Desarrollo de modelos predictivos en el carcinoma de celulas renales

Resumen

Renal cell carcinoma, malignant tumor derived from the renal parenchyma tubular epithelium, represents 2-3% of adult cancers. Because of the widespread use of imaging techniques, its incidence has increased in recent decades, mainly the tumors in less advanced stages. However, although in organ-confined stage surgical treatment is considered curative, it is a tumor that maintains high mortality and progresses in 30% of the cases. The heterogeneous evolution of the tumor evidences that there are distinct subpopulations of RCC and that knowledge of their characterization would be of particular interest to select patients according to risk and plan a follow-up or treatment after surgery. Multiple variables were analyzed as potential prognostic factors and have been part of predictive models to clarify the natural course of this disease. However, their evolution is still unpredictable. OBJECTIVES Primary: To create a predictive model for overall survival, cancer-specific survival and progression free survival for patients with RCC treated by surgery with curative intent, including clinical and pathological variables. Secondary: 1.-To identify prognostic factors for overall survival, cancer-specific survival and progression-free survival in these patients. 2.-To communicate, ffectively, the results of the models created by using nomograms. 3.-To assess the predictive power of the developed models. 4.-Externally validate two existing models in the literature assessing their predictive performance. MATERIAL AND METHOD Study population: This is a retrospective cohort of 363 patients with RCC in all its histopathologic variants who underwent a surgery with curative intent, in the period between January 1983 and December 2010 at the University Hospital Principe de Asturias (Alcalá de Henares) and between January 1984 and December 2006 in San Pedro de Alcántara Hospital (Cáceres). Exclusion criteria considered were: Patients with metastatic disease at diagnosis, family disease or family with RCC syndromes, bilateral renal tumors at diagnosis, stage pT4 (TNM 2009 classification), patients requiring adjuvant therapy and patients with unresectable lymph nodes. Variables analyzed: Independent variables were related to characteristics of the patient, tumor and treatment. Clinical outcomes were time to global overall mortality, cancer-specific mortality and relapse. Statistical analysis: Comparison of key features of both cohorts was performed using chi 2 tests. Descriptive analysis of qualitative variables by frequency distribution, and quantitative variables as means and SD, when normal distribution were found, or by median and interquartile range (IQR), in case of asymmetrical distributions. Homogeneity of frequencies was analyzed by means of chi2 test or Fisher exact. Agreement between different observers (image and pathology) was assessed by Kappa index. Length of follow-up was characterized by means of media time, media time of cohort after excluding deaths and the inverse Kaplan-Meier method. Survival functions ere estimated using Kaplan-Meier method. Between groups comparison of survival curves were done using the Log-Rank test. Proportional hazards assumption was checked using Schoenfeld residual tests. Univariate analysis was made using Cox regression. Cox multivariate models were fitted following manual backward strategy. Final predictive models were represented as nomograms. Evaluation of the predictive power of the models was done using calibration and discrimination, and was internally validated by bootstrap. External validation of other models was make computing Harrell concordance index. All hypothesis used tests a type I error < 0.05. Statistical packages used were SPSS v15, STATA v13 and R v3.0