p21-activated kinase 1 (PAK1) expression correlates with prognosis in solid tumors: A systematic review and meta-analysis

p21 protein (Cdc42/Rac)-activated kinase 1 (PAK1) expression appears to be predictive of prognosis in various solid tumors, though the evidence is not yet conclusive. We therefore performed a meta-analysis to explore the relationship between PAK1 and prognosis in patients with solid tumors. Relevant publications were searched in several widely used databases, and 15 studies (3068 patients) were included in the meta-analysis. Pooled hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated to evaluate the strength of the association between PAK1 and prognosis. Associations between PAK1 expression and prognosis were observed for overall survival (HR = 2.81, 95% CI = 1.07-7.39) and disease-specific survival (HR = 2.15, 95% CI = 1.47-3.16). No such association was detected for time to tumor progression (HR = 1.78, 95% CI = 0.99-3.21).Our meta-analysis thus indicates that PAK1 expression may be a predictive marker of overall survival and disease-specific survival in patients with solid tumors.

In the present study, we used a statistical approach to systematically investigate the association between PAK1 and the prognosis of solid tumors.Over the past decade, a series of studies have focused on the relationship between PAK1 expression and solid cancer prognosis [3,[6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21], but the results of those individual studies were not conclusive.We therefore performed a meta-analysis using a relatively large sample from 15 studies (3068 patients) with the aim of conclusively determining the relationship between PAK1 and prognosis in patients with solid tumors.

Meta-analysis
In the meta-analysis, three outcome endpoints including DFS, PFS, and RFS that were similar in meaning were combined to use a unified prognostic parameter, time to tumor progression (TTP) instead.The meta-analysis of PAK1 expression was therefore based on three outcome endpoints: OS, DSS and TTP.Eight studies were included in the meta-analysis of OS.A random effects model was used to calculate the pooled hazard ratio (HR) and 95% confidence interval (CI) because the heterogeneity test reported a P value of less than 0.01.No significant association was observed between PAK1 expression and OS (pooled HR = 2.08, 95% CI = 0.93-4.64)(Supplementary Figure S1).Because some individual HRs were indirectly estimated (see Materials and Methods) and were therefore less reliable, we also performed a meta-analysis of OS using only the individual HRs extracted directly from the original articles.Six studies were included in that analysis, and again the heterogeneity test reported a P value of less than 0.01.We therefore used a random effects model to calculate the pooled HR and 95% CI.In this analysis, a significant relationship between PAK1 expression and OS among patients with solid tumors was detected (pooled HR = 2.81, 95% CI = 1.07-7.39)(Figure 2A).Four studies were included in the meta-analysis of DSS.A fixed effects model was used to calculate the pooled HR and 95% CI because the heterogeneity test reported a P value of 0.570.The result provided evidence of an association between PAK1 expression and DSS (pooled HR = 2.15, 95% CI = 1.47-3.16)(Figure 2B).Seven studies were used in the meta-analysis for TTP.The heterogeneity test reported a P value of less than 0.01, so a random effects model  was used to calculate the pooled HR and 95% CI.No significant association between PAK1 expression and TTP was detected (pooled HR = 1.78, 95% CI = 0.99-3.21)(Figure 3).The results of our meta-analysis thus suggest that PAK1 expression may be a predictive marker of OS and DSS in patients with solid tumors, but it is not predictive of TTP.

Publication bias test results
The Begg's funnel plot (Figure 4) and Egger's test showed there was no publication bias for DSS (P = 0.901) or for TTP (P = 0.062).However, publication bias may exist for OS (P = 0.032) in the analysis of high versus low PAK1 expression.

DISCUSSION
The results of our meta-analysis suggest that higher tumoral PAK1 expression is associated with an unfavorable prognosis and is predictive factor associated with OS and DSS in patients with solid tumors.PAK1 is an effector of small Rho GTPases (Cdc42 and Rac1) [2].PAK1 and Rac1 reportedly play important roles within cancer cell signaling networks and contribute to invasive and metastatic phenotypes [22,23].On the other hand, our meta-analysis indicates that PAK1 expression is not significantly associated with TTP in patients with solid tumors.The heterogeneity across the included studies is one potential reason for this.In addition, the combined effects of PAK1 with other molecular and environmental factors likely differ among cancer types.
Our meta-analysis has several limitations, so the results should be considered with a degree of caution.One limitation is that the sample size was not sufficient, particularly for the analysis of DSS.A second limitation is the heterogeneity caused by the diverse methods used to detect PAK1 expression and the varied cutoff values used in individual studies.The third limitation is that the patient data were not adjusted to account for details of the patients' characteristics, such as age and lifestyle.In addition, subgroup meta-analysis based on cancer type, PAK1 nuclear localization and p-PAK1 expression could not be carried out with the existing data.To achieve a more convincing conclusion, further analysis using a larger sample size, a unified detection method and adjusted individual data will be required, along with a stratified analysis based on cancer type, PAK1 nuclear localization and p-PAK1 expression.

Literature search, selection and data collection
For this study, we searched for papers published before May 6, 2015 using the keywords "p21 protein (Cdc42/Rac)-activated kinase 1" / "PAK1" / "PAKalpha", "cancer" / "tumor" / "neoplasm" / "carcinoma", and "survival" / "prognosis" / "mortality" / "death" independently in PubMed and Web of Science.Among the papers identified, were further selected for the metaanalysis using the following selection criteria.1) The full text of the study was in English.2) The study provided adequate data for individual HRs and 95% CIs to be extracted or calculated [24].3) When studies sharing the same patient sample were compared, the most complete study among them was included in our meta-analysis.
Three investigators independently collected data from each eligible paper.The data collected included the name of first author, publication year, patients' country of origin, cancer type, number of patients, cancer stage or grade, detection method, percentage exhibiting high PAK1 expression and the corresponding cutoff value, median follow-up months, outcome endpoints, outcome definition, survival analysis method, and the HR and 95% CI for the high PAK1 expression group versus low PAK1 expression group.Individual HRs and 95% CIs were estimated [24] if only Kaplan-Meier survival plots were available.Multivariate HRs and 95% CIs were selected if both univariate and multivariate results were reported in an individual study.By checking among the three investigators, the final data collected was determined.

Meta-analysis methods
Using the data collected from each eligible paper, we performed a meta-analysis of the outcomes to evaluate the relationship between PAK1 and solid cancer prognosis.Stata version 14.0 (Stata Corporation, College Station, TX, USA) was used to carry out the statistical analysis.Because the outcome endpoints DFS, PFS and RFS are similar in meaning, they were combined and a unified prognostic parameter, TTP, was used for the metaanalysis.Pooled HRs and 95% CIs for three outcome endpoints (OS, DSS, and TTP) were calculated.All the pooled HRs and 95% CIs were calculated using a fixed effects or random effects model.The model was chosen using a heterogeneity test.For the heterogeneity test, we performed the χ 2 -based Q-test [25].When the Q-test reported a P value of more than 0.10, a fixed effects model was used to calculate the pooled HRs [26], otherwise random effects model was used [27].
Publication bias was tested using Begg's funnel plot and the Egger's test [28].If the funnel plot was asymmetric and the Egger's test reported a P value of less than 0.05, publication bias was deemed to probably exist.

Figure 1 :
Figure 1: Flow chart of the study selection.

Figure 2 :
Figure 2: Forest plots of the meta-analysis of the association between PAK1 expression and the prognosis of patients with solid tumors.A. Overall survival (using only individual HRs extracted directly from the original articles) B. Disease-specific survival.Abbreviations: HR, hazard ratio; CI, confidence interval.

Figure 3 :
Figure 3: Forest plot of the meta-analysis of the association between PAK1 expression and solid tumor progression.Abbreviations: HR: hazard ratio; CI: confidence interval.

Figure 4 :
Figure 4: Begg's funnel plots for the studies involved in the meta-analysis of PAK1 expression and the prognosis of patients with solid tumors.A. Overall survival.B. Disease-specific survival.C. Time to tumor progression.Abbreviations: loghr, logarithm of hazard ratios; s.e., standard error.

Table 1 : Studies and data included in this meta-analysis
bStudy investigated the prognostic effect of p-PAK1 only and was excluded from quantitative analysis.Abbreviations; UC-UUT, urothelial carcinoma of the upper urinary tract; NA, not available; IHC, immunohistochemistry; RFS, recurrence-free survival; OS, overall survival; PFS, progression-free survival; DSS, disease-specific survival; DFS, disease-free survival; M, multivariate cox proportional hazard regression; KM, Kaplan-Meier method.