OBJECTIVES: The National Cancer Institute (NCI)’s refined Charlson comorbidity score (developed specifically for cancer) and the Agency for Healthcare Research and Quality’s Elixhauser comorbidity scores are two popular methods to control confounding due to comorbidities in observational studies. The relative performance of these scores in cancer studies is unknown. The objective was to compare the performance of the Elixhauser and the Charlson comorbidity score in predicting survival in four cancers (breast, colorectal, prostate and lung). METHODS: This cohort study used the Texas Cancer Registry linked Medicare claims data from 2005-2011. Four cancer-specific cohorts were created: breast (n=19,082), colorectal (n=16,963), prostate (n=23,044) and lung (n=26,047) cancer. Baseline one year diagnosis claims were used to define Charlson and Elixhauser comorbidity score. Consistent with Charlson and NCI methodology, the outcome was 2-year non-cancer mortality; cancer mortality was treated as a competing risk. Competing risk models were created to determine the performance of Charlson and Elixhauser comorbidity score in predicting 2-year survival while controlling for age, gender and stage of cancer. Models were compared using Akaike information criterion (AIC), Bayesian information criterion (BIC) and c-statistics (c). RESULTS: The 2-year non-cancer mortality was 5.7% (breast), 11.5% (colorectal), 4.1% (prostate) and 14.5% (lung). Elixhauser (breast: AIC=9084, BIC=9101, c=0.776; colorectal: AIC=17977, BIC=18002, c=0.681) performed slightly better than Charlson (breast: AIC=9112, BIC=9129, c=0.769; colorectal: AIC=17992, BIC=18016, c=0.679) for breast and colorectal cancer. Whereas, Charlson (prostate: AIC=8379, BIC=8391; c=0.780, lung: AIC=34915, BIC=34943, c=0.581) had slightly better performance than Elixhauser (prostate: AIC=8387, BIC=8400, c=0.779; lung: AIC=34927, BIC=34955, c=0.577) for prostate and lung cancer. CONCLUSIONS: Performance of both scores were comparable, with the slightly better performance of Elixhauser for breast and colorectal, and Charlson for prostate and lung. Evidence from this study can be used for selecting appropriate comorbidity score for cancer-specific observational study.