The relationship between long non-coding RNAs and clinical metastasis as well as prognosis of renal cell carcinoma patients: a meta-analysis


 Background: Recently, some lncRNAs in studies were also regarded as the potential prognostic biomarkers for renal cell carcinoma. The aim of this meta-analysis is to evaluate the prognostic significance of lncRNAs for RCC and explore their correlation with lymph node metastasis (LNM) and distant metastasis (DM).Methods: A series of electronic databases were chosen to retrieve the eligible articles. A total of 35 studies involving 3535 patients with RCC were included finally. The statistical analysis was performed using STATA 16.0 and review manager 5.3 software. The relative risk (RR) and hazard ratio (HR) with 95% confidence interval (CI) were calculated to assess the significance. The stratified subgroup analysis was conducted by the up-regulated group and down-regulated group. Additionally, the sensitivity analysis was performed by the sequential omission of individual studies, and the publication bias was detected using Begg’s test.Results: The results indicated lncRNAs expression levels were correlated with LNM for up-regulated lncRNAs (RR=2.06, 95% CI: 1.53-2.78, p<0.00001) and down-regulated lncRNAs (RR=0.49, 95% CI: 0.31-0.76, p=0.002). The expression levels of lncRNAs were closely related to DM (RR=1.67, 95% CI: 1.21-2.29, p=0.002). Moreover, the expression levels of these lncRNAs were also associated with the overall survival (HR=1.71, 95% CI: 1.28-2.28, P=0.0003).Conclusions: This meta-analysis suggested that lncRNAs could be regarded as biomarkers for clinical lymph node metastasis or distant metastasis and also may serve as the potential predictive factors of prognosis for RCC.


Inclusion and exclusion criteria
The inclusion criteria were as follows: (1) studies that investigated RCC patients or cancer patients including RCC; (2) the expression levels of lncRNAs in tumor tissues were measured by quantitative real time polymerase chain reaction (qRT-PCR) or other methods and divided into two forms: high and low; (3) the relationship between lncRNAs expression and clinical metastatic features as well as survival outcomes were reported; (4) The hazard ratio (HR) for overall survival (OS) were provided in studies; (5) HR with 95% CI and p value could be obtained or calculated from data in articles.
The exclusion criteria were as follows: (1) letters, editorials, expert opinions, case reports and reviews; (2) non-English articles; (3) non-human species; (4) only cell or serum without human tissue was investigated; (5) lacking usable data or original data; (6) studies with duplicated data; (7) lncRNAs expression levels were not distinguished; (8) studies without clinical metastasis characteristics and survival outcome; (9) HR was calculated in the form of multiple lncRNAs or estimated from insu cient data.

Data extraction
Data from these eligible studies were extracted independently by two investigators (Su and Xu). These data retrieved from each article contained the following information: the last name of the rst author, the publication year, the investigated lncRNA name, nationality of research objects, number of patients, type of cancer, stage of cancer, sample style, lncRNA detection method, number of high and low lncRNA expression group as well as patients with lymph node metastasis and distant metastasis in each group and HR with 95% CI for overall survival or other survivals. Normally, these data were obtained from the original article, if unavailable, we contacted the corresponding author to collect them.

Statistical analysis
The relationship between lncRNAs expression levels and lymph node metastasis or distant metastasis as well as prognosis of RCC patients was assessed by HR and RR with 95% CI. The heterogeneity was evaluated by Q test and I 2 test. For the Q test and I 2 statistic respectively, it was considered signi cant if P value was less than 0.1 or I 2 was more than 50%. A random-effect model was used to calculate the pooled HR or RR if I 2 ≥50% or P<0.1, otherwise, a xedeffect model was utilized. Furthermore, in order to reduce the in uence of existing heterogeneity, we performed subgroup analysis. In this meta-analysis, statistical analyses of HR for prognosis and the RR for LNM and DM were calculated by Review Manager Version 5.3 (RevMan, the Cochrane Collaboration).
Meanwhile, the Stata Software 16.0 (Stata, College Station) was also used to assess the sensitivity and publication bias. P value less than 0.05 was considered to be statistically signi cant.

Characteristics of eligible studies
As shown in Figure 1, a total of 672 records about lncRNA expression and human cancer were retrieved in PubMed, Web of Science, Embase and Medline electronic databases according to our settings. 619 records among them were excluded after reviewing the titles and abstracts. Subsequently, remaining 53 full-text articles were assessed for eligibility, and 18 studies were further excluded because of the irrelevant analysis. Finally, 35 articles were chosen for this meta-analysis. Among these 35 studies, the speci c cancer we concerned was human renal cell carcinoma; 33 articles came from China and 2 came from Japan; all the metastasis were diagnosed by pathology; all the tissue specimens were well preserved before RNA extraction.

Clinical features
The basic research features and related results about metastasis and prognosis from the 35 including studies were listed in Table 1. A total of 3535 participants from China and Japan were involved in these 35 studies. Among these 35 articles, 28 studies reported the lymph node metastasis or distant metastasis or both. However, only 11 studies reported overall survival (OS).
As for the methods used to detect lncRNA expression levels in tumor tissues, all studies used qRT-PCR. The cut-off values were various in these studies due to the different cut-off de nitions. Finally, all these studies did not report that these lncRNAs were signi cantly associated with age and gender of patients. However, some studies showed that lncRNAs were signi cantly correlated to tumor size [17], histological grade [18] or tumor stage [14].

Relationship between lncRNAs expression levels and LNM
The 28 selected studies reported a total number of 921 patients with LNM in the form of different lncRNA expression levels. These studies have passed the Q test and the I 2 test (I 2 =79%, P<0.00001), which indicated that the heterogeneity was signi cant. Labbe Graph and Galbraith Plot for heterogeneity were shown in Figure 2A and Figure 2B. The random-effects models were selected, and the RR was 1.43 (95% CI:1.10-1.86, P=0.008) ( Figure 2C), which means the incidence of LNM in the lncRNA high expression group was 1.43 times that in the low expression group, and it was statistically signi cant. Visual inspection of the Begg's funnel plot revealed symmetry ( Figure 2D). The symmetry of the above funnel plot was tested, and the result showed no publication bias (P=2.31) ( Figure 2E).
Due to the signi cant heterogeneity, we explored the reasons for heterogeneity via sensitivity analysis ( Figure 2F) and meta regression ( Figure 2G). It was suggested that the up-regulated group and the down-regulated group were the sources of heterogeneity. Therefore, we performed subgroup analysis to these studies ( Figure 2H). As for the up-regulated lncRNAs subgroup (RR=2.06, 95% CI: 1.53-2.78, P<0.00001), high lncRNA expression had a signi cant elevated LNM rate compared with low lncRNA expression. On the contrary, low lncRNA expression had a signi cant increased LNM rate compared with high lncRNA expression in the down-regulated lncRNAs subgroup (RR=0.49, 95% CI: 0.01-0.53, P=0.002).
The publication bias test was performed to the subgroups. As shown in Figure 2I and Figure 2J, a publication bias was found in the up-regulated group (P=0.000). For the asymmetric funnel plot of the up-regulated group, the trim and ll analysis was selected for correction, and 8 articles were nally virtualized. After the trim and ll analysis, no publication bias was found in the up-regulated group ( Figure 2K and Figure 2L).

Relationship between lncRNAs expression levels and DM
Likewise, all chosen studies reported a total number of 730 patients with DM according to the different lncRNA expression levels. Labbe Graph and Galbraith Plot for heterogeneity were shown in Figure 3A and Figure 3B. As the heterogeneity was also signi cant (I 2 =78%, P=0.00001), a random-effects model was still adopted. As shown in Figure 3C, the results indicated that high lncRNA expression versus low lncRNA expression, had a statistic signi cant elevated DM rate for the up-regulated lncRNAs subgroup (RR=1.67, 95% CI: 1.21-2.29, P=0.002). Visual inspection of the Begg's funnel plot revealed asymmetry ( Figure 3D). The asymmetry of the above funnel plot was tested, and the result displayed a publication bias (P=0.002) ( Figure 3E). For the asymmetric funnel plot of the eligible studies, the trim and ll analysis was selected for correction, and 5 articles were nally virtualized. After the trim and ll analysis, no publication bias was found in the up-regulated group ( Figure 3F and Figure 3G).
We also explored the reasons for heterogeneity via sensitivity analysis ( Figure 3H) and meta regression ( Figure 3I). The results suggested that up/down regulation was not the source of heterogeneity and was not statistically signi cant.

Relationship between lncRNAs expression levels and prognosis 11 studies including 1850 patients reported the relationship between lncRNAs expression levels and prognosis (OS) of RCC patients in this meta-analysis. Due
to the signi cant heterogeneity (I 2 =73%, P<0.0001), the random-effects model was used. As shown in Figure 4A, lncRNA expression levels were closely related to OS (HR=1.71, 95% CI: 1.28-2.28, P=0.0003) of RCC patients. Moreover, high lncRNA expression had a signi cant reduced clinical survival compared with low lncRNA expression. Visual inspection of the Begg's funnel plot revealed symmetry ( Figure 4B). The symmetry of the above funnel plot was tested, and the result showed no publication bias (P=0.448) ( Figure 4C).
We also explored the reasons for heterogeneity via sensitivity analysis ( Figure 4D) and meta regression ( Figure 4E). The results suggested that up/down regulation was not the source of heterogeneity and was not statistically signi cant.

Publication bias, sensitivity analysis and meta regression
We used Stata 16.0 software to evaluate the sensitivity, publication bias and meta regression. In this meta-analysis, Begg's test was used in publication bias due to the possible excessive assessment. Meanwhile, the sensitivity analysis was also used to assess the reliability and stability of the combined results and to see whether the individual study affected the overall results. Finally, meta regression was used to explore the causes of heterogeneity.

Discussion
In recent years, a large number of studies have focused on lncRNAs and explored their biological mechanism and molecular function [47,48]. Increasing evidences showed the potential relationship between the aberrant expression of lncRNAs and cancers [49,50]. That is to say, the up-regulation or downregulation of some lncRNAs contributed to tumorigenesis and progression by affecting their cellular processes. However, the underlying mechanism involved in regulating cancers metastasis and prognosis remain unclear. In order to nd some effective treatments and molecular biomarkers for RCC, this systematic meta-analysis was performed. It was rstly to systematically analyze the association between lncRNA expression and clinical metastatic features as well as prognosis of RCC patients.
In this meta-analysis, we investigated the metastatic features and prognostic roles of lncRNAs in RCC by using the subgroup analysis. Meanwhile, the sensitivity analysis, publication bias and meta regression were also performed to optimize the statistical credibility. We found that dysregulated NEAT1 [14,22], HOTAIR [13,23], LOC389332 [45], MALAT1 [15][16][17], TCL6 [12], NBAT1 [46] could be viewed as biomarkers for lymph node metastasis while H19 [18], RCCRT1 [20], SPRY4-IT1 [21], TUG1 [9,39] could be regarded as biomarkers for distant metastasis in RCC patients. That indicated high or low lncRNA expression in RCC tissues was signi cant correlated with either LNM or DM. In addition, we also found that 10 high expressed lncRNAs were associated with the poor OS, while 1 low expressed lncRNA was correlated to the poor OS in RCC patients. Therefore, all of these results suggested that these lncRNAs could serve as the novel independent predictive factors of clinical metastasis and poor prognosis for RCC.
Strengths and limitations of this study: (1) This is the rst meta-analysis reporting the relationship between long non-coding RNAs and clinical metastasis as well as prognosis of renal cell carcinoma patients. (2) We used strict, broad search strategy of the internet databases to minimise any potential publication bias. (3) The number of eligible studies may be small. (4) All research objects were Asian. (5) Due to lack of follow-up, the time subgroup analysis could not be performed.

Conclusions
In summary, this meta-analysis summarized current related researches. It was also the rst one to evaluate the relationship between lncRNAs aberrant expression and RCC patients metastasis (LNM, DM) as well as prognosis (OS). The analysis revealed that lncRNAs not only could be considered as guideline for lymph node metastasis and distant metastasis, but could be regarded as potential prognostic biomarkers for RCC patients. However, more and larger-scale and higher quality researches should be performed to con rm the clinical values of lncRNAs for RCC due to the existing limitations here. The ow diagram of study process.