Long non‐coding RNA as a potential diagnostic and prognostic biomarker in melanoma: A systematic review and meta‐analysis

Abstract Recently, long noncoding RNAs (lncRNAs) have been applied as biomarkers for melanoma patients. In this systematic review and meta‐analysis, we investigated the diagnostic and prognostic value of lncRNAs. We used the keywords ‘lncRNA’ and ‘melanoma’ to search databases for studies published before June 14th, 2023. The specificity, sensitivity and AUC were utilized to assess diagnostic accuracy and the prognostic value was assessed using overall survival, progression‐free survival and disease‐free survival hazard ratios. After screening 1191 articles, we included seven studies in the diagnostic evaluation section and 17 studies in the prognosis evaluation section. The Reitsma bivariate model estimated a cumulative sensitivity of 0.724 (95% CI: 0.659–0.781, p < 0.001) and specificity of 0.812 (95% CI: 0.752–0.859, p < 0.001). The pooled AUC was 0.780 (95% CI: 0.749–0.811, p < 0.0001). The HR for overall survival was 2.723 (95% CI: 2.259–3.283, p < 0.0001). Two studies reported an HR for overall survival less than one, with an HR of 0.348 (95% CI: 0.200–0.607, p < 0.0002). The HR for progression‐free survival was 2.913 (95% CI: 2.050–4.138, p < 0.0001). Four studies reported an HR less than one, with an HR of 0.457 (95% CI: 0.256–0.817). The HR for disease‐free survival was 2.760 (95% CI: 2.009–3.792, p < 0.0001). In conclusion, the expression of lncRNAs in melanoma patients affects survival and prognosis. LncRNAs can also be employed as diagnostic biomarkers.

treatment of melanoma.However, significant progress in genetic, epigenetic and transcriptomic fields has shown great promise for developing potential biomarkers for diagnosing and determining patients' prognoses. 4lanoma occurs due to abnormalities in multiple genes and signalling pathways controlling cell proliferation and function, which itself arises from the alternation in either gene sequence or expression. 5Besides the genetic predisposition, so far, extensive attention is being paid to epigenetic events involved in the initiation or progression of melanoma. 5Noncoding RNAs (ncRNAs) are a new class of regulatory molecules associated with diseased conditions like different types of cancers.Long noncoding RNAs (lncRNAs) are noncoding transcripts longer than 200 nucleotides involved in much of the gene life cycle, including transcriptional, posttranscriptional and epigenetic mechanisms of gene regulation. 6rious types of RNAs have recently been applied as a biomarker for disease detection, 7,8 but still, there are scarce specified diagnostic panels.Identifying shared lncRNA dysregulation may lend insight into the early patient's diagnosis and prognosis and find potentially novel targets for treatment.0][11] However, indicating its diagnostic accuracy is important for the clinical application of these biomarkers.Also, the prognostic accuracy of these biomarkers must be identified for future clinical applications. 10views of the literature describe the disruption of lncRNA expression within cancer types, but lncRNA use as a diagnostic and prognostic biomarker has not been systematically reviewed across melanoma patients.To the best of our knowledge, the present study is the first systematic review and meta-analysis concerned with this issue.We recorded lncRNA-related diagnostic and prognostic values from articles that extracted lncRNAs from human tissue specimens retrieved from melanoma patients.

| ME THODS
We conducted a systematic review and meta-analysis in accordance with the PRISMA guidelines. 12Our systematic review and metaanalysis protocol has been registered at PROSPERO with the registration number CRD42023441549.

| Literature search
An in-depth search was performed until 14th June 2023, in PubMed, Web of Science (ISI), Scopus and Embase to identify English publications without any limitations on publication year.Databases were searched by the following medical subject headings (MeSH) terms and free keywords: 'long non-coding RNA' and 'melanoma' and their expansions.Table S1 provides the search query.

| Selection criteria
This study incorporated original research that had previously been reviewed by peers and presented the sensitivity, specificity or area under the curve (AUC) values of lncRNAs in the diagnosis of melanoma, as well as their association with prognosis in terms of overall survival (OS), progression-free survival (PFS), disease-free survival (DFS), recurrencefree survival (RFS) and event-free survival (EFS).While the diagnostic part of our research consisted of novel case-control human studies, the prognostic part employed cohort studies.The research studies were carried out in either a prospective or a retrospective manner, and they used samples acquired from patients who had been pathologically diagnosed with melanoma as well as those who were healthy as controls.In diagnostic accuracy studies, the comparison of lncRNA to an adequate reference control should have been performed regardless of the test assay time in order to evaluate sensitivity, specificity and AUC.
There were no limitations placed on eligibility based on the healthcare settings in which the research was carried out, nor were there any limitations placed on eligibility based on the total number of participants in the studies that were included.Non-English studies, studies on datasets or animal models, letters, comments, reviews, editorials, conference abstracts, case reports and case series were considered ineligible and were therefore excluded from the analysis.
Following the removal of any duplicates, SK and PF went through the remaining identified papers and evaluated their eligibility based on the inclusion and exclusion criteria that had been previously outlined.After compiling a list of studies that satisfied the eligibility requirements, both authors proceeded to independently conduct a comprehensive review of the full texts of the studies.During the review process, any conflicts that arose were effectively resolved through the formation of a consensus.

| Data extraction
Two reviewers (SMH and PF) independently extracted data from the included studies in a dedicated electronic spreadsheet.The following data were extracted from each when available: author, publication year, specimen type, sample size, control population, lncRNA name, change in levels of lncRNA in patients compared to the control group, diagnostic or prognostic performance measures, including sensitivity, specificity, AUC with corresponding 95% confidence interval (CI) and p-value, as well as mean, median and hazard ratio (HR) for survival outcomes with corresponding 95% CI and p-Value.
Discrepancies were resolved through discussion and consensus.

| Quality assessment
The quality of the included studies was assessed using an appropriate tool, the Newcastle-Ottawa Scale (NOS), for cohort and case-control studies. 13Two reviewers (PF and SMH) independently assessed the quality of each study based on predefined criteria.Any discrepancies in the quality assessment were resolved through discussion or consultation with a third reviewer.Selection, comparability and outcome are the three main categories of bias in NOS.Scores of 7 and above, 2-6, 1 and below were considered 'good', 'fair' and 'poor', respectively.

| Statistical analysis
We used the bivariate random effect model that was developed by Reitsma et al. (2005) in order to compile the research that provides diagnostic specificity and sensitivity. 14A bivariate model uses logit transformation to aggregate test sensitivity and specificity across studies.This is done by taking into account the interdependency of the two variables.This model also determines the summary receiver operating characteristic (sROC) curve and the AUC, both representing the accuracy of the diagnosis.For the studies that reported AUCs, the inverse variance method was used to meta-analyse AUC values.Because it was expected that there would be heterogeneity across the research included, the random effects model was used.
We employed the inverse variance method with logarithmic HR values to conduct a meta-analysis of prognostic values, which were reported as HRs.The random effects model was used to account for the observed heterogeneity in the reported values.Because HR less than one indicates that the explored variable (lncRNAs) has a cancerprotective function and HR more than one shows that the lncRNA has a cancer-promoting function, these two groups were split as distinct categories independent of the regulation of the studied lncRNA.
The standard error of the AUC and HRs for meta-analysis was calculated using either the 95% CI or the AUC value and sample size if the CI was not available.The study employed I 2 and the DerSimonian-Laird estimator of tau2 statistics to assess research heterogeneity.A subgroup analysis was conducted based on the sample type obtained to further explore the heterogeneity.The statistical analysis and visualizations were conducted using R version 4.2.2.Statistical significance was determined by an I 2 value exceeding 50% and a p-value below 0.05.

| Basic characteristics
After performing the initial database search, a total of 2603 titles were obtained.After eliminating duplicate articles, a total of 1191 articles were screened.After a comprehensive review of titles and

Include
abstracts, 1122 articles were excluded, and a total of 69 articles were considered appropriate for full-text review.A total of seven studies met the inclusion criteria for the diagnostic accuracy section, while 17 studies met the inclusion criteria for the prognosis section.
The excluded studies are listed in Table S2.The PRISMA flowchart, shown in Figure 1, outlines the process of selecting and excluding studies.
Table 1 provides a concise overview of the fundamental characteristics of the studies that were included.Overall, 647 melanoma patients and 721 controls were studied in the included studies for the diagnostic values, and 1453 melanoma patients were studied for the prognostic values.The papers included in the diagnosis section were published from 2016 to 2022, while those included in the prognosis section were published from 2014 to 2023.One study reported the diagnostic and prognostic value of more than one lncRNA. 10The meta-analysis of diagnosis accuracy comprised a total of 1579 melanoma samples and 805 healthy samples obtained from China and Poland.The prognosis section analysed a total of 1563 melanoma samples from China and Poland to assess OS.Additionally, 580 cases from Poland were examined to evaluate PFS, and 576 cases from China were used to assess DFS.The meta-analysis of diagnostic evaluations consisted of 24 evaluations, which involved two types of specimens: four tumour tissue samples and 20 blood samples.
The prognostic evaluations encompassed two types of specimens: tumour tissue samples and blood samples.
Among the 24 diagnostic evaluations conducted in the studies analysed, there were 23 distinct lncRNAs.Eleven diagnostic evaluations reported upregulation of lncRNAs, while 13 evaluations reported their downregulation.A total of 21 diagnostic evaluations have reported the sensitivity and specificity metrics for diagnosing TA B L E 1 Basic characteristics of the included studies.The meta-analysis of prognostic evaluations consisted of 19 evaluations.Nineteen distinct lncRNAs were assessed in the meta-analysed studies.In 10 prognostic evaluations, upregulation of lncRNAs was observed, while downregulation was observed in nine evaluations.

| Quality assessment
The studies included in the analysis were assessed for quality using the NOS by independent investigators (Table 2).Fourteen studies received a 'good' score, seven studies received a 'fair' score, and no studies received a 'poor' score, indicating a low risk of bias for included studies.

| Meta-analysis of diagnostic value of lncRNAs in melanoma patients
The Reitsma bivariate model estimated a cumulative sensitivity of 0.724 (95% CI: 0.659-0.781,p < 0.001) and a pooled specificity of 0.812 (95% CI: 0.752-0.859,p < 0.001) for lncRNAs in diagnosing melanoma involving 1407 melanoma cases and 681 controls (Figure 2).The estimated I 2 value using the Holling sample size unadjusted approach was 14.9%-32.5%.The test for equality of sensitivities among the studies had a p-value of <2e-16, and the test for equality of specificities had a p-value of 0.000599.The sROC curve was generated, and the overall pooled AUC for all specimen types was determined to be 0.837 (Figure 3).For lncRNAs  4).The studies were classified into subgroups based on the specimen type used to measure lncRNA expression.The AUC for the blood specimen subgroup (n = 20), involving 1147 cases and 397 controls, was 0.772 (95% CI: 0.735-0.808;I 2 = 64.9%).For the tissue specimen subgroup (n = 4), involving 432 cases and 408 controls, the pooled AUC was 0.808 (95% CI: 0.740-0.876;I 2 = 76.3%).The test for between subgroup differences was not statistically significant (p = 0.3563) (Table 3).5).

| Meta-analysis of the prognostic value of lncRNAs in melanoma patients
Among the 10 prognostic evaluations that reported HRs for PFS, six indicated HR greater than 1 and involved 348 cases, all of which utilized blood samples as specimens.The pooled HR for the included studies was 2.913 (95% CI: 2.050-4.138,p < 0.0001; I 2 = 0.0%).The four studies that reported an HR smaller than 1, involving 232 cases, had a pooled HR of 0.457 (95% CI: 0.256-0.817,p = 0.0083).These studies employed blood samples as specimens (Figure 6).

| DISCUSS ION
Recently, accumulating evidence has shown that lncRNAs play a key role in various biological processes in different malignancies, including melanoma.6][17][18] In this systematic review and meta-analysis, we aimed to summarize the results of individual studies on human samples and investigate the diagnostic and prognostic value of lncRNAs in melanoma.Our meta-analysis showed a cumulative sensitivity of 0.724, a pooled specificity of 0.812 and an overall AUC of 0.837 for lncRNAs in diagnosing melanoma.Regarding the type of specimen, there was no significant difference in the AUC of lncRNAs derived from tissue samples and those from serum.In the prognostic section, the combined HR for OS, PFS and DFS was 2.723  CI: 2.009-3.792),respectively.In our subgroup analysis, there was no significant difference in HR between tissue and blood samples.
Studies have shown that lncRNAs are involved in different cellular functions. 19Detection of large numbers of lncRNAs, their expression patterns in various types of malignancies, and their specificity and stability in body fluids suggest their potential role in developing novel diagnostic, prognostic and therapeutic tools for cancer. 20Currently, biopsy is the gold standard method of melanoma diagnosis. 21,22It has been known that lncRNAs are secreted in body fluids.Despite histopathological biopsy, which is an invasive and uncomfortable method, F I G U R E 2 Diagnostic accuracy of lncRNAs.

F I G U R E 3
The summary receiver operating characteristics (sROC) curve was plotted using the sensitivities and false positive rates of included studies.
lncRNAs can be easily obtained from patients.Thus, analysis of ln-cRNAs can be used as suitable diagnostic biomarkers for melanoma.
Multiple studies on melanoma cell lines have studied the mechanism of lncRNAs in melanoma.For example, in the study of Bian et al., NKILA, which was downregulated in melanoma tissue, suppressed the progression of the cell cycle and proliferation.
Further, NKILA significantly induced apoptosis and inhibited invasion in melanoma cell lines through regulation of the nuclear factor kappa B (NF-ĸB) signalling pathway. 9HOTAIR downregulation has been associated with inhibiting cellular proliferation and inducing apoptosis in melanoma cells through the regulation of NF-ĸB. 23HOTAIR leads to melanoma cell growth and metastasis by sponging miR-152-3p and activating the PI3k/Akt/mTOR signalling pathway. 24Through investigation of the lncRNA PVT1 mechanism in uveal melanoma (UM) cell lines, it was revealed that the clonogenic capacity of cells significantly decreased after silencing lncRNA PVT1.Furthermore, it was demonstrated that PVT1 knockdown represses the proliferation and increases the apoptosis of UM cells through the downregulation of EZH2 expression. 25A study by Zhang et al. using the data from TCGA and GEO databases has suggested PRRT3-AS1 as a potential diagnostic and prognostic biomarker for melanoma. 26sed on our results, lncRNAs could be used as a prognostic biomarker in melanoma patients.Similar to our results, studies on cell lines have shown that altered expression of TUG1, 27 HOTAIR 28,29 and BANCR 30  melanoma effectively.Thus, from these investigations, it could be concluded that using combinations of different lncRNAs with each other or other biomarkers could help clinicians make a more precise prediction of melanoma patients' prognosis.
To the best of our knowledge, our study is the first systematic review and meta-analysis on the prognostic and diagnostic value of lncRNAs in melanoma.However, several limitations in this study should be mentioned.Firstly, only a few lncRNAs appeared in more than one study; thus, we were unable to conduct a meta-analysis on the diagnostic and prognostic values of one type of lncRNA.
Secondly, differences in methods, including different sample types, outcome measures, sample sizes and follow-up periods in the studies, induced heterogeneity.The cut-off value varied between studies, and although RT-qPCR was used as the standard method to measure the expression level of lncRNAs, this may have caused heterogeneity in the results.Thirdly, we included the studies that obtained their samples directly from human subjects; studies that used data from databases were excluded from this study.Additionally, because most of the selected literature came from China, the results need further verification in other ethnicities.
Although physical examination and biopsy are considered the most reliable methods for diagnosing melanoma, the difficulties in F I G U R E 5 Forest plot of overall survival hazard ratios; prognostic evaluations were divided into two groups based on HR (less than or greater than one) and subgroups based on the specimen type.
7][38] The timely identification and proactive measures of sun protection play a crucial role in mitigating the adverse health F I G U R E 7 Forest plot of disease-free survival hazard ratios; prognostic evaluations were divided into two groups based on HR (less than or greater than one) and subgroups based on specimen type.
0][41][42] The findings presented in this study suggest that lncRNAs represent a newly recognized class of regulatory molecules that may have the potential to influence various aspects of melanoma, including proliferation, invasion, migration and apoptosis.Moreover, these molecules might play a direct role in the development of melanoma and contribute to the acquisition of drug resistance.As a result, lncRNAs hold promise as diagnostic and prognostic biomarkers for melanoma, and they may also serve as potential therapeutic targets in the future.

| CON CLUS ION
In conclusion, the current evidence shows that expression levels of some lncRNAs may vary in melanoma patients.Additionally, upregulation or downregulation of various lncRNAs is related to a patient's survival and melanoma prognosis.These findings suggest that lncR-NAs should be considered as novel diagnostic and prognostic biomarkers for better management of patients in the future.However, more investigations are required to determine the prognostic and diagnostic value of lncRNAs for clinical use.

9 *
indicate whether the study in a row received a score for each of the columns.
outcomes and fatalities linked to melanoma.Hence, the identification of disease-associated biomarkers holds significant therapeutic and prognostic implications, particularly in the context of advancedstage melanoma.Early detection and intervention in this form of F I G U R E 6 Forest plot of progression-free survival hazard ratios; prognostic evaluations were divided into two groups based on HR (less than or greater than one) and subgroups based on specimen type.
blood specimens (n = 19), involving 1107 melanoma cases and 381 controls, the Reitsma bivariate model indicated a cumulative Summary of findings in the meta-analysis.
identified a model based on the 12 pyroptosis-related lncRNA signature with the ability to predict the prognosis of cutaneous F I G U R E 4 Forest plot of pooled AUCs, showing blood specimen and tissue specimen subgroups and combined diagnostic values.TA B L E 3