Frailty phenotype and long-term mortality after curative colorectal cancer surgery: a retrospective multicenter cohort study

Background Frailty is associated with poor short-term surgical outcomes, but whether the Fried frailty phenotype is associated with long-term mortality after curative colorectal cancer (CRC) surgery, independently of age and tumor stage, is uncertain. We examined this association using restricted mean survival time (RMST), which does not require proportional hazards. Methods This retrospective multicenter cohort study included 304 patients aged ≥60 years who underwent curative CRC surgery at three centers in China in 2018–2019. The preoperative Fried phenotype was analyzed as robust (n=131), pre-frail (n=57), or frail (n=116); binary (robust versus non-robust) and ordinal codings were secondary. The primary outcome was 5-year all-cause mortality. Because proportional hazards were violated (Schoenfeld test P = 0.011), the RMST difference at 60 months was the primary effect measure, adjusted by pseudo-observation regression for age, sex, and TNM stage, the minimal adjustment set from a directed acyclic graph. Stage-specific and time-varying-effect analyses were also performed. Results Over a median follow-up of 60 months, 27 patients (8.9%) died. Five-year survival was 96.2%, 84.2%, and 88.8% in robust, pre-frail, and frail patients (log-rank P = 0.015). Unadjusted RMST differences versus robust patients were −2.38 months (95% CI −4.45 to −0.78) for pre-frail and −2.73 months (−4.58 to −1.09) for frail patients. After adjustment for age, sex, and TNM stage they were −1.48 months (−3.42 to 0.45; P = 0.13) and −0.81 months (−2.56 to 0.93; P = 0.36); for non-robust versus robust patients, −1.16 months (−2.50 to 0.19; P = 0.092). Estimates were in the same direction within every stage but imprecise. The hazard ratio for non-robust patients appeared higher in the first 36 months than later, but period-specific estimates were unstable and did not differ significantly (P = 0.15). Conclusions The Fried frailty phenotype identified older patients with lower 5-year survival after curative CRC surgery. After adjustment for age and tumor stage the association was attenuated and imprecise, so an independent effect could not be established with 27 deaths. The apparent early concentration of excess mortality in non-robust patients is hypothesis-generating and requires confirmation in larger cohorts with treatment and cause-of-death data.