Characteristics of eligible studies
This study was conducted following the PRISMA Checklist, as shown in Table S1. A total of 121 studies were initially identified as potential articles. After removing the duplications, 65 studies were screened through titles and abstracts. Afterwards, three articles including review, meeting abstract or irrelevant topic were excluded. The remaining eleven full-text articles were further evaluated. Thirteen studies were excluded as a result of irrelevant topics or insufficient data. Finally, ten articles compromising 690 patients were included to carry out qualitative and quantitative synthesis. As demonstrated in Figure 1, the selection procedure was presented by a flow diagram.
The main characteristics of the included studies were demonstrated in Table 1. These articles were published between 2017 and 2019 with a sample size ranging from 42 to 126. Generally, the enrolled patients were distributed in two groups (high or low NNT-AS1 expression), considering the levels of NNT-AS1 as measured by qRT-PCR. All of these investigations were carried out in China. Eight divergent types of cancers were analyzed in our meta-analysis, including osteosarcoma, breast cancer, gastric cancers, bladder cancer, cholangiocarcinoma, hepatocellular carcinoma, colorectal cancer and cervical cancer. The follow-up months for survival outcome ranged from 39 to 80 months. Seven studies adopted univariate analysis for the survival analysis method and the other three articles performed multivariate analysis. Furthermore, these studies also investigated other clinicopathologic parameters, such as age, gender, clinical stage, vascular invasion, LNM and DM. As to clinical stage, it should be noted that most studies adopted the tumor node metastasis (TNM) classification system, while two studies used the Enneking (12) or the International Federation of Gynecology and Obstetrics (FIGO) staging (20). All of these eligible studies are of high quality with a NOS score ≥7. Details of the NOS scoring were reported in the supplementary file (Table S2).
Association between NNT-AS1 and OS
We used fixed-effects model to analyze the pooled HR and corresponding 95% CI since heterogeneity among these studies was not obvious (I²=0.0%, p=0.932). As presented in Figure 2A, the pooled result showed that high expression of NNT-AS1 predicted unfavorable OS in cancers (HR=2.08, 95% CI: 1.84-2.36, P<0.001).
In addition, stratified analyses were performed to investigate the relevance between NNT-AS1 expression with OS in different subgroups according to tumor type (digestive system or others), sample size (more or less than 60), follow-up months (more or less than 60), and survival analysis method (univariate or multivariate analysis). The results revealed that all stratified analyses recapitulated the predictive potential of NNT-AS1 for OS in malignancies (Figure 3 and Table 2).
Association between NNT-AS1 and other clinicopathologic parameters
In addition, ORs with corresponding 95% CIs were applied to detect the association between NNT-AS1 and other clinicopathological parameters. The results of these analyses were summarized in Figure 4 and Table 3. Notably, fixed-effects model was applied in analyzing the association between NNT-AS1 and several clinicopathologic characteristics including age, gender, vascular invasion, and DM, since no obvious heterogeneity was observed. High expression of NNT-AS1 was significantly correlated to vascular invasion (OR=3.98, 95% CI: 2.06-7.71) and DM (OR=2.45, 95% CI: 1.39-4.30), but not age and gender.
By contrast, the random-effects model was used to analyze the correlation between NNT-AS1 and clinical characteristics including clinical stage and LNM due to the apparent between-study heterogeneity. Significantly, upregulated expression of NNT-AS1 predicted worse clinical stage (OR=4.25, 95% CI: 1.71-10.56) and LNM (OR=3.92, 95% CI: 1.35-11.41).
Sensitivity analysis and publication bias
In order to assess the stability of the aforementioned results, sensitivity analysis was performed. When each eligible study was removed, the result of NNT-AS1 for OS was not obviously changed, indicating the conclusion is reliable (Figure 2B).
Besides, the publication bias, regarding correlation between expression level of NNT-AS1 and OS, was evaluated via conducting Begg`s funnel plot and Egger`s regression test. The Begg’s funnel plot was symmetry, and Egger’s test showed P= 0.369, suggesting no obvious publication bias was measured (Figure 2C).
Validation of the results in TCGA dataset
Furthermore, the expression levels of NNT-AS1 in related cancers were explored by utilizing the data originated from TCGA. As demonstrated in Figure 5, NNT-AS1 showed aberrant expression in sarcoma, stomach adenocarcinoma, liver hepatocellular carcinoma, colon adenocarcinoma, and rectum adenocarcinoma when compared with normal control, but the difference was not significant. Moreover, NNT-AS1 expression level was markedly correlated with clinical stage in human cancers. Besides, we merged the expression data and OS (DFS) data of carcinomas from TCGA dataset deriving from GEPIA, which including 9,488 patients categorized in high or low expression group. These results suggested that the upregulated NNT-AS1 expression predicted worse OS (p=0.029), but not DFS, confirming that overexpression of NNT-AS1 was significantly correlated to unfavorable OS in cancer patients.