Integrated multitrajectory modeling of inflammatory, autonomic, sleep, and activity dynamics reveals distinct postoperative recovery endotypes associated with fatigue and sleep disturbance.

BACKGROUND

Recovery after colorectal cancer surgery reflects interacting inflammatory, autonomic, sleep, and behavioral processes, yet these domains are typically evaluated in isolation. Whether they form reproducible multidomain recovery patterns and how such patterns relate to symptoms and disease-free survival (DFS) remain unclear.

METHODS

In this prospective cohort, 292 patients (mean age 63.5 ± 9.2 years; 63.7% male) undergoing curative resection for stage I-III colorectal adenocarcinoma underwent repeated assessments of interleukin-6 (IL-6), C-reactive protein, nocturnal RMSSD, sleep efficiency, and daily step counts on postoperative days 1, 3, 7, 14, and 30. Group-based trajectory modeling identified domain-specific patterns, and multitrajectory modeling derived multidomain recovery endotypes. Fatigue and sleep were assessed at 1, 3, and 6 months using the Brief Fatigue Inventory (BFI) and Pittsburgh Sleep Quality Index (PSQI). DFS was followed for 36 months.

RESULTS

Three trajectory classes were identified within each domain. Multitrajectory modeling yielded four endotypes: Coordinated Recovery (n = 104), Inflammation-Persistent (n = 83), Behavioral-Lagging (n = 63), and Globally Impaired (n = 42). Major complications increased across endotypes (10.6%, 15.7%, 15.9%, 28.6%; p = 0.018). At 6 months, clinically significant fatigue and poor sleep increased across endotypes (both p < 0.001). During follow-up, 64 DFS events occurred. Three-year DFS estimates were 86.5%, 78.2%, 74.6%, 61.3% across endotypes (log-rank p = 0.014), although survival analyses were exploratory.

CONCLUSIONS

Postoperative recovery follows coherent multidomain trajectories, forming distinct endotypes associated with symptom burden and survival. Multidomain phenotyping may support systems-level characterization of postoperative recovery and inform future personalized approaches.