The modifying effect of patient location on stage-specific survival following colorectal cancer using geosurvival models

Abstract

Colorectal cancer (CRC) is the third leading cause of cancer death in the US, and stage at diagnosis is the primary prognostic factor. To date, the interplay between geographic place and individual characteristics such as cancer stage with CRC survival is unexplored. We used a Bayesian geosurvival statistical model to evaluate whether the spatial patterns of CRC survival at the census tract level varies by stage at diagnosis (in situ/local, regional, distant), controlling for patient characteristics, surveillance test use, and treatment using linked 1991-2005 SEER-Medicare data of patients ≥ 66 years old in two US metropolitan areas. The spatial pattern of survival varied by stage at diagnosis for both cancer sites and registries. Significant spatial effects were identified in all census tracts for colon cancer and the majority of census tracts for rectal cancer. Geographic disparities appeared to be highest for distant-stage rectal cancer. Compared to those with in situ/local stage in the same census tracts, patients with distant-stage cancer were at most 7.73 times and 4.69 times more likely to die of colon and rectal cancer, respectively. Moreover, frailty areas for CRC at in situ/local stage more likely have a higher relative risk at regional stage, but not at distant stage. We identified geographic areas with excessive risk of CRC death and demonstrated that spatial patterns varied by both cancer type and cancer stage. More research is needed to understand the moderating pathways between geographic and individual-level factors on CRC survival.

Fig. 1. Census tracts of Detroit and Atlanta, according to Google Maps, retrieved May 18, 2012 from http://maps.google.com

Fig. 2. Distribution of median survival time at the census tract level after diagnosis by stage in Detroit and Atlanta

Fig. 3. Maps of relative risks (RR) for colon cancer by stage at diagnosis in both SEER registries, calculated by the exponential spatial function. Brown represents higher RR, and blue represents lower RR. The geographic disparity (GD) for each map is calculated by exponentiating the square root of the spatial variance

Fig. 4. Maps of relative risks (RR) for rectal cancer by stage at diagnosis in both SEER registries, calculated by the exponential spatial function. Brown represents higher RR, and blue represents lower RR. The geographic disparity (GD) for each map is calculated by exponentiating the square root of the spatial variance