Overexpression of stathmin promotes metastasis and growth of malignant solid tumors: a systemic review and meta-analysis

Stathmin has been investigated to be involved in development and progress of malignant tumors. This study was to clarify the relationship between expression of stathmin and tumors and assess its clinical significance. We identified 25 studies with a total of 3,571 individuals from the electronic bibliographic databases and strictly evaluated the quality and heterogeneity of included studies. We analysed the relationship between expression of stathmin and clinical characteristics by the fixed-effects and random-effects of meta-analysis and constructed a summary receiver-operator characteristic curve to estimate the test characteristics. The results showed that patients with cancer displayed a higher stathmin expression than those of non-cancer individuals (OR, 0.31), and overexpression of stathmin correlated with tumor cell differentiation (OR, 0.73), lymph node invasion (OR, 0.80) and high TNM stage (OR, 0.67). The pooled sensitivity of stathmin for distinguishing malignant tumors was 0.73 and the specificity was 0.77. The maximum balance joint for sensitivity and specificity (the Q-value) was 0.7566 and the area under the curve (AUC) was 0.8234. In conclusion, these results showed that overexpression of stathmin intimately correlated with malignant behavior of tumors, suggesting it could be a risk factor of malignant tumors. Stathmin had great sensitivity and specificity indicated it should be a significant molecular biomarker for malignant tumors.


INTRODUCTION
With increasing incidence and mortality, cancer has become a leading cause of death and major public health problem all over the world. It is reported that a total of 1,658,370 new cancer cases and 589,430 cancer deaths are projected to occur in the United States in 2015 [1]. Another investigation from China reports that an estimated 4,292,000 new cancer cases and 2,814,000 cancer deaths would occur in China in 2015 [2]. Medical studies suggest that understanding the molecular mechanism of tumors is very critical for improving the diagnosis and treatment. Especially, the level of some certain protein expression is associated with the prognosis and treatment of malignant tumors [3]. Therefore, it is very common and urgent medical problem to disclose new gene and protein expression level and molecular mechanism.
Stathmin (also known as Op18, p18, p19, stathmin 1 or metablastin), is upregulated in a variety of cancers and correlates with cell proliferation and migration of cancers, especially in malignant solid tumors [4,5]. The main function of stathmin seems to be a major cytosolic phosphoprotein that regulates microtubule dynamics by preventing tubulin polymerization and promoting microtubule destabilization [6]. Recent studies support a role for stathmin in the growth regulation of malignant tumor cells and indicate it is involved in malignant biological behavior of cancers [4,[7][8][9][10][11][12][13]. Thus, stathmin may be an attractive molecular biomarker and target for the diagnosis and treatment of malignant tumors. We aimed to review the evidence in the literature to date in Research Paper www.impactjournals.com/oncotarget order to (1) disclose the expression patterns of stathmin in malignant tumors, (2) clarify the relationship between stathmin expression and malignant tumors, and (3) evaluate the clinical value of stathmin in diagnosing and monitoring of malignant tumors.

Heterogeneity assessment of studies
The analysis of fixed effects model showed the value of Chi-square was 17.21 with 24 degree of freedom (P = 0.08), and the I-square value was 45.6, which indicated that there was no heterogeneity between the studies. Through reviewing the current literature, we also found that there was a very good clinical homogeneity. In the premise of homogeneity, we used fixed models to combine all indices.

Comparison of stathmin expression between cancer and normal tissues
Eleven studies showed the detailed data pertaining to the expressions of stathmin in cancer and normal tissues with control design [8-10, 13, 18-20, 23-25, 32], which included 783 cancer cases and 496 normal controls ( Table 3). As shown in Figure 2A, the overall odds ratio (OR) was 0.31 (95% CI = 0.25-0.39) via a fixed model analysis (Z = 9.70, p < 0.001), suggesting that expression of stathmin was remarkably higher in cancer tissues than in normal tissues.

Comparison of stathmin expression between early and advanced stage of cancers
Eighteen studies [6, 8-10, 12-15, 17, 18, 20, 21, 23, 25-28, 31] showed the detailed data pertaining to the expressions of stathmin in different clinical stages of cancers. Patients with cancer were divided into two groups according to the tumor-node-metastasis (TNM) classification: I-II stage (early stage) versus III-IV stage (advanced stage), including 1174 cancer cases of early stage and 1100 cases of advanced stage ( Table 3). As shown in Figure 2B, the overall OR was 0.67 (95% CI = 0.58-0.78) via a random model analysis (Z = 5.26, p < 0.001), suggesting that expression of stathmin was remarkably higher in patients with advanced stage than in those of early stage.

Comparison of stathmin expression between different differentiated degree of cancer tissues
As shown in Figure 3A, nineteen studies [6, 8-10, 12-15, 17, 18, 20, 21, 23, 25-28, 31] showed the detailed data pertaining to the expressions of stathmin in different differentiated degree of cancer tissues (well and moderate versus poor differentiated cancer tissues), including 1599 cancer cases of well and moderate differentiation and 775 cases of poor differentiation ( Table 3). The overall OR was 0.73 (95% CI = 0.62-0.86) (Z = 3.83, p < 0.0001), showing that expression of stathmin was significantly higher in poor differentiated cancer tissues than in those of well and moderate differentiated tissues.

Analysis of correlation between overexpression of stathmin and overall survival of cancer patients
Four studies reported the correlation between overexpression of stathmin and overall survival of cancer patients [16,18,21,23]. The results showed that patients who exhibited high expressions of stathmin had a significantly shorter post-surgical survival time (27.93 ± 11.54 months) compared with patients who exhibited moderate and low expressions of stathmin (44.81 ± 15.82 months). The t-value was 5.687 and the degree of freedom was three (P = 0.01). Table 2 showed the principal characteristics on study quality of the 25 studies. All studies were retrospective and all adopted the cancer tissues to detect the expression of stathmin. As shown in Table 4, most of 25 studies had added up to more than 6 stars of scores in evaluating with Newcastle-Ottawa Scale (NOS), which suggested that higher quality studies rigorously controlled for potential confounders [8,20,32] and did better with respect to selection of cases and controls [9,10,13,18,19,23,24]. Eleven studies showed the detailed data pertaining to the expressions of stathmin for diagnosis of malignant tumors. Figure 4A and 4B showed summary of methodological quality of included studies according the four best differentiating items from quality assessment of studies of diagnostic accuracy (QUADAS) checklist. Most of studies were evaluated as low risk, and high risk only accounted for less than 10%.

Sensitivity analysis and assessment of publication bias
Sensitivity analysis was performed by omitting each study from the estimated pool at every step, the exclusion of any one study individually did not substantially modify the estimators, with OR values vacillating between 0.23 and 2.21 ( Figure 4C) [33]. The shape of the funnel appeared to be approximately symmetrical ( Figure 4D). In addition, Begg's test indicated the Std. Dev. of Score was 28.58 and Pr > |z| = 0.421. ( Figure 4E). Therefore, the funnel plot and Begg test all suggested that publication biases did not have a significant influence on the results [33].

Sensitivity and specificity of stathmin expression in tumor tissues for the diagnosis of cancers
In order to assess the pooled diagnostic value based on the sensitivity and specificity of stathmin test in tissues for the diagnosis of cancers, we made the forest plots of sensitivity (true positive rate) and 1−specificity (false positive rate) for the 11 research reports [8-10, 13, 18-20, 23-25, 32]. The sensitivity of stathmin in tissues for the diagnosis of cancer was 0.73 (0.70 to 0.76) ( Figure 5A Figure 5C) and the negative likelihood ratio (NLR) was 0.35 (0.28 to 0.44) ( Figure 5D).

Diagnostic accuracy of stathmin test in tissues for the diagnosis of cancers
In our present meta-analysis, the mean diagnostic odds ratio (DOR) was 10.92, indicating that stathmin assay in cancer tissues could be helpful in the diagnosis of malignant tumors ( Figure 6A). The summary receiveroperator characteristic (SROC) curve help negotiate the balance between sensitivity and specificity of a concrete index in differentiating concrete disease [34]. In our analysis, the maximum balance joint for sensitivity and specificity (the Q-value) was 0.7566. The area under the curve (AUC) was 0.8234 ( Figure 6B).

DISCUSSION
Although stathmin is usually over-expressed in many human cancers, and intimately correlates with the development and progression of malignant tumors [4-11, 13-29, 31, 32]; up to now, we still know relatively little about the role of stathmin in malignant tumors. However, reports on the relationship between abnormal protein expression and cancers, as known, are becoming numerous. The identification of tumor marker has been proposed as a useful strategy to help deeply understanding of human tumorigenesis [33]. In present study, we disclosed the internal connection between abnormal expression of stathmin and malignant tumors and assessed the clinical value of stathmin for discerning malignant tumors.
We used the QUADAS-2 tool to assess included studies, which specially used for evaluating the quality of non-randomized control studies [35]. In addition, Newcastle-Ottawa Scale (NOS) is also a validated technique for assessing the quality of observational and non-randomized studies, which is considered an internationally recognised standard [36]. We employed these two tools to perform the evaluation of study quality and found most of included studies were low risk, suggesting that the conclusion from those studies were reliable and stable. Identifying heterogeneity is important to systematic reviews [37], in our study, we found included studies had a very good clinical homogeneity and statistical analysis showed that the heterogeneity was acceptable. In analysis of sensitivity, we found that the exclusion of any one study did not shake the overall conclusion. In assessment of publication biases, the shape of the funnel plot and Begg's test did not show the statistical significance on publication biases.
In our study, eleven studies compared the expressions of stathmin in cancer with normal tissues [8-10, 13, 18-20, 23-25, 32] and the results suggested that cancer tissues displayed a higher expression of stathmin than normal tissues. Thus, we may infer that higher stathmin expression Figure 1: Selection of studies for review. Studies were retrieved from the electronic bibliographic databases such as PubMed, Embase, Cochrane Library and SCI database. www.impactjournals.com/oncotarget plays a potentially role in the occurrence and development of cancer. Subsequently, we performed a series of subgroup analysis and found that stathmin displayed a higher expression in advanced cancers, poor differentiated cancers and cancers with lymphatic metastasis, which implied that stathmin intimately involved in tumor cell differentiation, proliferation and invasion of tumors. Metastasis and invasion are the critical factors in the progression of cancer, stathmin has been reported intimately correlating with malignant behavior of tumors [5,8,9,22,[38][39][40][41]  and affect the survival of tumor patients. In our study, we noticed that patients who exhibited high expressions of stathmin had a significantly shorter post-surgical survival time (27.93 ± 11.54 months) compared with patients who exhibited moderate and low expressions of stathmin (44.81 ± 15.82 months). This phenomena strongly indicated that stathmin may have a potential to become prognostic index for predicting the survival rate of patients with cancer.
In meta-analysis of diagnostic accuracy, we found that the overexpression of stathmin had impressive sensitivity (0.73, 0.70 to 0.76) and specificity (0.77, 0.73 to 0.81) when discriminating the cancer from normal cases, which meant stathmin had a potential to discern cancers. The positive likelihood ratio (PLR) of 3.31 (2.35 to 4.66) suggested that patients with cancer have a nearly 3.31-fold higher chance of being stathmin test positive compared with patients without cancer. This ratio suggested a potential role for stathmin in discerning malignant tumors. Similarly, the negative likelihood ratio (NLR) of 0.35 (0.28 to 0.44) also indicated that low expression of stathmin may help exclude non-cancer individuals. Theoretically, the value of a DOR ranges from zero to infinity, with higher values indicating better discriminatory test performance (higher accuracy) [42]. In our meta-analysis, the value of DOR was 10.92, indicating that the detection of stathmin could be useful to distinguish malignant tumors. The summary measure of test characteristics derived from the SROC curve where sensitivity equalled specificity was 82%, which suggested that the test performance of stathmin in discerning cancer tissues is reasonably good, and it would be a useful tool for the judgement of malignant tumors.
So far, we still know relatively little about how the stathmin regulates tumor proliferation, motility, migration and occurence of metastasis at the molecular level. However, the abnormal expression of stathmin in human tumors has provided a clear possibility for the  development of stathmin-dependent molecular diagnosis and targeting therapy. However, we also found some limitations. First, the bias from different tumors may exist because stathmin is highly expressed in many kinds of human cancers. Second, diagnostic review bias may occur because stathmin itself may affect the determination of diagnosis in reports using the clinical course as the gold standard [43]. In the future, it is very crucial to investigate stathmin expression in cases with large samples and multiple clinical centers. Although some deficiencies existed in the studies reviewed, they still provided credible evidence that the stathmin plays an important role in malignant tumors and its detection may be useful for the diagnosis of malignant tumors. In summary, overexpression of stathmin involved in tumor differentiation, lymph node invasion and high TNM stage, suggesting that high expression of stathmin play an important role in malignant tumors. Stathmin had a   relatively good sensitivity and specificity suggests that it should be a molecular biomarker for the diagnosis and target for therapy of malignant tumors.

Searching of literature
We searched the published literature from the databases of PubMed, Embase, Cochrane Library and SCI database (from the start of each database up to May 2016). We used the following free text and Medical Subject Heading terms such as "cancer," "tumor," "carcinoma," "tumors," "neoplasms," "malignant neoplasms," "malignant tumors," "Op18," "p18," "p19," "stathmin 1," "stathmin", "metablastin", "expression," and "diagnosis." We also performed hand searches from the references of the articles. If necessary, we contacted with the authors of articles by e-mail for detailed information.

Inclusion and exclusion criteria of literature
The exclusion criteria: (1) must have showed the data on stathmin expression in malignant tumors; (2) must be case-control or cohort association studies; (3) must have clearly described the control; and (4) the number of cases must be greater than or equal to 50. The exclusion criteria: (1) not original articles such as abstracts, letters, editorials and expert opinions and case reports; (2) did not clearly report sufficient data of stathmin expression; (3) patients who were involved in studies had received chemotherapy, radiotherapy and other drug treatment before taking samples; (4) control group was absent; and (5) non-human studies.

Extraction procedure of key data from literature
Two researchers extracted the useful data independently. They checked the validity of original data, and contacted the authors of articles if it was necessary. When stathmin expression was tested by different measures in one study, we adopted the results from acknowledged and common measurement. If the paper showed stratum data, we collected all strata data to utilize as full as possible. If two abstracters have disputes on some data, the third abstracter would check the data again, and finally reached consensus through discussion.

Data of extraction
General information included: authors, countries, publication date, gender and age of patients, type of tumors, testing method of stathmin and study design. Critical data included: case number of different groups, histological classification, tumor node metastases (TNM) classification, tumor differentiation degree, concrete data of stathmin expression (number of true positives, true negatives, false positives, and false negatives).

Quality assessment of studies
We evaluated the quality of primary studies using the Newcastle-Ottawa Scale (NOS), which uses a star system to evaluate observational studies based on three criteria: participant selection, comparability of study groups and assessment of outcome or exposure [44]. We also used the Quality Assessment of Diagnostic Accuracy Studies (QUADAS -2) tool [35] to assess the quality of diagnostic accuracy. The QUADAS -2 tool uses 14 items to assess risk of bias of study, each item is rated "yes," "no," or "unclear" [35], and is phrased such that "yes" indicates low risk of bias.

Statistical analysis
We conducted a series of analysis according to the standard methods recommended for a meta-analysis of diagnostic test evaluations [45]. The odds ratios (OR) and their 95% confidence intervals (CI) was calculated directly using two different meta-analysis approaches (fixed-effects model and random effects model) according to heterogeneity. We calculated the Chi-square value and I 2 to evaluate heterogeneity between studies. In the absence of statistically significant heterogeneity, we used the fixed effects method to combine the results. If the heterogeneity existed, we used the random effects method. The overall effect was tested using Z-scores, with significance being set at p <0.05. We run sensitivity analysis to evaluate whether single study should affect the overall effects and evaluated publication bias through funnel plots, Begg's and Egger's test respectively. If the data were available, we would combine the following variables of test accuracy: sensitivity; specificity; positive likelihood ratio (PLR); negative likelihood ratio (NLR); and diagnostic odds ratio (DOR), which were derived from summary receiveroperator characteristic (SROC) curve. Statistical analyses were performed using SPSS (SPSS Institute, version 22.0, Chicago, USA), RevMan 5.3.3 (Cochrane Collaboration), Meta DiSc statistical software (Version 1.4, Madrid, Spain), and Stata version 14.0 (Stata Corporation, TX, USA). All the tests were two-sided and the significant level was 0.05.