Prognostic models and nomograms for predicting survival of patients with maxillary sinus carcinomas. Issue 7 (23rd May 2017)

• Record Type: Journal Article
• Authors: Shen, Weidong; Sakamoto, Naoko; Yang, Limin
• Is Part Of: International forum of allergy & rhinology. Volume 7:Issue 7(2017:Jul.)
• Page Start: 741
• Page End: 748
• Publication Date: 2017-05-23
• Subjects: maxillary sinus carcinoma -- nomogram -- population‐based -- predictive tools -- survival prognosis; 617.51005
• DOI: 10.1002/alr.21950
• Languages: English
• ISSNs: 2042-6976

Abstract

Abstract : Background: Maxillary sinus carcinoma is an uncommon malignancy. Most reports on prognosis of this disease are from single institutions and include few patients. We used data from the United States National Cancer Institute's Surveillance Epidemiology and End Results (SEER) program to construct models and nomograms for predicting outcomes of patients with maxillary sinus carcinomas. Methods: We used records from 668 patients with primary maxillary sinus carcinomas reported to the SEER program from 2004 to 2013 to build nomograms based on stratified multivariable Cox proportional hazard models for predicting 5‐year overall survival (OS) and cause‐specific survival (CSS). Model building was internally validated with the bootstrap approach. Results: Five‐year survival was 39.7% (95% confidence interval [CI], 35.5% to 44.5%) and 46.8% (42.3% to 51.8%) for OS and CSS, respectively. The final Cox model included the variables of age at diagnosis, tumor size, histologic type, TNM stage, and surgery. Radiotherapy was a stratification factor in the models. The models demonstrated good accuracy for predicting survival with a bootstrap‐corrected Somers D xy of 0.44 for both OS and CSS models. Calibration curves indicated acceptable model calibration. Conclusion: We developed tools for predicting prognosis that incorporate TNM stage and other readily available variables for patients with maxillary sinus carcinomas. The model performance was validated as good. These models can help clinicians to offer improved patient counseling in terms of clinical outcomes and make optimal treatment plans.