Combine P53, E-cadherin And VEGF-C For Prediction Of Urinary Bladder Cancer Prognosis After Radical Cystectomy

ObjectiveUrothelial carcinoma of the bladder (UCB) is an especially complex and heterogeneous disease with a broad spectrum of histologic findings and potentially lethal behavior. Despite advances in surgical techniques, as well as intra-vesical and systemic therapies, up to30%of patients with non-muscle-invasive UCB and50%of patients with muscle invasive UCB experience disease progression, recurrence, and eventual death. Standard prognostic features, such as pathological stage and grade, have limited ability to predict the outcomes of this heterogeneous population. Current risk-stratification algorithms using clinical and pathologic parameters are limited in their prognostic ability. Molecular medicine holds the promise that clinical outcomes will be improved by more accurate prognostication and directing therapy towards the mechanisms and targets associated with the growth of an individual patient’s tumor. Immunohistochemical analysis of biomarker expression has provided insight into the molecular pathogenesis of UCB and offers the potential for improving clinical decision making. Numerous candidate immunohistochemical biomarkers for patients with UCB have been identified, with those relating to the cell cycle and apoptosis/cell proliferation being the most extensively studied. To our knowledge, there was no report about markers related to UCB genesis for prediction of patient’s prognosis. Our study tested three biomarkers involved in urothelial carcinogenesis, including proteins of P53, E-cadherin and VEGF-C, whether the combination of three biomarker could improve the ability to predict clinical outcomes (disease-free survival and over-all survival, DFS and OS) of patients who had been done the radical cystectomy in our center. We also assessed whether the combination of molecular markers is superior to any individual biomarker. MethodsA total of106urothelial bladder carcinoma patients treated by radical cystectomy in our center from Mar2003to Feb2009were enrolled in this study, who had received radiotherapy or neoadjuvant chemotherapy before surgery and those died during perioperative period were excluded from this study. We collected clinical information as follows:age and gender, TNM stage and pathological grade (2004WHO grading system), lymphovascular invasion (LVI), hydronephrosis and adjuvant chemotherapy. One follow-up endpoint is the day of discovery of recurrence or metastasis, which can be used to calculate disease-free survival (DFS) and the other endpoint, is the day of patients’death, used for overall survival (OS).Pathologists with expertise in genitourinary pathology examined all specimens without knowledge of clinical information, and chosen the106specimens and other20normal urothelial epithelium for control. We undertook immunohistochemical staining for the three markers in all paraffin-embedded tissue.All markers were placed in one of two categories-altered or normal. We used the χ2test and Fisher exact test to measure the association between molecular markers and clinicopathological variables and the χ2test for trend to compare ordinal variables. The Kaplan-Meier method was used to calculate survival functions, and differences were assessed with the log-rank statistic. We undertook univariate and multivariate survival analyses with the Cox proportional-hazards regression model. When calculating hazard ratios, we used the lowest category as the reference. In Cox regression models, the area under the curve is substituted with Harrell’s concordance index. We did internal validation of predictive accuracies with bootstrapping analysis, for which the study cohort was resampled200times with replacement to reduce overfit bias. We then calculated predictive accuracy for every model for all200bootstrap samples and, finally, the average bootstrap-corrected predictive accuracy of every model. Predictive accuracy estimates were expressed as percentages and compared with the Mantel-Haenszel test. We judged p<.0.05to be significant; all reported p values are two-sided. We undertook all analyses with either SPSS version19.0or Stata Professional12.0. ResultsPatients’median age was59(range:25to75), and median follow up time was40months (range:2～97months). Only6(5.7%) patients were lost,34(32.1%) and25(23.6%) patients were discovered cancer recurrence or metastasis and death during our follow-up, respectively.Altered expression of P53, E-cadherin and VEGF-C were found in31(29.2%),61(57.5%) and51(48.1%) patients, respectively. P53expression had no association with clinical parameters above(all p>0.05), while E-cadherin expression were associated with T-stage(p=0.000), Grade (p=0.000) and LVI (p=0.009), and VEGF-C expression were correlated with Grade (p=0.001), N-stage(p=0.003) and LVI (p=0.006). But the three markers had no correlation with each other (all p>0.05).Patient with more altered markers had shorter DFS and OS (both long-rank p<0.05), who owned3altered markers only had28months DFS and30months OS according to our study. By univariable and multivariable regression analysis along with bootstrapping internal validation, we confirmed that combination of molecular markers was superior to any individual biomarker (all p<0.05). Our study with combination of P53, E-cadherin and VEGF-C could improve the prediction accuracy by3.56%and2.17%for DFS and OS, respectively (both p<0.05).ConclusionsCombined altered markers of P53, E-cadherin and VEGF-C, associated with clinical prognosis parameters, could improve the prediction accuracy of DFS and OS. Combination of molecular markers was superior to any individual biomarker. This could eventually guide our clinical decision making about personal follow-up schedule and treatment choice.