This retrospective analysis enrolled 1,436 CRC patients who underwent surgical resection at the First Affiliated Hospital of Guangxi Medical University between 2015 and 2017. The inclusion criteria were as follows: histological confirmation of CRC; availability of complete data on albumin, neutrophil percentage, and other relevant clinicopathological factors; and age ≥ 18 years.​ Exclusion criteria included: prior neoadjuvant chemotherapy before surgery; concurrent malignancies; pre-existing autoimmune diseases; acute or chronic inflammatory conditions that might affect neutrophil or albumin levels at the time of data collection; lack of follow-up data; and incomplete medical records.​ This study was approved by the Institutional Review Board of the First Affiliated Hospital of Guangxi Medical University, and informed consent was obtained from all participants.

Clinicopathological data were comprehensively collected from the hospital’s electronic medical record system. Specifically, this included patients’ baseline information: gender, age, height, weight, body mass index (BMI), presence of hypertension, presence of diabetes, and contact details. Tumor-related information encompassed tumor location (colon or rectum), TNM staging (determined in accordance with the 8th Edition of the American Joint Committee on Cancer (AJCC) staging system), tumor size, differentiation grade, perineural invasion, and vascular invasion. All laboratory parameters were measured under fasting conditions within 1 week before surgery, including neutrophil count, platelet count, lymphocyte count, albumin levels, and carcinoembryonic antigen (CEA) levels. The neutrophil-to-lymphocyte ratio (NLR) was calculated as neutrophil count (×10⁹/L) divided by lymphocyte count (×10⁹/L). The platelet-to-lymphocyte ratio (PLR) was defined as platelet count (×10⁹/L) divided by lymphocyte count (×10⁹/L). The Prognostic Nutritional Index (PNI) was defined as albumin (g/L) +5×lymphocytes (×10⁹/L). The ANLR was computed using the formula: ANLR = Albumin (g/L)/NLR.

Follow-up assessments were scheduled every 3 months for the first 2 years, every 6 months for the subsequent 3 years, and annually thereafter. Each follow-up included a detailed inquiry into symptoms and signs to detect potential recurrence or metastasis, supplemented by essential examinations: routine blood tests, biochemical analyses, tumor marker measurements, colonoscopy, and imaging studies (CT or MRI) for a comprehensive health evaluation.​ All 1,436 eligible patients from the First Affiliated Hospital of Guangxi Medical University (treated between 2015 and 2017) underwent long-term follow-up via telephone interviews and outpatient visits. The follow-up duration ranged from 1 to 107 months (mean: 61 months; median: 65 months; interquartile range: 41–78 months). Recurrence was confirmed by imaging and pathological findings. During the follow-up period, 580 patients (40.4%) died, and 398 (27.2%) experienced recurrence.​ The primary endpoints were progression-free survival (PFS) and overall survival (OS). PFS was defined as the time from surgical resection to the first occurrence of local recurrence, distant metastasis, or death. OS was measured as the time from surgery to death from any cause or the date of the last follow-up. Sarcopenia was diagnosed in accordance with the criteria established by the Asian Working Group for Sarcopenia (AWGS 2019): skeletal muscle mass index (SMI) < 6.92 kg/m² for males and < 5.13 kg/m² for females. According to previous literature (22), the SMI derived from the anthropometric equation was calculated as follows: [0.193 × weight (kg) + 0.107 × height (cm) - 4.157 × sex (1 = male, 2 = female) - 0.037 × age (year) - 2.631]/height squared (m²). For postoperative complications (modified Clavien Classification System), the time window was defined as 90 days after surgery, and we analyzed complications of Grade II or higher.

Statistical analyses were performed using SPSS 25.0 and R 4.0.2 software. Continuous variables were presented as mean ± standard deviation (SD) or median (interquartile range, IQR), while categorical variables were expressed as counts and percentages. The optimal cutoff value for ANLR was determined via receiver operating characteristic (ROC) curve analysis. Kaplan–Meier survival curves combined with log-rank tests were used to compare survival outcomes between the low- and high-ANLR groups. Cox proportional hazards models with three-knot restricted cubic spline (RCS) were used to illustrate ANLR as a continuous predictor of survival. Cox proportional hazards models were employed to identify independent prognostic factors. Logistic regression analysis was employed to determine the independent association between ANLR and sarcopenia/complications in CRC patients. Nomograms were constructed based on the significant variables identified in the Cox regression model using the rms package in R. The predictive performance of the nomograms was evaluated using the concordance index (C-index) and calibration curves. Decision curve analysis (DCA) was used to compare the clinical utility of the ANLR-based nomogram with that of the traditional TNM staging system. A two-tailed p-value < 0.05 was considered statistically significant.

This retrospective study analyzed 1,436 CRC patients with a median age of 58.17 years (SD: 13.12), including 903 males (62.9%). The cohort exhibited a median BMI of 22.04 kg/m² (IQR: 19.96–24.35), with preoperative comorbidities comprising hypertension (16.7%, n=240) and diabetes (6.1%, n=88). Tumor profiling revealed advanced disease: 74.4% (n=1,069) presented T3–T4 stage lesions, 47.0% (n=675) had TNM stage III–IV disease, and 9.5% (n=136) showed distant metastases (M1). Nodal involvement distribution was N0 (56.0%, n=804), N1 (27.6%, n=397), and N2 (16.4%, n=235). Perineural and vascular invasion occurred in 10.4% (n=149) and 17.2% (n=247) of cases, respectively, while 51.0% (n=733) had rectal tumors. Pathologically, 13.1% (n=188) demonstrated poor differentiation. Systemic biomarker assessment showed median CEA levels of 3.87 ng/mL (IQR: 2.07–10.73). Treatment modalities included radiotherapy (9.3%, n=133) and chemotherapy (45.6%, n=655). During follow-up, mortality and recurrence rates reached 40.4% (n=580) and 27.2% (n=398), respectively. Resource utilization metrics indicated a median hospitalization duration of 17.00 days (IQR: 11.00–21.00) with associated costs of ¥49,541.85 (IQR: ¥44,564.47–55,983.79).

The optimal ANLR cutoff was established at 15.51 (AUC = 0.582; sensitivity = 0.644, specificity = 0.507), which we used to stratify 1,436 patients into low ANLR (<15.51, n=599) and high ANLR (≥15.51, n=837) groups (Supplementary Figure 1). The bootstrap-corrected C-index was 0.561, and the bootstrap-corrected calibration slope was 0.679. The distribution of 1,000 Bootstrap-derived cut-off values was concentrated around the original threshold (Supplementary Figure 2), with no significant skewness, indicating minimal single-center bias. Notably, lower ANLR values demonstrated significant associations with multiple adverse clinical features: male gender, advanced age, reduced BMI, hypertension, poorly differentiated tumors, rectal tumor location, larger tumor diameter, elevated CEA levels, prolonged hospitalization, and higher treatment costs. Critically, the low ANLR group exhibited substantially worse clinical outcomes, with significantly higher mortality (49.1% vs. 34.2%, p<0.001) and recurrence rates (31.4% vs. 25.1%, p<0.001) compared to the high ANLR group (Table 1).

Spearman correlation analysis showed moderate correlations between ANLR and established inflammation-nutrition indices (NLR, PLR, PNI), especially with PNI (Supplementary Figure 3), which may be attributed to overlapping components of the indicators. We conducted ROC analysis to evaluate the prognostic utility of the ANLR against established biomarkers—NLR, PLR, and PNI. Consistently across survival endpoints, ANLR demonstrated superior discriminative ability. For PFS, ANLR achieved higher AUC values than comparator indices at both 3 years (0.553 vs. NLR: 0.544, PLR: 0.533, PNI: 0.556) (Supplementary Figure 4A) and 5 years (0.554 vs. NLR: 0.545, PLR: 0.540, PNI: 0.559) (Supplementary Figure 4B). This performance advantage extended to OS prediction, where ANLR similarly outperformed other markers at 3-year follow-up (AUC: 0.571 vs. NLR: 0.565, PLR: 0.555, PNI: 0.566) (Supplementary Figure 4C) and maintained superior accuracy at 5 years (0.559 vs. NLR: 0.551, PLR: 0.544, PNI: 0.561) (Supplementary Figure 4D). C-index analysis supplemented with the “ANLR + traditional inflammation-nutrition indices” combined model revealed that the AUC of the combined model was slightly higher than that of single traditional indices. Although the difference was not statistically significant, it confirmed that ANLR can provide incremental predictive information rather than merely duplicating the value of existing indicators (Supplementary Tables 2, S3).

Kaplan-Meier analysis demonstrated significantly worse survival outcomes in low-ANLR patients versus high-ANLR counterparts, with markedly reduced 5-year rates for both PFS (48.2% vs. 63.0%, p<0.001) and OS (50.9% vs. 65.8%, p<0.001; Figures 1A, B). Critically, this prognostic pattern persisted across all subgroups. In TNM staging subgroups, low-ANLR patients exhibited compromised survival regardless of disease stage (Stage I-II: PFS 66.9% vs. 76.3%, p=0.014; OS 69.2% vs. 79.1%, p=0.006; Stage III-IV: PFS 29.3% vs. 46.8%, OS 32.3% vs. 49.7%, both p<0.001; Figures 2A–D). Similarly, for tumor location subgroups, colon cancer patients with low ANLR had inferior PFS (43.9% vs. 61.5%, p<0.001) and OS (46.6% vs. 65.4%, p=0.001) (Supplementary Figures 5A, B), while rectal cancer patients showed parallel deficits (PFS: 51.4% vs. 65.0%, OS: 54.0% vs. 66.4%, both p<0.001) (Supplementary Figures 5C, D).

A non-linear relationship was found between ANLR and PFS/OS risk, revealing an L-shaped pattern (p < 0.001). As the ANLR increased, the HR gradually decreased (Figure 3). Multivariate Cox regression analysis revealed that for every 1 standard deviation (SD) increase in ANLR, the risk of adverse PFS in CRC patients decreased by 9.5% (HR = 0.905, 95% CI: 0.831–0.986, p = 0.023). Compared with patients in the low-ANLR group, those in the high-ANLR group had a significantly lower risk of adverse PFS (HR = 0.745, 95% CI: 0.630–0.880, p = 0.001). Quartile analysis of ANLR showed that patients in the second, third, and fourth quartiles had adverse PFS rates 0.817, 0.634, and 0.635 times those of patients in the first quartile, respectively (Table 1).​ For OS, multivariate Cox regression analysis indicated that for every 1 SD increase in ANLR, the risk of adverse OS decreased by 9.5% (HR = 0.905, 95% CI: 0.828–0.988, p = 0.026). The low-ANLR group had a 26.1% higher risk of adverse OS than the high-ANLR group (HR = 0.739, 95% CI: 0.622–0.878, p = 0.001). As ANLR increased, the HR for adverse OS gradually decreased: quartile 2 (Q2: 0.787), quartile 3 (Q3: 0.632), and quartile 4 (Q4: 0.602) were all associated with a reduced risk of adverse OS (Table 2). Sensitivity analysis showed that excluding patients who died within 3 months after surgery did not alter the prognostic effect of ANLR (Supplementary Tables 4, S5). Multivariable forest plot analysis demonstrated that ANLR served as an independent prognostic factor for PFS across most patient subgroups (Supplementary Figure 6A). Similarly, in terms of OS, patients with low ANLR generally had poorer prognoses compared to those with high ANLR in the majority of subgroups (Supplementary Figure 6B).

For sarcopenia, multivariate analysis showed that each 1 SD increase in ANLR was associated with an 18.0% reduction in risk (OR = 0.820, 95% CI: 0.710–0.960, p = 0.009). The high-ANLR group had a 40.7% lower risk of sarcopenia than the low-ANLR group (Odds Ratio [OR] = 0.593, 95% CI: 0.442–0.796, p < 0.001). As ANLR increased, the OR for sarcopenia gradually decreased: compared with Q1, Q2 (0.816), Q3 (0.472), and Q4 (0.541) were all associated with a reduced risk of sarcopenia (Table 3).​ Regarding complications, each 1 SD increase in ANLR was linked to a 22.0% reduction in risk (OR = 0.780, 95% CI: 0.670–0.890, p < 0.001). The high-ANLR group had a 43.6% lower risk of complications than the low-ANLR group (OR = 0.564, 95% CI: 0.429–0.742, p < 0.001). With increasing ANLR, the HR for complications gradually decreased, with Q2 (0.561), Q3 (0.378), and Q4 (0.476) all associated with a reduced risk of adverse complications compared with Q1 (Table 4).

Multivariate Cox regression analysis identified seven independent prognostic factors influencing PFS in CRC patients: age, T stage, N stage, M stage, vascular invasion, CEA levels, and ANLR (Supplementary Table 6). For OS, the independent risk factors were age, T stage, N stage, M stage, vascular invasion, differentiation, CEA levels, and ANLR (Supplementary Table 7). The full coefficients of the Cox proportional hazards models for PFS and OS in the Supplementary Tables 8, S9. Based on these key factors, nomograms were developed to predict 1-, 3-, and 5-year PFS (Figure 4) and OS (Figure 5) in CRC patients. The predicted probabilities of PFS and OS at these time points were calculated by summing the scores assigned to each variable. To facilitate personalized prognostication for CRC patients, we deployed the PFS and OS nomogram models on a Shiny server, developing dedicated webpages for predicting PFS and OS in CRC patients. These webpages enable more convenient and individualized prognostic prediction for patients, with the URLs available at https://hailun.shinyapps.io/ANLR_PFS/(Accessed: [12/28/2025]) and https://hailun.shinyapps.io/ANLR_OS/(Accessed: [12/28/2025]) respectively (Supplementary Table 10).

The C-indices for the PFS and OS nomograms were 0.719 (95% CI: 0.698–0.740) and 0.727 (95% CI: 0.705–0.749), respectively. Calibration curves for 1–5-year postoperative PFS (Supplementary Figure 7A) and OS (Supplementary Figure 7B) demonstrated optimal consistency between the predicted survival probabilities and actual observed values. The 1-, 3-, and 5-year area under the ROC curve (AUC) values for the PFS nomogram were 0.802, 0.773, and 0.762, respectively, while those for the OS nomogram were 0.771, 0.774, and 0.762 (Supplementary Figures 8A, B). Patients were stratified into high- and low-score groups based on the median scores derived from the nomogram. Analysis showed that the high-score group had significantly poorer PFS and OS compared to the low-score group (Supplementary Figure 9). The DCA revealed that the ANLR-based nomograms provided greater clinical utility than the traditional TNM staging system for both PFS and OS at the 1-, 3-, and 5-year time points (Supplementary Figures 10A, B).