3.1 Study characteristics
A total of 386 studies were identified from the databases; among them, 174 duplicate studies were excluded, and 158 studies were omitted after reading the abstracts and full texts. Furthermore, 16 publications did not investigate the association between BANCR expression and the prognosis of patients, 6 publications did not divide patients into high and low BANCR expression groups, and 12 publications lacked usable data. Finally, 20 eligible studies were included for qualitative and quantitative synthesis (Figure 1).
Of these 20 studies with 1997 patients, 19 studies with 1847 patients were from China, and 1 study comprising 150 patients was from Iran [22]. The publication years ranged from 2014 to 2019, and the expression levels of BANCR were all detected by qRT-PCR for the following cancer types: lung cancer [19], hepatocellular carcinoma [15-17], osteosarcoma [23], papillary thyroid cancer [24-27], gastrointestinal cancer [28, 29], bladder cancer [30], malignant melanoma [31], breast cancer [32, 33], clear cell renal cell carcinoma [18], esophageal squamous cell carcinoma and endometrial cancer (details in Table 1) [22, 34, 35]. The NOS scores are presented in Table 2.
3.2. The association of BANCR with OS
A total of 10 studies comprising 1151 patients were included in the analysis of the relationship between BANCR and OS. The random-effects model was applied due to marked heterogeneity (I2=60%, P=0.008). The pooled results supported the conclusion that patients with high BANCR expression tended to have shorter overall survival (HR=1.60, 95% CI: 1.19-2.15, P=0.002, Figure 2A). Moreover, subgroup analysis was conducted to explore the sources of heterogeneity based on cancer type, the level of BANCR expression (high BANCR expression vs. low BANCR expression), the method of HR extraction (direct / indirect extraction), sample size (less / more than 100 patients) and NOS score (score of 9 / less than 9). A strong correlation was revealed between high BANCR expression and poor OS for cancers in the digestive system (HR=1.94, 95% CI, 1.38-2.73; P=0.0001), for HRs extracted directly from articles (HR=1.69, 95% CI, 1.44-1.99; P<0.00001), for HRs from multivariate analysis (HR=1.71, 95% CI, 1.47-2.02; P<0.00001), for high BANCR expression group (HR=1.72, 95% CI, 1.48-1.98; P<0.00001), for studies with less than 100 patients (HR=1.62, 95% CI, 1.11-2.35; P=0.05) and for studies with more than 100 patients (HR=1.57, 95% CI, 1.07-2.31; P=0.02). No correlation between BANCR expression and OS was found for non-digestive system cancers (HR=1.35, 95% CI, 0.86-2.13; P=0.20), for HRs from univariate analysis (HR=0.84, 95% CI, 0.41-1.75; P=0.65) or HRs extracted indirectly from articles (HR=1.15, 95%CI, 0.52-2.56; P=0.73). Detailed results are shown in Table 3. The poor prognosis related to BANCR was also identified by the positive association between high BANCR expression and short DFS (HR=1.21, 95% CI: 0.33-4.41, P=0.77) and RFS (HR=1.53, 95% CI: 1.27-1.85, P<0.00001) (Figure 2B).
3.3. The association of BANCR with TNM stage
Fourteen studies including 1378 patients were enrolled to investigate the association of BANCR expression level with TNM stage. The random-effects model was adopted, and subgroup analysis was carried out due to significant heterogeneity (I2=83.9%, P<0.00001). The pooled OR showed a strong association between high BANCR expression and advanced tumor stage (HR=2.39, 95% CI: 1.26-4.53, P<0.001). According to the results of the subgroup analysis, a strong association between high BANCR expression and advanced TNM stage for digestive system cancers (HR=4.01, 95% CI: 2.45-6.57, P<0.00001) and female reproductive system cancers (HR=12.25, 95% CI: 1.27-118.37, P=0.03) was found; a negative association for non-small cell lung cancer (HR=0.26,95% CI: 0.11-0.61, P=0.002) was found; And no association was found for other system cancers (HR=1.30, 95% CI: 0.40-4.27, P=0.15) (Figure 3).
3.4. The association of BANCR with other clinicopathological parameters
Other prognostic parameters were also assessed, and obvious correlations between increased BANCR expression and advanced lymph node metastasis (OR=2.03, 95% CI=1.08-3.83, P<0.05) (Figure 4), distant metastasis of tumor cells (OR=3.08, 95% CI: 1.92-4.96, P<0.001) (Figure 5A), advanced invasion depth (OR=1.54, 95% CI: 1.06-2.24, P=0.02) (Figure 5B), worse histological grade (OR=1.54, 95% CI: 1.00-2.383, P=0.05) (Figure 5C), larger tumor size (OR=1.63, 95% CI: 1.09-2.46, P=0.02) (Figure 6) and more local tumor nodes (multiple / single) (OR=1.78, 95% CI: 1.12-2.83, P=0.01) were found. However, no associations were found for smoking status (smoker vs. nonsmoker) (OR=1.01, 95% CI: 0.65-1.56, P=0.98), age (old vs. young) (OR=0.88, 95% CI: 0.71-1.09, P=0.236) and sex (female vs. male) (OR=0.91, 95% CI: 0.72-1.16, P=0.469) (Table 4).
3.5. Publication bias and sensitivity analysis
Sensitivity analysis was performed to assess the OS outcome stability among the included studies. We found that removing each study successively did not influence the overall results significantly (The overall HR value of the sensitivity analysis is: HR=0.47, 95%CI: 0.18-0.77. The detail HR value with removing each study successively could be seen in Figure 7, and no HR value exceeds the confidence interval of the combining result (95%CI: 0.18-0.77)), indicating that the results of each publication were almost consistent with the combined results, in other words, the merged results have high robustness and reliability (Figure 7). Potential publication bias was estimated by Begg’s test. As shown in Figure 8, slight publication bias was revealed among the included studies for OS (Pr > |z| =0.245), TNM stage (Pr > |z| =0. 477), LNM (Pr > |z| =0. 493), DM (Pr > |z| =0. 042), histological grade (Pr > |z| = 0.245) and tumor size (Pr > |z| =0.497). Consequently, there was no significant publication bias in this meta-analysis.