Prognostic role of galectin-3 expression in patients with solid tumors: a meta-analysis of 36 eligible studies

Galectin-3 as a β-galactoside-binding protein, has been found to be involved in tumor cell growth, anti-apoptosis, adhesion, angiogenesis, invasion, and distant metastases, indicating that it may play a pivotal role in cancer development and progression. However, their results remain debatable and inconclusive. Hence, this meta-analysis was performed to clarify the precise predictive value of galectin-3 in various cancers. PubMed, Web of Science, Embase, Cochrane Library, CNKI and Wanfang databases were searched comprehensively for eligible studies up to July 15, 2018. Pooled hazard ratios (HRs) with 95% confidence intervals (CIs) of OS or DFS/PFS/RFS were calculated to demonstrate their associations. A total of 36 relevant studies were ultimately enrolled in this meta-analysis. Our results shed light on the significant association of elevated galectin-3 expression with reduced OS or DFS/RFS/PFS in overall cancer patients (pooled HR = 1.79, 95% CI 1.42–2.27, I2= 67.3%, p < 0.01; pooled HR = 1.57, 95% CI 1.04–2.37, I2= 67.1%, p = 0.001). In tumor type subgroup analysis, we found high expression of galectin-3 was correlated with shorter OS or DFS/RFS/PFS in colorectal cancer (pooled HR = 3.05, 95% CI 2.13–4.35, I2= 0.0%, p = 0.734; pooled HR = 2.49, 95% CI 1.82–3.41, I2 = 0.0%, p = 0.738; respectively) and meanwhile it merely associated with reduced OS in ovarian cancer or non-small cell lung cancer (pooled HR = 2.24, 95% CI 1.38–3.64, I2= 0.0%, p = 0.910; pooled HR = 2.07, 95% CI 1.48–2.88, I2= 0.0%, p = 0.563; separately). Taken together, our results suggested that galectin-3 played an oncogenic role in colorectal cancer, ovarian cancer and non-small cell lung cancer, indicating it could be a promising biomarker and a novel therapeutic target for them. Further studies were warranted to validate our findings.

Galectin-3 has been found to be differently expressed in various normal and malignant tissues. Previous studies indicated that down-regulation of galectin-3 was associated with loss of the transformed phenotypes in thyroid papillary carcinoma cells, but up-regulation of it could induce the transformed phenotype in normal thyroid follicular cell lines [9]. Accumulating data have demonstrated that different galectin-3 expression in tumor tissues was associated with unfavorable survival in cancer patients [10][11][12][13][14]. These studies concentrated on colorectal carcinoma, cervical carcinoma, breast cancers, gastric carcinoma, laryngeal squamous-cell carcinoma and so on. However, their results remained inconsistent. The discrepancies among these studies highlighted the importance of evaluating the prognostic significance of galectin-3 in multiple human malignant neoplasms. Hence, this meta-analysis was conducted to clarify the relationship between galectin-3 expression and the prognosis of patients with carcinoma. Last but not least, it is the first time for us to shed light on their relationship and galectin-3 is anticipated to be a prognostic marker in clinical applications.

Literature search strategy
We conducted a comprehensive search of online databases PubMed, EMBASE and Web of Science, Cochrane Library, Chinese National Knowledge Infrastructure (CNKI) and Wanfang database (Chinese) to identify relevant literature published before July 15, 2018. The search strategy was mainly consisted of the following keywords in combination with Medical Subject Headings (MeSH) terms and text words: ("cancer" or "carcinoma" or "neoplasm" or "tumor" or "tumour") and ("galectin-3" or "GAL3" or "LGALS3" or "L31" or "MAC2" or "CBP35" or "GALBP" or "GALIG"). In addition, potentially eligible articles were identified via meticulously searching from the reference lists of relevant reviews and original literature.

Inclusion and exclusion criteria
The eligible studies needed to meet the following four inclusion criteria: (1) English or Chinese publications; (2) patients with carcinoma; (3) a relationship of galectin-3 expression with cancer prognosis; (4) sufficient data could be extracted. Additionally, the exclusion criteria included the following points: (1) non-English or non-Chinese research; (2) duplicates of the previous publication; (3) reviews or letters or case reports or comments or editorials; (4) unrelated to galectin-3 or human patients; (5) absence of key information.

Quality assessment
The following information should be extracted from included articles before being evaluated: (1) the study population and country; (2) the study design; (3) assay method to determine galectin-3 expression; (4) the prognosis or survival assessment; (5) the detected tumor and pathology information; (6) the cutoff point of galectin-3; and (7) the follow-up duration. In addition, Newcastle-Ottawa Scale (NOS), as one of the most useful scale to evaluate the quality of non-randomized studies (http://www.ohri.ca/progr ams/clini cal_epide miolo gy/ oxfor d.htm), was independently evaluated by two blind reviewers [15]. The criteria of quality assessment were as follows: (1) representativeness of the exposed cohort; (2) selection of the non-exposed cohort; (3) ascertainment of exposure; (4) outcome of interest not present at start of study; (5) control for important factor or additional factor; (6) assessment of outcome; (7) follow-up long enough for outcomes to occur; (8) adequacy of follow up of cohorts. Total quality score of NOS was ranged from 0 to 9, which was regarded as high quality with the final score > 6. Details were presented in Table 1.

Data extraction
All available data from the identified studies were extracted respectively by two reviewers (Y.W and SW.L). If any disagreement achieved, a third reviewer (Y.T) would join in and reached a consensus. Extracted data were recorded in a standardized form including following items: first author's surname, publication year, patients' median or mean age, nationality, dominant ethnicity, number of patients, investigating method, cutoff value, follow-up time, and hazard ratios (HRs) for prognostic outcomes (overall survival [OS] and disease/recurrence/progression-free survival [DFS/RFS/ PFS]) along with their 95% CI and p-values. Data were extracted from Kaplan-Meier curves to extrapolate HRs with 95% CIs by using previously described methods, when it could not be directly obtained from each article [16,17]. Details of the aforementioned data were displayed in Tables 2 and 3.

Statistical analysis
Based on available data, the relationship between galectin-3 and multiple human malignant neoplasms was conducted by OS or DFS/RFS/PFS and the pooled hazard ratios (HRs) with 95% confidence intervals (CIs) were utilized to evaluate their efficacy. The effect of heterogeneity was quantified via I 2 = 100% × (Q − df )/Q. If significant heterogeneity (p < 0.1 or I 2 > 50%) existed, the random-effects model (DerSimonian-Laird method) would be applied; otherwise, a fixed-effects model (Mantel-Haenszel method) would be utilized [18]. Moreover, in the case of significant heterogeneity, subgroup analysis was carried out by the type of malignant disease and dominant ethnicity to further minimize the influence. Sensitivity analysis was conducted to access the stability of results by deleting one single study each time to reflect the impact of the individual to overall. Publication bias was evaluated by the Begg's funnel plot and Egger linear regression test with a funnel plot [19]. If p < 0.05, it indicated the existence of publication bias. All p-values were calculated using a two-sided test and p < 0.05 was considered statistically significant. Besides, all statistical data were conducted by Stata software  Table 1 Newcastle-Ottawa quality assessments scale 1. Representativeness of the exposed cohort; 2. Selection of the non-exposed cohort; 3. Ascertainment of exposure; 4. Outcome of interest not present at start of study; 5. Control for important factor or additional factor; 6. Assessment of outcome; 7. Follow-up long enough for outcomes to occur; 8. Adequacy of follow up of cohorts

0.01
AdenoCa adenocarcinoma, SqCa squamous carcinoma, OS overall survival, DFS disease-free survival, PFS progression-free survival, RFS recurrence-free survival, NM not mentioned, IRS immunoreactivity score, Hscore the intensity and respective percentage cells that stain at each intensity were multiplied to reach a Hscore that ranged from 0 to 300, OS overall survival, HR hazard ratio, CI confidence interval a, b Data extracted from one study due to different malignant disease (biliary cancer and pancreatic cancer)

Summary of enrolled studies
The literature search yielded 1109 citations through online databases by means of previous search strategy. Amongst them, 970 records were excluded because of reviews, letters, case-reports, duplicates and so on, after screening the tittles and abstracts. The full texts of the remaining 139 articles were evaluated by the reviewers. Among them, 103 potentially suitable studies were excluded because of lacking sufficient survival data (HRs and 95% CIs), not related to OS or DFS/RFS/PFS, absence of key information. Ultimately, 36 studies were considered to be eligible for this meta-analysis (Fig. 1) [4,[10][11][12][13][14].
Detailed quality assessments of each eligible article were presented in Table 1

DFS/RFS/PFS associated with galectin-3 expression
A total of 11 original studies were included to evaluate the role of elevated galectin-3 expression in patients with various solid tumors by DFS/RFS/PFS, within a random-effects model. Our results successfully identified the significant association of high galectin-3 expression with reduced DFS/RFS/PFS in overall cancer patients (pooled HR = 1.57, 95% CI 1.04-2.37, I 2 = 67.1%, p = 0.001; Fig. 3a). In the subgroup analysis of specific cancer type, we found that high expression of galectin-3 was correlated with shorter DFS/RFS/PFS in colorectal cancer (pooled HR = 2.49, 95% CI 1.82-3.41, I 2 = 0.0%, p = 0.738; Fig. 3b). However, In terms of dominant ethnicity subgroup analysis, both the Asian and Caucasian ethnicity were not statistically significant (Fig. 3c).

Sensitivity analyses
In order to determine the robustness and the stability of our results, sensitivity analysis was conducted to access the stability of results by deleting one single study each time, to reflect the impact of the individual to overall. Our results indicated that no single study significantly influenced the pooled OR and 95% CIs. Namely, our results are comparatively reliable and stable (Fig. 4).

Publication bias
The combined application of Begg's and Egger's test was utilized to evaluate the publication bias and meanwhile the funnel plots were displayed in Fig. 5. In the pooled analysis of OS or DFS/RFS/PFS, the p values of Begg's test and the p values of Egger's test were all above 0.05, indicating no publication bias in this study.

Discussion
Up to now, elaborate efforts have been made to establish reliable and convincing evidence to detect promising biomarkers for patients with solid tumors. Galectins, as a family of animal carbohydrate-binding proteins, which had the ability to agglutinate cells, were considered to be potential biomarkers of cancer prognosis given their unique structure and functions into consideration [50,51]. Over the past years, galectins have been implicated in the development of cancer, the pathogenesis of heart failure and ventricular remodeling, infectious processes, and inflammatory processes [52]. Amongst them, due to its differential expression between cancer and normal tissues, galectin-3 was regarded as one important member of galectins family. However, the definite role of galectin-3 in various human malignant neoplasms remained inconsistent. Hence, this meta-analysis was conducted to clarify this question.
It was the first time for us to shed light on the association between elevated galectin-3 expression and the prognosis of patients with solid tumors. Meanwhile, our results were the systematic evaluation of the prognostic outcomes (OS or DFS/RFS/PFS) in a larger population. Our results did suggest that galectin-3 play an oncogenic role in overall cancer patients. Moreover, we found that high expression of galectin-3 was correlated with shorter OS or DFS/RFS/PFS in colorectal cancer and meanwhile it merely associated with reduced OS in ovarian cancer or non-small cell lung cancer, indicating that it could be a promising biomarker and a novel therapeutic target for them. Furthermore, in subgroup analyses of dominant ethnicity, we observed that both the Asian and Caucasian ethnicity were statistically significant for OS, suggesting that the detection of high galectin-3 expression in these patients might be useful for prognosis prediction. Besides, the outcomes of us shed light on that no matter galectin-3 in the tissue or in the plasma, its role remained stable, indicating it could be a promising biomarker and a novel therapeutic target. Meanwhile, according to the results of sensitivity analyses and publication bias, no single study significantly influenced the pooled OR and 95% CIs and no obvious publication bias was detected in this meta-analysis, indicating the robustness and the stability of our results.
Previous researches indicated that increased expression of galectin-3 often predicted unfavorable outcomes and the level of galectin-3 was positively correlated with invasion of depth, vessel invasion, lymph node metastasis, distant metastasis, and TNM stages of various cancers [26,53]. Tao et al. [37] demonstrated that the positive expression of galectin-3 was associated with more malignant biological behavior of colorectal cancer and it could be used as a predictor of poor prognosis for patients. As for tongue carcinomas, Honjo showed that cytoplasmic galectin-3 expression increased during the progression from normal to cancerous states, whereas nuclear galectin-3 expression decreased during the progression from normal to cancerous states, indicating that enhanced expression of cytoplasmic galectin-3 could serve as a predictor of disease recurrence in these patients [33].
As for its relevant mechanisms, several studies found that galectin-3 was expressed in both cytosol and nucleus [10,54]. Therein as an important regulator of the Wnt/ β-catenin signaling pathway, galectin-3 could activate the epithelial-mesenchymal transition (EMT) in tumor cells to promote the invasion and metastasis of cancer [55,56]. Furthermore, it could subsequently activate the Ras-mediated Akt signaling pathway to inhibit cell apoptosis by interacting with the activated GTP-bound K-Ras [57]. Besides, it could also modulate VEGF-and bFGFmediated angiogenesis by binding its carbohydrate recognition domains (CRDs) to integrate αvβ3, and then promote the growth of new blood vessels [58]. As for the effects on heterogeneity, subgroup analysis was a way to discover their potential sources and even decrease the huge heterogeneity. As presented by our results, we could easily find that there might be the existence of significant heterogeneity of elevated galectin-3 expression in the overall cancer patients. So we conducted a subgroup analysis based on the specific cancer types and found that most of their heterogeneity decreased significantly, even with no heterogeneity. However, subgroup analysis of dominant ethnicity was not associated with significant reduction of heterogeneity, indicating that the dominating source of heterogeneity might be the different cancer types.
Sometimes, galectin-3 combined with another biomarker was often utilized simultaneously in prognostic outcome analyses, showing it might not be an independent factor affecting the prognosis of cancer patients. As indicated by Li et al. [11] the expressions of ezrin and galectin-3 were correlated with the development of cervical cancer, and over-expressions of those proteins were indicative of poor prognosis in patients with cervical cancer. Galectin-3 associated with cyclin D1 expression was also studied in non-small cell lung cancer. As a result,  no important correlations with clinicopathological findings and no prognostic values were revealed between them. However, higher cyclin D1 expression was found in galectin-3 negative tumor tissues and the differences in correlations between their expressions in two main histopathological types of non-small cell lung cancer were also discovered [38].
The strength of this study was our broad search strategy with few restrictions to minimize any potential publication bias. Moreover, this was the first meta-analysis reporting the prognostic value of galectin-3 for cancers in the medical literature, which could provide some references for clinical work. Although this meta-analysis was performed with rigorous statistics, our conclusion still had several limitations for the following reasons. Firstly, different studies had their own varied expression cut-off values, which brought many difficulties for us to define the standard cutoff value, resulting in bias in the results of the effectiveness of galectin-3 as a prognostic factor in cancer patients. Secondly, heterogeneity existed in the total OS and DFS/RFS/PFS group and it was likely due to the different characteristics of the patients, such as the age, cancer type, different method in detecting samples and the varied cut-off values of galectin-3 expression. Thirdly, due to the insufficient studies, correlation between galectin-3 and OS or DFS/RFS/PFS in other tumor types has not been further analyzed. Fourthly, some essays studied galectin-3 combined with another biomarker in prognostic outcome analyses, showing galectin-3 was not an independent factor affecting the prognosis of cancer patients. Last but not least, all of these enrolled studies were derived from retrospective or observational data, which could not have a clear impact on group baseline features as RCTs. Upcoming prospective RCTs were required to provide more available data. Taking these aforementioned limitations into consideration, our results could be interpreted rigorously and meanwhile more well-designed studies were required to verify our findings.

Conclusions
In summary, it was the first time for us to shed light on the prognostic role of elevated galectin-3 expression in various cancers. Our results did suggest that galectin-3 played an oncogenic role in colorectal cancer, ovarian cancer and non-small cell lung cancer, indicating that it could be a promising biomarker for predicting the prognosis of patients with malignant neoplasms, and the biological functions of galectin-3 were of great research value of the subject. Due to the aforementioned limitations, larger samples of more strictly designed studies were required to provide more high-quality data to elaborate their associations.