Association between preoperative neutrophil percentage-to-albumin ratio and survival outcomes in patients with colorectal cancer undergoing radical surgery: a retrospective study with external validation
Background Accurate preoperative risk stratification remains a major clinical challenge in patients with colorectal cancer (CRC) undergoing radical surgery. Simple and readily available biomarkers that reflect systemic inflammation and nutritional status may improve prognostic assessment in routine clinical practice. Methods This retrospective study included 581 patients with CRC who underwent radical surgery as the training cohort. Overall survival (OS) and disease-free survival (DFS) were the primary endpoints. NPAR was analyzed primarily as a continuous variable standardized per 1-SD increase. A pre-specified final model including NPAR, age, sex, tumor site, pathological stage, and log10-transformed CEA was used consistently across Cox regression, nomogram development, time-dependent ROC analysis, calibration, decision-curve analysis, and external validation. Dichotomized NPAR was retained only for descriptive and supplementary analyses. External validation was performed in 483 patients from an external clinical database. Results In the training cohort, higher preoperative NPAR was associated with adverse clinicopathological features. After adjustment for age, sex, tumor site, pathological stage, and log10-transformed CEA, each 1-SD increase in NPAR was associated with shorter OS (HR 2.19, 95% CI 1.90–2.53, p < 0.001) and DFS (HR 1.88, 95% CI 1.62–2.19, p < 0.001). The final model showed optimism-corrected C-indexes of 0.835 for OS and 0.752 for DFS in the training cohort. In the external cohort, the association between continuous NPAR and outcomes was directionally consistent but attenuated for both OS (HR 1.21, 95% CI 1.08–1.35, p < 0.001) and DFS (HR 1.24, 95% CI 1.12–1.37, p < 0.001). External C-indexes were 0.690 for OS and 0.705 for DFS. Adding NPAR to the baseline clinicopathological model did not improve external time-dependent discrimination for OS and showed only limited incremental value for DFS. Low external recalibration slopes suggested possible optimism in the training model. Conclusion Preoperative NPAR showed an adjusted association with OS and DFS, and the direction of this association was reproduced in the external cohort. However, its incremental predictive value beyond conventional clinicopathological factors was limited, particularly for OS. NPAR may be considered a simple adjunctive marker of inflammatory and nutritional status, but its clinical utility for individualized risk prediction requires further validation in larger prospective cohorts.