Perioperative immune trajectories in severe scoliosis surgery: a retrospective study of 2,736 patients undergoing corrective surgery for severe scoliosis.

Background: Severe scoliosis surgery is accompanied by substantial perioperative immune perturbation, yet the heterogeneity of postoperative immune recovery and its relationship with adverse clinical outcomes remain insufficiently defined. Methods: We performed a single-center retrospective longitudinal cohort study in one of China's largest care centers for severe scoliosis, including 2,736 patients who underwent corrective surgery for severe scoliosis between January 1, 2018, and December2024, 2024 and had serial immune measurements at preoperative baseline, 6 h, postoperative day (POD) 1, POD3, POD5, and discharge. Latent class mixed modeling based on absolute lymphocyte count, neutrophil-to-lymphocyte ratio, IL-6, and monocyte HLA-DR was used to identify perioperative immune trajectory phenotypes. The primary endpoint was major postoperative infection within 30 days. Models were adjusted for demographic, deformity-related, operative, and perioperative-management covariates, including surgical approach, steroid exposure, and opioid requirement. Results: Three distinct immune trajectory phenotypes were identified: adaptive recovery (n = 1,290, 47.1%), intermediate recovery (n = 903, 33.0%), and persistent dysregulated immunity (n = 543, 19.8%). Compared with the other two classes, the persistent dysregulated class was characterized by lower perioperative lymphocyte counts, higher neutrophil-to-lymphocyte ratio and IL-6 levels, more sustained monocyte HLA-DR suppression, higher C-reactive protein levels, lower CD4/CD8 ratio, and delayed albumin recovery. Major postoperative infection occurred in 8.3%, 16.1%, and 24.5% of the three classes, respectively (P <0.001), while prolonged postoperative length of stay occurred in 7.1%, 25.5%, and 66.9%, respectively (P <0.001). After multivariable adjustment, intermediate recovery and persistent dysregulated immunity remained associated with major infection (adjusted odds ratio [aOR] 2.06, 95% CI 1.54-2.76, and 3.31, 95% CI 2.22-4.94, respectively; both P <0.001). Corresponding associations with prolonged length of stay were 3.71 (95% CI 2.80-4.91) and 17.84 (95% CI 12.34-25.79), respectively (both P <0.001). The infection model showed moderate discrimination (area under the receiver operating characteristic curve, 0.668). Conclusion: Severe scoliosis surgery is followed by clinically meaningful perioperative immune recovery phenotypes. These trajectories are associated with infection and prolonged hospitalization but should be interpreted as dynamic markers reflecting both host response and perioperative stress rather than as purely causal immune endotypes.