RESUMO
BACKGROUND: The use of home parenteral nutrition (HPN) in incurable cancer patients is extremely varied across different countries and institutions. In order to assess the clinical impact implied, we previously conducted a survey of incurable cancer patients receiving HPN, which shows that survival was markedly affected by Karnofsky performance status (KPS), tumor spread, Glasgow prognostic score (GPS) and tumor site. The aim of this study was to develop a nomogram incorporating the above factors for survival prediction. PATIENTS AND METHODS: We gathered a series of 579 patients, all receiving HPN, which was randomly split into a training and a testing sample. Using Cox proportional hazard regression modeling, a nomogram was built in the training sample, in order to estimate median survival or survival probability at 3 and 6 months according to individual patient characteristics. The nomogram performance was then verified in the testing sample. RESULTS: In the training sample, median survival was 3.2 (95% CI 3.0-3.7) months. GPS, KPS, tumor site and spread were confirmed to be significant prognostic factors. A significant interaction was also shown between the site and spread while weight loss (WL), adjusted for body mass index, failed to provide any substantial prognostic contribution. In the testing sample, nomogram performance was good in terms of calibration and discreet regarding discrimination. CONCLUSION: With the growing availability of new oncological treatments and their tendency to transform the trajectory of the advanced cancer into a chronic condition characterized by progressive WL and poor nutrients intake, an increasing number of patients are expected to receive HPN. In such a setting, tools for predicting the survival outcome may play a role toward personalized medicine and for investigating novel experimental therapies. Our proposed nomogram is a step forward in this direction but needs to be made stronger in order to definitely have clinical utility.