Resumen de Area-based method for assessing survival benefit in Kaplan-Meier curves
Emilio Jesús Alegre del Rey, J. Díaz Navarro, S. Fénix Caballero, María del Carmen Pardo, Teresa Pérez Pérez 
, Estrella Figueroa Murillo, M. Ángeles Moreno Santos, Carmen Martínez Díaz
Introduction and objectives: The difference in medians is the typically used measure of absolute survival benefit in cancer patients. However, it shows the time between curves at a single point, and may over- or underestimate the survival benefit. A complementary, more reliable method is needed. The aim of this study was to test a new method for estimating the absolute survival benefit based on area under Kaplan-Meier's curves. Method: A search was carried out for clinical trials examining overall survival in patients with advanced breast, lung or colorectal cancer. Articles with a survival figure showing patients at risk, both curves reaching the median and p <0.05 for HR, were included. The relative increase in global survival was calculated using both the area under the curve (AUC) method and the difference in median survival method. The increase in survival was calculated using both AUC and medians-based method. A correlation study between relative increment in survival (A vs B) and the inverse (B vs A) relative increase in death risk based on HR was conducted. The two correlations were compared using Steiger's Z-test. Results: A total of 41 articles were selected. The correlations between the increase in death risk and the increase in global survival by AUC-based method and difference in median survival-based method were 0.92 and 0.85, respectively, both of them statistically significant. The former was significantly better (p = 0.036). Discussion: The AUC-based method provides supplementary data to the difference in medians. This method is especially useful when the later over- or un-derestimates overall survival, taking into account the variations in distance between the curves throughout the graph. It is important to know absolute survival benefit in cancer patients with poor prognosis, rather than the proportion of patients who will suffer the event. A good estimate of overall survival is relevant for decision-making on the optimal treatment for the patient
