, P=0.049) among cohort studies. The delayed bleeding rate between the two groups was statistically different. The RR was 0.72(P<0.05) with a 95% CI (0.67-0.77), and the weight reached 99.17%. In contrast, there was basically no heterogeneity (I2=0.0%,P=0.76) in RCT articles. There was no statistically significant difference in the delayed bleeding rate between the Vonoprazan group and PPIs group [RR = 0.56, 95%CI (0.27, 1.15), P>0.05].
The sensitivity analysis was done by merging all the remaining articles every time an article was deleted, as shown in Figure 5A. Abe et al. was a source of heterogeneity (15), but the overall results were still robust.
A funnel plot analysis showed no publication bias (Figure 6A) and Egger’s regression test was non-significant indicating symmetry in the funnel plot (P=0.725).
blood transfusion
There are 3 literatures that reported blood transfusion (20, 23, 26). Basically there was no heterogeneity between the three articles (I2=0.0%, P=0.566), so the fixed effects model was used for analysis. There was no statistically difference in the incidence of blood transfusion events between these two groups (Figure 7). The overall mean RR for blood transfusion following the administration of Vonoprazan was 0.76 (P>0.05) with a 95% CI (0.41-1.40). The sensitivity analysis showed the overall results were robust (Figure 5B). The funnel plot analysis showed no publication bias (Figure 6B). Egger's regression test was not significant (P=0.749), also suggesting no publication bias.
shrinkage ratio
There were 7 articles (16, 18, 19, 21, 22, 25, 27) reporting the shrinkage ratio of artificial ulcers whose data can be extracted for analysis. Because of the high heterogeneity (I2=78.5%, P<0.05), the random effects model was used. There was a statistical difference in the shrinkage ratio between the Vonoprazan group and the PPIs group. The weighted mean difference was 2.68(P<0.05) with a 95% CI (0.44-4.91). Certainly, the subgroup analysis was carried out according to the postoperative recovery time. The subgroups at 1, 4, and 8 weeks after ESD had lower heterogeneity. However, the heterogeneity within the subgroup at 2 weeks after the operation was still high (I2=92.1%, P<0.05). The detailed subgroup and total forest maps were shown in Figure 8. Therefore, we further conducted sensitivity analysis, finding the research of Horikawa et al. was the source of heterogeneity (18), but the overall results were still robust (Figure 5C). The funnel plot analysis showed no publication bias (Figure 6C) and Egger’s regression test was non-significant indicating symmetry in the funnel plot (P=0.0725).