Prognostic and clinicopathological value of high expression of TIM‐3 in different cancer types: A meta‐analysis

This meta‐analysis was performed to clarify the prognostic role of the expression of T‐cell immunoglobulin mucin‐3 (TIM‐3) in different cancer types.

biomarkers to predict prognostic outcomes for cancer survivors are urgently needed.

T-cell immunoglobulin mucin-3 (TIM-3) is mainly localized on
the T-cell surface as an immune checkpoint (co-inhibitory signal receptor). 3 Lots of studies demonstrated that TIM-3 played a pivotal role in immune modulation of cancers. [4][5][6] Moreover, previous research showed that aberrant TIM-3 expression was central to carcinogenesis and progression. 7 Furthermore, binding of TIM-3 to its ligand galectin-9 triggers Th1 cells died and induces peripheral tolerance, suggesting an inhibitory effect of TIM-3 in immune response. 7,8 In fact, TIM-3 is identified as an important negative regulatory factor of CD4+ and CD8+ T-cells. 9 Evidence shows that TIM-3 suppresses CD4+ T-cells activation by multiple mechanisms. 10 Moreover, overexpression of TIM-3 has a relationship with cancerspecific antigen CD8+ T-cell dysfunction in patients with malignant tumor. 11 These observations made us speculate TIM-3 as a new biomarker.
Accumulating studies have suggested that TIM-3 is overexpressed and is related to poor prognosis in different kinds of tumor, consisting of hepatocellular carcinoma, 12 lung carcinoma, 13 cervical carcinoma, 14 gastric carcinoma, 15 renal cell carcinoma, 16 bladder urothelial carcinoma, 17 colon carcinoma, 18 and so on. A previous meta-analysis including these cancers showed that TIM-3 was a potential therapeutic target and prognostic marker of patients with solid tumors. 19 However, more and more researches have been published to explore the function of TIM-3 in cancers over the past 5 years, and new evidences have been contrary to previously published studies. Specifically, Burugu et al found that high TIM-3 expression cancer tissue was associated with good prognosis in patients. [4][5][6] Meanwhile, previous research only discussed the prognostic significance of patients' overall survival (OS), 19 and so far, the relationship between high expression of TIM-3 and certain tumor types remains unclear.
Therefore, it is necessary to conduct an updated study to evaluate the prognostic value of TIM-3 in various cancer types. The objective of current research was to comprehensively and systematically assess the effects of high TIM-3 expression on OS and progressionfree survival (PFS) in different cancer types.

| Search strategy
Our research was performed basing on the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines. 20 The PubMed, EMBASE and Web of Science databases were thoroughly screened up to December 31, 2019. Keywords employed in the search were "TIM-3" AND "tumor OR carcinoma OR cancer OR neoplasm OR leukemia OR lymphoma" AND "survival OR prognosis OR progress OR progression OR outcome OR death," by searching "all fields." Furthermore, the references of qualified articles were also checked to identify eligible studies.

| Selection criteria
In our study, the inclusion criteria for a study's selection were: (a) studies analyzed the prognostic role of TIM-3 in different kinds of cancers, (b) cancers were confirmed by histologic or pathologic examinations, (c) with enough data to extract hazard ratios (HRs) and their 95% confidence intervals (CIs) for OS and/or disease-free survival (DFS) and/or PFS and/or recurrence-free survival (RFS), and (d) researches involved human participants only. In addition, the criteria for exclusion were as follows: (a) editorials or conference abstracts, letters to the editor, case studies or nonclinical studies, (b) articles were not in English languages, and (c) articles had duplicate data or repeat analysis. Apart from the above, only the most informative recent publication or the largest sample size manuscript were eligible for inclusion if the same samples were reported in different studies.

| Data collection and endpoints
All eligible articles were reviewed independently by two investigators (W.B.X. and R.D.J.) and any disagreement was resolved through a third investigator (F.Q.). Then, the following elements were gathered: year of publication, name of first author, patient's nationality, cancer types, number of patients, sex ration, age at diagnosis, number of patients with TIM-3 overexpression, detection methods, the site of TIM-3 0 expression, cut-off value, therapeutic strategy, follow-up, HRs, and 95% CIs for endpoints. The primary endpoints of this study were OS and DFE/PFS/RFS. If HRs with 95% CIs were not available directly, they were estimated from Kaplan-Meier survival curves by appropriate software (Engauge Digitizer version 10.9) according to the method described by Tierney's. 21 Importantly, multivariate analysis results were preferred because of the higher precision when compared with univariate analysis.

| Quality assessment
Two investigators (W.B.X. and R.D.J.) independently evaluated the quality of the included researches by the Newcastle-Ottawa-Scale (NOS) (http://www.ohri.ca/programs/clinical_epidemiology/oxford. htm). 22 The NOS criteria were scored based on three categories: subject selection (0-4 scores), comparability of subjects (0-2 scores), and outcome assessment (0-3 scores). Typically, a score of 6 or more points was deemed to indicate a high-quality research.

| Statistical analyses
Cochran Q test and inconsistency index (I 2 ) statistics were undertaken to assess the heterogeneity between enrolled researches. 23 If P < .10 or I 2 > 50%, showing that the heterogeneity was significant, a random-effect model (DerSimonian and Laird) was utilized. Otherwise, a fixed-effect model was applied instead. Possible sources of heterogeneity were analyzed by conducting subgroup analysis, sensitivity analysis and meta-regression. We evaluated the publication bias graphically using Begg's funnel plot and Egger's publication bias plot. 24,25 P < .05 (two-sided) was generally considered as statistically significant. All data analyses in this study were conducted using STATA v. 15.0 (StataCorp, Texas).

| TIM-3 expression and the association with clinicopathological factors
To determine the clinical prognostic value of TIM-3, we investigated the association between TIM-3 expression and clinicopathological factors in different cancer types (

| Meta-regression
For the sake of exploring the sources of heterogeneity between studies, a meta-regression was performed, 49 including outcome, cancer types, publication year, sample size, region, type of TIM-3, expression site of TIM-3, analysis of survival, and treatment modality. Table 2 displayed the results of meta-regression. And our results suggested these variables did not significantly contribute to the heterogeneity.

| Sensitivity analyses and publication bias
To confirm the reliability and stability of the results, sensitivity analy-  What is more interesting, in both solid and hematologic cancer, coblockade of the TIM-3 and Programmed death 1 protein (PD-1) pathways is better than block PD-1 pathway alone at promoting antitumor effect of the general immune system. [52][53][54] These reports pointed out high expression of TIM-3 to be a putative predictor related to poor OS and PFS. Therefore, above results indicated that TIM-3 may be a potential target candidate in different cancer types.
In the present study, our data showed that high TIM-3 expression had significantly shorter OS and PFS in cancer patients, but was insig-

| CONCLUSION
This study indicated that high TIM-3 expression might be a potential biomarker which can be used to predict the poor prognosis of different cancer types, especially osteosarcoma, gastric cancer, liver cancer, esophageal cancer, and lymphoma. However, further exploration was restricted due to the limited studies. Hence, randomized controlled trials and large sample size studies are urgently needed to verify our findings.

CONFLICT OF INTEREST
The authors have no conflicts of interest to declare.

ETHICS STATEMENT
Ethical approval or informed consent was not required for this metaanalysis.

DATA AVAILABILITY STATEMENT
The datasets used in this study are available from the corresponding author upon reasonable request.