Interaction between age and race alters predicted survival in colorectal cancer

Background: Racial disparities in colorectal cancer persist. Late stage at presentation and lack of stage-specific treatment may be contributing factors. We sought to evaluate the magnitude of disparity remaining after accounting for gender, stage, and treatment using predicted survival models. Methods: We used institutional tumor registries from a public health system (two hospitals) and a not-for-profit health system (nine hospitals) from 1995 to 2011. Demographics, stage at diagnosis, treatment, and survival were recorded. Hazard ratios (HRs) and predicted HRs were determined by Cox regression and postestimation analyses. Results: There were 6,990 patients: 55.7 % white, 23.6 % African American, 15.1 % Hispanic, and 5.6 % Asian/other. Predictors of survival were surgery (HR 0.57, 95 % confidence interval [CI] 0.46-0.70), chemotherapy (HR 0.7, 95 % CI 0.62-0.79), female gender (HR 0.87, 95 % CI 0.83-0.90), age (HR 1.04, 95 % CI 1.03-1.05), and African American race (HR 3.6, 95 % CI 1.5-8.4). Balancing for stage, gender, and treatment reduced the predicted HRs for African Americans by 28 % and Hispanics by 17 %. In this model, African American and Hispanics still had the worst predicted HRs at younger ages, but whites had the worst predicted HR after age 75. Conclusions: Gender, stage, and treatment partially accounted for worsened survival in African Americans and Hispanics at all ages. At younger ages, race-related disparities remained which may reflect tumor biology or other unknown factors. Once gender, stage, and treatment are balanced at older ages, the increased mortality observed in whites may be due to factors such as comorbidities. Further system- and patient-level study is needed to investigate reasons for colorectal cancer survival disparities.