The prognostic value of long non coding RNAs in cervical cancer: A meta-analysis

Cervical cancer is one of the most common malignancies in women worldwide. Numerous literatures demonstrate that aberrantly expressed lncRNAs are involved in tumorigenesis and development, and may have the potential to be prognostic markers. However, their prognostic functions in cervical remain controversial. Therefore, we performed this meta-analysis to evaluate the prognostic significance of lncRNAs in cervical cancer. We searched databases to identify relevant articles. Pooled hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated. 15 studies involving 1868 patients with cervical cancer and 12 lncRNAs were included. Our results indicated that the levels of lncRNAs were associated with the overall survival ((HR = 1.56, 95%CI = 1.02–2.38, P < 0.001, random-effect) and event free survival (HR = 1.33, 95%CI = 0.77–2.28, P < 0.01, random-effect). High HOTAIR expression was associated with shorter overall survival in cervical cancer (HR = 3.93, 95% CI = 2.34–6.62, P < 0.001, fixed-effect). In conclusion, our meta-analysis suggested that lncRNAs may serve as novel predictive factors for prognosis of cervical and high expression HOTAIR was associated with shorter overall survival in cervical cancer.


INTRODUCTION
Cervical cancer is the second most deadly gynecological cancer and caused about 260,000 women's death each year [1]. Radical surgery is recommended as the preferred treatment for patients with cervical cancer in early stage, producing a relatively favorable prognosis [2]. However, patients with cervical cancer are usually diagnosed at advanced or recurrent stages with relatively poor prognosis [3]. Hence, exploring sensitive and specific biomarkers for prognosis is critical for the research and clinical treatment of cervical cancer.
Long noncoding RNA (lncRNA) are a kind of non-coding transcripts longer than 200 nucleotides [4]. LncRNAs can regulate various biological processes, such as gene expression, transcription, and cellular proliferation, through epigenetic silencing, mRNA splicing, and lncRNA-microRNA interaction [5,6]. Furthermore, by comparing their expression of tumors and normal cells, lncRNAs are abnormally expressed in the various tumors, functioning as oncogenes or tumor suppressors [7,8]. Recent studies have shown that lncRNAs also play important role in cervical cancer, including occurrence, progression, metastasis and prognosis [9,10]. However, due to the sample size and research programs, singe study may be inaccurate and insufficient. With the aim to obtain a better understanding of the prognostic value of lncRNAs in patients with cervical cancer, we performed a meta-analysis to explore the prognostic value of lncRNAs through larger sample size of patients.
The basic information and data from the included studies are shown in Table 1 and Table 2. 15 studies enrolling 1868 participants, with a maximum sample size of 218 and a minimum sample size of 49 patients. Among these 15 studies, 14 focused on Asians and 1 evaluated Caucasians. 9 studies reported patient overall survival (OS), 2 focused on OS as well as disease free survival (DFS), and 5 investigated OS as well as relapse free survival (RFS). DFS and RFS were combined together into event free survival (EFS) as prognosis parameter of our study. All studies investigated patients with cervical cancer and qRT-PCR was performed to detect lncRNAs expression in tumor tissues. Because the cut-off definitions were various, the cut-off values were different in these studies.

Analysis between lncRNAs expression level and OS
A total of 15 studies were reported that the expression levels of lncRNAs were related to OS. The random-effect model was used to assess the HRs and 95%CI. As shown in Figure 2, the expression levels of lncRNAs were associated with the OS of cervical cancer patients (HR = 1.56, 95%CI = 1.02-2.38, P < 0.001, random-effect). From the forest plot, the high expressions of HOTAIR, PVT1, GAS5, MALAT1, CCHE1, HULC, HOXA11, ANRIL were associated with poor prognosis. Besides, XLOC_010588, XIST, LET, MEG3 were correlated to poor prognosis with the low expressions of lncRNAs in cervical cancer. With all the lncRNAs HOTAIR generate the highest HR of 5.28 [16]; by contrast, XIST exhibited the lowest HR of 0.27 [17]. Moreover, stratified analyses were performed using studies with the ethnicity. 14 articles reported the relationship between lncRNAs and Asian patents' OS (HR = 1.54, 95%CI = 0.96-2.47, P < 0.01, random-effect), and 1article was about Caucasian (HR = 1.73, 95%CI = 1.05-2.85, P = 0.03, fixed-effect).
In the enrolling studies, HOTAIR and PVT1 were investigated in two or more articles, other lncRNAs were performed in single study. Then we carried out a metaanalysis on the relationship between the expressions of HOTAIR/PVT1 and the OS of patients with cervical cancer. We found that the high levels of HOTAIR were associate with a poor OS (HR = 3.93, 95% CI = 2.34-6.62, P < 0.001, fixed-effect) ( Figure 3A). Meanwhile, a poor prognosis in cervical cancer was found in the increased of PVT1 (HR = 1.84, 95%CI = 1.27-2.65, P = 0.01, fixed-effect) ( Figure 3B)

Analysis between lncRNAs expression level and EFS
A total of 6 studies (912 patients) included in the EFS analysis revealed a protective role increased lncRNAs expression (HR = 1.33, 95%CI = 0.77-2.28, P < 0.01, random-effect) ( Figure 4). From the forest plot, the increased expressions of HOTAIR, MALAT1 and CCHE1 correlated with a worse prognosis, the decreased expressions of XLOC_010588 and MEG3 were associated with a worse prognosis. With all the lncRNAs HOTAIR generate the highest HR of 2.98 [12]; by contrast, MEG3 exhibited the lowest HR of 0.16 [24].

Publication bias and sensitivity analysis
As shown in Figure 5, Begg's test was used to perform the publication bias, respectively. In our meta-analysis, Begg's test suggested there were no publication bias in all groups, owing to the values of P > 0.05. Meanwhile, we used Stata11.0 software to carry  out sensitivity analysis to assess whether the individual studies affected the overall results. The results suggested that individual study had little influence on our eventual outcome (Figure 6), and proved that our analysis was relatively stable and credible.

DISCUSSION
Cervical cancer is one of the leading causes of cancer related death in women worldwide [26]. Identification of effective disease biomarkers for prognosis is urgently required. In recent years, mounting epidemiological and molecular biological studies have demonstrated that lncRNAs dysregulation was involved in cancers [27][28][29]. Furthermore, abnormal expression of multiple lncRNAs was found to be related with the tumorigenesis and may have the potential to be prognostic biomarkers and even therapeutic targets of cervical cancer [15,30]. Therefore, we perform this meta-analysis to evaluate the prognostic ability of lncRNAs in cervical cancer.
Up to present, two meta-analysis [30,31] assessed the correlation between lncRNAs and cancer survival involved cervical cancer as one of the cancer sites. Both  of these two studies focused on exploring a single lncRNA (H19 or HOTAIR) in relation to cervical cancer. With very different goal, our article mainly aimed to evaluate the prognosis ability of all lncRNAs in cervical cancer. Therefore, the current meta-analysis is the first to evaluate the relation between lncRNAs expression and prognosis of patients with cervical cancer comprehensively.
In current meta-analysis, we evaluated the prognostic value of lncRNAs in cervical cancer. Our results suggested that high expression of lncRNAs represented a risk factor   for OS and cervical cancer. Furthermore, we found that the high expressions of 8 lncRNAs were associated with poor prognosis. Besides, 4 lncRNAs were correlated to poor prognosis with the low expressions of lncRNAs in cervical cancer. Moreover, stratified analyses were performed using studies with the ethnicity. 14 articles reported the relationship between lncRNAs and Asian patents' OS (HR = 1.54,), and 1article was about Caucasian (HR = 1.73). Besides, EFS analysis also revealed a protective role decreased lncRNAs expression (HR = 1.33). Our results suggested that the expression of these lncRNAs may have a significantly prognosis value in cervical cancer. Due to the limitation of the study number, these conclusions need more clinical trials for verification.
In the enrolling studies, PVT1 and HOTAIR were investigated in two or more articles, and with the increased expression of the two lncRNAs, the prognosis rate was poor in cervical cancer. Our results revealed that PVT1 is increased in the cervical cancer patients, and the high expression of PVT1 was related with the poor prognostic outcome of cancer patients. Summary of Hazard ratios (HRs) of lncRNAs in cervical cancer patients showed that HOTAIR generate the highest HR of 5.28. HOTAIR was the first lncRNA found to be involved in tumor and has been proven to be raised in a variety of human cancers [32]. Meanwhile, HOTAIR is a tumorigenic factor and can be adopted as a diagnosis or predictive biomarker in various cancer types [33,34], those findings in consist with our results. Notably, HOTAIR was the most investigated lncRNA in cervical cancer, therefore, articles explored HOTAIR as a new biomarker in the diagnosis and treatment of cervical cancer is possibly the most promising.
It should be stressed that there were limitation in our analysis. First, the number size of eligible articles is relatively small, which restrict our ability to evaluate the prognosis of lncRNAs in subgroup analyses. Second, the main ethnicities of the patients in our analysis were Asian, which also might affect the results. Third, due to positive results could be published more easily than negative results, this may lead to hidden publication bias.
In summary, the current meta-analysis was firstly to evaluate the correlation between lncRNAs and prognosis of patients with cervical cancer. Despite these limitations, there was a relationship between lncRNAs levels and OS and EFS in cervical cancer, which demonstrated the strong prognostic value of lncRNAs in cervical cancer. However, large-scale and comprehensive studies were needed to confirm our findings and thus promote the clinical utility of lncRNAs in cervical cancer prognosis evaluation.

Search strategy
A comprehensive search was performed independently by two researchers (Xiangrong Cui and Xuan Jing) via PubMed, EMBASE and Web of Science for literatures published up to February 2017 to obtain relevant articles for the meta-analysis. The search strategy used both MeSH terminology and free text words to increase the sensitivity of the search. The keywords for the search in these databases included: "Long noncoding RNA", "lncRNA", "LincRNA", "Long ncRNA", "survival" and cervix (or cervical) cancer/neoplasm/ tumour/carcinoma. Meanwhile, we screened the references of retrieved relevant articles to identify potentially eligible literatures.

Inclusion and exclusion criteria
Literatures included in this meta-analysis had to meet the following inclusion criteria: studies about the association between lncRNAs expression in tissues or blood samples and prognosis of patients with cervical cancer; the survival outcomes were performed with OS or EFS including disease free survival DFS and RFS; patients were divided into high and low expression groups; sufficient published data were provided to calculate hazard ratios (HR) and 95% confidence interval (CI). Exclusion criteria were as follow: studies without usable or sufficient data; laboratory articles, reviews, letters, unpublished data and conference abstracts.

Data extraction
Two investigators (Xiangrong Cui and Xuan Jing) extracted relevant data independently using predesigned and standardized form from the eligible studies. Extracted information included title, authors' names, the nationality and ethnicity of study population, lncRNAs, case number, methodological information and cut-off value. Disagreements were resolved through group discussion.

Statistical methods
HRs and 95% CIs were calculated to assess to the relation between lncRNAs and survival in cervical cancer, with a significance level of α = 0.05. An observed HR > 1 implied a worse survival for the group with elevated lncRNAs expression. Conversely, HR < 1 implied a worse survival for the group with decreased lncRNAs expression [35,36]. We used Revman5.3 Software (Revman, the Cochrane Collaboration) to perform the meta-analysis and evaluate heterogeneity between studies by Cochrance Q-test and P-values. If heterogeneity was present (I 2 ≥ 50% or P ≤ 0.05), random-effect model was used to calculate pooled HRs or ORs, otherwise, the fixed-effect model was utilized [37,38]. Furthermore, we minimized the influence of heterogeneity through classifying the included studies into subgroups accorded to similar features. The Stata 11.0 Software (Stata, College Station) was performed to evaluate the sensitivity and publication bias of the studies. Publication bias was estimated using