K A Pathak1, P Lambert2, R W Nason3, T Klonisch4. 1. CancerCare Manitoba, Winnipeg, Manitoba, Canada; Canada University of Manitoba, Winnipeg, Manitoba, Canada. Electronic address: apathak@cancercare.mb.ca. 2. CancerCare Manitoba, Winnipeg, Manitoba, Canada. 3. CancerCare Manitoba, Winnipeg, Manitoba, Canada; Canada University of Manitoba, Winnipeg, Manitoba, Canada. 4. Canada University of Manitoba, Winnipeg, Manitoba, Canada.
Abstract
OBJECTIVE: Thyroid prognostic nomogram can be applied across different histological types for predicting the individualized risk of death from thyroid cancer. The objective of this study was to compare the strength of our recently published thyroid prognostic nomogram with 12 existing staging systems to predict the risk of death from thyroid cancer. METHOD: This study included 1900 thyroid cancer patients, from a population based cohort of 2296 patients, on whom adequate staging information was available. Competing risk sub-hazard models were used to compare 12 pre-existing prognostic models with the nomogram model. Their relative strengths for prediction of patients' individualized risks of death from thyroid cancer were compared using Akaike information criterion (AIC), delta AIC, and concordance index. R version 3.2.2 was used to analyze the data. RESULTS: Our cohort of 450 males and 1450 females included 1796 (93.4%) differentiated thyroid cancers. Amongst the compared models, thyroid prognostic nomogram model appeared to be better than other models for predicting the risk of death from all non-anaplastic thyroid cancer (concordance index = 94.4), differentiated thyroid cancer (concordance index = 94.1) and papillary thyroid cancer (concordance index = 94.7). The difference from next best staging systems was most pronounced in non-anaplastic thyroid cancer (delta AIC = 114.8), followed by differentiated thyroid cancer (delta AIC = 35.6) and papillary thyroid cancer (delta AIC = 8.4). CONCLUSIONS: Thyroid prognostic nomogram model was found to be better than the other models compared for predicting risk of death from thyroid cancer.
OBJECTIVE: Thyroid prognostic nomogram can be applied across different histological types for predicting the individualized risk of death from thyroid cancer. The objective of this study was to compare the strength of our recently published thyroid prognostic nomogram with 12 existing staging systems to predict the risk of death from thyroid cancer. METHOD: This study included 1900 thyroid cancerpatients, from a population based cohort of 2296 patients, on whom adequate staging information was available. Competing risk sub-hazard models were used to compare 12 pre-existing prognostic models with the nomogram model. Their relative strengths for prediction of patients' individualized risks of death from thyroid cancer were compared using Akaike information criterion (AIC), delta AIC, and concordance index. R version 3.2.2 was used to analyze the data. RESULTS: Our cohort of 450 males and 1450 females included 1796 (93.4%) differentiated thyroid cancers. Amongst the compared models, thyroid prognostic nomogram model appeared to be better than other models for predicting the risk of death from all non-anaplastic thyroid cancer (concordance index = 94.4), differentiated thyroid cancer (concordance index = 94.1) and papillary thyroid cancer (concordance index = 94.7). The difference from next best staging systems was most pronounced in non-anaplastic thyroid cancer (delta AIC = 114.8), followed by differentiated thyroid cancer (delta AIC = 35.6) and papillary thyroid cancer (delta AIC = 8.4). CONCLUSIONS: Thyroid prognostic nomogram model was found to be better than the other models compared for predicting risk of death from thyroid cancer.
Authors: Min Ji Jeon; Won Gu Kim; Tae Hyuk Kim; Hee Kyung Kim; Bo Hyun Kim; Hyon Seung Yi; Eun Sook Kim; Hosu Kim; Young Nam Kim; Eun Heui Kim; Tae Yong Kim; Sun Wook Kim; Ho Cheol Kang; Jae Hoon Chung; Young Kee Shong; Won Bae Kim Journal: Endocrinol Metab (Seoul) Date: 2017-11-22