Prognostic and clinicopathological significance of MUC expression in head and neck cancer: a systematic review and meta-analysis

The prognostic value of mucins expression in patients with head and neck cancer (HNC) remains controversial. To address this, a meta-analysis was performed to systematically evaluate prognostic significance of mucins expression in HNC. Electronic and manual searches were performed and a total of 20 studies including 2046 patients were selected for the final analysis. Increased mucins expression was associated with unfavorable overall survival in HNC patients (HR=1.83, 95% CI: 1.43-2.33, p=0.000). Mucins overexpression was also in correlation with more advanced TNM stage (RR=0.84, 95% CI: 0.73-0.97, p=0.017), higher risk of lymph node metastasis (RR=0.69, 95% CI: 0.57-0.84, p=0.000) and deeper invasion (RR=0.58, 95% CI: 0.44-0.76, p=0.000). These results suggested that elevated mucins expression was significantly associated with worse prognosis and more detrimental clinicopathological outcomes, revealing the promising potential of mucins as biomarkers for HNC management.


INTRODUCTION
Head and neck cancer (HNC) is a group of biologically similar cancers originating from the oral cavity, nasopharyngeal, oropharynx, hypopharynx and larynx, which mainly behave as squamous cell carcinoma histologically. HNC is the sixth most frequent type of malignant tumor, causing more than 400,000 deaths annually worldwide [1][2][3][4][5]. Its high mortality rate and the disfigurement or functional deficiency that survivors may suffer result in a considerable global public health burden. Despite great advance in multidisciplinary combined diagnosis and treatment, only 30-50% patients with HNC survive over 5 years after initial diagnosis worldwide [6]. Clinically, the classic tumor, node, and metastasis (TNM) staging system is widely used for the initial diagnosis but failed to reflect the inherent biological heterogeneity especially in atypical early symptoms or concealed metastasis patients. Therefore, novel biomarkers involved in cancer development are greatly needed to stratify patients with poor prognosis of HNC in order to make optimal individualized therapy.
Much attention has been focused on the involvement of mucins (MUC) in tumor carcinogenesis and metastasis recently. MUC is a family of high O-glycosylated protein, characterized by a basic structure including a central polymorphic tandem repeat region [7][8]. Only expressing on the apical surfaces of various luminal and glandular normal epithelial cells, MUC play an important role in cell-cell adhesion, immune response and alteration of intracellular signaling [9]. However, the tightly regulated homeostatic expression may be disrupted by various factors such as cancer cells, in particular. The observation of subcellular distribution and biochemical features Meta-Analysis changes during malignant transformation and tumor progression suggest that MUC may be the key point in carcinogenesis and subsequent metastasis of cancers [10][11][12][13][14]. Therefore, aberrant MUC expression may be predictive biomarkers in HNC.
Over the past decade, numerous independent studies have evaluated the clinical and the prognostic value of MUC protein expression in HNC. Yet, the results of these reports remain controversial and no clear consensus has been achieved so far [15][16][17][18][19]. Limited to small sample size some publications may draw inconsistent results due to potential random errors. Therefore, we conducted a systematic review and meta-analysis to address the association between MUC expression and prognostic value and the common clinicopathological parameters of HNC.

Demographic characteristics of included studies
Altogether, the sample-size ranged from 28 to 357, with a median of 103.2 participants. The median age of patients in all studies was 55.6 years old (range, 42.3-66.3) and the mean proportion of male patients was about 52.9%. The matching criteria varied among the enrolled studies. Studies concerning MEC and PTC occupied the largest proportion of cancer types among all primary literatures (n = 5, respectively), followed by HNSCC (n = 4), OSCC (n = 3) and remaining types of solid neoplasm. The MUC1 and MUC4 expression were evaluated by IHC staining in paraffin-embedded tissue blocks using several monoclonal antibodies as DF3, Ma695, VU4H5 for MUC1 and 1G8, 8G7, 15H10 for MUC4 respectively. As for the clinicopathological factors, seventeen eligible studies were divided into seven subgroups: 5 for age, 11 for gender, 10 for TNM tumor stage, 11 for tumor size, 13 for lymph node metastasis, 12 for tumor differentiation and 6 for mode of invasion. With respect to the quality of the including studies, none of the eligible entries scored less than six stars by the Newcastle-Ottawa Scale system, indicating a relative high methodological quality across all studies. Other detailed information were extracted and summarized in Table 1.

Impact of MUC expression on survival rates of patients in HNC
Seven observational trials including 993 participants offered original data on survival rates in terms of different MUC expression (Table 2). It demonstrated that increased MUC activity was associated with unfavorable survival rate (HR = 1.83, 95%CI: 1.43-2.33, P < 0.001). There was no significant heterogeneity among these studies (I 2 = 0.0%, p (Q-test) = 0.577), so that a fixed-effect model was used to combine the HR and 95% CI ( Figure 2).
Furthermore, subgroup analysis stratified by nationality, cancer type, adjuvant therapy, sample size, methods, cut-off value, MUC subtype, tumor location, antibody for MUC1 and antibody for MUC4 were performed respectively.

Sensitivity analysis
Pooled estimates of the relation of MUC overexpression to prognostic outcomes and clinicopathological parameters were not substantially altered according to the 'leave-one-out' method, demonstrating the reliability of our results ( Figure 4A-4H).

Publication bias
Analysis of survival rate and clinicopathological features demonstrated no obvious asymmetry in the funnel plots for publication bias (p > 0.05) ( Figure 5A-5H). More sensitive Egger's regression test confirmed these results, indicating that our pooled results had no significant publication bias.

Trial sequential analysis (TSA)
Data from twenty trials (2046 cases) were used to investigate the reliability of MUC status as a candidate for predicting the prognosis of HNC ( Figure 6A-6C). Using the relevance between MUC expression and cervical lymph node metastasis of HNC (including 13 trials with 1101 patients) as an example, the required information size (RIS) for adequate power was 1790 subjects. The cumulative z-curve crossed both the conventional boundary and the trial sequential monitoring boundary before reaching the RIS, indicating that our findings were conclusive and further trials seem to be unnecessary ( Figure 6B). Similar method was applied for other groups which were not shown. On the whole, potent evidence suggested that the overexpression of MUC predicted a worse outcome.

DISCUSSION
As revealed by the laboratorial evidence, aberrant overexpression, mislocation profiles and truncated glycans of tumor-associated MUC (TA-MUC) were commonly observed in variety of epithelial cancers. The structural and functional complexity of TA-MUC emphasizes its pivotal value in the pathogenesis and progression of cancer. However, some conflicting conclusions were reported in clinic. Although the majority of studies demonstrated that high level of MUC indicated worse clinicopathological parameters and poor prognosis, there are still a number of studies which showed the opposite conclusion [16,17,19]. Thus, a comprehensive study is urgently demanded.
To our knowledge, the present study is the first and most full-scale meta-analysis systemically exploring the possible prognostic role of MUC up-regulation in HNC. Here, 20 eligible studies including 2064 cases were combined to yield statistics, indicating that higher MUC expression was strongly related to poorer overall survival (HR = 1.83, 95% CI: 1.43-2.33, p = 0.000) in patients with HNC. Subgroup analysis was consistent with the pooled OS regardless of cancer type, sample size, staining cut-off value and MUC subtypes. Among patients accepted adjuvant therapy (AR and/or AC) along with surgery operation, no significant association was identified between positive MUC expression and poor OS, exhibiting a sharp contrast with the patients without AT (HR = 1.96, 95% CI: 1.51-2.53, p = 0.000). The discrepancy may be, partially, explained by the advanced radiotherapy technique and platinum-based therapeutic regimen for advanced HNC, achieving a significant improvement in 5-year OS [41]. As for western countries and ISH subgroups stratified by nationality and detecting methods respectively, no significant association was identified though a tendency was shown. This situation was possibly due to the relatively limited studies enrolled in the subgroups. As shown in Table 2, only three qualified studies for western countries and one for ISH detecting method studies were subjected to subgroup analysis, which may affect the real results. Based on these points, large-scaled studies are still needed to strength our results in the future. Stratified by the location of tumor, aberrant expression of MUC was significantly associated with poor OS in OSCC subgroup, which exhibited a sharp contrast with the patients that suffering from LSCC. The discrepancy may be, partially, explained by the morphologic diversity between oral cavity and laryngeal regions combined with different carcinogens. The inherent biological heterogeneity induces different disease evolution process though both of them originate from epithelial cells. Varying results were noted with regard to different antibodies used for detecting MUC1 or MUC4 specificity. DF3 for MUC1 may be a more tumorspecific antibody in immunohistochemistry than others  (i.e. VU4H5) in HNC especially for paraffin-embedded tissues, as the same to 8G7 for MUC4. Perhaps the best explanation for staining sensitivity differences among these antibodies may be the different binding epitopes. For example, aberrantly underglycosylated TA-MUC1 tissues may somehow mask the carbohydrate epitopes recognized by VU4H5, while still permitting reaction with DF3 which binds to peptide epitopes within the variable number of tandem repeat (VNTR) domain [42]. Although different survival rate were exhibited by these antibodies, the same tendency revealed that the increased expression of MUC protein was associated with unfavorable OS in patients with HNC.
Consistent with the theoretical inference, our results also demonstrated that overexpression of MUC was tightly associated with advanced TNM stage (I+II vs.III+IV, RR = 0.84, 95%CI: 0.73-0.97, p = 0.017 ), high risk of cervical lymph node metastasis (negative vs. positive, RR = 0.69, 95%CI: 0.57-0.84, p = 0.000 ) and the depth of invasion (1 to 3 vs. 4c+4d, RR = 0.58, 95%CI: 0.44-0.76, p = 0.000 ). Given that more advanced TNM stage, more positive cervical lymph node metastasis and deeper invasion are adverse prognostic features, the pooled analysis may explain why positive MUC expression was associated with poor survival rate in patients with HNC. Several studies have uncovered unique roles of the MUC on the oncogenic and pro-metastasis effects, including promoting proliferation, metabolism, angiogenesis, invasion, metastasis, epithelial to mesenchymal transition (EMT) and resistance to apoptosis [9,14,[43][44][45][46][47]. For example, multiple studies provide direct evidence for the role of MUC1 in EMT process, which enable cancer cells acquire their invasive and metastasis potential. Notably, MUC1 can directly inhibit the expression of E-cadherin, but upregulation the EMT inducers, such as Snail, Slug, Vimentin and Twist [43][44]. Similarly, MUC4 can specifically potentiate phosphorylation of the receptor tyrosine kinase ErbB2, which results in up-regulation of the cyclin-dependent kinase inhibitor p27 kip and silence of protein kinaseB/Akt pathways [47]. Taken together, these mechanisms can explain the association between high MUC expression and adverse clinicalpathological features in HNC patients. Therefore, MUC can serve as a good candidate for predicting the status and prognosis of HNC, making up the deficiency of the current physical detecting methods.
Apart from the inspiring outcomes, several limitations still existed in this meta-analysis. First, the lack of unified standardized protocol and evaluation system for detecting MUC status influence the accurate estimation of prognosis for HNC. Second, the numbers of studies and patients pooled into analysis were limited due to insufficient data. Third, the missing information from negative studies which were less frequently published could lead to publication bias, although significant heterogeneity was not detected from the current analysis.
Finally, despite the usage of random-effect model and subgroup analysis, some heterogeneity still existed, which may weaken our pooled conclusions.
In conclusion, despite the abovementioned limitations, our meta-analysis indicated that MUC expression was significantly associated with poorer OS, more advanced TNM stage, higher risk of cervical lymph node metastasis, and deeper invasion in patients with HNC for the first time. These findings revealed that the status of MUC could not only distinguish normal and precancerous cells but also could differentiate of a less metastatic cancer form its highly aggressive form. Therefore, patients with highly expression of MUC in HNC may need more radical treatment and rigorous monitoring, especially for those without typical symptoms or signs of metastasis. However, prospective clinical trials that are well-designed, using standardized methods, with long-term follow up, are required to verify the MUC hypothesis in HNC directly.

Search strategy
This meta-analysis was conducted according to the PRISMA statement [20]. A systematic literature search of PubMed, Embase, and Web of Science database was performed with the following strategy: ("mucin-1" OR "muc-1" OR "episialin") AND ("oral" OR "mouth" OR "tongue" OR "gingival" OR "pharynx" OR "larynx") AND ("tumor" OR "cancer" OR "carcinoma" OR "neoplasm" OR "malignant") AND ("prognostic" OR "prognosis" OR "outcome" OR "survival") up to September 1, 2016. Citation lists of retrieved articles, including review articles, were additionally reviewed to guarantee the sensitivity of the search process.

Inclusion/exclusion criteria
The criteria for inclusion were listed as following: (1) studies detected MUC expression by immunohistochemistry (IHC) or in situ hybridization (ISH) analysis in human cancer tissues; (2) sufficient information was provided to evaluate the relationship between MUC expression and clinicopathological parameters and/or cancer prognosis; (3) the hazard ratios (HRs) with 95% confidence intervals (CIs) of overall survival (OS) were provided or can be extracted from the Kaplan-Meier curves based on the method reported before [21].
Studies were excluded if they were: (1) reviews, case reports, meta-analyses, letters, conference abstracts without original data; (2) articles which could not extract the relevant data; (3) overlapping articles and those with duplicated data. www.impactjournals.com/oncotarget

Data extraction
The valuable data including surname of the first author, mean age, gender, cancer type, anti-body used, clinicopathological features, prognosis and other relevant data were extracted by two investigators (Lu and Liang) independently and illustrated in Table 1. When the prognosis was only plotted as a Kaplan-Meier curve in some articles, Engauge Digital 9.3 software (from https:// sourceforge.net/projects/digitizer/) was applied to digitize and extract the data. A joint decision was offered in the case of any disagreement.

Quality assessment
The quality of included studies was assessed by two independent reviewers (Zhu and Xu) on the basis of the Newcastle-Ottawa scale (NOS) system, which is a 9-point scoring system by judging three main categories: including selection, comparability and outcome. Studies with scores≥6 were regarded as high-quality studies [22][23].

Statistical analysis
Pooled HR with their 95% CI was used to assess the association between MUC expression levels and the cancer prognosis (OS). Meanwhile, the impact of MUC expression on clinicopathological parameters was performed by RR with their 95% CIs. The heterogeneity among the included studies was checked by the chisquared Q test. P > 0.10 or I 2 < 50% indicated that the differences between the results of various studies were due to chance, and then a fixed-effect model was used. Otherwise, a random-effects model was employed. Subgroup analyses were performed to explore the source of heterogeneity. The sensitivity analysis and funnel plots were carried out to evaluate the robustness and the possible publication bias respectively. All statistical tests were performed using Stata version 12.0 (Stata Corp, College Station, TX, USA) with two-tails p values. A p-value < 0.05 was considered to be statistically significant.

Trial sequential analysis (TSA)
TSA was employed to test whether the current findings were reliable and conclusive, based on diversityadjusted threshold for statistical significance. We conducted TSA with assumptions including a plausible overall 5% risk of type I error with a power of 80%. When the cumulative z-curve crosses the trial monitoring boundary or enters the futility zone before the required power is reached, future trials are superfluous considering a sufficient level of evidence is reached [24][25].
Otherwise, the current evidence is insufficient for drawing a conclusion. These analyses were performed using Trial Sequential Analysis Software version 0.9 beta (www.ctu. dk/tsa).