Abstract
Objective
To develop and validate a multidimensional prognostic model for predicting postoperative 1-year all-cause mortality in esophageal cancer patients by integrating clinical, hematological, tumor-related, and perioperative factors (including the Naples Prognostic Score, NPS), and to provide evidence for clinical risk stratification and individualized interventions.
Methods
A retrospective analysis was conducted on 625 patients (≥ 18 years) who underwent esophageal cancer resection (January 2018–January 2023) at the Cancer Hospital Affiliated to Xinjiang Medical University. Ultimately, 521 eligible patients were enrolled and split into a training cohort (n = 364, 70%) and validation cohort (n = 157, 30%). LASSO regression was used for feature selection, and Cox univariate/multivariate analyses identified independent risk factors for 1-year mortality. Kaplan-Meier analysis compared survival differences by NPS and Clavien-Dindo classification.
Results
The 1-year all-cause mortality rate was 20.0% (104/521). Independent predictors included anemia (assessed by red blood cell count), white blood cell count, NPS, Clavien-Dindo classification, and TNM stage (all P < 0.05). The model had an AUC of 0.82 (95% CI: 0.76–0.87) in the training cohort and 0.76 (95% CI: 0.66–0.86) in the validation cohort. Calibration curves and DCA confirmed good calibration and clinical utility. Kaplan-Meier analysis showed significant survival differences between NPS < 3 vs. ≥3 and Clavien-Dindo < 3 vs. ≥3 (both P < 0.0001).
Conclusion
This multidimensional model effectively predicts postoperative 1-year mortality in esophageal cancer patients. With NPS as a core component, it provides a practical tool for perioperative risk stratification and individualized management to identify high-risk patients and reduce mortality.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12893-025-03478-4.
Keywords: Postoperative Esophageal Cancer, NPS, Prognostic Model, 1-year mortality
Introduction
Globally, over 600,000 individuals are diagnosed with esophageal cancer annually, and surgical intervention remains a primary treatment modality for the disease. The overall incidence of complications after esophagectomy reaches 59%, with 30-day and 90-day mortality rates as high as 10% and 15%, respectively [1–3]. Despite significant advances in modern surgical techniques and perioperative care measures, challenges persist in identifying high-risk patients and selecting interventions based on individual risk profiles during the perioperative period for cancer patients [4, 17]. Recent studies indicate that perioperative nutritional support and inflammation management in esophageal cancer patients significantly impact prognosis, with these factors closely linked to postoperative complication rates and survival outcomes [5]. Studies demonstrate that elevated neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR) are significantly associated with poor prognosis. Assessing a patient’s inflammatory status enables clinicians to provide more accurate survival predictions for cancer patients [6]. Therefore, integrated strategies combining nutritional intervention and management of inflammatory markers hold potential significance for improving overall patient prognosis.
The Naples Prognostic Score (NPS) is a comprehensive indicator based on inflammation and nutrition, first proposed by Galizia et al. as a novel system for assessing the prognosis of patients with malignant tumors [7]. This score comprises the following factors: serum albumin level, total cholesterol level, neutrophil-to-lymphocyte ratio (NLR), and lymphocyte-to-monocyte ratio (LMR). The score ranges from 0 to 4 points, and different total score ranges generally correspond to varying prognostic outcomes, with higher scores typically indicating poorer patient prognosis [7]. Beyond the nutritional-inflammatory dimension covered by the NPS, hematological indices (e.g., red blood cell count, white blood cell count), tumor clinicopathological characteristics (e.g., T stage, TNM stage), and the severity of perioperative complications (e.g., Clavien-Dindo classification) are also closely associated with postoperative survival in esophageal cancer patients from distinct dimensions [9, 15]. These multidimensional indicators complement each other, forming a complex risk network affecting patient prognosis. However, there is currently a lack of prognostic assessment tools that systematically integrate these key factors with the NPS as the core, making it difficult to meet the clinical demand for precise risk stratification. This study aims to develop and validate a prognostic model based on the Naples Prognostic Score (NPS) for predicting postoperative 1-year mortality risk in esophageal cancer patients. The model will serve as a tool for perioperative risk stratification and individualized management in clinical practice, thereby reducing patients’ 1-year postoperative mortality risk.
Method
Data source
Data on esophageal cancer patients were retrieved through the Specialized Disease Research Data Analysis Platform (Specialized Disease Database) of the Cancer Hospital Affiliated to Xinjiang Medical University. This database system primarily features data statistics, retrieval, management, and analysis export functions. It assists medical researchers in efficiently and conveniently conducting specialized disease research, statistics, management, and analysis, thereby enhancing research efficiency and expanding research scope.
For this retrospective study, the data extraction was performed in March 2025. During and after data collection, the authors did not have access to any information that could identify individual participants, as all retrieved data were fully anonymized prior to analysis.
Sample size calculation
Based on previous studies, the postoperative 1-year mortality rate of esophageal cancer patients is 18%. Assuming a margin of error of 0.05 for the true outcome, we estimated the sample size using the following clinical prediction model formula: n = (1.96/δ)² φ(1-φ), where n is the sample size, φ is the event occurrence rate, and δ is the margin of error. n = (1.96/0.05)² × 0.18 × (1–0.18) = 227 (cases). Considering a 20% dropout rate, the required sample size is at least 227 + 227 × 20% = 273 (cases).The events per variable (EPV) for the final model was approximately 14.9 (104 events / 7 variables), exceeding the recommended minimum of 10, thus indicating a sufficient sample size for reliable model development.
Calculation of the Naples Prognostic Score (NPS)
The cut-off values for serum albumin, total cholesterol, lymphocyte-to-monocyte ratio (LMR), and neutrophil-to-lymphocyte ratio (NLR) used for calculating the Naples Prognostic Score (NPS) in this study were identical to those reported in the original study by Galizia et al. Specifically, the scoring criteria were defined as follows: serum albumin concentration < 4.0 g/dL = 1 point, ≥ 4.0 g/dL = 0 points; total cholesterol concentration < 180 mg/dL = 1 point, ≥ 180 mg/dL = 0 points; LMR < 4.44 = 1 point, ≥ 4.44 = 0 points; NLR < 2.96 = 0 points, ≥ 2.96 = 1 point. The total NPS was calculated as the sum of the four individual scores, ranging from 0 to 4.
Study population
Patients 18 years or older who underwent esophageal cancer resection at the Cancer Hospital Affiliated to Xinjiang Medical University between January 2018 and January 2023 were selected as study subjects (Fig. 1). Inclusion criteria: (1) Patients with pathologically confirmed esophageal cancer who underwent surgical treatment, including radical esophagectomy or combined thoraco-laparoscopic radical esophagectomy. (2) Patients aged ≥ 18 years with the capacity to participate in the study. (3) Patients with complete medical records containing detailed preoperative examination results (e.g., physiological indicators, imaging findings), surgical records, and at least one year of follow-up information after surgery. Exclusion criteria: (1) emergency surgery patients due to acute medical conditions. (2) Patients experiencing severe complications during surgery (e.g., intraoperative hemorrhage, significant organ injury), preventing completion of standard surgical procedures. (3) Incomplete clinical data or missing follow-up records.Among the 104 excluded patients, 12 underwent emergency surgery, 8 did not complete standard surgery due to severe intraoperative complications, 20 had missing key variables, and 64 lacked ≥ 1 year of follow-up data. Ultimately, 521 patients were included in this study.This research adheres to the Declaration of Helsinki. Informed consent was waived due to the retrospective and anonymous nature of the clinical data. The study was approved by the Ethics Committee of the Third Clinical Medical College (Affiliated Cancer Hospital) of Xinjiang Medical University (Ethics Approval No. K-2024258).
Fig. 1.
Flowchart of inclusion-exclusion criteria for study participants
Variables and variable definitions
The primary objective of this study is to predict postoperative 1-year mortality, with the outcome measure defined as 1-year postoperative all-cause mortality. Based on prior clinical experience and available data in the specialized disease registry, we selected the following variables: gender, age, actual length of hospital stay, coronary heart disease, hypertension, diabetes, blood transfusion history, tumor recurrence status, tumor metastasis status, anemia (assessed by RBC), white blood cell count (WBC), platelet count, surgical approach, duration of surgery, duration of anesthesia, tumor location, histological type, 1-year mortality status, neutrophil-to-lymphocyte ratio (NLR), lymphocyte-to-monocyte ratio (LMR), T stage, N stage, TNM stage, ASA classification, Naples Prognostic Score (NPS), Clavien-Dindo classification.
Classification of surgical approaches
Open surgery
Esophagectomy was performed solely via open thoracotomy and/or laparotomy incisions, with no endoscopic assistance throughout the procedure (e.g., resection of lower esophageal cancer via combined right thoracotomy and upper abdominal incisions).
Endoscopic surgery
Esophagectomy was completed entirely through minimally invasive endoscopic ports via thoracoscopy and laparoscopy, without any open incisions (e.g., totally thoracoscopically-laparoscopically assisted partial esophagectomy).
Hybrid endoscopic-open surgery
Esophagectomy was carried out by combining endoscopic minimally invasive techniques in one surgical field with open surgical approaches in another (e.g., thoracoscopically guided esophageal resection combined with open surgical procedures).
This classification was determined based on the primary surgical modality documented in the operative records.
Clavien–Dindo classification
Postoperative complications were graded using the Clavien–Dindo classification applied to events occurring within 30 days after surgery or during the index hospitalization, whichever was longer. Grades were assigned by two independent reviewers (a surgical fellow and an attending surgeon) using the original Clavien–Dindo definitions: Grade I — deviation from normal postoperative course not requiring pharmacological treatment beyond allowed drugs; Grade II — requiring pharmacological treatment; Grade III — requiring surgical, endoscopic or radiological intervention (IIIa without general anaesthesia; IIIb under general anaesthesia); Grade IV — life‑threatening complication requiring ICU management; Grade V — death. In our analysis Clavien–Dindo ≥ 3 denotes complications that required invasive intervention (Grade III) or were life‑threatening (Grade IV) or fatal (Grade V). Disagreements in grading were resolved by consensus.
Anemia (assessed by RBC)
Preoperative anemia status was defined using red blood cell count (RBC). Based on the distribution of our cohort and clinical references, an RBC count < [3.5] ×10¹²/L was classified as indicating anemia.
Statistical analysis
All statistical analyses were conducted in R version 4.5.0. The Shapiro-Wilk test assessed the normality of continuous variables. Missing data were handled by mean imputation (for normally distributed continuous variables) and median imputation (for non-normally distributed continuous variables), respectively. For categorical variables with missing values, the mode was used for imputation. This approach was adopted to maintain data integrity while minimizing potential biases, as the proportion of missing data in the study dataset was less than 5%.Continuous variables meeting normality criteria were reported as mean ± standard deviation (mean ± SD), while those failing normality were reported as median (interquartile range) [M (IQR)]. Categorical variables were summarized as n (%) [number (percentage)]. For intergroup comparisons, the following methods were used: For categorical variables, either the χ² test or Fisher’s exact test was selected based on data characteristics; for continuous variables, independent samples t-tests were used for normally distributed data, and Mann-Whitney U tests for non-normally distributed data. The study cohort was randomly split into a training set (70%) and a validation set (30%) via computer-generated random numbers. First, Lasso regression identified potential predictors of postoperative 1-year mortality in the training set. Before performing univariate and multivariate Cox proportional hazards regression analyses, the proportional hazards assumption was verified.Subsequently, these selected variables were included in univariate and multivariate Cox proportional hazards regression models to further define independent predictors of postoperative 1-year mortality. A nomogram-based prognostic model was constructed based on these independent predictors. Internal validation of the nomogram utilized a 500-replicate bootstrap sampling. To enhance validation reliability and mitigate overfitting risks from single-data partitioning, 5-fold cross-validation was further implemented in the training set: the training cases were randomly allocated into 5 mutually exclusive subsets of similar size. Four subsets served for model parameter optimization, and the remaining subset for performance evaluation. This process was repeated 5 times to ensure all training samples underwent validation, and the mean of the 5 validation results was used as a supplementary indicator of model stability. In both training and validation sets, model discrimination was evaluated via the Area Under the Receiver Operating Characteristic Curve (AUC), model consistency with actual outcomes via calibration curves, and clinical utility via Decision Curve Analysis (DCA). Survival curves were generated with the Kaplan-Meier method, and log-rank tests compared survival differences between groups. All statistical tests were two-sided, with P < 0.05 indicating statistical significance.
Results
Baseline analysis
We collected data from 625 patients undergoing esophageal cancer resection. Of these, 114 patients met exclusion criteria, resulting in 521 patients who underwent esophageal cancer resection included in this analysis. Among them, 104 (20%) died within one year after surgery. There were 352 males (67.6%), and the mean age was 64 years (Median (IQR): 58.0, 70.0) (Table 1).
Table 1.
Patients’ baseline characteristics
| Variables | Total (n = 521) | Training group(n = 364) | Validation group(n = 157) | p |
|---|---|---|---|---|
| Gender, n (%) | 0.694 | |||
| Male | 352 (67.6) | 244 (67) | 108 (68.8) | |
| Female | 169 (32.4) | 120 (33) | 49 (31.2) | |
| Age, Median (IQR) | 64.0 (58.0, 70.0) | 64.0 (58.0, 70.0) | 64.0 (57.0, 70.0) | 0.843 |
| Number of Hospitalization Days, Mean ± SD | 36.1 ± 13.8 | 36.3 ± 13.9 | 35.7 ± 13.6 | 0.609 |
| Heart Disease, n (%) | 0.874 | |||
| No | 463 (88.9) | 324 (89) | 139 (88.5) | |
| Yes | 58 (11.1) | 40 (11) | 18 (11.5) | |
| Hypertension, n (%) | 0.796 | |||
| No | 438 (84.1) | 307 (84.3) | 131 (83.4) | |
| Yes | 83 (15.9) | 57 (15.7) | 26 (16.6) | |
| Diabetes, n (%) | 0.692 | |||
| No | 417 (80.0) | 293 (80.5) | 124 (79) | |
| Yes | 104 (20.0) | 71 (19.5) | 33 (21) | |
| History of Blood Transfusion, n (%) | 0.889 | |||
| No | 453 (86.9) | 316 (86.8) | 137 (87.3) | |
| Yes | 68 (13.1) | 48 (13.2) | 20 (12.7) | |
| Alcohol Consumption, n (%) | 0.321 | |||
| No | 166 (31.9) | 119 (32.7) | 47 (29.9) | |
| Yes | 355 (68.1) | 245 (67.3) | 110 (70.1) | |
| Recurrence of tumor, n (%) | 1 | |||
| No | 519 (99.6) | 362 (99.5) | 157 (100) | |
| Yes | 2 ( 0.4) | 2 (0.5) | 0 (0) | |
| Tumor metastasis, n (%) | 0.41 | |||
| No | 503 (96.5) | 353 (97) | 150 (95.5) | |
| Yes | 18 ( 3.5) | 11 (3) | 7 (4.5) | |
| RBC, Mean ± SD | 3.8 ± 0.5 | 3.8 ± 0.5 | 3.9 ± 0.5 | 0.621 |
| WBC, Mean ± SD | 7.1 ± 2.9 | 7.1 ± 2.8 | 7.2 ± 3.0 | 0.59 |
| PLT, Mean ± SD | 298.8 ± 105.0 | 300.2 ± 105.7 | 295.5 ± 103.4 | 0.635 |
| Surgical Approach, n (%) | 0.696 | |||
| Open Surgery | 215 (41.3) | 147 (40.4) | 68 (43.6) | |
| Laparoscopy | 111 (21.3) | 77 (21.2) | 34 (21.8) | |
| Laparoscopy Combined with Open Surgery | 194 (37.3) | 140 (38.5) | 54 (34.6) | |
| Duration of Surgery (minutes), Mean ± SD | 288.0 ± 83.2 | 286.2 ± 80.0 | 292.1 ± 90.4 | 0.457 |
| Duration of anesthesia(minutes), Mean ± SD | 318.0 ± 83.2 | 316.2 ± 80.0 | 322.1 ± 90.4 | 0.457 |
| Tumor Location, n (%) | 0.331 | |||
| Upper | 76 (14.6) | 50 (13.7) | 26 (16.7) | |
| Lower | 286 (54.9) | 208 (57) | 78 (50) | |
| Middle | 159 (30.5) | 107 (29.3) | 52 (33.3) | |
| Histological Type, n (%) | 0.012 | |||
| Squamous Cell Carcinoma | 492 (94.4) | 348 (95.3) | 144 (92.3) | |
| Adenocarcinoma | 25 ( 4.8) | 17 (4.7) | 8 (5.1) | |
| Others | 4 ( 0.8) | 0 (0) | 4 (2.6) | |
| Death, n (%) | 0.837 | |||
| No | 417 (80.0) | 293 (80.3) | 124 (79.5) | |
| Yes | 104 (20.0) | 72 (19.7) | 32 (20.5) | |
| ASA, n (%) | 0.472 | |||
| < 2 | 58 (11.1) | 43 (11.8) | 15 (9.6) | |
| ≥ 2 | 463 (88.9) | 322 (88.2) | 141 (90.4) | |
| Clavien-Dindo classification, n (%) | 0.757 | |||
| < 3 | 474 (91.0) | 333 (91.2) | 141 (90.4) | |
| ≥ 3 | 47 ( 9.0) | 32 (8.8) | 15 (9.6) | |
| NLR, Mean ± SD | 3.9 ± 3.2 | 3.8 ± 3.0 | 4.0 ± 3.4 | 0.52 |
| LMR, Mean ± SD | 3.4 ± 5.8 | 3.4 ± 5.6 | 3.5 ± 6.4 | 0.809 |
| TyG, Mean ± SD | 8.5 ± 0.5 | 8.5 ± 0.5 | 8.5 ± 0.5 | 0.889 |
| NPS Score, n (%) | 0.985 | |||
| < 3 | 184 (35.3) | 129 (35.3) | 55 (35.3) | |
| ≥ 3 | 337 (64.7) | 236 (64.7) | 101 (64.7) | |
| T Stage n (%) | 0.526 | |||
| < 3 | 197 (38.0) | 141 (38.8) | 56 (35.9) | |
| ≥ 3 | 322 (62.0) | 222 (61.2) | 100 (64.1) | |
| N Stage, n (%) | 0.803 | |||
| < 2 | 360 (69.1) | 251 (68.8) | 109 (69.9) | |
| ≥ 2 | 161 (30.9) | 114 (31.2) | 47 (30.1) | |
| TNM Stage, n (%) | 0.084 | |||
| < 3 | 327 (63.0) | 220 (60.6) | 107 (68.6) | |
| ≥ 3 | 192 (37.0) | 143 (39.4) | 49 (31.4) | |
Data are presented as n (%), mean ± SD, or median (IQR). Normality of continuous variables was tested by Shapiro-Wilk test. Categorical variables were compared using χ² test or Fisher’s exact test; continuous variables were compared using independent samples t-test or Mann-Whitney U test
Abbreviations: RBC Red blood cell count, WBC White blood cell count, PLT Platelet count, ASA American Society of Anesthesiologists, NLR Neutrophil-to-lymphocyte ratio, LMR Lymphocyte-to-monocyte ratio, TyG Triglyceride-glucose index, NPS Naples Prognostic Score
Data splitting and variable selection
The study cohort was divided into training and validation sets at a 7:3 ratio using computer-generated random numbers. All subsequent model building and variable selection were performed on the training set: Lasso regression was applied to screen potential predictor variables, ultimately retaining those with non-zero coefficients corresponding to λ-se (Figure2a, Fig. 2b). These included: anemia (assessed by RBC), WBC, NLR, ASA, Clavien-Dindo classification, NPS, T stage, N stage, and TNM stage Supplementary Table 1.
Fig. 2.
LASSO regression-based variable selection for postoperative 1-year mortality risk prediction in esophageal cancer patients. 2a presents the partial likelihood deviance plot for LASSO regression, illustrating the relationship between partial likelihood deviance and log(λ) to determine the optimal regularization parameter. 2b) displays the coefficient profile plot, showing the trajectory of each coefficient as a function of log(λ) for variable selection in LASSO regression
Training set Cox regression analysis
Variables screened by Lasso regression were first subjected to the proportional hazards test (Supplementary Table 2) and then incorporated into univariate and multivariate Cox proportional hazards regression models in the training set to assess their correlation with the risk of postoperative 1-year mortality.Univariate Cox Regression (Training Set): NPS score (HR = 2.08, 95% CI: 1.21–3.59, P = 0.008), anemia (assessed by RBC) (HR = 0.45, 95% CI: 0.29–0.70, P < 0.001), WBC (HR = 1.09, 95% CI: 1.03–1.15, P = 0.003), NLR (HR = 1.05, 95% CI: 1.00–1.10, P = 0.05), ASA classification (HR = 1.26, 95% CI: 0.58–2.74, P = 0.565), Clavien-Dindo classification (HR = 2.48, 95% CI: 1.33–4.61, P = 0.004), and TNM staging (HR = 2.85, 95% CI: 1.77–4.59, P < 0.001).Multivariate Cox regression (training set): After adjusting for potential confounders: NPS score (HR = 2.33, 95% CI: 1.35–4.03, P = 0.002), anemia (assessed by RBC)(HR = 0.47, 95% CI: 0.30–0.72, P = 0.001), WBC (HR = 1.10, 95% CI: 1.03–1.17, P = 0.002), TNM staging (HR = 3.05, 95% CI: 1.87–4.96, P < 0.001),Clavien-Dindo classification (HR = 1.78,95% CI: (0.93 ~3.38),P < 0.001) (Table 2).
Table 2.
Univariate and multivariate Cox proportional hazards regression analyses for postoperative 1-year mortality
| Variable | HR (95%CI) | P value | adj HR (95%CI) | adj P value |
|---|---|---|---|---|
| NPS Score | ||||
| < 3 | (Reference) | (Reference) | ||
| ≥ 3 | 2.08(1.21 ~ 3.59) | 0.008 | 2.33(1.35 ~ 4.03) | 0.002 |
| RBC | 0.45 (0.29 ~ 0.7) | < 0.001 | 0.47(0.3 ~ 0.72) | 0.001 |
| WBC | 1.09 (1.03 ~ 1.15) | 0.003 | 1.12(1.05 ~ 1.19) | 0.005 |
| NLR | 1.05 (1 ~ 1.11) | 0.05 | 1.05 (0.99 ~ 1.11) | 0.124 |
| ASA | ||||
| < 2 | (Reference) | |||
| ≥ 2 | 1.26 (0.58 ~ 2.74) | 0.565 | 1.19 (0.54 ~ 2.6) | 0.67 |
| TNM stage | ||||
| < 3 | (Reference) | (Reference) | ||
| ≥ 3 | 2.85(1.77 ~ 4.59) | < 0.001 | 3.05 (1.87 ~ 4.66) | < 0.001 |
| Clavien-Dindo classification | ||||
| < 3 | (Reference) | (Reference) | ||
| ≥ 3 | 2.48 (1.33 ~ 4.61) | < 0.001 | 1.78(0.93 ~ 3.38) | < 0.001 |
Variables with P < 0.05 in univariate analysis were included in multivariate Cox regression
HR Hazard ratio, CI Confidence interval, adj HR Adjusted hazard ratio
Construction of a nomogram
A nomogram was constructed based on the predictors selected by LASSO regression that were statistically significant in the multivariate Cox model. The ASA classification was included to enhance the clinical face validity and utility of the model, as it is a well-established perioperative risk indicator. This model assigns corresponding scores to each predictor, sums the scores for all factors to obtain a total score, and then directly maps the total score to the patient’s probability of 1-year postoperative mortality, enabling convenient prediction of individual risk (Fig. 3).
Fig. 3.
Nomogram for predicting 1-year overall survival in patients with esophageal cancer. Variables include NPS, RBC, WBC, ASA, Clavien-Dindo, TNM stage, with total points converted to survival probability
Validation of the nomogram prognosis model
Model validation employed a combined approach of internal validation (training set) and external validation (validation set), using the area under the receiver operating characteristic curve (AUC), calibration curve, and decision curve analysis (DCA) as evaluation metrics. Internal validation of the training set was conducted via 500 bootstrap resampling iterations. Results showed a model AUC of 0.82 (95% confidence interval [95% CI]: 0.76–0.87), indicating great discriminatory ability for 1-year postoperative mortality risk (Fig. 4). Five-fold cross-validation confirmed robust model performance, with a mean area under the curve (AUC) of 0.7446 and a mean Brier score of 0.1407. Boxplots of AUC-K-CV and Brier-K-CV showed that AUC values (0.70–0.80) and Brier scores (0.12–0.18) were tightly clustered across validation folds, with minimal fluctuation and low inter-fold heterogeneity. Supplementary Fig. 1) Calibration curves demonstrated good agreement between the model-predicted mortality risk probabilities and actual incidence rates (Fig. 5a). DCA results indicate that within the high-risk threshold range of 0–0.7, the model’s net benefit significantly exceeds both “treat all patients” and “treat none” strategies, demonstrating strong clinical utility (Fig. 6a). When validated on the validation set, the model achieved an AUC of 0.76(95% CI: 0.66–0.86) (Fig. 4). Although slightly lower than the training set, this remains high, confirming the model’s robust generalization ability. The calibration curve of the validation set demonstrated minimal deviation between predicted values and actual outcomes (Fig. 5b). DCA results indicate that within the high-risk threshold range of 0–0.4, the model’s net benefit significantly exceeded the “treat all patients” and “treat none” strategies (Fig. 6b). This suggests reliable clinical applicability across diverse patient cohorts, providing valuable reference for developing individualized postoperative management plans.
Fig. 4.
ROC curve analysis of the nomogram for predicting 1-year overall survival in esophageal cancer patients. Receiver operating characteristic (ROC) curves illustrate the discriminative ability of the nomogram in training and validation cohorts
Fig. 5.
Calibration plots of the nomogram for 1-year overall survival in esophageal cancer patients. 5a: Training Cohort 5b: Validation Cohort
Fig. 6.
Decision curve analysis of the nomogram for predicting 1-year overall survival in esophageal cancer patients. 6a: Training cohort 6b: Validation cohort
Impact of NPS score and Clavien-Dindo classification on 1-year survival outcomes in esophageal cancer patients postoperative (Kaplan-Meier analysis)
Kaplan-Meier survival curve analysis based on the Naples Prognostic Score (NPS, as a nutritional and inflammatory indicator) and Clavien-Dindo classification revealed a highly significant statistical difference (p < 0.0001) in survival curves between the NPS < 3 group (NPS = 1) and the NPS ≥ 3 group (NPS = 2) among esophageal cancer patients after surgery. Patients in the NPS < 3 group exhibited a slower decline in survival probability over time and demonstrated significantly higher overall survival rates compared to the NPS ≥ 3 group (Fig. 7a). Simultaneously, in the CD grading groups, the survival curves for the CD < 3 group (CD = 1) and the CD ≥ 3 group (CD = 2) also showed a highly significant statistical difference (p < 0.0001). Patients in the CD < 3 group exhibited a slower decline in survival probability and a significantly higher overall survival rate than the CD ≥ 3 group (Fig. 7b). This indicates that the NPS score and Clavien-Dindo classification influence the postoperative 1-year mortality risk in esophageal cancer patients.Patients with NPS ≥ 3 and Clavien-Dindo ≥ 3 had a 1-year survival rate of only 21.7%, indicating that combined assessment of preoperative systemic status and postoperative recovery can better identify high-risk individuals requiring intensive follow-up.
Fig. 7.
Kaplan-Meier Survival Curves for 360-Day Survival in Esophageal Cancer Patients. 7a shows the Kaplan-Meier survival curve stratified by NPS. 7b shows the Kaplan-Meier survival curve stratified by Clavien-Dindo classification
Discussion
This study developed and validated a risk stratification model for 1-year postoperative all-cause mortality in patients undergoing esophagectomy. Its visual scoring system allows clinicians to assess 1-Year Mortality Risk survival outcomes more objectively and conveniently. We identified the preoperative Naples Prognostic Score (NPS) as a robust predictive variable for perioperative management of esophageal cancer; integrating preoperative inflammatory and nutritional status via the NPS facilitates early identification of high-risk individuals, offering a reference for optimizing individualized perioperative interventions.
Systemic inflammation correlates closely with cancer-related symptoms (fatigue, anorexia, pain), severely impairing patients’ quality of life [8]. Adequate nutrition can regulate inflammation, alleviating its effects on tumor initiation and progression. Multiple studies confirm the neutrophil-to-lymphocyte ratio (NLR) and lymphocyte-to-monocyte ratio (LMR) as reliable prognostic markers in colorectal, gastric, and esophageal cancer [9, 10]. The mechanism may involve M2 polarization of tumor-associated macrophages in the tumor microenvironment (TME)—promoting tumor growth, metastasis, and immune evasion—and tumor-associated neutrophils releasing inflammatory/angiogenic factors to accelerate progression and induce metabolic disorders [11]. Reduced total lymphocytes suppress antitumor immunity, worsening prognosis [12]. Thus, we integrated the clinically feasible Naples Prognostic Score (NPS, reflecting inflammation and nutrition) into our model, strengthening risk assessment and guiding early identification of high-risk patients and personalized interventions.
Against a backdrop of continuous advances in surgical techniques and perioperative care, 30/90-day mortality — a conventional short-term endpoint for cancer patients — has fallen to relatively low levels. This metric, however, may not fully capture the prognostic implications of delayed postoperative complications and adjuvant therapy delays. In contrast, 3/5-year overall survival, the gold-standard long-term endpoint, is often confounded by distant metastasis, subsequent treatment modifications and other extraneous factors, potentially masking the specific prognostic value of perioperative variables. By integrating short-term surgical safety outcomes and early oncological efficacy, 1-year mortality may effectively bridge this critical gap between short- and long-term endpoints.
Postoperative mortality is recognized as one of the leading causes of death worldwide. Surgical complications not only tend to elevate mortality risk to varying degrees but also potentially drive a substantial increase in healthcare costs, adding to the burden on healthcare systems and making this a global priority [13]. In the perioperative management of esophageal cancer patients, 1-year mortality is a prognostic indicator that has not been fully explored [14]. For instance, Xavier et al. identified ten key variables influencing postoperative prognosis after esophageal cancer resection using multivariable logistic regression, and developed and validated a predictive model for 90-day postoperative mortality, providing actionable insights for clinical decision-making [15]. Shuang Li et al. constructed a nomogram to predict postoperative pulmonary infection in esophageal cancer patients, incorporating length of hospital stay, serum albumin levels and other metrics; this work confirmed that multivariate models offer significantly higher prognostic accuracy than their univariate counterparts [16, 17]. Notably, the 1-year mortality predictive model can effectively complement existing mainstream models for in-hospital or 90-day postoperative prognosis in esophageal cancer cohorts.
A key strength of our model lies in its integration of multidimensional factors, encompassing preoperative laboratory tests, tumor biological characteristics, and postoperative recovery status. Preoperative hematological indices, including anemia (assessed by RBC) and white blood cell (WBC) counts, are integral components of multidimensional assessment. They not only reflect patients’ baseline nutritional reserves and inflammatory status but also provide objective evidence for evaluating physiological tolerance to surgery. When combined with the NPS (which integrates nutritional and inflammatory markers), these preoperative indicators form a solid foundation for risk stratification. We further incorporated tumor-related factors (T stage and TNM staging), which directly reflect the extent of tumor invasion and metastasis, and the Clavien-Dindo classification— a measure of postoperative complication severity that mirrors the quality of postoperative recovery and organ function. Collectively, these elements construct a comprehensive multidimensional risk assessment system that captures systemic health, disease aggressiveness, and postoperative recovery trajectories. This multidimensional approach addresses the limitations of single-factor evaluation, enabling a more holistic and accurate assessment of patient risk.
Our 1-year mortality prediction model complements existing mainstream prognostic tools for in-hospital mortality or 90-day postoperative outcomes in esophageal cancer patients. While most current models focus on 1-Year Mortality Risk postoperative endpoints, our study targets 1-year all-cause mortality—an indicator that better reflects long-term patient recovery and disease progression. By integrating multidimensional factors (preoperative laboratory tests, tumor staging, and postoperative complication severity), the model delivers a more clinically actionable risk stratification framework. For clinical translation, we defined risk thresholds based on the baseline 1-year mortality rate: low-risk (predicted mortality < 10%), intermediate-risk (10–30%), and high-risk (> 30%). This threshold system aligns closely with clinical practice and provides a reference for patient risk stratification in subsequent decision-making.
The nomogram integrates preoperative systemic status (NPS, anemia (assessed by RBC), WBC), tumor staging, and anticipated postoperative risk (Clavien-Dindo classification, though postoperative, enables post-intervention risk updating), offering targeted guidance for perioperative management. Preoperatively, patients identified as high-risk by the model (elevated predicted 1-year mortality) undergo dietitian-led formal nutritional assessment, targeted nutritional support (e.g., oral nutritional supplements, immunonutrition, or enteral nutrition regimens), and correction of modifiable laboratory abnormalities (e.g., anemia management) where feasible. Multidisciplinary consultation (involving surgery, anesthesia, oncology, nutrition, and physiotherapy) optimizes surgical timing and informs alternative or staged approaches for frail patients, while enhanced informed consent discussions address individualized risk profiles and expected postoperative trajectories. Intraoperatively, surgical teams prioritize low-physiological-stress techniques (e.g., minimally invasive or hybrid approaches when appropriate) for high-risk individuals, shorten operative duration within oncologically acceptable limits, and ensure involvement of senior anesthetic and surgical personnel. Postoperatively, high-risk patients are triaged to higher-intensity monitoring (planned ICU/HDU admission, invasive monitoring), early aggressive physiotherapy, and continued targeted nutritional support initiated preoperatively.
The model further refines implementation pathways for nutritional intervention, enhanced monitoring, and treatment escalation. For nutritional support, high-risk patients (with elevated NPS or predicted 1-year mortality) may undergo 7–14 days of preoperative nutritional optimization (with supplementation) in elective cases, where time permits; postoperatively, early enteral feeding via jejunostomy or enteral tube—guided by surgical and nutritional teams—is prioritized for those at high nutritional risk. For enhanced monitoring and escalation, high-risk patients warrant planned ICU/HDU admission, lower thresholds for imaging and invasive diagnostics in suspected clinical deterioration, and early involvement of interventional radiology or endoscopy services if complications are suspected. Additionally, reinforcement of standardized postoperative pathways (e.g., enhanced recovery after surgery [ERAS] elements) and close monitoring of adherence are critical, particularly for high-risk patients, to optimize outcomes.
This multidimensional integration empowers clinicians to identify high-risk patients early via routine preoperative assessments, enabling tailored perioperative management strategies. Ultimately, this approach may reduce postoperative 1-year mortality, confirming that multidimensional integration—specifically the synergistic inclusion of preoperative systemic status, tumor biological characteristics, and postoperative recovery outcomes—represents a feasible and effective strategy to improve prognostic assessment in esophageal cancer patients.
Using LASSO regression combined with Cox univariate and multivariate analyses, we identified RBC count, WBC count, NPS, Clavien-Dindo classification and TNM staging as independent predictors of 1-year postoperative all-cause mortality. Although the American Society of Anesthesiologists (ASA) classification did not reach statistical significance in the multivariate model, it was retained due to its recognized value in clinical anesthetic risk assessment, enhancing the model’s clinical applicability.Notably, updated multivariate Cox analysis confirmed WBC as an independent predictor of postoperative 1-year mortality (adjusted HR = 1.12, 95% CI: 1.05–1.19, P = 0.001), whereas the neutrophil-to-lymphocyte ratio (NLR) — despite approaching statistical significance in univariate analysis (HR = 1.05, 95% CI: 1.00–1.10, P = 0.05) — did not reach significance in the multivariate model (adjusted HR = 1.05, 95% CI: 0.99–1.11, P = 0.124). Despite potential mathematical redundancy (neutrophils being a major subset of WBC), this discrepancy aligns with clinicopathological logic:1. Biological dimensional difference: As a global marker of systemic inflammation, WBC encompasses neutrophils, lymphocytes, monocytes, and other subsets, comprehensively reflecting the systemic inflammatory response to perioperative surgical trauma and the body’s immune status. In contrast, NLR focuses solely on the proportional relationship between two cell types, failing to capture the synergistic roles of other leukocyte subsets in inflammation regulation and tissue repair.2. Model redundancy reduction: A core advantage of LASSO regression lies in selecting effective predictors while mitigating multicollinearity. Additionally, the Naples Prognostic Score (NPS) — a key component of our model — already integrates inflammation-nutrition balance information, indirectly encompassing part of the predictive value of NLR. WBC was prioritized by the model for its broader coverage of inflammatory dimensions.3. Clinical scenario adaptability: Short-term prognosis (1-year mortality) after radical esophagectomy is closely linked to acute perioperative inflammatory stress and infection risk. As a routine, rapidly detectable indicator, WBC directly reflects such acute pathological changes. In contrast, NLR is more closely associated with chronic inflammation and tumor progression, and its predictive efficacy is overshadowed by WBC’s global inflammatory assessment in the context of acute surgical trauma.This finding underscores the predictive specificity of inflammatory markers across different clinical scenarios. It does not negate the clinical value of NLR but confirms that the combination of global inflammatory indicators and inflammation-nutrition integrated scores offers greater practical value for predicting short-term perioperative prognosis, aligning with the clinical need for comprehensive risk assessment.
The prognostic model constructed based on the aforementioned independent predictors demonstrated an area under the ROC curve (AUC) of 0.82 (95% CI: 0.76–0.87) in the training cohort and 0.76 (95% CI: 0.66–0.86) in the validation cohort, indicating robust risk discrimination capability. Calibration curves and decision curve analysis (DCA) further confirmed that the model’s predicted values closely aligned with patients’ mortality risks. Moreover, it provided net benefit for decision-making within commonly used clinical risk thresholds, fully demonstrating its clinical utility. To determine individualized 1-year postoperative mortality risk with the nomogram, we summed scores for each clinical variable: for a patient with NPS = 3 (61 points), Anemia(RBC) = 3.5 × 10¹²/L (50 points), WBC = 12 × 10⁹/L (32points), ASA grade 2 (9 points), Clavien-Dindo grade 3 (14 points) and TNM stage 3 (35 points), the total score was 201. Projecting this score vertically onto the nomogram’s 1-Year Overall Survival axis yielded a corresponding 1-year overall survival probability of 28%, enabling rapid and convenient individualized risk stratification using routine clinical data.Future multicenter studies are needed to validate the model’s generalizability [18].
Limitations
This study has limitations: (1) Single-center data limitation: As a single-center retrospective study without external validation, the model’s applicability across different healthcare settings may be compromised. Constrained by inter-institutional data-sharing protocols, limited collaborative networks, and time and resource restrictions during the study period, we were unable to perform external validation, which further limits the model’s generalizability. (2) Incomplete variable coverage: Potential prognostic factors such as lifestyle and psychological status were not included, leading to information bias. (3) Risk of overfitting: The predictive efficacy of the model in real-world clinical settings requires further verification. (4) Single outcome measure: The study only focused on 1-year post-surgery mortality, with no analysis of patients’ long-term survival status and quality of life. In the future, we will conduct external validation using de-identified data from two collaborating tertiary centers within the regional medical network. Active discussions are underway, and we anticipate finalizing the harmonization of variable definitions and obtaining data usage agreements within 1 year.
Conclusion
The prognostic model developed in this study effectively predicts the 1-year after-surgery mortality risk in esophageal cancer patients. Combined with the application of the NPS score, it provides a tool for perioperative risk stratification and individualized management. This facilitates clinicians’ early identification of high-risk patients and implementation of interventions, thereby reducing postoperative mortality risk.
Abbreviations
| Category | English Full Name | Abbreviation | Unit (if applicable) | Notes |
|---|---|---|---|---|
| Prognostic Scores | Naples Prognostic Score | NPS | — (Score range: 0–4 points) | Comprehensive index integrating nutritional and inflammatory status |
| Surgical Classification | Clavien-Dindo Classification | CDC | — (Grade: Ⅰ–Ⅴ) | System for assessing postoperative complication severity |
| Hematological Indices | Red Blood Cell Count | RBC | ×10¹²/L | Measured via preoperative blood routine (within 1 week before surgery) |
| White Blood Cell Count | WBC | ×10⁹/L | Measured via preoperative blood routine (within 1 week before surgery) | |
| Neutrophil-to-Lymphocyte Ratio | NLR | — (Ratio) | Derived index reflecting systemic inflammatory status | |
| Lymphocyte-to-Monocyte Ratio | LMR | — (Ratio) | Derived index reflecting immune and inflammatory balance | |
| Tumor Staging | Tumor Stage (TNM Classification) | T stage | — (Stage: Ⅰ–Ⅳ) | Reflects primary tumor invasion depth (per AJCC TNM staging system) |
| Node Stage (TNM Classification) | N stage | — (Stage: 0–Ⅲ) | Reflects regional lymph node metastasis status (per AJCC TNM staging system) | |
| TNM Stage (Tumor-Node-Metastasis) | TNM stage | — (Stage: Ⅰ–Ⅳ) | Comprehensive tumor staging system (per AJCC 8th edition) | |
| Perioperative Assessment | American Society of Anesthesiologists Classification | ASA | — (Class: Ⅰ–Ⅴ) | System for evaluating preoperative anesthetic risk |
| Statistical Metrics | Area Under the Receiver Operating Characteristic Curve | AUC | — (Value range: 0.5–1.0) | Indicates model discriminative ability (higher = better discrimination) |
| Hazard Ratio | HR | — (Ratio) | Reflects relative risk of postoperative 1-year mortality between groups | |
| Confidence Interval | CI | — (95% CI reported) | Quantifies uncertainty of HR estimates | |
| Study Design | Training Cohort | Training set | — (n = 364) | 70% of total sample, used for model development |
| Validation Cohort | Validation set | — (n = 157) | 30% of total sample, used for model performance verification |
Supplementary Information
Acknowledgements
Not applicable.
Authors’ contributions
A.K. (Adilai Kurexi) and M.M. (Maimaiti Mijiti) performed the data analysis. R.Y. (Rui Yan) supervised the entire study. A.K., M.M., T.Y. (Tingting Yuan), Y.Z. (Yaozhong Zhang), F.W. (Fanrong Wei), and R.Y. collectively contributed to the study design, conception, and data collection. A.K. drafted the manuscript. All authors reviewed the manuscript and approved the final version for submission.
Funding
Not applicable.
Data availability
The data that support the findings of this study are available from the Science and Technology Management Section of the Affiliated Cancer Hospital of Xinjiang Medical University but are not publicly accessible due to local policy restrictions and patient privacy protection requirements. Data can be obtained upon reasonable request to corresponding author and with formal review and approval by the section.
Declarations
Ethics approval and consent to participate
This study was approved by the Ethics Committee of the Third Clinical Medical College (Affiliated Cancer Hospital) of Xinjiang Medical University (Ethics Approval No. K-2024258). Adhering to the Declaration of Helsinki, informed consent was waived due to the retrospective nature of the study and the full anonymization of all clinical data, which precluded identification of individual participants.
Given the retrospective design and anonymized data collection, formal written informed consent from individual participants was not required and was therefore waived by the Ethics Committee of the Third Clinical Medical College (Affiliated Cancer Hospital) of Xinjiang Medical University.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Li J, Xu J, Zheng Y, et al. Esophageal cancer: Epidemiology, risk factors and screening. Chin J Cancer Res. 2021;33(5):535–47. 10.21147/j.issn.1000-9604.2021.05.01. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Zhu H, Ma X, Ye T, et al. Esophageal cancer in china: practice and research in the new era. Int J Cancer. 2023;152(9):1741–51. 10.1002/ijc.34301. [DOI] [PubMed] [Google Scholar]
- 3.Jackson JC, Molena D, Amar D. Evolving perspectives on esophagectomy care: clinical update. Anesthesiology. 2023;139(6):868–79. 10.1097/ALN.0000000000004720. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Ko CY, Hall BL, Hart AJ, Cohen ME, Hoyt DB. The American college of surgeons National surgical quality improvement program: achieving better and safer surgery. Jt Comm J Qual Patient Saf. 2015;41(5):199–204. 10.1016/s1553-7250(15)41026-8. [DOI] [PubMed] [Google Scholar]
- 5.Gong S, Rao X, Yuan Y, et al. Clinical-pathological features and perioperative outcomes of mediastinoscopy vs. thoracoscopy esophagectomy in esophageal cancer: A meta-analysis. Front Surg. 2023;10:1039615. 10.3389/fsurg.2023.1039615. Published 2023 Feb 14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Feng JF, Huang Y, Chen QX. Preoperative platelet lymphocyte ratio (PLR) is superior to neutrophil lymphocyte ratio (NLR) as a predictive factor in patients with esophageal squamous cell carcinoma. World J Surg Oncol. 2014;12:58 10.1186/1477-7819-12-58. Published 2014 Mar 19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Galizia G, Lieto E, Auricchio A, et al. Naples prognostic Score, based on nutritional and inflammatory Status, is an independent predictor of Long-term outcome in patients undergoing surgery for colorectal cancer. Dis Colon Rectum. 2017;60(12):1273–84. 10.1097/DCR.0000000000000961. [DOI] [PubMed] [Google Scholar]
- 8.Paulsen Ø, Laird B, Aass N, et al. The relationship between pro-inflammatory cytokines and pain, appetite and fatigue in patients with advanced cancer. PLoS ONE. 2017;12(5):e0177620. 10.1371/journal.pone.0177620. Published 2017 May 25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Inamoto S, Kawada K, Okamura R, et al. Prognostic impact of the combination of neutrophil-to-lymphocyte ratio and Glasgow prognostic score in colorectal cancer: a retrospective cohort study. Int J Colorectal Dis. 2019;34(7):1303–15. 10.1007/s00384-019-03316-z. [DOI] [PubMed] [Google Scholar]
- 10.Yamamoto T, Kawada K, Obama K. Inflammation-related biomarkers for the prediction of prognosis in colorectal cancer patients. Int J Mol Sci. 2021;22(15):8002. 10.3390/ijms22158002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Li L, Yu R, Cai T, et al. Effects of immune cells and cytokines on inflammation and immunosuppression in the tumor microenvironment. Int Immunopharmacol. 2020;88:106939. 10.1016/j.intimp.2020.106939. [DOI] [PubMed] [Google Scholar]
- 12.Gooden MJ, de Bock GH, Leffers N, et al. The prognostic influence of tumour-infiltrating lymphocytes in cancer: A systematic review with meta-analysis. Br J Cancer. 2011;105(1):93–103. 10.1038/bjc.2011.189. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Nepogodiev D, et al. Global burden of postoperative death. Lancet. 2019;393:401. 10.1016/S0140-6736(19)30178-1. [DOI] [PubMed] [Google Scholar]
- 14.Feng JF, Zhao JM, Chen S, Chen QX. Naples prognostic score: A novel prognostic score in predicting Cancer-Specific survival in patients with resected esophageal squamous cell carcinoma. Front Oncol. 2021;11:652537. 10.3389/fonc.2021.652537. Published 2021 May 28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.D’Journo XB, Boulate D, Fourdrain A et al. Risk Prediction Model of 90-Day Mortality After Esophagectomy for Cancer. JAMA Surg. 2021;156(9):836–845. 10.1001/jamasurg.2021.2376. [published correction appears in JAMA Surg. 2021 Sep 1;156(9):894. https://doi.org/10.1001/jamasurg.2021.4340]. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Li S, Su J, Sui Q, Wang G. A nomogram for predicting postoperative pulmonary infection in esophageal cancer patients. BMC Pulm Med. 2021;21(1):283. 10.1186/s12890-021-01656-7. Published 2021 Sep 6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Hao X, Wang Y, Li K, et al. Applying machine learning for perioperative adverse event prediction: a narrative review toward better clinical efficacy and usability. Anesthesiol Perioper Sci. 2025;3:51. 10.1007/s44254-025-00136-z. [Google Scholar]
- 18.Yap WK, Shih MC, Kuo C et al. Development and Validation of a Nomogram for Assessing Survival in Patients With Metastatic Lung Cancer Referred for Radiotherapy for Bone Metastases. JAMA Netw Open. 2018;1(6):e183242. 10.1001/jamanetworkopen.2018.3242. Published 2018 Oct 5. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available from the Science and Technology Management Section of the Affiliated Cancer Hospital of Xinjiang Medical University but are not publicly accessible due to local policy restrictions and patient privacy protection requirements. Data can be obtained upon reasonable request to corresponding author and with formal review and approval by the section.