Search Results
After the search in several international databases, we initially included 1,404 articles. We screened titles or abstracts, and 808 duplicates were excluded. Then, 536 articles - reviews, not for CSN5, not digestive system cancer, and not full-texts - were eliminated for meta-analysis. Besides, another 38 records were further excluded by screening the full texts, since they did not present sufficient data for analysis. Thus, the remaining 22 studies were utilized for further analysis. The selection process is described in Figure 1.
Study Characteristics
The principal features of the included researches are listed in Table 1. The studies analyzed were in the 2008-2020 publication range. The sum of patients in the included studies reached 2,193 with a range of 40–286. All studies included were performed in Asian countries, including 21 in China, and one in Japan. The Kaplan-Meier curves were adopted to calculate HRs and 95% CIs indirectly, due to an absence of HRs and 95% CIs in some articles.
Table 1
Characteristics of studies included in the meta-analysis
First author, year
|
nation
|
Cancer type
|
Case number(High/Low)
|
Cut-off value
|
Detection method
|
Outcome
|
Follow-up time
|
Liu C,2020(14)
|
China
|
CRC
|
189(92/97)
|
Positive cells: +
|
IHC
|
OS
|
>140months
|
Wang L,2020(16)
|
China
|
GC
|
90(55/35)
|
Score=8
|
IHC
|
OS
|
>70months
|
Zhou R,2018(17)
|
China
|
CRC
|
116(69/47)
|
Positive cells: +
|
IHC
|
OS
|
>125months
|
Shen Q,2020(18)
|
China
|
ESCC
|
124(65/59)
|
NA
|
IHC
|
OS
|
>60months
|
Pan YB,2017(19)
|
China
|
CRC
|
286(143/143)
|
NA
|
cDNA
|
RFS
|
192months
|
Mao LX,2019(20)
|
China
|
PC
|
106(70/36)
|
NA
|
IHC
|
NA
|
NA
|
Kugimiya N,2017(21)
|
Japan
|
CRC
|
50(17/33)
|
ROC
|
RT-PCR
|
RFS
|
>38months
|
Liu HL,2018(22)
|
China
|
HCC
|
102(73/29)
|
NA
|
IHC
|
OS
|
>80months
|
Wang Y,2014(23)
|
China
|
HCC
|
67(41/26)
|
Score=3
|
IHC
|
OS
|
60months
|
Hsu MC,2008(24)
|
China
|
HCC
|
99(37/62)
|
Staining color: T=N
|
IHC
|
NA
|
NA
|
Chen L,2010(15)
|
China
|
HCC
|
76(43/33)
|
Positive cells=69%
|
IHC
|
OS
|
60months
|
Wang F,2009(25)
|
China
|
ESCC
|
90(75/15)
|
Positive cells=10%
|
IHC
|
OS
|
60months
|
Zheng L,2016(26)
|
China
|
ESCC
|
187(122/65)
|
Positive cells=50%
|
IHC
|
OS
|
>45months
|
Guo ZQ,2014(27)
|
China
|
CRC
|
80(66/14)
|
Positive cells=30%
|
IHC
|
NA
|
NA
|
Yang F,2013(28)
|
China
|
GC
|
80(57/23)
|
Score=1
|
IHC
|
NA
|
NA
|
Zhang SW,2014(29)
|
China
|
CRC
|
94(81/13)
|
Score=1
|
IHC
|
NA
|
NA
|
Cao Y,2013(30)
|
China
|
HCC
|
40(28/12)
|
Positive cells=25%
|
IHC
|
NA
|
NA
|
Yang SH,2013(31)
|
China
|
CRC
|
74(60/74)
|
Score=1
|
IHC
|
OS
|
60months
|
Shi H,2010(32)
|
China
|
ESCC
|
60(47/13)
|
Positive cells=25%
|
IHC
|
NA
|
NA
|
Gu GJ,2017(33)
|
China
|
GBC
|
65(39/26)
|
Score=3
|
IHC
|
NA
|
NA
|
Zhang LY,2011(34)
|
China
|
ESCC
|
58(37/21)
|
Positive cells=25%
|
IHC
|
NA
|
NA
|
Li S,2012(35)
|
China
|
GC
|
60(43/17)
|
Score=1
|
IHC
|
OS
|
>60months
|
Survival Analysis
After a pooled analysis of 22 studies with 2,193 patients, a combined HR of 2.28 (95% CI: 1.71–3.03; p < 0.00001; Figure 2A) was acquired to verify the significant association between poor OS of digestive system carcinomas and high expression of CSN5. We detected non-significant heterogeneity (c2 = 0.36; freedom degrees = 11; p = 0.95; I2 = 0%). A fixed‐effect model was applied when the study presented low heterogeneity. The subgroup analysis of the relation between CSN5 expression and OS in tumor types indicated that high expression of CSN5 was correlated with poor OS in CRC (HR = 1.83, 95% CI: 1.05–3.19; p = 0.03; Figure 2B). Additionally, CSN5 overexpression was shown to be obviously related to poor OS in HCC (HR = 2.80, 95% CI: 1.76–4.45; p < 0.00001; Figure 2C). Moreover, a worse OS was discovered in ESCC patients with CSN5 high‐expression as well (HR = 2.52, 95% CI: 1.23–5.15; p = 0.01; Figure 2D). However, we did not detect a significant correlation between CSN5 expression and RFS (Figure S1).
Association of CSN5 Expression with Clinical Parameters
Correlation analysis outcome between clinicopathologic features and CSN5 level is presented in Table 2. A high CSN5 expression was found to be significantly associated with poorer invasion depth (OR = 0.49, 95% CI: 0.25-0.96, p = 0.04; Figure 3A), positive lymphatic metastasis (OR = 0.28, 95% CI: 0.16-0.47, p = 0.00001; Figure 3B), positive distant metastasis (OR = 0.32, 95% CI: 0.13-0.76, p = 0.01; Figure 3C) and poorer differentiation degree (OR = 0.34, 95% CI: 0.19-0.60, p = 0.0003; Figure 3D). However, the CSN5 level did not significantly correlate with age, gender, tumor stage, tumor size or vascular invasion (Table 2). In tumor-types subgroup analysis, patients with positive lymphatic metastasis in the groups of CRC (OR = 0.21, 95% CI: 0.07-0.66, p = 0.008), GC (OR = 0.28, 95% CI: 0.16-0.51, p < 0.0001) and ESCC (OR = 0.24, 95% CI: 0.12-0.48, p < 0.0001) presented a correlation with high CSN5 expression. A correlation was detected between the level of CSN5 and differentiation degree in GC (OR = 0.21, 95% CI: 0.08-0.53, p = 0.001) and CRC (OR = 0.39, 95% CI: 0.17-0.89, p = 0.03). However, the expression of CSN5 had no significant connection with invasion depth in CRC and ESCC, as well as differentiation degree in ESCC. Some tumor types were not available for analysis.
Publication bias and Sensitivity Analyses
To evaluate potential publication bias, Begg’s funnel plot and Egger’s test were conducted. We successively omitted one study at a time in the sensitivity analysis. The results of publication bias and sensitivity analysis within the included studies are demonstrated in Figures 4, S2, and S3. No significant publication bias was detected in OS analysis (Egger’s test: p = 0.112). Sensitivity analysis results indicated the robustness and reliability of our estimates since the pooled results for OS could not be significantly altered by only one trial. Publication bias results and sensitivity analysis in RFS are in Figure S2. Moreover, the analysis of the clinicopathological parameters (Table 2) demonstrated that no remarkable publication bias existed. Regarding sensitivity analysis, none of the pooled ORs for invasion depth, lymphatic metastasis and differentiation degree was remarkably affected by eliminating any single study (Figure 5). However, the sensitivity analysis of ORs for distant metastasis indicated a lack of stability. The results of the other clinicopathological characteristics in the analysis of publication bias and sensitivity are in Figure S3.
Table 2
Correlation of high CSN5 expression with clinicopathological parameters
Parameters
|
Studies
|
Case number
|
Pooled OR(95%CI)
|
P
|
Heterogeneity
|
Model
|
Publication bias
|
|
|
|
|
|
I2
|
P
|
|
|
Age
|
5
|
406
|
1.37 [0.89, 2.13]
|
0.16
|
0%
|
0.72
|
Fixed
|
0.462
|
Gender
|
17
|
1487
|
1.00 [0.78, 1.27]
|
0.97
|
34%
|
0.08
|
Fixed
|
0.343
|
TNM stage
|
9
|
721
|
0.81 [0.34, 1.91]
|
0.63
|
80%
|
<0.00001
|
Random
|
0.536
|
Tumor size
|
8
|
749
|
0.83 [0.60, 1.16]
|
0.27
|
37%
|
0.14
|
Fixed
|
0.536
|
Invasion depth
|
6
|
591
|
0.49 [0.25, 0.96]
|
0.04
|
56%
|
0.04
|
Random
|
0.26
|
Lymphatic metastasis
|
15
|
1294
|
0.28 [0.16, 0.47]
|
<0.00001
|
68%
|
<0.00001
|
Random
|
0.701
|
Distant metastasis
|
3
|
246
|
0.32 [0.13, 0.76]
|
0.01
|
0%
|
0.42
|
Fixed
|
1
|
Differentiation degree
|
11
|
984
|
0.34 [0.19, 0.60]
|
0.0003
|
55%
|
0.01
|
Random
|
1
|
Venous invasion
|
4
|
322
|
1.11 [0.22, 5.53]
|
0.9
|
82%
|
0.0009
|
Random
|
1
|