Estimating breast cancer recurrence in a population-based registry in Georgia, US

Abstract

Although the descriptive epidemiology of primary breast cancer is well characterized in the US, breast cancer recurrence rates have not been measured in an unselected population. The number of breast cancer survivors at risk for recurrence is growing each year, so recurrence surveillance is a pressing need. We used missing data methods to impute breast cancer recurrence and estimate the risk of recurrence in the Cancer Recurrence Information and Surveillance Program (CRISP) cohort in the Georgia Cancer Registry. The imputation model was based on an internal validation substudy and indicators recorded in the registry (e.g., pathology reports, imaging claims), prognostic variables (e.g., stage at diagnosis), and characteristics associated with missing data (e.g., insurance coverage). We pooled hazard ratios (HR) and 95% Confidence Intervals (CI) across 1000 imputed datasets, adjusted for age, stage, grade, subtype, race and ethnicity, marital status, and urban/rural county at diagnosis. There were 1,606 patients with a validated outcome (75% with breast cancer recurrence) and we imputed the outcome for the remaining 23,439 patients. We estimated an overall 7.2% incidence of recurrence between at least 1 year after diagnosis and up to 5 years of follow up. When comparing the hazards pooled across imputations, we found that some patterns differed from established patterns in mortality or survival, notably by race and ethnicity, underscoring the need for continued research on the descriptive epidemiology of breast cancer recurrence. These results provide new insights into surveillance for breast cancer survivors in Georgia, especially those with higher stage and grade tumors, of Hispanic ethnicity, and who may be lacking social support.