Predictors of late recurrence of colorectal cancer (CRC) among older long-term survivors using claims-based evidence.

Abstract

66 Background: There are established features that predict early recurrence of CRC in the first 5 years from diagnosis, but predictors of late recurrence 5-10 years after diagnosis are less well understood. Long-term CRC survivors require multidisciplinary follow-up accounting for both cancer and non-cancer related conditions. Identifying predictors of cancer recurrence after 5 years of disease-free survival may help inform survivorship care. Methods: This was a retrospective cohort study of patients aged ≥66 in the SEER-Medicare database with non-metastatic CRC diagnosed between 2003 and 2011. As SEER does not collect recurrence data, we used a validated algorithm to identify treated recurrence in Medicare claims. Patients with early recurrence or a second primary malignancy in years 0-5 were excluded. Demographic and cancer-specific clinical data were collected at time of cancer diagnosis. Comorbidities were identified between 4 -5 years after cancer diagnosis and recurrence was modeled starting at year 5 using restricted mean survival time (RMST) regression. Estimates from RMST models were converted to RMST ratios (ratio <1 indicates shorter time to event). The Fine-Gray model was used to estimate cumulative incidence of CRC recurrence or mortality between 5 and 10 years after cancer diagnosis. Results: 20,255 patients with non-metastatic CRC were included: 16,158 with colon cancer (33.4% stage I, 41.7% stage II, 25.0% stage III), and 4,097 with rectal cancer (43.1% stage I, 30.1% stage II, 26.7% stage III). Mean age at cancer diagnosis was 77 years. In those who remained disease free five years after initial diagnosis, late recurrence developed in 4.8% (973/20,255) of the combined CRC cohort, 4.2% (679/16,158) of those with colon cancer, and 7.2% (294/4,097) of those with rectal cancer. In colon cancer, stage III disease was associated with a shorter time to recurrence (RMST ratio = 0.97, CI = 0.96-0.98). In both colon (RMST ratio = 0.98, 95% CI 0.96-0.99) and rectal cancer (RMST ratio = 0.97, 95% CI 0.94-0.99) having ≥3 comorbidities was associated with a shorter RMST, consistent across disease stages. For colon cancer, the cumulative incidence of CRC-specific mortality vs. all-cause mortality in years 5-10 from diagnosis was 3% (95% CI 2.8-3.0) vs. 35% (95% CI 34.0-36.0), respectively. For rectal cancer, during the same period, the corresponding rates were 6% (95% CI 5.4-7.0) vs. 30% (95% CI 28.4-31.0). Conclusions: Late CRC recurrence is rare in the SEER-Medicare population. Primary risk factors are advanced stage at diagnosis and multiple co-morbidities. Patients who survive 5 years from their diagnosis have a higher risk of all-cause mortality compared to CRC-specific mortality. Interpretation of data is limited by the validated algorithm used to detect recurrence, as it may underestimate recurrence in patients who did not undergo treatment or workup for recurrent disease.