Clinical and Prognostic Implications of P21 (WAF1/CIP1) Expression in Patients with Esophageal Cancer: A Systematic Review and Meta-Analysis

Background Previous studies have demonstrated that P21 (WAF1/CIP1) is a valuable prognostic factor in several malignant tumors. However, it is not known whether P21 can predict the prognosis in patients with esophageal cancer (EC). The aim of this research was to investigate the contribution of P21 expression to the clinicopathological characteristics and of EC. Methods A systematic review and meta-analysis of study focusing on P21 expression, clinicopathological characteristics, and clinical outcomes in patients with EC was performed using seven databases (PubMed, Embase, Web of Science, and four Chinese databases). Pooled hazard ratios and odds ratios were used to explore the association between P21 expression, clinicopathological characteristics, and outcomes in patients with EC. The heterogeneity of the studies was classified by the I2 statistic. The sensitivity analysis was then utilized to assess the robustness of the results. Finally, the funnel plot and Begg's test were used to evaluate the publication bias. Results Forty-five studies with 3098 patients were eligible for inclusion in the meta-analysis. Thirty of these studies reported on clinicopathological characteristics and 15 on clinical outcomes. The pooled hazard ratio of 1.456 (95% confidence intervals 1.033–2.053, P = 0.032) for overall survival indicated that a low P21 expression level was an unfavorable prognostic factor for a clinical outcome in patients with EC. Furthermore, the pooled odds ratio confirmed an association between decreased P21 expression and poor clinicopathological characteristics, including differentiation, lymph node metastasis, invasion, and higher grade and clinical stage. Notably, high P21 expression was a significant predictor of a favorable response to chemotherapy. There was no evidence of publication bias. Conclusion Reduced P21 expression is associated with a poor outcome in patients with EC.


Introduction
Esophageal cancer (EC) is the seventh leading cause of cancer mortality worldwide and in 2016 accounted for 15,690 deaths in the United States alone [1]. EC is a complex disease that includes squamous cell carcinoma, adenocarcinoma, and other rarer histologic types. Risk factors are slightly different between the two major types but include sex, race, alcohol consumption, diet, and genetics [2][3][4]. Several genetic biomarkers are effective in predicting the prognosis of patients with EC, including TP53, CYCLIN D1, VEGF, COX-2, and HER-2 [5]. Moreover, treatment based on these molecular targets has improved survival outcomes in patients with this disease. For example, inhibitors of c-MET [6], EGFR [7], HER2 [8], and VEGR [9] have been demonstrated to extend survival in these patients. However, drug resistance remains a major concern, and not all patients benefit from targeted therapy. Therefore, novel biomarkers are required to provide insight into the molecular mechanism of EC, identify novel diagnostic methods, and increase the number of treatment options available.
P21 (WAF1/CIP1), a member of the P21/P27/P57 family, is a universal cell cycle inhibitor regulated by P53. P21 plays an essential role in the control of cell growth, terminal differentiation, stem cell phenotypes, apoptosis, and cellular stress response. P21 has also been reported to participate in the proliferation of all types of cells. The expression of P21 is altered by wild-type P53 when DNA is damaged, resulting in cell cycle arrest or apoptosis at the G1 checkpoint. P21 plays a vital role in limiting proliferation and tumor growth, and abnormal expression of this gene has been observed in various types of malignancy. Recent research by Xie and colleagues [10] suggests that overexpression of P21 is associated with a poor prognosis in patients with non-small-cell lung cancer, while the loss of P21 protein expression could be a significant predictor of disease progression in patients with pancreatic cancer [11]. A further study demonstrated that aberrant expression of the P21 protein is associated with vascular invasion, pathological disease stage, and overall survival in patients with gastric cancer [12]. Interestingly, Goan et al. reported that overexpression of P21 predicted an unfavorable survival outcome in patients with esophageal squamous cell carcinoma [13] while other researchers found a significant association of low P21 expression with shorter survival in patients with the disease [14,15]. Furthermore, P21 was found to regulate apoptosis in acute myeloid leukemia cells and malignant glioma cells [16,17]. Thus, although there is an association of P21 expression with various types of cancer, the impact of the P21 level on the disease progression and prognosis of EC remains controversial. Therefore, we performed a systematic review and meta-analysis to assess the potential contribution of P21 expression to the clinicopathological characteristics and prognosis of EC.

Inclusion and Exclusion
Criteria. Studies were eligible for inclusion in the meta-analysis if they met the following criteria: (1) the subjects were patients diagnosed with any type of EC; (2) P21 expression in tissue or serum was detected by Western blot, quantitative real-time polymerase chain reaction (PCR), immunohistochemistry, or RNA sequencing; (3) the association of the P21 expression level with clinicopathological characteristics or the prognosis of EC was investigated; (4) the study population included more than 20 patients with EC; and (5) publication was written in the Chinese or English language. The following exclusion criteria were applied: publication as a review, abstract, experimental study, or letter and no key data provided for the evaluation of the relationship between differential expression of P21 Full-test screen (n = 77) Figure 1: Flowchart of the study selection.   Disease Markers and the clinicopathological characteristics and survival outcomes in patients with EC.

Data Extraction and Quality Assessment.
The following data were collected and tabulated: the surname of the first author, year of publication, histologic type, sample size, country, specimen type, P21 detection assay used, and the Newcastle-Ottawa Scale (NOS) score. The NOS score was used to assess the quality of the included studies as follows: >6, high quality; 5-6, medium quality; and <5, low quality.

Statistical
Analysis. The pooled hazard ratio (HR) and 95% confidence intervals (CIs) were used to estimate the impact of P21 expression level on the survival outcome in patients with EC. The individual HRs and 95% CIs were extracted directly from the text by two investigators (JW and LL). A pooled HR > 1 and 95% CIs that did not overlap indicated a positive association between a lower P21 expression level and a poorer survival outcome. When the HR and 95% CIs for survival were not provided, estimates were calculated from the Kaplan-Meier curves according to the method described by Tierney et al. [18]. All data were extracted by two of the authors working independently (FW and LQ). The pooled ORs and associated 95% CIs were used to determine the association between the P21 expression level and the clinicopathological characteristics of patients with EC according to specimen type (tumor sample vs. normal control), age (younger vs. older), sex (male vs. female), differentiation (poor vs. well or moderate), tumor stage (III-IV vs. I-II), distant metastasis (yes vs. no), lymph node metastasis (yes vs. no), grade (G3-4 vs. G1-2), depth of invasion (III-IV vs. I-II), tumor size (large vs. small), tumor location (upper-middle vs. low), and clinical stage (III/IV vs. I/II). We also explored the relationship between P21 expression and other better-studied biomarkers of EC, including P53 and the apoptosis index. The ability of the P21 level to predict the efficacy of chemotherapy was analyzed by combining the ORs. As with the HRs, an OR > 1 indicated a positive correlation between decreased P21 expression and poor clinicopathological characteristics. The I 2 statistic was used to classify the heterogeneity of the studies as low (I 2 < 30%), moderate (30% ≤ I 2 < 60%), substantial (61% ≤ I 2 < 75%), or high (I 2 ≥ 75%) [19]. A P value for the I 2 statistic less than 0.10 or I 2 larger than 50% was defined as having statistically significant heterogeneity, and thus, a random-effect model was used. In contrast, a fixed-effect model was used when heterogeneity was not significant. Publication bias was quantified by Begg's test and funnel plot analyses [20]. All statistical analyses were performed using Stata (version 12, StataCorp, College Station, TX, USA).  (Table 1). Thirty studies focused on the association of the P21 expression level with clinicopathological characteristics, and 15 assessed the ability of the P21 expression level to predict overall survival (Table 2). Twenty-seven studies were performed in China and 11 in Japan. Most of the included studies detected the P21 level by immunohistochemistry with cutoff values ranging from 1% to 50%, while the remaining studies used real-time PCR or Western blotting. Six studies with a score of 9 and 13 studies with a score of 8 were considered high quality, and 7 studies with NOS scores < 7 were considered low quality.

Prognostic Value of P21 in Patients with EC.
A metaanalysis of the 15 studies that reported overall survival yielded a pooled HR of 1.456 (95% CI: 1.033-2.053, P = 0:032, z = 2:14), indicating a significant association between the P21-negative group and decreased survival time when compared with the P21-positive group. Significant heterogeneity was noted across the studies (I 2 = 81:2%, P < 0:05, Figure 2). Therefore, a random-effects model was used. Next, subgroup analyses of publication country, continent, sample size, cutoff value, and the statistical methods used to calculate the HRs were performed to explore the origin of the heterogeneity Ultimately, the country of publication might be a source of heterogeneity. The pooled HR of 2.05 indicated that a low P21 expression level was correlated with shorter survival time in the Japanese studies. Notably, the degree of Note: HR = hazard ratio; LL = lower confidence interval limit; UL = upper confidence interval limit; OS = overall survival. 5 Disease Markers heterogeneity in this group was reduced in the fixed-effects model (I 2 = 0:0%, P = 0:425). However, obvious heterogeneity was also detected in the other subgroup analyses.

Sensitivity Analysis and Publication Bias.
A sensitivity analysis confirmed that the results were not obviously impacted by any individual study, suggesting that the meta-   Disease Markers analysis had good stability. The results of Begg's test and the funnel plot did not indicate any publication bias ( Figure 6, Table 3).

Discussion
There is accumulating evidence to suggest that abnormal expression of P21 is present in various types of malignancy, including gastric [63], lung [64], and tonsillar [65] cancers. However, the results of studies that have investigated the potential role of P21 expression are not consistent. To our knowledge, this meta-analysis contains the largest number of studies that have evaluated the association between P21 expression and the clinicopathological characteristics and outcomes in patients with EC. We found that overall survival in patients with EC was likely to be longer in those with higher P21 expression than in those with lower P21 expression. Our finding that decreased P21 expression was correlated with disease progression, that is, differentiation, lymph node metastasis, and invasion, as well as an advanced disease grade and clinical stage, indicates that P21 has a sup-pressor role in EC. A particularly important finding in this study was that P21 might be a valuable predictor of the effectiveness of chemotherapy. In this study, there was a significant association between low P21 expression and a poorer outcome of EC. In contrast, high P21 expression has been reported to be an unfavorable prognostic factor in patients with prostate cancer [66] and breast cancer [67]. However, the results of yet other studies in patients with cervical adenocarcinoma [68] and bladder cancer [69] are consistent with our finding that P21 might act as a tumor suppressor. Like in our study, previous research showed significant associations between a low P21 level and advanced clinical stage and grade of bladder cancer, indicating that P21 has an important role in tumor progression [70]. A previous study in prostate cancer showed that P21 inhibits cell growth by targeting E2F1 [71]. It has been confirmed that P21 expression could be reduced by DDX3 in lung cancer, leading to inhibition of the growth of cancer cells. Wu and colleagues demonstrated that inhibition of P21 via the P53-DDX3 pathway may promote the proliferation of cancer cells and tumor growth in vitro and in vivo   Disease Markers [69]. Moreover, it has been shown that P21 interacts with subunits of cyclin-dependent kinases [72], resulting in inhibition of tumor growth and progression. Finally, the tumor suppressor activity of P21 can be promoted by interaction with tumor-related factors like MYC, proliferating cell nuclear antigen (PCNA), and signal transducer and activator of transcription (STAT3) [73][74][75][76].
This study has several limitations that should be taken into account when interpreting its results. First, according to the NOS criteria, the quality of the included studies was variable (ranging from a score of 6 to a score of 9). Second, several HR values and their respective 95% CIs were obtained from Kaplan-Meier curves, potentially leading to inaccurate results. The inclusion of univariate HRs without adjustment could also have contributed to heterogeneity. Third, the methodological differences between the studies may have resulted in the underestimation of the effect size. For example, most of the studies detected P21 expression by immunohistochemistry, but some used diverse methods, including Western blot and real-time quantitative PCR. The use of the streptavidin-peroxidase conjugate in some studies and the streptavidin-biotin complex method in others was also    Figure 5: Forest plots of odds ratios for P21 expression and clinicopathological parameters including chemotherapy effectivity in esophageal cancer patients. 8 Disease Markers a potential source of heterogeneity. Another methodological difference was that the most common cutoff values for the detection of P21 were 10% and 50%, but this was not completely consistent across the studies. Inclusion of research published only in Chinese or English may have been another source of bias, given that negative results tend to be published in local journals. Furthermore, the number of studies included in this analysis was limited and we only restrict the patient number for the enrolled studies by a threshold of 20.

Conclusion
In this study, the results suggest that low P21 expression has a clinically important negative clinicopathological and prognostic impact in patients with EC. Well-designed studies that include larger patient cohorts are required to identify the mechanisms underlying how P21 is involved in the tumorigenesis and progression of EC.