The present meta-analysis was performed in line with the guidelines of the Meta-analysis of Observational Studies in Epidemiology and Preferred Reporting Items for Systematic Reviews and Meta-Analyses groups (27,28). The PubMed/Medline, Web of Science and The Cochrane Library databases were systematically searched (up to January, 2015) for articles assessing the prognostic value of MALAT1 in various types of cancer. The key words and related Medical Subject Headings for lncRNA, MALAT1, cancer, prognosis, death and survival were used. In addition, experts were consulted, the reference lists of retrieved articles were reviewed, and the ‘see related articles’ links were searched for key publications in PubMed. In addition, the authors of the articles were contacted if necessary.

The inclusion criteria for the present analysis were as follows: i) Studies investigating the prognostic role of MALAT1 in patients with various types of cancer; ii) providing enough information to estimate hazard ratio (HR) and 95% confidence interval (CI) for overall survival (OS), disease-free survival (DFS) or recurrence-free survival (RFS); and iii) studies conducted on adults and published in English. Duplicate studies, non-original articles and animal experiments were excluded. Two reviewers (J.Y. and X.Y.Z.) independently scanned all titles and abstracts identified during the search. In addition, we obtained full-text reports of articles that indicated or suggested eligibility, resolving disagreements on exclusion through consensus adjudication. Finally, a total of 10 studies were included in the analysis (14,18–26).

Data were independently extracted by two investigators (J.Y. and X.Y.Z.) trained to interpret information to ensure homogeneity in data collection and entry. These data included name of first author, year of publication, demographic characteristics of the patients included in the study (number, age, gender and ethnicity), cancer characteristics (tumor type, tumor size, differentiation, invasion, lymph node metastasis and stage), study design (specimen, measuring method, cut-off point defining high MALAT1 expression and follow-up period) and survival analysis. HRs were preferentially extracted from multivariate or univariate analyses; if these were not available, Engauge 4.0 was used to calculate HRs and corresponding 95% CIs.

Study quality was rated by the Newcastle-Ottawa Quality Assessment Scale for Cohort Studies (29), in which the quality of the selected study was determined by 8 questions in 3 domains: selection (0–4 points) and comparability (0–2 points) of the study groups, and ascertainment of the outcome of interest (0–3 points). Based on previous recommendations, studies with 5 points were considered to be of high quality. The two aforementioned investigators independently evaluated the quality of each study. Disagreements were resolved by consensus.

Estimates of odds ratios (ORs) and HRs were weighted and pooled using the generic inverse-variance (30). ORs with 95% CIs were used to estimate the association between MALAT1 expression and clinicopathological characteristics. Pooled HRs with 95% CIs were used to estimate the prognostic role of MALAT1 in various cancers. I² was calculated as a measure of statistical heterogeneity, with values of 25, 50 and 75% representing mild, moderate and severe heterogeneity, respectively. Random-effects models were applied in cases with significant heterogeneity. Stratified analyses were conducted to assess potential confounder contribution to heterogeneity. In addition, publication bias was investigated using Begg's funnel plots and Egger's linear regression test. Analyses were performed with STATA statistical software, version 12.0 (Stata Corporation, College Station, TX, USA). The results were considered significant at two-sided P-value of 0.05.

A total of 296 articles were identified during the electronic search through PubMed/Medline, Web of Science and The Cochrane Library. Duplicates, animal experiments, non-MALAT1-related studies, non-English articles, and studies lacking data on the association of MALAT1 expression with prognosis were excluded, leaving 10 eligible studies to be included in this meta-analysis (Fig. 1). The characteristics of the included studies are summarized in Table I. All the studies were published from 2011 onwards. The participants in 8 studies were Chinese and in the other 2 studies German and American. Various types of cancers were recorded, including colorectal, non-small-cell lung and bladder cancer, pancreatic ductal adenocarcinoma, hepatocellular carcinoma, pancreatic cancer, clear cell renal cell carcinoma, gastric cancer and glioma. The expression level of MALAT1 was detected by reverse transcription quantitative polymerase chain reaction or in situ hybridization (ISH). The sample size ranged between 45 and 222 patients. There were 8 studies for OS, 2 for DFS and 1 for RFS. HRs with the corresponding 95% CIs were extracted from univariate analysis and Kaplan-Meier curves in 10 studies, and from multivariate analysis in 7 studies. The NOS scores of 90% of the included studies were ≥5.

The main results of the association between MALAT1 expression and clinicopathological characteristics are summarized in Table II. The results demonstrated that MALAT1 was not associated with clinicopathological parameters such as age, gender, diffferentiation, depth of invasion, lymph node metastasis, distant metastasis or tumor stage, but was significantly associated with tumor size (HR=2.70, 95%CI: 1.23–5.92, P=0.013).

A total of 8 studies including 1,008 patients reported HRs for OS. Of the 8 eligible studies, one (12.5%) reported a non-statistically significant HR (i.e., the 95% CIs crossed 1). A forest plot of all studies is presented in Fig. 2. It was suggested that increased MALAT1 expression predicted a poor outcome for OS (HR=2.07, 95% CI: 1.67–2.56, P=0.000). There appeared to be no heterogeneity between the HRs of MALAT1 among these studies (I²=0.0%, P=0.488). However, subgroup analysis was also conducted to investigate the association between HRs and these variables, including type of cancer, region of subjects, sample size, analysis methods and quality scores (Table III). Stratified analysis by cancer type indicated a significant prognostic effect of MALAT1 for digestive (HR=1.86, 95% CI: 1.37–2.53) and non-digestive system cancers (HR=2.28, 95% CI: 1.70–3.07). Although the differences in HRs between subjects from China (HR=2.33, 95% CI: 1.79–3.03) and Western countries (HR=1.66, 95% CI: 1.16–2.37) was not statistically significant (P=0.134), high MALAT1 expression in Chinese subjects was associated with numerically higher HR values. Moreover, the HRs were significant for studies including <110 subjects (HR=2.67, 95% CI: 1.69–4.23) as well as those with >110 subjects (HR=1.93, 95% CI: 1.52–2.45). When different analysis methods were considered, MALAT1 was a strong prognostic marker by univariate (HR=1.83, 95% CI: 1.81–2.85) as well as multivariate (HR=2.15, 95% CI: 1.68–2.73) analyses. In addition, when performing subgroup analyses stratified by quality score, increased MALAT1 expression was significantly associated with poor prognosis in the studies with a score of <7 (HR=3.40, 95% CI: 1.85–6.22), as well as in those with a score of ≥7 (HR=1.93, 95% CI: 1.54–2.42).

Begg's funnel plots and Egger's linear regression tests were applied to evaluate publication bias. The shape of the funnel plot exhibited no significant asymmetry (Fig. 3). Subsequently, Egger's test also indicated no evidence of publication bias (P=0.174).

A total of 2 studies including 224 patients reported HRs for DFS (Table III). Overall, upregulation of MALAT1 was associated with an HR for DFS of 2.60 (95% CI: 1.69–4.00). In addition, 1 study including 112 patients reported HRs for RFS (Table III). High MALAT1 expression also predicted a poor clinical outcome for RFS (HR=3.28, 95% CI: 1.52–7.09).