An overview of the study selection process is shown in Fig. 1. Our systematic literature review identified 1828 articles, of which 61 were full texts. Finally, 15 studies with 8550 patients were included. Table 1 summarizes the key characteristics of these studies. Overall, most were from South Korea; two were from China, and one was from the United States. In the adult trials, the mean age of the participants ranged from 60–67.6. Two of these studies[7, 8] include patients who underwent ME-NBI after white light endoscopy. Table 2 shows the quality of these studies using the NOS.
Table 1
Characteristics of studies included
Author
|
Publication Time
|
Country
|
Case Number
|
Mean Age
|
Sex (Male/Female)
|
Lesion Size
|
J. W. Jeon, et al
|
2020
|
Korea
|
1774
|
63.89 ± 9.36
|
1493/657
|
13.94 ± 11.28
|
Yang, L., et al.*
|
2018
|
China
|
104
|
|
72/32
|
12.8 ± 8.3
|
Kang, D. H., et al.
|
2018
|
Korea
|
1006
|
63.7 ± 9.1
|
707/299
|
12.7 ± 8.6
|
Ryu, D. G., et al.
|
2017
|
Korea
|
1410
|
63.6
|
1104/337
|
12.7 ± 8.6
|
Wu, S., et al. **
|
2017
|
China
|
105
|
63.86 ± 9.34
|
45/33
|
|
Hwang, J. W., et al.
|
2017
|
Korea
|
266
|
65.21 ± 8.55
|
165/101
|
14.31 ± 7.59
|
Kim, M. K., et al.
|
2014
|
Korea
|
285
|
67.6 ± 8.5(case)
63.1 ± 8.8(control)
|
34/12(case)
167/72(control)
|
20.00 ± 18(case)
12 ± 9(control)
|
Choi, C. W., et al.
|
2014
|
Korea
|
413
|
62 ± 9.27
|
151/67
|
12.2 ± 8
|
Cho, S. J., et al.**
|
2010
|
Korea
|
228
|
64.6 ± 8.6(case)
60.0 ± 9.2(control)
|
56/24(case)
118/38(control)
|
|
Kim, Y. J., et al. **
|
2010
|
Korea
|
258
|
62.9 ± 8.7
|
181/92
|
|
Lim, H., et al.
|
2014
|
Korea
|
1845
|
61.7(case)
61.5(control)
|
958/295(case)
347/93
|
2.5
1.9
|
Tsuji, Y., et al.
|
2012
|
Japan
|
137
|
67.6 ± 10.4(case)
67.7 ± 10.3(control)
|
48/16 (case)
53/20(control)
|
18.5 ± 11.4(case)
13.8 ± 7.1(control)
|
Won, C. S., et al.**
|
2011
|
Korea
|
241
|
63.9 ± 9.7 (case)
61.8 ± 10.1(control)
|
107/39 (case)
68/27(control)
|
|
Kim, H.-S., et al.***
|
2006
|
Korea
|
54
|
|
|
|
Joo, J. S., et al.
|
2021
|
Korea
|
726
|
63.42 ± 8.76
|
493/233
|
1.87 ± 0.77
|
* Mean age is not available. |
** Average lesion size is not available |
***Mean age, gender compositions, and lesion sizes were not available |
Table 2
New castle-ottawa Quality assessment Scale(NOS).
Author
|
Publication time
|
NOS
|
Selection
|
Comparability
|
Exposure
|
Score
|
J. W. Jeon, et al
|
2020
|
☆☆☆
|
☆☆
|
☆☆
|
7
|
Yang, L., et al.
|
2018
|
☆☆
|
☆☆
|
☆☆
|
6
|
Kang, D. H., et al.
|
2018
|
☆☆☆
|
☆☆
|
☆☆
|
7
|
Ryu, D. G., et al.
|
2017
|
☆☆☆
|
☆☆
|
☆☆
|
6
|
Wu, S., et al.
|
2017
|
☆☆
|
|
☆☆
|
4
|
Hwang, J. W., et al.
|
2017
|
☆☆
|
☆☆
|
☆☆
|
6
|
Kim, M. K., et al.
|
2014
|
☆☆☆
|
|
☆☆
|
5
|
Choi, C. W., et al.
|
2014
|
☆☆☆
|
☆☆
|
☆☆
|
7
|
Cho, S. J., et al.
|
2010
|
☆☆☆
|
☆☆
|
☆☆
|
7
|
Kim, Y. J., et al.
|
2010
|
☆☆
|
☆☆
|
☆☆
|
6
|
Lim, H., et al.
|
2014
|
☆☆
|
☆☆
|
☆☆
|
6
|
Tsuji, Y., et al.
|
2012
|
☆☆☆
|
☆☆
|
☆☆
|
7
|
Won, C. S., et al.
|
2011
|
☆☆
|
☆☆
|
☆☆
|
6
|
Kim, H.-S., et al.
|
2006
|
☆☆
|
|
☆☆
|
4
|
Joo, J. S., et al.
|
2021
|
☆☆☆
|
☆☆
|
☆☆
|
7
|
Erythema
The association between erythema and the LGIN upgrade diagnostic risk was assessed in 12 articles[3,5,7−15]. Most of these studies were conducted in Korea. Figure 2 summarizes the pooled results for studies comparing erythema and LGIN upgrade diagnosis risk, revealing a significantly increased risk, OR 2.87 (95% CI,1.94–4.25) with high heterogeneity (I2 = 74%), with no publication bias (p = 0.2204). Sensitivity analysis excluding individual studies and meta-regression analysis did not markedly alter these associations. (Supplementary Fig. 1)
Lesion Diameters
The association between lesion diameter and the risk of upgraded diagnosis was assessed in 10 studies[5, 7, 9, 10, 12–14, 16, 17], the majority of which were conducted in Korea and one in China. A significantly increased upgrade risk was observed in larger lesions, OR 2.50 (95%CI [1.85, 3.371]), with moderate heterogeneity (I2 = 54%) (Fig. 3) with no publication bias (p = 0.8092). A significant increase in the risk when a lesion diameter > 20 mm was observed in two studies[7, 13]. Three studies showed risk increases in lesions with diameters > 15 mm[9, 13, 14], and four studies showed a risk increase in lesions > 10 mm. Subgroup analysis (Fig. 4) showed strong heterogeneity among studies, with a 1 cm diameter as the cut-off value. Relatively low heterogeneity was observed among studies with 1.5 cm diameter and 2 cm diameter as the cut-off values. Sensitivity analysis yielded results similar to those of the main analysis. (Supplementary Fig. 2).
Depressed lesions
The association between gross lesion type and the risk of upgraded diagnosis was assessed in 14 studies. A significantly increased upgrade risk was observed in depressed lesions (OR, 1.61 [1.00; 2.59]), with substantial heterogeneity (I2 = 0.73) (Fig. 5). Sensitivity analysis showed that after excluding Tsuji, Y., et al. study, I2 decreased to 65.3%, with an OR of 1.87 (95%CI [1.25, 2.78]). (Supplementary Fig. 3).
Nodularity
Six studies investigated the risk of diagnosis upgrade with regard to the presence of nodularity[5, 10, 12, 14, 15, 18], of which two studies were published after 2015. Figure 6 illustrates the significantly increased risk of an upgraded diagnosis in the presence of nodularity. High heterogeneity was observed between the studies, and sensitivity analysis did not markedly alter the associations. Based on the results of the meta-regression, a subgroup analysis was carried out by year of publication. The results showed that the publication year was a source of heterogeneity. Publication bias was not considered because the number of studies was relatively small. The results were consistent across the sensitivity analyses. (Supplementary Fig. 4)
Ulceration
The association between ulceration and LGIN upgrade diagnosis risk was assessed in six articles[5, 8, 10, 11, 15, 16, 18]. The pooled OR between ulceration and LGIN upgrade diagnosis was 1(95%CI). Moderate heterogeneity was observed among studies. (Fig. 7) Results were consistent across sensitivity analyses. (Supplementary Fig. 5)
Helicobacter pylori infection
The risk of Hp infection was assessed in eight articles[3,5,7,11,13−16]. Pooled OR was 1.00 (95% CI [0.80, 1.241] with moderate heterogeneity observed. (Fig. 8). When the study conducted by Hwang et al. was excluded, the heterogeneity decreased remarkably. (Supplementary Fig. 6)
Location
The association between lesion location and the risk of upgraded diagnosis was assessed in six articles. Figure 9 summarizes the pooled results of studies comparing the tumors located in the upper 1/3 and middle 1/3 of the stomach versus tumors located in the lower 1/3, revealing a significantly decreased risk of tumors located in the middle 1/3 compared to the lower 1/3, with moderate heterogeneity. (Fig. 9)