Abstract
BACKGROUND: Frailty has been globally recognized as a predictor of adverse postoperative outcomes. Frailty assessment using the five-factor modified frailty index (5-mFI) has recently gained traction; however, long-term outcomes are unknown in colorectal cancer (CRC) surgery. This study aimed to investigate whether the 5-mFI predicted long-term survival and cause of death on the basis of frailty severity in elderly patients who underwent CRC surgery and to determine the risk factors for mortality. METHODS: A total of 299 patients underwent CRC surgery with curative intent between January 2013 and December 2017. Patients were divided into three groups by the 5-mFI score: group 1 (5-mFI: 0 or 1; n = 164): no frailty; group 2 (5-mFI: 2; n = 91): moderate frailty; and group 3 (5-mFI: ≥ 3; n = 44): severe frailty. Clinicopathological variables, namely comorbidities, 5-mFI, prognostic nutrition index, operative/postoperative data, and outcome, including cause of death, were compared between the three groups. To identify factors associated with death from CRC- and non-CRC-related causes, univariate and multivariate analyses using a Cox regression model were performed. RESULTS: The immediate postoperative morbidity of patients with Clavien-Dindo grade ≥ III complications (9.1%) in group 3 was not significantly different from that in group 1 (9.1%) or group 2 (14.3%); however, the 30-day mortality rate (4.5%) in group 3 was significantly higher. Long-term disease-free survival was similar between frailty groups, suggesting that CRC surgery provides oncological benefit to patients irrespective of frailty. The 5-year survival rates in groups 1, 2, and 3 were 83.5%, 71.2%, and 47.9%, respectively, showing a significantly lower survival rate as frailty advanced. Sixty percent of the deaths in frail patients were due to respiratory failure and cardiovascular diseases. Multivariate analysis demonstrated that advanced age, higher 5-mFI score, and longer postoperative hospital stay were risk factors for mortality unrelated to CRC. Multivariate analysis also revealed that advanced tumor stage, carcinoembryonic antigen ≥ 5 ng/ml, undifferentiated tumor, and R1 resection were risk factors for CRC-related mortality. CONCLUSIONS: The 5-mFI score can predict postoperative short- and long-term outcomes and risk factors for mortality unrelated to CRC. Additionally, long-term survival was negatively associated with the 5-mFI score.