The registry was established with PROSPERO prior to the start of the research (registration number: CRD42021286741). A structured literature search of Web of Science, Embase, PubMed and The Cochrane Library was performed from their inception through 12 March 2022. “Long intergenic non-protein coding RNA 662,” “lncRNA LINC00662,” “LINC00662,” “neoplasm,” “cancer,” “malignancy,” “neoplasia,” “melanoma,” “tumor,” “sarcoma,” “carcinoma” or “adenoma” were the search terms. Relevant study citation lists were also thoroughly reviewed. To ensure accuracy and consistency, two investigators individually assessed the database search methodologies and debated the results.

The inclusion criteria were as follows: 1) discussing LINC00662 expression in cancer tissues; 2) in which patients were classified into two groups by the LINC00662 expression levels; 3) reporting the connection between LINC00662 expression and cancer prognosis or clinicopathological characteristics; and 4) in which hazard ratios (HRs) and 95% confidence intervals (CIs) for survival data were presented or could be obtained from survival curves.

The exclusion criteria were as follows: 1) conference reports, in vitro or in vivo experiments, and reviews; 2) duplicate publications; and 3) articles whose data could not be extracted and articles from non-English publications.

Two researchers separately explored and chose articles based on the aforementioned criteria, extracting the following data: first author’s name, publication year, country of study, types of cancer, number of patients, detection methods, cut-off criteria, clinical parameters, overall survival (OS), relapse-free survival (RFS), disease-free survival (DFS), and progression-free survival (PFS). Studies were used immediately if they contained precise survival data. When publications supplied only Kaplan Meier (KM) curves without precise survival data, we used Engauge Digitizer V.4.1 software to obtain data from the KM curves and computed HRs and 95% CIs. Literature quality was evaluated using the Newcastle-Ottawa Scale (NOS). The study was graded on a nine-item scale, with one point assigned for each item completed. Aggregate scores varied from 0 to 9. The NOS score of 7 and above denoted high-quality research findings.

Gene Expression Profiling Interactive Analysis (GEPIA), an online web-based tool based on TCGA and GTEx data, is designed to verify the LINC00662 expression levels in pan-cancer. LINC00662-related survival analysis was conducted using the KM method and log-rank test.

The MEM-Multi Experiment Matrix database, which is based on the Affymetrix Gene Chip Human Genome U133 Plus 2.0 Array technology, was used to identify the relevant genes of LINC00662. Then, gene ontology (GO) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were performed using R software. Finally, Cytoscape software was utilized for constructing the signaling pathway network.

Stata 12.0 was used to estimate all statistical data. To investigate the relationship between LINC00662 expression and survival outcome, HRs with 95%CIs were determined. To investigate the link between LINC00662 expression and clinicopathological features, researchers estimated pooled odds ratios (ORs) with 95% CIs. If heterogeneity was moderate (P _Q > 0.1, I ² < 50%), a fixed-effects model was applied. Otherwise, a random-effects model was applied. Significant heterogeneity was addressed using methods such as subgroup analysis or sensitivity analysis. All data were presented in the form of Forest plots. Egger’s test and Begg’s funnel plot were used to determine possible publication bias, while sensibility analysis was employed to determine the causes of heterogeneity and to verify the stability of the results. p < 0.05 was considered statistically significant.

In total, 115 studies were initially obtained according to the search strategy (Web of Science = 35, Embase = 44, PubMed = 36, Cochrane = 0). After removing 38 duplicate studies, 45 articles were removed as extraneous papers by scanning the titles and abstracts. The remaining 32 studies were then carefully reviewed by reading the full article. Of them, 9 were eliminated due to the lack of eligible data, 9 were eliminated since they were physical experimental studies, and 14 studies (Gong et al., 2018; Li et al., 2019; Wang et al., 2019; Cheng B. et al., 2020; Cheng L. et al., 2020; Guo et al., 2020; Liu and Meng, 2020; Tian et al., 2020; Wei et al., 2020; Zhang et al., 2020; Ma et al., 2021; Wu et al., 2021; Yu et al., 2021; Wang et al., 2022) were finally included for meta-analysis. The study screening procedure is presented in Figure 1.

The 14 enrolled articles contained a total of 960 patients, with sample sizes between 39 and 105. All of these articles were conducted in China and published between 2018 and 2022. The above articles included ten tumors, including three cases of OSA (Liu and Meng, 2020; Yu et al., 2021; Wang et al., 2022), two cases of HCC (Guo et al., 2020; Tian et al., 2020), two cases of CRC (Wang et al., 2019; Cheng B. et al., 2020), one case of bladder cancer (Ma et al., 2021), one case of lung cancer (Gong et al., 2018), one case of CC (Wei et al., 2020), one case of ESCC (Zhang et al., 2020), one case of BC (Cheng L. et al., 2020), one case of PCa (Li et al., 2019), and one case of OC (Wu et al., 2021). Nine of these articles (Gong et al., 2018; Li et al., 2019; Wang et al., 2019; Cheng B. et al., 2020; Guo et al., 2020; Tian et al., 2020; Ma et al., 2021; Wu et al., 2021; Yu et al., 2021) reported OS, two (Guo et al., 2020; Tian et al., 2020) reported RFS, and all studies reported clinical parameters. Three (Li et al., 2019; Guo et al., 2020; Wu et al., 2021) OS data and one (Guo et al., 2020) RFS data were obtained directly from the original text, and six (Gong et al., 2018; Wang et al., 2019; Cheng B. et al., 2020; Tian et al., 2020; Ma et al., 2021; Yu et al., 2021) OS data and one (Tian et al., 2020) RFS data were extracted by KM curves. All studies employed real-time quantitative polymerase chain reaction (RT-qPCR) to detect the expression levels of LINC00662. The characteristics of the included articles are summarized in Table 1.

Nine articles (Gong et al., 2018; Li et al., 2019; Wang et al., 2019; Cheng B. et al., 2020; Guo et al., 2020; Tian et al., 2020; Ma et al., 2021; Wu et al., 2021; Yu et al., 2021) including 689 cancer patients reported a relationship between abnormal LINC00662 expression levels and OS. A fixed-effects model was used because no obvious heterogeneity existed (I ² = 0%, P _Q = 0.807). As shown in Figure 2A, the HR of the pooled OS was 1.91, indicating that LINC00662 overexpression was notably interrelated with poor OS. Furthermore, two articles (Guo et al., 2020; Tian et al., 2020), with a total of 156 patients with HCC, provided suitable data for RFS. A fixed-effects model was used as no obvious heterogeneity existed (I ² = 25.7%, P _Q = 0.246). LINC00662 overexpression was markedly interrelated with shorter RFS in patients with HCC (HR = 2.12, 95% CI 1.19–3.76, p = 0.010) (Figure 2B).

In addition, we conducted a subgroup meta-analysis based on tumor type (digestive tract tumors and non-digestive tract tumors), follow-up time (≥60 months or <60 months), HR availability (reported in text and survival curve), and NOS score (NOS scores ≥7 or <7) to investigate further the correlation between LINC00662 expression levels and OS. According to all subgroup analyses, revealed that LINC00662 overexpression was notably interrelated with poor OS (Figure 3; Table 2). The aforementioned results imply that LINC00662 expression levels is probably a prognostic factor for OS in patients with tumors.

In total, 14 articles (Gong et al., 2018; Li et al., 2019; Wang et al., 2019; Cheng B. et al., 2020; Cheng L. et al., 2020; Guo et al., 2020; Liu and Meng, 2020; Tian et al., 2020; Wei et al., 2020; Zhang et al., 2020; Ma et al., 2021; Wu et al., 2021; Yu et al., 2021; Wang et al., 2022) presented usable data regarding the link between LINC00662 expression levels and clinicopathological parameters, which included age, sex, tumor stage, tumor size, LNM, differentiation grade, and distant metastasis (DM). The analysis was conducted using ORs and their 95% CIs. According to the results, LINC00662 overexpression was notably interrelated with advanced tumor stage (OR = 4.23, 95% CI 2.50–7.17, p < 0.001), larger tumor size (OR = 1.49, 95% CI 1.11–1.99, p = 0.008), earlier LNM (OR = 2.40, 95% CI 1.25–4.59, p = 0.008), and earlier DM (OR = 4.78, 95% CI 2.57–8.88, p < 0.001). However, the differences in terms of age (OR = 1.16, 95% CI 0.90–1.51, p = 0.246), gender (OR = 1.10, 95% CI 0.79–1.53, p = 0.578), and differentiation grade (OR = 1.53, 95% CI 0.71–3.33, p = 0.280) were not statistically significant. The results were presented in Figure 4 and Table 3.

To assess the reliability of LINC00662 expression and OS relevance findings, we conducted sensitivity analysis. No single study was found to alter the results (Figure 5A), and therefore, high LINC00662 expression was reliably associated with poor OS. Meanwhile, we performed sensitivity analysis by using the results of clinicopathological parameters. The heterogeneity of the relevance of LINC00662 overexpression to the tumor stage (Figure 5B) was completely eliminated after excluding the study of Cheng et al. (Cheng B. et al., 2020) and Zhang et al. (Zhang et al., 2020) (P _Q = 0.427, I ² = 1.2%), without affecting the results (OR = 3.84, 95% CI 2.76–5.35, p < 0.001). In a sensitivity analysis of LNM (Figure 5C), excluding the study of Wu et al. (Wu et al., 2021) significantly reduced heterogeneity (P _Q = 0.337, I ² = 12.2%), but did not affect the final results (OR = 3.02, 95% CI 2.02–4.50, p < 0.001). For the sensitivity analysis on differentiation grade (Figure 5D), heterogeneity was significantly reduced with the elimination of the studies of Wei et al. (Wei et al., 2020) and Zhang et al. (Zhang et al., 2020) (P _Q = 0.265, I ² = 22.5%), and a significant correlation was observed between high LINC00662 expression and the degree of tumor differentiation (OR = 2.29, 95% CI 1.49–3.52, p < 0.001). Furthermore, sensitivity analysis of the relevance of LINC00662 overexpression to DM (Figure 5E) displayed complete elimination of heterogeneity after the exclusion of the study of Cheng et al. (Cheng B. et al., 2020) (P _Q = 0.578, I ² = 0.0%), but the results were not affected (OR = 3.54, 95% CI 2.25–5.56, p < 0.001).

Additionally, we used Begg’s funnel plot and Egger’s regression test to analyze the publication bias of the included studies on the correlation between LINC00662 expression and OS. Neither Begg’s funnel plot (Figure 5F) nor Egger’s regression test (Pr > |t| = 0.307) found significant publication bias, suggesting that our results are plausible.

To further confirm the reliability of our results, we assessed the relationship between LINC00662 expression in pan-cancer and prognosis of cancer patients by using the GEPIA database. Regarding aberrant LINC00662 expression, LINC00662 was highly expressed in cholangiocarcinoma (CHOL), lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), pheochromocytoma and paraganglioma (PCPG), skin cutaneous melanoma (SKCM), and thymoma (THYM) (|log₂FC| Cutoff: 1, p-value: 0.01, Figure 6A). Regarding the relationship between LINC00662 expression and prognosis, LINC00662 overexpression was notably interrelated with worse OS in adrenocortical carcinoma (ACC; Figure 6B), kidney chromophobe (KICH; Figure 6C), brain lower grade glioma (LGG; Figure 6D), liver hepatocellular carcinoma (LIHC; Figure 6E), sarcoma (SARC; Figure 6F), and thyroid carcinoma (THCA; Figure 6G), and shorter DFS in ACC (Figure 6H), KICH (Figure 6I), LGG (Figure J), and LIHC (Figure 6K).

We screened 200 LINC00662-related genes from the MEM database for GO and KEGG pathway analyses (Figure 7; Table 4). The GO analysis revealed that these genes were remarkably correlated with the cellular amino acid biosynthetic process, mitochondrial matrix, and transaminase activity. Additionally, the KEGG pathway analysis demonstrated that these genes were strongly interrelated with amino acid metabolic pathways. Furthermore, we built a signaling pathway network with Cytoscape (Figure 8).

The molecular mechanisms of LINC00662-related carcinogenesis are recapitulated as follows.

• In HCC, LINC00662 facilitated the progression of HCC and the polarization of M2 macrophages by binding miR-107, miR-15a and miR-16 competitively to upregulate WNT3A expression and secretion, thereby activating the Wnt/β-catenin signaling pathway (Tian et al., 2020). Additionally, LINC00662 also exerted tumorigenic effects through simultaneous regulation of intracellular S-adenosylhomocysteine (SAH) and S-adenosylmethionine (SAM) levels resulting in altered methylation profiles (Guo et al., 2020).