MUC1 overexpression predicts worse survival in patients with non-small cell lung cancer: evidence from an updated meta-analysis

Background Previous studies on the prognostic role of MUC1 expression in non-small cell lung cancer (NSCLC) remain controversial. We conducted a meta-analysis to appraise the clinicopathological and prognostic effect of MUC1 in NSCLC patients. Materials and Methods Searches of PubMed, EMBASE and CNKI (Chinese National Knowledge Infrastructure) were conducted and relevant studies were extracted. The pooled hazard ratio (HR) or odds ratio (OR) with 95% confidence intervals (CIs) were used to estimate effects. Heterogeneity among studies and publication bias were also evaluated. Results A total of 15 studies with 1,682 patients were included in this meta-analysis. The pooled HRs indicated that elevated MUC1 expression was associated with poorer overall survival (HR = 2.12, 95% CI: 1.47–3.05; P < 0.001) and progression-free survival (HR = 2.00, 95% CI: 1.53-2.62; P < 0.001) in patients with NSCLC. Significant associations were also found in patients treated with epidermal growth factor receptor tyrosine kinase inhibitors (EGFR-TKIs) (HR = 3.16, 95% CI: 2.21–4.52, P < 0.001) and with a platinum-based regimen (HR = 4.35, 95% CI: 2.45–7.72, P < 0.001). Additionally, MUC1 overexpression was significantly associated with performance status (OR = 2.32, 95% CI: 1.13–4.73, P = 0.021). Conclusions MUC1 could be a valuable biomarker of the prognoses of NSCLC patients.


INTRODUCTION
Lung cancer is the most common type of cancer and the leading global cause of cancer-related death [1,2]. Non-small cell lung cancer (NSCLC) accounts for 80-85% of lung cancer cases. Although progress has been achieved in the past decades, the prognosis for NSCLC is still poor, with an estimated survival rate of only 15% at 5 years [1]. Several markers, including tumor stage, tobacco smoking [3], ki-67 expression [4], cyfra21-1 [5] and XRCC1 (X-ray repair cross-complementing protein 1) polymorphism [6] have been reported as prognostic indicators of outcomes

Meta-Analysis
Oncotarget 90316 www.impactjournals.com/oncotarget in NSCLC patients. However, it is still difficult to predict patients' outcomes before treatment.
Mucin-1, previously called KL-6, EMA and CA15-3, is a glycoprotein present in normal epithelial tissue and in various cancers, including NSCLC [7,8]. Mucin1 is capable of increasing the invasive and metastatic capability of tumor cells by reducing cell-cell adhesion [9] and cell-extracellular matrix adhesion [10]. Mucin1 can also interact with the family of epidermal growth factor receptors (EGFRs) and participate in the progression of carcinogenesis [11]. Therefore, Mucin1 has been extensively studied in a variety of neoplasms, including breast [12], gastric [13] and colorectal [14]. The first report of high MUC1 expression as a valuable prognostic marker for NSCLC was presented in 1998 [15]. Subsequently, numerous studies have been performed to validate this result [15][16][17][18][19][20][21][22][23][24], but it remains controversial [25][26][27][28]. A previous meta-analysis reported the prognostic value of high MUC1 expression in NSCLC patients [29] but included relatively few studies (n = 4). In addition, subgroup analysis based on ethnicity, method of detection and choice of therapy was not performed. Therefore, we conducted an updated meta-analysis to reappraise the effect of MUC1 expression on the prognosis of NSCLC patients.
Fourteen studies investigated the prognostic role of MUC1 on overall survival (OS), and 5 studies explored the prognostic impact of MUC1 on progression-free survival (PFS). Nine studies were from Japan, three from China, two from Germany, and one from Greece. The sample sizes ranged from 41 to 185. HRs and 95% CIs were extracted directly from the 11 studies. HRs in 4 studies were estimated by Kaplan-Meier survival curves [15,20,23,27]. MUC1 expression was divided into high and low levels, and different cut-off values were selected in each study. Most studies performed experiments using the manufacturer's instructions; some applied the median or mean levels as cut-off values, and the remaining studies defined the cut-off value independently or by using a ROC curve. Detailed characteristics of the included studies are listed in Table 1.

Publication bias
Begg's funnel plot and the Egger's linear regression test were conducted to evaluate publication bias in the literature. No significant publication bias was detected by both Begg's test (P = 0.208 for OS and P = 0.327 for PFS) and the Egger's test (P = 0.604 for OS and P = 0.514 for PFS) (Figure 4). Therefore, no evidence of publication bias was noted.

Sensitive analysis
We adopted the "leave-one-out" scheme (i.e., the analysis is conducted using all studies except one) to explore the influence of individual studies on the pooled HRs. The results showed that the pooled HRs were not materially altered, which suggested that no individual study significantly affected the pooled results ( Figure 5).

DISCUSSION
To the best of our knowledge, only one metaanalysis on the prognostic value of MUC1 expression in NSCLC had previously been performed [29]. Our metaanalysis included three times more patients than the previous study, and the studies included in our analysis used more detailed information and patients with longer follow-up intervals. As a result, we were able to obtain more relevant results.
Our meta-analysis combined the results from 15 individual studies with 1,682 NSCLC patients and found that MUC1 overexpression had significantly prognostic value for OS (HR = 2.12, 95% CI: 1.47-3.05; P < 0.001) and PFS (HR = 2.00, 95% CI: 1.53-2.62, P < 0.001) in NSCLC patients. This link was observed in both surgical and non- Oncotarget 90319 www.impactjournals.com/oncotarget surgical treatment groups. Subgroup analysis by ethnicity indicated the result was significant for the Asian subgroup (HR = 2.49, 95% CI: 1.73-3.59, P < 0.001), but not for the Caucasian subgroup (HR = 1.10, 95% CI: 0.45-2.73, P = 0.832). Considering the limited number of Caucasian patients in our analysis, more studies should be conducted. When stratified by sample type, a significant risk was found in the sera group (HR=2.38, 95% CI: 1.47-3.82, P < 0.001), indicating that MUC1 may be a convenient tumor marker for use in clinical practice. We found that 500 U/ml is the most frequently used cut-off value and is associated with significant risk (HR = 2.20, 95% CI: 1.19-4.10, P = 0.012).
We verified the poor prognostic role of high MUC1 expression in patients treated with a platinum-  [40]. However, its efficacy varies among individuals [41]. The prognostic or predictive roles of a series of tumor markers were reported in NSCLC patients treated with  Oncotarget 90321 www.impactjournals.com/oncotarget platinum-based chemotherapy [42,43], but until now, none was recommended for clinical practice. Based on our results, MUC1 might be a promising biomarker. EGFR-TKI therapy significantly improves the survival of NSCLC patients who harbor an EGFR mutation [44]. Unfortunately, there is no indicator that predicts the efficacy of EGFR-TKI therapy. Our findings indicate that MUC1 may be such an indicator, but as the sample size of our analysis is limited, large-scale prospective studies are needed to further confirm our results.
There are some limitations to our meta-analysis. First, the heterogeneity was moderately significant in the pooled HRs of OS (I 2 = 75.7%, P < 0.001). Although we performed subgroup analysis and sensitivity analysis to find the source of heterogeneity, none could completely explain it. Second, this meta-analysis was limited to articles published in English or Chinese, indicating that language bias likely existed. Third, most of the studies selected were conducted on Asian populations; thus, standardized analyses should be used to apply our results to other populations. Fourth, several HRs were extracted from Kaplan-Meier curves, which might have biased our results. Finally, NSCLC consists of several subtypes, such as adenocarcinoma, squamous Oncotarget 90322 www.impactjournals.com/oncotarget cell carcinoma and others. The prognosis and selection of therapy for each type are dissimilar, but detailed information on NSCLC subtypes was lacking, and we did not conduct subgroup analysis by subtypes. More studies on the association between MUC1 and NSCLC subtypes are needed.
In conclusion, our results indicate that high MUC1 expression may be a marker of poor prognosis in NSCLC Oncotarget 90323 www.impactjournals.com/oncotarget patients and a promising therapeutic target. Large, welldesigned prospective studies are needed to confirm our findings.

Inclusion and exclusion criteria
All included studies had to meet the following criteria: (1) evaluation of the association between MUC1 expression and NSCLC prognosis; (2) case-control studies; (3) sufficient data for estimating the hazard ratio (HR) with a 95% confidence interval (CI). The major reasons for exclusion were (1) duplicate studies; (2) case reports, comments or review articles; (3) studies lacking detailed data.

Data extraction
Two investigators (XH and QS) performed searches and identified articles independently using a standard approach [45]. The following information was extracted: first author, publication year, nationality, ethnicity, quantitative method, cut-off value, follow-up months, hazard ratios (HR) with corresponding 95% confidence intervals (CI) for overall survival (OS) or progressionfree survival (PFS) and the total number of participants, respectively. In case of discrepancies, another investigator (CC) was invited to check and discuss the original data until a consensus was reached. Quality assessment for each study included in the final analysis was performed by the same two reviewers according to the Newcastle-Ottawa quality assessment scale (NOS) [46]. NOS scores ranged from 0 to 9, and a score ≥ 6 indicates good quality in the present study.

Statistical analysis
The intensity of the relationship between MUC1 expression and survival was expressed as HRs, and the strength of the association between MUC1 and clinical parameters was expressed as an odds ratio (OR). In some studies, HR and the 95% CI were directly obtained using univariate or multivariate survival analysis. Otherwise, a method reported by Tierney was used to reconstruct the HR and its variance from Kaplan-Meier survival curves [47]. Heterogeneity among eligible studies was estimated using a Chi-square-based Q test and considered statistically significant when I 2 > 50% or P < 0.1 [48]. A fixed effects model (Mantel-Haenszel method) was used if there was no significant heterogeneity; otherwise, a random effects model (Der Simonian and Laird method) was used [49]. Publication bias was evaluated using Egger's test and Begg's test, and P < 0.05 was considered significant [50]. All statistical tests were conducted with STATA software version 12.0 (STATA Corporation, College Station, TX, USA) and P < 0.05 was considered significant.

Author contributions
This study was conceived of and designed by XH. The data were extracted and analyzed by XH and QS. XH and CC wrote the manuscript. This manuscript was approved by all of the authors prior to submission.