ABSTRACT
Background
The ratio of red blood cell distribution width (RDW) to serum albumin (RAR) combines hematologic and nutritional information from routine blood tests and has shown potential value for prognostication in several diseases. This study aimed to evaluate the prognostic significance of preoperative RAR in patients with gastric cancer undergoing curative gastrectomy and to explore whether RAR could provide additional prognostic information beyond conventional indicators such as TNM stage.
Aims
This study evaluated the prognostic value of preoperative RAR in patients with gastric cancer undergoing curative resection. Specifically, we aimed to (i) determine the association between RAR and clinicopathological characteristics, (ii) assess its predictive value for overall survival (OS) and disease‐free survival (DFS), and (iii) evaluate the complementary prognostic value of RAR beyond conventional indicators such as TNM stage and other markers.
Methods
This is a retrospective cohort study including 6988 patients with gastric cancer who underwent curative resection at a tertiary cancer center between July 2011 and December 2022. We explored associations between preoperative RAR and a range of clinicopathological characteristics. Overall survival was estimated by the Kaplan–Meier method, and Cox proportional hazards models were used to assess factors influencing prognosis. The discriminative ability of RAR was evaluated using time‐dependent ROC analysis.
Results
An elevated baseline RAR was associated with older age, larger tumor dimensions, higher TNM classification, lymphatic spread, deeper tumor invasion, and the presence of vascular and perineural invasion, with poorer overall survival (OS) and disease‐free survival (DFS). Time‐dependent receiver operating characteristic analyses indicated that RAR consistently achieved higher discriminative performance than neutrophil‐to‐lymphocyte ratio (NLR), lymphocyte‐to‐monocyte ratio (LMR), and platelet‐to‐lymphocyte ratio (PLR) across follow‐up. However, the absolute AUC values remained modest, indicating that RAR alone has limited predictive capacity.
Conclusion
RAR may facilitate postoperative risk stratification as an accessible prognostic marker.
Keywords: curative gastrectomy, gastric cancer, overall survival, prognostic biomarker, red blood cell distribution width to albumin ratio
1. Introduction
Globally, stomach cancer ranks as the fifth most prevalent malignant neoplasm and constitutes the fourth leading cause of cancer‐related mortality [1, 2]. Despite advances in therapy, the prognosis for affected individuals remains poor, particularly in advanced stages, where recurrence and metastatic spread are the principal drivers of cancer death [3, 4].
After curative surgery, prognostic assessment continues to rely on the AJCC tumor–node–metastasis (TNM) framework, which remains the benchmark method for predicting postoperative outcomes [5]. However, survival outcomes vary substantially among patients with the same TNM stage, highlighting the need for additional prognostic markers.
Previous studies have highlighted a connection between the host's inflammatory response in the host and cancer prognosis, particularly in breast cancer and other malignancies [6, 7]. Moreover, individuals with malignant disease frequently present with higher circulating white blood cell counts, elevated concentrations of acute‐phase proteins, and lower serum albumin levels [8].
Increasing evidence links systemic inflammation to cancer progression and prognosis. Blood‐based inflammatory indicators including NLR, LMR, and PLR have been studied extensively in gastric cancer [9, 10, 11]. Prevailing evidence, including findings from our own study, indicates that the systemic inflammation score (SIS)—a composite of serum albumin and LMR—demonstrates superior prognostic accuracy compared with alternative inflammatory indices [12, 13]. In addition, lipid‐related biomarkers have also been investigated in earlier work [14].
RAR, which integrates RDW with serum albumin, has emerged as a valuable prognostic biological indicator across diverse conditions, notably cognitive impairment and intracerebral hemorrhage [15, 16, 17, 18]. However, despite its demonstrated prognostic value in these conditions, a large cohort research on RAR's role in tumors—particularly gastric cancer—is still limited. The specific prognostic significance of RAR in gastric cancer patients, as well as its comparative performance against established biomarkers, remains unclear and warrants further investigation.
This investigation intends to elucidate the relationship between RAR and both clinicopathological parameters and survival outcomes in post‐curative gastrectomy gastric cancer patients. In addition, it evaluates whether RAR provides incremental prognostic information beyond established biomarkers and the TNM staging system, particularly for postoperative risk stratification.
2. Methods
2.1. Patient Population
Ethical clearance and institutional oversight for this work were provided by the Institutional Review Board of Fudan University Shanghai Cancer Center. This analysis cohort consisted of 15,307 patients who underwent curative gastrectomy for gastroesophageal junction tumors or gastric adenocarcinoma at Fudan University Shanghai Cancer Center from July 2011 to December 2022. The exclusion criteria are as follows: age below 18 years (n = 2); presence of concurrent malignancies (n = 207); evidence of distant metastases (n = 560); history of previous oncologic surgery (n = 453); retrieval of fewer than 16 lymph nodes (n = 484); lack of preoperative laboratory assessments within 7 days before surgery (n = 172); missing key clinicopathological or follow‐up information (n = 4769); administration of anti‐inflammatory drugs (e.g., antibiotics) or immunosuppressants (e.g., corticosteroids) within 3 months prior to surgery (n = 720); diagnosis of chronic inflammatory conditions, including autoimmune diseases (n = 19); receipt of neoadjuvant treatments, such as chemotherapy or chemoradiotherapy (n = 865); undergoing emergency radical surgery due to complications such as hemorrhage or perforation (n = 68) (The flow of patient selection is shown in Figure 1). Among 15,307 eligible patients, 6988 were analyzed (8319 excluded). To assess selection bias, we compared baseline characteristics (age, sex, BMI, clinical stage) between included and excluded cohorts for Table 1, reporting SMDs with p‐values. We estimated inclusion probabilities using a multivariable logistic regression model and applied stabilized inverse probability of inclusion weighting (IPIW). The propensity model included variables available in both the included and excluded cohorts: age, sex, body mass index (BMI), tumor size, tumor location, histology, clinical TNM stage, and year of surgery. The distribution of stabilized weights was symmetric and centered near 1.0, and post‐weighting covariate balance was acceptable for all variables (all weighted |SMD| < 0.10; Figure 2). Detailed annual distributions of exclusion reasons and missing data were further summarized to evaluate whether data incompleteness was concentrated in specific enrollment periods.
FIGURE 1.
Diagrammatic flow of participant cohort delineation.
TABLE 1.
Baseline demographics and clinicopathologic parameters of Gastric Cancer patients.
| Variables | RAR | p | ||
|---|---|---|---|---|
| Total (n = 6988) | Low (n = 4343) | High (n = 2645) | ||
| Age (years) | ||||
| < 70 | 5671 (81.2%) | 3766 (86.7%) | 1905 (72.0%) | < 0.0001 |
| ≥ 70 | 1317 (18.8%) | 577 (13.3%) | 740 (28.0%) | |
| Sex | ||||
| Male | 4807 (68.8%) | 3014 (69.4%) | 1793 (67.8%) | 0.1667 |
| Female | 2181 (31.2%) | 1329 (30.6%) | 852 (32.2%) | |
| Anemia | ||||
| No anemia | 4621 (66.1%) | 3719 (85.6%) | 902 (34.1%) | < 0.0001 |
| Anemia | 2367 (33.9%) | 624 (14.4%) | 1743 (65.9%) | |
| BMI | ||||
| Normal | 4795 (68.6%) | 2945 (67.8%) | 1850 (70.0%) | < 0.0001 |
| Underweight | 479 (6.9%) | 235 (5.4%) | 244 (9.2%) | |
| Overweight | 1590 (22.8%) | 1079 (24.8%) | 511 (19.3%) | |
| Obese | 124 (1.8%) | 84 (2.0%) | 40 (1.5%) | |
| Histology | ||||
| Well and moderately | 1966 (28.1%) | 1260 (29.0%) | 706 (26.7%) | 0.0378 |
| Poorly and others | 5024 (71.9%) | 3087 (71.0%) | 1937 (73.3%) | |
| Tumor location | ||||
| Upper | 1271 (18.2%) | 747 (17.2%) | 524 (19.8%) | 0.0728 |
| Middle/lower | 5717 (81.9%) | 3596 (82.8%) | 2121 (80.2%) | |
| Tumor size (mm) | ||||
| < 30 | 2601 (37.2%) | 1993 (45.9%) | 608 (23.0%) | < 0.0001 |
| ≥ 30 | 4387 (62.8%) | 2350 (54.1%) | 2037 (77.0%) | |
| Borrmann classification | ||||
| 0/1/2 | 335 (4.8%) | 200 (4.6%) | 135 (5.1%) | 0.3636 |
| 3/4 | 6653 (95.2%) | 4143 (95.4%) | 2510 (94.9%) | |
| T stage | ||||
| T1 | 2279 (32.6%) | 1736 (40.0%) | 543 (20.5%) | < 0.0001 |
| T2 | 899 (12.9%) | 586 (13.5%) | 313 (11.8%) | |
| T3 | 1681 (24.1%) | 941 (21.7%) | 740 (28.0%) | |
| T4 | 2129 (30.5%) | 1080 (24.8%) | 1049 (39.6%) | |
| N stage | ||||
| N0 | 3107 (44.5%) | 2179 (50.2%) | 928 (35.1%) | < 0.0001 |
| N1 | 1126 (17.1%) | 663 (15.3%) | 463 (17.5%) | |
| N2 | 1113 (16.0%) | 648 (14.9%) | 465 (17.6%) | |
| N3 | 1642 (23.5%) | 853 (19.6%) | 789 (29.8%) | |
| pStage | ||||
| I | 2557 (36.6%) | 1906 (43.9%) | 651 (24.7%) | < 0.0001 |
| II | 1658 (23.7%) | 995 (22.9%) | 663 (25.1%) | |
| III | 2773 (39.7%) | 1442 (33.2%) | 1331 (50.3%) | |
| Lymphovascular invasion | ||||
| Absent | 2618 (37.5%) | 1877 (43.2%) | 741 (28.0%) | < 0.0001 |
| Present | 4370 (62.5%) | 2466 (56.8%) | 1904 (72.0%) | |
| Perineural invasion | ||||
| Absent | 3658 (52.3%) | 2462 (56.7%) | 1196 (45.2%) | < 0.0001 |
| Present | 3330 (47.7%) | 1881 (43.3%) | 1449 (54.8%) | |
Note: Body mass index (BMI) was categorized according to the World Health Organization (WHO) adult criteria: underweight (< 18.5 kg/m2), normal weight (18.5–24.9 kg/m2), overweight (25.0–29.9 kg/m2), and obese (≥ 30.0 kg/m2). Hemoglobin (Hb) levels were classified according to the Chinese diagnostic criteria for adult anemia (GB/T 26366–2022), defining anemia as Hb < 130 g/L in men or < 120 g/L in women; these thresholds are consistent with the World Health Organization criteria. All variables presented in Table 1 had complete data for the final cohort (n = 6988).
FIGURE 2.
Distribution of stabilized inverse probability of inclusion weights (IPIW) and covariate balance assessment after weighting. The plot displays weighted absolute standardized mean differences (SMDs) for all covariates following the application of stabilized IPIW.
2.2. Follow‐Up Investigation and Recurrence Assessment
Postoperative surveillance involved outpatient visits or telephone follow‐ups every 3–6 months. Each follow‐up included a physical examination, serum biomarker assays (CA19‐9, CEA, CA72‐4, and CA125), and imaging investigations when indicated. Imaging modalities comprised upper‐abdominal contrast‐enhanced CT, magnetic resonance imaging, or chest radiographs, performed at 3‐month intervals during the first 2 years after surgery and every 6 months thereafter. Overall survival (OS) was defined as the interval from date of curative resection to the time of demise or the most recent assessment. Disease‐free survival (DFS) was defined as the interval from surgery to the first occurrence of tumor recurrence or the latest tumor assessment; patients who died without documented recurrence were censored at the date of death.
2.3. Surgical Procedures and Quality Control
Gastrectomies were performed by gastrointestinal surgeons with extensive experience in gastric cancer at a high‐volume tertiary cancer center. Standard operative techniques were applied, and pathological staging followed the AJCC 8th edition. Data collection was centralized to maintain consistency. Postoperative management followed routine perioperative protocols. Follow‐up included physical examination, serum tumor markers (CA19‐9, CEA, CA72‐4, CA125), and imaging (MRI, chest radiographs, contrast‐enhanced CT) every 3 months during the first 2 years and every 6 months thereafter. Recurrence was confirmed by radiological and pathological findings.
2.4. Data Extraction
Patient demographic information, including age and sex, along with clinical and pathological data such as tumor site, size, type of surgical resection, macroscopic appearance, depth of invasion, histological classification, number of positive and examined lymph nodes, as well as lymphovascular and perineural invasion status, were collected. All gastric cancer cases were histologically confirmed by postoperative paraffin‐embedded hematoxylin–eosin staining and independently reviewed by two experienced pathologists. Preoperative laboratory test results and follow‐up data were also documented. Available postoperative adjuvant treatment information was additionally re‐retrieved from medical records and follow‐up documents. Staging was determined according to the AJCC TNM 8th edition guidelines. Blood samples for serum and peripheral counts of neutrophils, lymphocytes, and monocytes were obtained within 1 week before surgery. RAR was computed by dividing RDW (percentage) by serum albumin (g/dL). RDW was measured as the coefficient of variation (RDW‐CV), which is the routinely reported parameter in clinical practice. The optimal cutoff value for RAR was determined using maximally selected rank statistics (R package “maxstat”), which identified 3.18 as the threshold. Patients were classified into low‐RAR (≤ 3.18) and high‐RAR (> 3.18) groups for subsequent analyses. The optimal cutoff value for tumor size was determined using maximally selected rank statistics based on log‐rank tests, which identified 3.3 cm as the value that maximized survival separation. Considering clinical interpretability and common practice, tumor size was therefore dichotomized at 3 cm (< 3 cm vs. ≥ 3 cm).
2.5. Statistical Analysis
Statistical analyses were conducted with SPSS version 20.0 (SPSS Inc., Chicago, IL, USA). Categorical data were analyzed with chi‐square or Fisher's exact tests to assess associations; continuous data were analyzed with Wilcoxon rank‐sum or Kruskal–Wallis tests, as appropriate. Survival probabilities were generated by the Kaplan–Meier method, with intergroup differences assessed through the log‐rank test. Because patients were enrolled between 2011 and 2022, not all had completed 5 years of follow‐up. Therefore, 5‐year OS and DFS were estimated using Kaplan–Meier survival functions, with patients censored at their last follow‐up date. Potential prognostic factors for gastric cancer were first screened in univariate Cox regression analyses and subsequently integrated into a Multivariable Cox PH (Proportional Hazards) model. Variables with p < 0.05 in univariate analysis and other clinically relevant covariates—including age, sex, tumor size, tumor location, histology, TNM stage, BMI, and anemia—were entered into the multivariate model. T stage was primarily analyzed as a dichotomized variable (T3–T4 vs. T1–T2) to reflect clinically meaningful separation between advanced and early local invasion and to ensure model parsimony and stability. In sensitivity analyses, T stage was additionally modeled as a categorical variable (with T1 as the reference) and as an ordinal variable to evaluate potential dose–response effects. The direction and magnitude of the associations were consistent across modeling strategies; therefore, the dichotomized classification was retained for interpretability. No automatic stepwise variable selection was performed to avoid overfitting; instead, variables were retained based on clinical relevance and model stability. All VIF values were below 3.5, indicating no major multicollinearity. A mild correlation (VIF = 3.162) was observed between lymph node metastasis and pathological TNM stage, which is expected given the hierarchical structure of the AJCC staging system, in which pathological stage grouping is derived from T, N, and M components. Despite this structural correlation, both variables were retained in the final model due to their complementary clinical relevance, with TNM stage reflecting overall disease burden and nodal status providing specific prognostic information. The predictive performance of biomarkers such as NLR, RAR, LMR, and PLR was examined using time‐dependent receiver operating characteristic (ROC) analyses implemented in R (version 4.5.1), primarily using the timeROC package. Time‐dependent AUCs were estimated based on the Kaplan–Meier method, and bootstrap resampling was applied to compare the corresponding ROC curves. This cohort study has been reported in line with the STROCSS 2025 criteria [19]. A nomogram was constructed based on the final multivariable Cox model using the rms package in R. Discrimination of the nomogram was assessed by the concordance index (C‐index), and calibration was evaluated using bootstrap‐corrected calibration curves for 3‐year and 5‐year overall survival (OS). To evaluate the specific contribution of RAR to the nomogram, calibration curves were generated to compare the agreement between predicted and observed OS probabilities at 1, 3, and 5 years for models with and without RAR. For visual presentation, high‐resolution TIFF and PDF versions were exported. To further clarify whether RAR provides prognostic information beyond pathological stage, exploratory stage‐specific Cox analyses were first performed in patients with pStage I, II, and III disease separately. Considering that pStage I–II disease represents an anatomically lower‐stage population and that prognosis in pStage III disease may be largely driven by advanced anatomical tumor burden, we further conducted a focused analysis in the combined pStage I–II subgroup. In this subgroup, Kaplan–Meier curves were used to compare OS and DFS between the high‐ and low‐RAR groups, and multivariable Cox regression models were adjusted for pStage, age, tumor size, and lymphovascular invasion. To further assess the potential influence of missing data in earlier years, a sensitivity analysis was conducted by restricting the cohort to patients treated between 2015 and 2022, during which data completeness was relatively improved.
3. Results
3.1. Correlations Between RAR and Clinicopathological Characteristics
The study enrolled 6988 patients. Table 1 provides the clinical and pathological features of the cohort. Of these participants, 4807 were male (68.8%) while Females totaled 2181 (31.2%), with a median age of 60.0 years (range: 20–87). In line with the AJCC's eighth edition TNM classification, Stage I accounted for 39.3%, Stage II 20.8% and Stage III 39.9%.
As a considerable number of patients were excluded due to missing data and predefined exclusion criteria, we further summarized the annual distribution of exclusion reasons and missing variables. Missingness was more frequent in earlier years, particularly before 2015. In the sensitivity analysis restricted to patients treated between 2015 and 2022, the effect estimates for RAR were directionally consistent with those observed in the main cohort, although the associations did not reach statistical significance (Tables S1–S3). After re‐retrieval of available postoperative treatment records, 4190 of 4431 pStage II–III patients had available postoperative treatment information. Among them, 3838 patients (91.6%) received postoperative adjuvant chemotherapy. As presented in Table 1, several clinical and pathological features showed significant differences between the two RAR groups. High RAR linked to senior status (≥ 70 years: 28.0% vs. 13.3%, p < 0.0001), greater tumor burden (≥ 30 mm: 77.0% vs. 54.1%, p < 0.0001), advanced T stage (T4: 39.6% vs. 24.8%, p < 0.0001), nodal metastasis (N3: 29.8% vs. 19.6%, p < 0.0001), higher pathological stage (stage III: 50.3% vs. 33.2%, p < 0.0001), lymphovascular invasion (72.0% vs. 56.8%, p < 0.0001), perineural invasion (54.8% vs. 43.3%, p < 0.0001), and poor differentiation (73.3% vs. 71.0%, p = 0.0378). Conversely, differences were not statistically significant in sex distribution (p = 0.1667), tumor location (p = 0.0728), or Borrmann type (p = 0.3636) between the groups.
3.2. Link Between the RAR Score and Patient Survival Outcomes
The Kaplan–Meier survival analyses demonstrate that an elevated RAR is significantly tied to worse 5‐year OS and DFS. Specifically, among patients with pStage I disease (Figure 3A,D for OS and DFS, respectively), patients with high RAR scores experienced significantly poorer overall survival (OS) and disease‐free survival (DFS) (OS: p = 0.0062; DFS: p = 0.0003). This association remained evident among pStage II patients (Figure 3B,E), with higher RAR levels associated with reduced OS (p = 0.0002) and DFS (p = 0.0019). Similarly, in pStage III patients (Figure 3C,F), elevated RAR levels were consistently associated with significantly worse OS and DFS (both p < 0.0001). The optimal cutoff value of RAR (3.18) for survival stratification was determined by ROC analysis, as illustrated in Figure 4 and data extraction. Overall, Figure 3 illustrates that RAR stratified both overall and disease‐free survival within pathological stages, suggesting its potential value as a complementary prognostic factor to TNM staging.
FIGURE 3.
Survival analysis was conducted with Kaplan–Meier plots of OS (a–c) and DFS (d–f) for each pStage in the RAR‐high (> 3.18) or RAR‐low (≤ 3.18) groups. (a, d) 
, RAR‐high (n = 651); 
, RAR‐low (n = 1906); (b, e) 
, RAR‐high (n = 663); 
, RAR‐low (n = 995); (c, f) 
, RAR‐high (n = 1331); 
, RAR‐low (n = 1442).
FIGURE 4.
Dynamic ROC analysis was applied to evaluate the prognostic value of the RAR for five‐year overall survival. Using the maximally selected rank statistic, the RAR cutoff representing the optimum was 3.18. Consequently, patients with RAR ≤ 3.18 were assigned to the low‐RAR group, whereas those with RAR > 3.18 comprised the high‐RAR group for subsequent survival analyses.
To further assess whether RAR provides additional prognostic information beyond pathological stage, we first performed exploratory stage‐specific Cox analyses. The prognostic contribution of RAR was less evident in pStage III disease, where prognosis may be largely dominated by advanced anatomical tumor burden. Therefore, in accordance with the clinical question of identifying apparently lower‐stage but higher‐risk patients, we further focused on the combined pStage I–II subgroup as an anatomically lower‐stage population. In the pStage I–II subgroup, high RAR remained significantly associated with worse OS and DFS. After adjustment for pStage, age, tumor size, and lymphovascular invasion, high RAR was associated with an increased risk of death [HR, 1.271; 95% CI: 1.005–1.607; p = 0.045] and disease recurrence or death [HR, 1.265; 95% CI: 1.033–1.550; p = 0.023]. These findings suggest that RAR may help identify a subset of patients with apparently lower‐stage disease but relatively higher postoperative risk. These results are presented in Figure 5.
FIGURE 5.
Prognostic value of RAR in patients with pStage I–II disease. (A, B) Kaplan–Meier curves comparing overall survival and disease‐free survival between low‐ and high‐RAR groups among patients with pStage I–II disease. (C) Forest plot showing the adjusted association between high RAR and OS/DFS. Hazard ratios were adjusted for pStage, age, tumor size, and lymphovascular invasion. The vertical dashed line indicates HR = 1.
The results of the univariate and multivariable Cox regression analyses for OS are presented in Table 2. Univariate analysis demonstrated that several factors were linked to worse OS, including advanced age (≥ 70 vs. < 70), presence of anemia, higher BMI (≥ 25 vs. < 25), elevated RAR, tumor located in the upper stomach, higher T stage, poorly differentiated or other histological types, lymph node metastasis, perineural invasion, lymphovascular invasion, larger tumor size (≥ 3 cm vs. < 3 cm), and advanced TNM stage (III vs. I/II). According to the multivariate Cox model, independent predictors of poor overall survival were: elevated RDW‐to‐serum albumin ratio, older age, advanced T stage, adverse histology, lymph node metastasis, perineural invasion, and larger tumor size. Although anemia and higher BMI (≥ 25 kg/m2) were both significantly associated with poor survival in univariate analysis (p < 0.01), their effects did not remain statistically significant after adjustment for confounding variables in the multivariate model (p = 0.06 for both). The prognostic significance of RAR was retained, with HR 1.32 and 95% CI: 1.14–1.53 (p < 0.01). To summarize, the results indicate that an elevated RAR is closely linked with poorer overall survival and disease‐free survival, underscoring its putative prognostic value.
TABLE 2.
Multivariate and univariate analysis results for factors predictive of OS in gastric cancer patients who underwent curative‐intent surgery (n = 6988).
| Variable | Univariate analysis | Multivariate analysis | ||||
|---|---|---|---|---|---|---|
| Hazard ratio | CI (95%) | p | Hazard ratio | CI (95%) | p | |
| Age(years): ≥ 70 vs. < 70 | 2.12 | 1.89–2.37 | < 0.01* | 1.67 | 1.42–1.96 | < 0.01* |
| Sex: male vs. female | 1.04 | 0.93–1.17 | 0.456 | |||
| Anemia: present vs. absent | 1.86 | 1.68–2.07 | < 0.01* | 1.16 | 0.97–1.39 | 0.06 |
| BMI: ≥ 25 vs. < 25 | 1.24 | 1.09–1.46 | < 0.01* | 1.17 | 0.99–1.41 | 0.06 |
| RAR: high vs. low | 1.87 | 1.69–2.18 | < 0.01* | 1.32 | 1.14–1.53 | < 0.01* |
| Tumor location: upper vs. middle/low | 1.41 | 1.19–1.67 | < 0.01* | 1.06 | 0.89–1.26 | 0.54 |
| T stage: T3, T4 vs. T1, T2 | 7.25 | 6.20–8.47 | < 0.01* | 1.74 | 1.31–2.31 | < 0.01* |
| Histology: Poorly/others vs. well/moderately | 2.3 | 1.98–2.66 | < 0.01* | 1.33 | 1.09–1.63 | < 0.01* |
| Lymph node metastasis: present vs. absent | 6.22 | 5.35–7.23 | < 0.01* | 1.67 | 1.16–2.42 | < 0.01* |
| Perineural Invasion: present vs. absent | 4.06 | 3.62–4.58 | < 0.01* | 1.44 | 1.18–1.75 | < 0.01* |
| Lymphovascular invasion: present vs. absent | 6.58 | 5.53–7.82 | < 0.01* | 1.29 | 0.88–1.89 | 0.2 |
| Tumor size: ≥ 3 vs. < 3 | 3.76 | 3.26–4.33 | < 0.01* | 1.49 | 1.21–1.83 | < 0.01* |
| TNM stage: III vs. I/II | 6.72 | 5.94–7.60 | < 0.01* | 1.97 | 1.55–2.5 | < 0.01* |
Note: At our institution, the reference ranges for RDW‐CV and serum albumin are 11.7%–14.4% and 4.0–5.5 g/dL, respectively. T stage was dichotomized as T3–T4 vs. T1–T2 for primary analyses; alternative categorical and ordinal parameterizations yielded consistent results. Tumor size was dichotomized at 3 cm based on maximally selected rank statistics (optimal cutoff = 3.3 cm) and clinical relevance. *p‐values < 0.05 were considered statistically significant.
Abbreviations: AJCC/UICC, American Joint Committee on Cancer/Union for International Cancer Control staging manual, 8th edition (2017); CI, confidence interval; HR, hazard ratio; RAR, RDW‐to‐serum albumin ratio; TNM, tumor‐node‐metastasis.
3.3. Comparison of Prognostic Performance Between RAR and Its Components (Serum Albumin and RDW)
A dynamic ROC analysis was utilized to measure the prognostic capability of RAR and its individual components—serum albumin and RDW—for predicting overall survival. At 1 year, RAR showed the highest discrimination: AUC 0.666 (95% CI: 0.636–0.696), versus serum albumin (neg) 0.657 (0.625–0.688; p = 0.438 vs. RAR) and RDW 0.544 (0.525–0.568; p < 0.0001 vs. RAR). At 5 years, serum albumin (neg) modestly exceeded RAR: 0.638 (0.617–0.659) versus RAR 0.621 (0.600–0.642; p = 0.0125), while RDW remained lowest (0.546 [0.525–0.568]; p < 0.0001 vs. RAR) (see Table 3). These data indicate that RAR provides numerically better short‐term discrimination than either component and remains comparable to albumin at 1 year, whereas albumin is modestly superior at 5 years. Across both horizons, RDW alone performs consistently worse. Taken together, RAR captures complementary information to albumin in the short‐term, while albumin retains slightly stronger long‐term separation; therefore, analyses beyond single time points (e.g., multivariable models and time‐dependent metrics) are warranted to reflect their distinct prognostic contributions.
TABLE 3.
The AUC for 1‐year and 5‐year OS prediction based on RAR, serum albumin levels (neg), and RDW.
| Component | AUC | 95% CI | p |
|---|---|---|---|
| 1‐year | |||
| RAR | 0.666 | 0.636–0.696 | Ref. |
| Serum albumin (neg) | 0.657 | 0.625–0.688 | 0.4380 |
| RDW | 0.544 | 0.525–0.568 | < 0.0001 |
| 5‐year | |||
| RAR | 0.621 | 0.6–0.642 | Ref. |
| Serum albumin (neg) | 0.638 | 0.617–0.659 | 0.0125 |
| RDW | 0.546 | 0.525–0.568 | < 0.0001 |
Note: (neg) indicates that serum albumin was negatively transformed (−value) so that a higher value consistently reflects a worse prognosis, allowing direct comparison with RAR in AUC analyses.
3.4. Comparison of RAR, LMR (Neg), PLR, and NLR With Regard to Prognostic Discrimination for Overall Survival
Within a cohort of gastric cancer patients, we assessed the prognostic utility of several circulating biomarkers for 5‐year OS. Special attention was paid to the predictive value of RAR. Time‐dependent AUC analyses (Figure 6B) illustrated the evolution of predictive accuracy for these biomarkers throughout postoperative follow‐up. Notably, RAR consistently demonstrated a higher and more stable AUC than LMR, PLR, and NLR, underscoring its sustained prognostic robustness over time. The receiver operating characteristic (ROC) analysis at 60 months post‐surgery (Figure 6A) further corroborated RAR's superior discriminative ability at the 5‐year endpoint.
FIGURE 6.
Time‐dependent ROC analyses were conducted to evaluate the prognostic performance of RAR and the inflammatory markers neutrophil‐to‐lymphocyte ratio (NLR), lymphocyte‐to‐monocyte ratio (LMR), and platelet‐to‐lymphocyte ratio (PLR) in predicting 5‐year OS and DFS (panels A and C). Panels (B and D) depict the corresponding time‐dependent AUC curves for RAR, LMR, NLR, and PLR. The x‐axis charts the duration after surgery, while the y‐axis reflects the prognostic AUC for OS across time. The evolution of OS‐related AUC values is shown for up to 5 years following surgery.
As detailed in Table 4, RAR reached an AUC of 0.620 with a 95% CI of 0.599–0.641, serving as the reference standard. The inverted LMR showed a slightly lower AUC of 0.594 (95% CI: 0.573–0.616), with a borderline p‐value of 0.0517, suggesting inferior prognostic utility compared to RAR. PLR (AUC = 0.593, 95% CI: 0.572–0.614, p = 0.0356) and NLR (AUC = 0.574, 95% CI: 0.553–0.596, p = 0.0013) both exhibited significantly poorer prognostic accuracy relative to RAR. In summary, although the absolute AUC values were modest, RAR consistently outperformed LMR, PLR, and NLR across time points, indicating superior relative prognostic discrimination rather than high absolute predictive accuracy. Such modest absolute AUC values are commonly observed for inflammation‐ and nutrition‐based biomarkers in gastric cancer, reflecting the multifactorial and biologically heterogeneous determinants of long‐term oncologic outcomes. Therefore, AUC values alone should not be interpreted as definitive evidence of clinical irrelevance, particularly when a biomarker demonstrates independent prognostic significance in multivariable models. These findings support the role of RAR as a complementary prognostic marker for risk stratification, rather than a standalone survival prediction tool.
TABLE 4.
The AUC for forecasting 5‐year overall survival based on RAR, LMR (neg), PLR, NLR, and pStage.
| Marker | AUC | 95% CI | p |
|---|---|---|---|
| RAR | 0.620 | 0.599–0.641 | Ref. |
| LMR (neg) | 0.594 | 0.573–0.616 | 0.0517 |
| PLR | 0.593 | 0.572–0.614 | 0.0356 |
| NLR | 0.574 | 0.553–0.596 | 0.0013 |
| pStage | 0.767 | 0.751–0.784 | < 0.0001 |
Note: (neg) indicates that LMR was negatively transformed so that higher values represent worse prognosis, enabling direct comparison with RAR.
3.5. Nomogram Construction and Calibration
Based on the independent prognostic variables identified in multivariate analysis, a nomogram was constructed to estimate the 1‐, 3‐, and 5‐year overall survival (OS) probabilities for patients with gastric cancer (Figure 7). Each variable was assigned a corresponding score on the top axis, and the cumulative point total was mapped to the predicted survival probability on the bottom scale. This graphical tool provides an individualized estimate of postoperative prognosis and may assist in clinical risk stratification.
FIGURE 7.
Nomogram for predicting 1‐, 3‐, and 5‐year overall survival (OS) after curative gastrectomy.
The nomogram showed favorable discriminative performance for OS, with a C‐index of 0.7867 (95% CI: 0.7683–0.8050). By comparison, the traditional TNM staging system showed a lower C‐index of 0.6512 (95% CI: 0.6339–0.6686), indicating that the multivariable nomogram provided more comprehensive prognostic information than anatomical staging alone.
To further clarify the specific contribution of RAR to the nomogram, we compared calibration curves between the clinicopathological model without RAR and the expanded model incorporating RAR. As shown in Figure 8, the model incorporating RAR showed slightly improved agreement between predicted and observed survival probabilities, particularly for 3‐ and 5‐year OS. However, the magnitude of improvement was modest, suggesting that RAR provided incremental but complementary prognostic information rather than serving as the dominant driver of nomogram performance.
FIGURE 8.
Comparison of calibration curves for nomogram models with and without RAR. Calibration curves were generated to compare the agreement between nomogram‐predicted and observed overall survival probabilities at 1, 3, and 5 years. The clinicopathological model without RAR was compared with the expanded model incorporating RAR. The dashed diagonal line represents perfect calibration. The dotted lines indicate a ±10% margin of error. Error bars represent 95% confidence intervals. The addition of RAR resulted in only marginal changes in calibration, suggesting that RAR provided complementary rather than dominant prognostic information within the multivariable model.
4. Discussion
This investigation examined the link between systemic inflammatory status and the prognostic outlook for gastric cancer patients who underwent curative gastrectomy. Importantly, this study was not only designed to establish RAR as a stand‐alone survival prediction model, but also to assess its incremental prognostic value beyond conventional clinicopathological factors. A higher preoperative RAR was significantly tied to older age, larger tumor size, elevated T and N classifications, an overall advanced pathological stage, a greater frequency of lymphovascular and perineural invasion, and poorer tumor differentiation. Across all pathological stages, individuals with elevated RAR exhibited substantially poorer OS and DFS. Notably, RAR provided additional refinement of prognostic stratification even within the same pathological stage. When compared with other inflammatory markers, including NLR, LMR, PLR, and pathological stage, RAR consistently demonstrated superior relative discrimination compared with other inflammation‐based indices; however, its primary value lies in postoperative risk stratification rather than individual‐level survival prediction. Compared with the systemic inflammation score (SIS), which combines albumin and LMR, the RAR index provides a simpler yet broader reflection of the host status. By integrating red‐cell heterogeneity (RDW) and nutritional–inflammatory reserve (albumin), RAR captures both erythropoietic dysfunction and systemic inflammation, dimensions not covered by SIS. Moreover, RAR relies on two universally available laboratory parameters and can be modeled as a continuous variable, improving reproducibility and risk discrimination. To improve the clinical utility of inflammatory markers and address their limited standalone predictive performance, we constructed a multivariable nomogram incorporating RAR alongside other prognostically relevant variables, including pathological stage, tumor size, lymphovascular invasion, and age. This model demonstrated favorable discrimination for overall survival, with a C‐index of 0.743, and for disease‐free survival, with a C‐index of 0.718. The calibration plots further supported the consistency between predicted and observed survival probabilities at 3 and 5 years. These findings suggest that although the predictive ability of RAR alone may be suboptimal, its integration into a composite prognostic model yields improved clinical relevance. These features suggest that RAR may serve as a more accessible and physiologically comprehensive biomarker for prognostication in gastric cancer. Taken together, these results position RAR as a clinically relevant prognostic biomarker in gastric cancer. Although the AJCC TNM classification remains the foundational tool for postoperative prognostication, RAR provides supplementary, readily accessible prognostic insight in the preoperative setting. The clinical value of RAR should therefore be interpreted as complementary rather than substitutive to pathological staging. Pathological stage remains the dominant determinant of postoperative management and prognosis. In pStage III disease, the heavy anatomical tumor burden may outweigh the additional prognostic contribution of inflammatory or nutritional indicators such as RAR. By contrast, among patients with pStage I–II disease, who are generally considered anatomically lower‐stage, high RAR remained associated with worse OS and DFS after adjustment for clinicopathological factors. This suggests that RAR may be useful for identifying a subset of apparently lower‐stage patients with relatively higher postoperative risk. Therefore, RAR may serve as an auxiliary marker for postoperative risk stratification and follow‐up planning, rather than as a replacement for pathological staging.
RAR is a composite biomarker derived from red cell distribution width (RDW) and serum albumin concentration, both routinely measured in clinical practice. Its prognostic value likely reflects the combined biological significance of these two parameters. In addition to RAR, anemia and BMI were also evaluated as potential prognostic indicators. While both showed significant associations with overall survival in univariate analysis, neither remained independent after multivariable adjustment, suggesting that their prognostic impact may be partially mediated through systemic inflammation and nutritional status reflected by RAR. Serum albumin functions as a multifactorial prognostic biomarker, reflecting nutritional status, systemic inflammatory burden, and liver function—all of which are critical determinants of cancer progression and treatment response. Monitoring albumin levels therefore provides essential insights for prognostic assessment and treatment planning in oncology [20, 21, 22].
RDW quantifies erythrocyte size heterogeneity and serves as a marker of erythrocyte physiology disturbances and systemic health. Elevated RDW has been connected to adverse outcomes in various diseases, including diabetes, malignancies, and cardiovascular disorders. Earlier investigations have demonstrated that elevated RDW values are linked to higher mortality risk and poorer prognostic outlook in several cancers, such as colorectal carcinoma and Hodgkin lymphoma [23, 24]. Moreover, elevated RDW has been implicated in perioperative complications and cognitive decline among elderly patients. Potential underlying mechanisms include oxidative stress, chronic inflammation, and cellular senescence, contributing collectively to erythrocyte dysfunction [25].
Elevated RAR—combining RDW and serum albumin—has shown consistent prognostic value across malignancies. In a large HBV‐related hepatocellular carcinoma cohort, high RAR independently predicted poorer long‐term survival and enhanced risk discrimination within conventional staging systems [26]. Similarly, in cervical cancer, elevated preoperative RAR was linked to higher transfusion risk and worse overall survival, underscoring its perioperative and prognostic relevance [27]. In extensive‐stage small cell lung cancer, real‐world data further supported the predictive value of RDW and albumin as accessible indicators of systemic inflammation and nutritional status [28].
The prognostic impact of RAR in gastric cancer likely arises from intertwined biological pathways linking inflammation, erythropoiesis, and nutrition. Elevated RDW reflects impaired erythroid maturation, iron restriction, and IL‐6–hepcidin–mediated inflammation, resulting in functional anemia, oxidative stress, and altered red‐cell deformability [29] Concurrently, hypoalbuminemia mirrors chronic inflammation, catabolism, and reduced antioxidant capacity, leading to compromised immune surveillance and enhanced tumor aggressiveness [30, 31]. These processes collectively exacerbate tissue hypoxia, stimulate epithelial–mesenchymal transition, and facilitate invasion and metastasis, thereby providing a mechanistic explanation for the association between high RAR and unfavorable prognosis. This interpretation aligns with previous evidence that inflammation–nutrition–hematologic interactions broadly modulate cancer biology and clinical outcomes [32].
Compared with previous nomograms that incorporated multi‐indicator or psychosocial parameters (e.g., grip strength, depression status) [33] for predicting postoperative frailty or long‐term survival in gastric cancer patients, our model based solely on routine inflammatory biomarkers such as RAR achieved a comparable or even superior discrimination (C‐index = 0.7867). This highlights the robust prognostic potential of RAR as a simplified and cost‐effective marker.
Collectively, findings across hepatobiliary, gynecologic, and thoracic malignancies suggest that RAR captures interrelated inflammation, nutrition, and hematologic pathways that influence prognosis across tumor types. Accordingly, RDW serves as a valuable biomarker reflecting hematologic and systemic pathophysiological alterations.
4.1. Strength and Limitations
To our knowledge, our study represents the most sizeable evaluation of RAR as a prognostic marker in gastric cancer thus far. Nonetheless, limitations exist. Initially, the retrospective, single‐institution sampling of the study could lead to selection bias. In addition, a relatively large proportion of the initial cohort was excluded because of missing data and predefined exclusion criteria. Although inverse probability of inclusion weighting and additional sensitivity analyses were performed, missingness was more common in earlier years, which may have affected the representativeness of the final analytic cohort. In the sensitivity analysis restricted to patients treated between 2015 and 2022, the effect estimates for RAR remained in the same direction as those in the main cohort, but the associations were not statistically significant. Therefore, the prognostic value of RAR should be interpreted cautiously and further validated in external cohorts. Secondly, we only assessed RDW and albumin within 7 days prior to surgery, precluding evaluation of dynamic changes in these parameters and their impact on patient outcomes. In addition, although our pStage I–II subgroup analysis suggests that RAR may help identify higher‐risk patients within anatomically lower‐stage disease, whether RAR‐guided closer surveillance can improve clinical outcomes requires validation in prospective and external cohorts. Third, postoperative adjuvant therapy is an important determinant of survival, especially in pStage II–III patients. Following the reviewer's suggestion, we additionally re‐retrieved available postoperative treatment information and found that 3838 of 4190 pStage II–III patients with available records (91.6%) received adjuvant chemotherapy. However, detailed postoperative treatment information was not uniformly available; therefore, residual confounding related to postoperative treatment cannot be completely excluded.
5. Conclusion
Overall, this expansive study demonstrates that the preoperative RAR is closely linked with unfavorable clinicopathological features and poorer survival in gastric cancer patients undergoing curative gastrectomy. When benchmarked against established inflammatory indices, including the NLR, RAR provides complementary prognostic information that may assist in preoperative risk stratification and clinical decision‐making when integrated with established prognostic models, rather than serving as a standalone survival prediction tool. Given its accessibility from routine blood tests, RAR may serve as a complementary preoperative marker for risk stratification, but its clinical value should be interpreted cautiously because residual confounding related to postoperative treatment cannot be completely excluded. Further prospective, multicenter studies with more complete postoperative treatment information are needed to validate these findings.
Author Contributions
Xinhang Cao: methodology, data curation, investigation, formal analysis, visualization, writing – original draft, conceptualization. Wei Wang: investigation, writing – original draft, writing – review and editing, validation, software, formal analysis. Weiwei Shang: methodology, investigation, visualization, validation, formal analysis, writing – original draft. Yan Zhang: conceptualization, methodology, supervision, project administration, resources, writing – review and editing. Mingzhe Ma: conceptualization, supervision, funding acquisition, project administration, resources, writing – review and editing, methodology. Dandan Xiao: investigation, writing – original draft, resources, writing – review and editing, validation, methodology. Ziyu Wang: data curation, formal analysis, visualization, resources, writing – review and editing. Yangbai Sun: investigation, validation, resources, project administration.
Funding
This work was supported by the following grants: M.Z.M. discloses support for the research of this work from National Nature Science Foundation of China [grant number 82473176, 81972213] and Natural Science Foundation of Shanghai [grant number 22ZR1412700].
Ethics Statement
This study was approved by the Institutional Review Board of Fudan University Shanghai Cancer Center. All procedures involving human participants were conducted in accordance with the Declaration of Helsinki.
Consent
Written informed consent was obtained from all individual participants included in the study.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Table S1: Annual distribution of missing data and final analytic cohort.
Table S2: Annual distribution of exclusion reasons and final analytic cohort.
Table S3: Sensitivity analysis restricted to patients treated between 2015 and 2022.
Contributor Information
Yan Zhang, Email: yanyan0921@sina.com.
Mingzhe Ma, Email: mmz666@163.com.
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
References
- 1. Sung H., Ferlay J., Siegel R. L., et al., “Global Cancer Statistics 2022: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries,” CA: A Cancer Journal for Clinicians 74, no. 3 (2024): 229–263, 10.3322/caac.21834. [DOI] [PubMed] [Google Scholar]
- 2. Sundar R., Nakayama I., Markar S. R., et al., “Gastric Cancer,” Lancet 405, no. 10494 (2025): 2087–2102, 10.1016/S0140-6736(25)00052-2. [DOI] [PubMed] [Google Scholar]
- 3. Chen Y., Jia K., Xie Y., et al., “The Current Landscape of Gastric Cancer and Gastroesophageal Junction Cancer Diagnosis and Treatment in China: A Comprehensive Nationwide Cohort Analysis,” Journal of Hematology & Oncology 18 (2025): 42, 10.1186/s13045-025-01698-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Wang Y., Yan Q., Fan C., et al., “Overview and Countermeasures of Cancer Burden in China,” Science China. Life Sciences 66, no. 11 (2023): 2515–2526, 10.1007/s11427-022-2240-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Amin M. B., Greene F. L., Edge S. B., et al., “The Eighth Edition AJCC Cancer Staging Manual: Continuing to Build a Bridge From a Population‐Based to a More ‘Personalized’ Approach to Cancer Staging,” CA: A Cancer Journal for Clinicians 67, no. 2 (2017): 93–99, 10.3322/caac.21388. [DOI] [PubMed] [Google Scholar]
- 6. Savioli F., Morrow E. S., Dolan R. D., et al., “Prognostic Role of Preoperative Circulating Systemic Inflammatory Response Markers in Primary Breast Cancer: Meta‐Analysis,” British Journal of Surgery 109, no. 12 (2022): 1206–1215, 10.1093/bjs/znac319. [DOI] [PubMed] [Google Scholar]
- 7. Schuettfort V. M., D'Andrea D., Quhal F., et al., “A Panel of Systemic Inflammatory Response Biomarkers for Outcome Prediction in Patients Treated With Radical Cystectomy for Urothelial Carcinoma,” BJU International 129, no. 2 (2022): 182–193, 10.1111/bju.15379. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Rocco A., Sgamato C., Pelizzaro F., et al., “Systemic Inflammatory Response Markers Improve the Discrimination for Prognostic Model in Hepatocellular Carcinoma,” Hepatology International 45 (2025): 10806, 10.1007/s12072-025-10806-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Kim S. G., Eom B. W., Yoon H., Kim Y. W., and Ryu K. W., “Prognostic Value of Preoperative Systemic Inflammatory Parameters in Advanced Gastric Cancer,” Journal of Clinical Medicine 11, no. 18 (2022): 5318, 10.3390/jcm11185318. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Ding Y., Chen Y., Zhang J., et al., “Blood Biomarker‐Based Predictive Indicator for Liver Metastasis in Alpha‐Fetoprotein‐Producing Gastric Cancer and Multi‐Omics Tumor Microenvironment Insights,” Frontiers in Oncology 14 (2024): 1156324, 10.3389/fonc.2024.1156324. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Wang W., Xu Y., and Li G., “Noninvasive Imaging of the Tumor Immune Microenvironment Correlates With Response to Immunotherapy in Gastric Cancer,” Nature Communications 13, no. 1 (2022): 7285, 10.1038/s41467-022-32816-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Ma M., Weng M., Chen F., et al., “Systemic Inflammation Score Is a Prognostic Marker After Curative Resection in Gastric Cancer,” ANZ Journal of Surgery 89, no. 4 (2019): 377–382, 10.1111/ans.15103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Ma M., Wang J., Hu Y., Weng M., Liu X., and Wang Y., “Prognostic Value of Inflammatory Biomarkers in Gastric Cancer Patients and the Construction of a Predictive Model,” Digestive Surgery 36, no. 5 (2019): 433–445, 10.1159/000493432. [DOI] [PubMed] [Google Scholar]
- 14. Ma M., Yuan S., Chen Y., and Zhou Z., “Preoperative Apolipoprotein B/Apolipoprotein A1 Ratio: A Novel Prognostic Factor for Gastric Cancer,” Oncotargets and Therapy 11 (2018): 2169–2176, 10.2147/OTT.S156690. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Liu J., Wang X., Gao T. Y., et al., “Red Blood Cell Distribution Width to Albumin Ratio Associates With Prevalence and Long‐Term Diabetes Mellitus Prognosis: An Overview of NHANES 1999–2020 Data,” Diabetology & Metabolic Syndrome 16 (2024): 50, 10.1186/s13098-024-00985-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Yu B., Li M., Yu Z., et al., “Red Blood Cell Distribution Width to Albumin Ratio (RAR) is Associated With Low Cognitive Performance in American Older Adults: NHANES 2011–2014,” BMC Geriatrics 25 (2025): 157, 10.1186/s12877-025-05800-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Liu H., Xiang R., and Chen Z., “The Association Between Red Blood Cell Distribution Width‐To‐Albumin Ratio and Risk of Depression: A Cross‐Sectional Analysis of NHANES,” Journal of Affective Disorders 48 (2025): 40086477, 10.1016/j.jad.2025.03.037. [DOI] [PubMed] [Google Scholar]
- 18. Xu J., Wang G., Chen X., et al., “Association Between Red Blood Cell Distribution Width‐To‐Albumin Ratio and All‐Cause Mortality in Intracerebral Hemorrhage,” Frontiers in Nutrition 12 (2025): 1599104, 10.3389/fnut.2025.1599104. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Agha R. A., Mathew G., Rashid R., et al., “Revised Strengthening the Reporting of Cohort, Cross‐Sectional and Case‐Control Studies in Surgery (STROCSS) Guideline: An Update for the Age of Artificial Intelligence,” Premier Journal of Science 10 (2025): 100081. [Google Scholar]
- 20. Liao L., Zhang S., Li W., et al., “Serum Albumin and Atrial Fibrillation: Insights From Epidemiological and Mendelian Randomization Studies,” Frontiers in Cardiovascular Medicine 10 (2023): 1234567, 10.3389/fcvm.2023.1234567. [DOI] [Google Scholar]
- 21. Xia Z., Fu X., Yuan X., et al., “Serum Albumin to Globulin Ratio Prior to Treatment as a Potential Non‐Invasive Prognostic Indicator for Urological Cancers,” Frontiers in Oncology 14 (2024): 1178912, 10.3389/fonc.2024.1178912. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Toyoda H. and Johnson P. J., “The ALBI Score: From Liver Function in Patients With HCC to a General Measure of Liver Function,” Journal of Hepatology 66, no. 6 (2017): 1031–1042, 10.1016/j.jhep.2016.12.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Cheng K. C., Lin Y. M., Liu C. C., Wu K. L., and Lee K. C., “High Red Cell Distribution Width Is Associated With Worse Prognosis in Early Colorectal Cancer After Curative Resection: A Propensity‐Matched Analysis,” Cancers 14 (2022): 945, 10.3390/cancers14040945. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Herraez I., Bento L., Del Campo R., et al., “Prognostic Role of the Red Blood Cell Distribution Width (RDW) in Hodgkin Lymphoma,” Diagnostics 10, no. 11 (2020): 893, 10.3390/diagnostics10110893. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Salvagno G. L., Sanchis‐Gomar F., Picanza A., and Lippi G., “Red Blood Cell Distribution Width: A Simple Parameter With Multiple Clinical Applications,” Critical Reviews in Clinical Laboratory Sciences 52, no. 2 (2015): 86–105, 10.3109/10408363.2014.992064. [DOI] [PubMed] [Google Scholar]
- 26. Tan M., You R., Cai D., et al., “The Red Cell Distribution Width to Albumin Ratio: A Novel Prognostic Indicator in Hepatitis B Virus‐Related Hepatocellular Carcinoma,” International Journal of Medical Sciences 22, no. 2 (2025): 441–450, 10.7150/ijms.103125. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Sim J.‐H., Jang D.‐M., Cho H.‐S., Park J. Y., and Choi W.‐J., “Association of Red Cell Distribution Width/Albumin Ratio With Intraoperative Blood Transfusion in Cervical Cancer Patients,” PLoS One 17, no. 11 (2022): e0277481, 10.1371/journal.pone.0277481. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Li X., Tong L., Wang S., et al., “Integration of Clinical and Blood Parameters in Risk Prognostication for Patients Receiving Immunochemotherapy for Extensive‐Stage Small Cell Lung Cancer: Real‐World Data From Two Centers,” BMC Medicine 22 (2024): 381, 10.1186/s12916-024-03612-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Allen L. A., Felker G. M., Mehra M. R., et al., “Validation and Potential Mechanisms of Red Cell Distribution Width as a Prognostic Marker in Heart Failure,” Journal of Cardiac Failure 16, no. 3 (2010): 230–238, 10.1016/j.cardfail.2009.11.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Liu J., Wang X., Gao T., et al., “Red Blood Cell Distribution Width to Albumin Ratio Associates With Prevalence and Long‐Term Diabetes Mellitus Prognosis: An Overview of NHANES 1999–2020 Data,” Frontiers in Endocrinology (Lausanne) 15 (2024): 1362077, 10.3389/fendo.2024.1362077. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Eyiol A. and Ertekin B., “Association of Red Blood Cell Distribution Width to Albumin Ratio With Prognosis in Stroke Patients,” Biomarkers in Medicine 18, no. 7 (2024): 311–320, 10.2217/bmm-2023-0460. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Hanahan D., “Hallmarks of Cancer: New Dimensions,” Cancer Discovery 12, no. 1 (2022): 31–46, 10.1158/2159-8290.CD-21-1059. [DOI] [PubMed] [Google Scholar]
- 33. Liu Z., Cao X., Chen Y., et al., “Frailty Trajectory and Its Determinants in Older Patients With Gastric Cancer: A Prospective Cohort Study,” International Journal of Nursing Studies 152 (2024): 104697, 10.1016/j.ijnurstu.2024.104697. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1: Annual distribution of missing data and final analytic cohort.
Table S2: Annual distribution of exclusion reasons and final analytic cohort.
Table S3: Sensitivity analysis restricted to patients treated between 2015 and 2022.
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.