Abstract
Background
Early postoperative infection (EPI) is common in elderly patients with hepatocellular carcinoma (HCC) and is associated with poor prognosis. Effective risk prediction tools are urgently needed for early identification. This study aimed to develop and validate an inflammation-, nutrition-, and surgery-based INSEP nomogram model for predicting EPI-related adverse events in this population.
Methods
A total of 988 elderly HCC patients from two medical centers were retrospectively enrolled and randomly divided into a training dataset (n = 692) and an internal validation dataset (n = 296), while 381 patients from a third medical center served as an external validation dataset. Restricted cubic spline (RCS), receiver operating characteristic (ROC), least absolute shrinkage and selection operator (LASSO), and multivariable logistic regression analyses were performed to identify independent predictors of EPI and develop the INSEP nomogram. An online risk calculator was established, and model performance was assessed using ROC, calibration, decision curve, and clinical impact curve analyses.
Results
The INSEP model incorporated five independent predictors: SII ≥600 (OR 2.73, p = 0.023), CONUT ≥ 3 (OR 7.84, p < 0.001), ECOG 2–4 (OR 4.96, p < 0.001), laparoscopic surgery (OR 0.83, p = 0.025), and surgical duration ≥180 min (OR 7.44, p < 0.001). The model achieved AUCs of 0.870 in the training dataset, 0.803 in internal validation, and 0.798 in external validation, with good calibration and clinical utility. Two representative case demonstrations confirmed the model's accurate risk stratification: a low-risk patient (predicted EPI 3.3%) had no infection, while a high-risk patient (predicted 97.0%) developed pneumonia. Usability testing by three independent hepatobiliary surgeons (6–10 years of experience) showed efficient operation (25–60 s per case), intuitive interface, and clear output, with no system errors identified.
Conclusion
The INSEP model is a reliable and practical tool for individualized EPI risk assessment in elderly HCC patients undergoing curative hepatectomy, demonstrating good predictive performance, clinical applicability, and user acceptability.
Keywords: early postoperative infection, elderly hepatocellular carcinoma, INSEP model, nomogram, nutritional status, risk prediction, systemic inflammation
Introduction
Hepatocellular carcinoma (HCC) is one of the most common malignancies worldwide, with both its incidence and disease burden continuing to increase (1, 2). Curative hepatectomy remains the primary treatment for patients with resectable HCC and offers the best chance for long-term survival (3, 4). However, with the accelerating aging of the global population, an increasing number of elderly patients are undergoing hepatectomy for HCC. Compared with younger patients, elderly individuals often have a higher burden of comorbidities, impaired immune function, and reduced physiological reserve, resulting in diminished postoperative recovery and an increased risk of postoperative complications (3). Among these, early postoperative infection (EPI) is one of the most common and clinically significant adverse events following hepatectomy. It not only prolongs hospital stay and increases healthcare costs but may also adversely affect both short-term recovery and long-term prognosis (5–7). Therefore, accurate preoperative identification of patients at high risk of EPI and implementation of targeted perioperative management strategies are of great clinical importance for improving postoperative outcomes in elderly patients with HCC.
Growing evidence has demonstrated that systemic inflammatory status and nutritional condition play pivotal roles in tumor progression and postoperative recovery (8–10). Several peripheral blood-based inflammatory and nutritional biomarkers, including the systemic immune-inflammation index (SII), controlling nutritional status (CONUT) score, neutrophil-to-lymphocyte ratio (NLR), and prognostic nutritional index (PNI), have been shown to be closely associated with postoperative complications and long-term outcomes in patients with malignant tumors. Notably, in elderly patients, immunosenescence and age-related impairment in metabolic regulation may further amplify the detrimental effects of chronic inflammation and inadequate nutritional reserves on postoperative recovery, making inflammatory and nutritional biomarkers promising tools for predicting adverse postoperative outcomes in this population (11–13).
In addition to patient-related characteristics, perioperative factors also play a critical role in the development of postoperative infections. Surgical approach, operative duration, and intraoperative blood loss may influence the risk of infection by affecting the extent of tissue injury, the systemic stress response, and postoperative immune function. Previous studies have suggested that open surgery is associated with a higher risk of major postoperative complications than minimally invasive procedures (14, 15). However, studies specifically evaluating the risk of EPI in elderly HCC patients remain limited. Most existing studies have focused on a single inflammatory biomarker, nutritional indicator, or perioperative factor, making it difficult to comprehensively characterize the complex clinical profile of elderly patients. Moreover, many currently available prediction models have been developed using single-center datasets with relatively small sample sizes, and some lack rigorous internal validation or independent external validation, limiting their predictive robustness and generalizability across different clinical settings (10, 16, 17). Therefore, there remains an unmet clinical need to develop a comprehensive prediction model that integrates baseline patient characteristics, age-related physiological status, inflammatory response, nutritional status, and surgery-related factors to achieve accurate and clinically applicable risk stratification for elderly HCC patients.
Nomograms have been widely adopted in oncology for individualized risk prediction by integrating multiple clinical variables into an intuitive and practical predictive tool. However, a comprehensive model specifically developed and multicenter-validated for predicting the risk of EPI in elderly HCC patients is still lacking (18–20). Therefore, in the present study, we developed and validated a novel inflammation–nutrition–surgery–based (INSEP) nomogram by integrating inflammatory, nutritional, and surgery-related factors. This model aims to improve the prediction of postoperative EPI in elderly HCC patients through comprehensive risk assessment, thereby facilitating individualized risk stratification, optimizing perioperative management, and providing a theoretical basis for personalized prevention and intervention in high-risk patients.
Materials and methods
Study design and patient selection
This multicenter retrospective study collected clinical data from three medical centers in China. The inclusion criteria were as follows: elderly patients (aged ≥60 years, with a consistent definition of “elderly” across all centers) who were pathologically diagnosed HCC between January 2015 and December 2025. Patients with evidence of preoperative infection, hematological disorders, or unexplained fever were excluded.
Initially, 1,347 elderly HCC patients from the First Hospital of Jilin University and the First Affiliated Hospital of Zhengzhou University, as well as 455 patients from Huaihe Hospital of Henan University, were screened for eligibility. After strict application of the exclusion criteria, 988 patients from the two model-development centers were finally included for model construction and internal validation. Meanwhile, 381 patients from Huaihe Hospital of Henan University were enrolled as an independent external validation dataset to evaluate the generalizability of the model.
This study was approved by the Ethics Committee of the First Hospital of Jilin University and was conducted in accordance with the principles of the Declaration of Helsinki. All data were collected and processed under strict privacy protection measures. Given the retrospective design of this study, all patient data were anonymized, and no personally identifiable information was disclosed; therefore, the requirement for informed consent was waived. Patients were excluded if they met any of the following criteria: (1) receiving preoperative HCC-related treatment; (2) recurrent HCC; (3) preoperative tumor rupture; or (4) missing key variables required for analysis, including the systemic immune-inflammation index (SII), controlling nutritional status (CONUT) score, operative time, or intraoperative blood loss.
Finally, 988 patients were included for model development and internal validation. These patients were randomly assigned to the training dataset (n = 692, 70%) and the internal validation dataset (n = 296, 30%). A total of 381 patients from Huaihe Hospital of Henan University constituted the independent external validation dataset (Figure 1).
Figure 1.
Flowchart showing patients screening process.
To ensure consistency in variable definitions, all candidate predictors and outcome measures were standardized across different study centers. The clinical data were extracted from electronic medical records by trained researchers and integrated into a comprehensive database for statistical analysis. Only patients with complete data for all study variables were included to ensure data integrity and the reliability of the predictive model.
Data collection and variable definitions
The variables collected in this study included demographic information [sex, age, and body mass index (BMI)], performance status and medical history [Eastern Cooperative Oncology Group (ECOG) performance status, smoking history, alcohol consumption, diabetes, liver disease history, and cirrhosis], and liver function and tumor-related parameters [Child-Pugh grade, Barcelona Clinic Liver Cancer (BCLC) stage, and preoperative alpha-fetoprotein (AFP)]. Preoperative laboratory parameters, including platelet count, neutrophil count, lymphocyte count, albumin, and total cholesterol, were also recorded. Tumor- and surgery-related variables included tumor location, maximum tumor diameter, tumor number, surgical approach, surgical duration, intraoperative blood loss, and intraoperative transfusion.
Composite inflammation- and nutrition-related indices were calculated as follows: the SII index was defined as platelet count × neutrophil count/lymphocyte count, and the CONUT score was calculated as the sum of albumin score, lymphocyte score, and cholesterol score (20).
In this study, operative duration was defined as the actual intraoperative time, referring to the period from the initiation of the surgical procedure to its completion. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the predictive performance of the SII and operative duration for EPI, and the optimal cutoff values were determined using the Youden index. Furthermore, to minimize potential confounding bias associated with the surgical learning curve, early surgical cases performed by surgeons who had not yet acquired sufficient independent operative experience were excluded from the analysis (18).
The primary outcome of this study was EPI, defined as any newly developed infectious event occurring within 7 days after surgery. Infectious events included those diagnosed during postoperative hospitalization and those identified during follow-up within 7 days after surgery that met the predefined diagnostic criteria.
All postoperative infectious events were adjudicated according to a prespecified standardized diagnostic protocol, which was consistently applied across all participating centers. The diagnostic criteria for infections were based on internationally recognized surveillance systems, particularly the definitions and principles established by the Centers for Disease Control and Prevention (CDC)/National Healthcare Safety Network (NHSN). The diagnosis of infection was determined based on a comprehensive assessment of clinical manifestations, microbiological findings, laboratory parameters, and imaging examinations (21, 22).
The types of infections included, but were not limited to, surgical site infection (SSI), pulmonary infection, intra-abdominal infection, urinary tract infection, and bloodstream infection. Detailed diagnostic criteria and adjudication procedures for each type of infection are provided in Supplementary Table 1 to improve methodological transparency and reproducibility. All suspected infection cases were independently reviewed by two experienced physicians according to the predefined diagnostic criteria. In cases of disagreement between the two reviewers, a third senior clinician independently reassessed the case and made the final decision to ensure consistency and accuracy in infection classification.
Other clinical outcomes included length of hospital stay, postoperative intensive care unit (ICU) admission, occurrence of postoperative complications, and complication severity. The severity of postoperative complications was assessed according to the Clavien–Dindo classification system, with grade I–II complications defined as minor complications and grade ≥III complications defined as major complications (23).
Statistical analysis
All statistical analyses were performed using IBM SPSS Statistics and R software (version 4.6.1). Continuous variables were expressed as mean ± standard deviation (M ± SD) according to their distribution characteristics, and comparisons between groups were performed using the independent-samples t-test. Categorical variables were presented as frequencies and percentages and compared using the chi-square test (χ2 test). Initially, a correlation heatmap was generated to visualize the pairwise correlations among candidate predictors and to explore potential correlation patterns. In addition, the variance inflation factor (VIF) was calculated to quantitatively assess multicollinearity among candidate variables. Generally, higher VIF values indicate a greater possibility of problematic multicollinearity.
To further investigate potential dose–response relationships between continuous variables and the risk of EPI, RCS regression models were applied to evaluate the associations of the SII index and operative duration with EPI risk. Nonlinearity tests were performed to determine whether significant nonlinear associations existed between continuous predictors and outcomes. Subsequently, receiver operating characteristic (ROC) curve analysis was conducted to assess the predictive performance of SII and operative duration for EPI, and the area under the curve (AUC) was calculated. The optimal cutoff values were determined using the Youden index, and continuous variables were dichotomized according to these cutoff values for subsequent model development. To reduce redundant variables and identify clinically relevant predictors associated with EPI, least absolute shrinkage and selection operator (LASSO) regression was performed. All candidate variables were entered into the LASSO model. Ten-fold cross-validation was applied to determine the optimal penalty parameter (λ), and variables selected based on the cross-validation results were retained as candidate predictors.
The variables selected by LASSO regression were subsequently incorporated into multivariable logistic regression analysis to identify independent predictors of EPI and develop the INSEP prediction model. Considering the multicenter design of this study, potential heterogeneity among participating centers, including differences in baseline characteristics, perioperative management strategies, and surgical procedures, was considered. Therefore, study center was included as a categorical covariate in the multivariable logistic regression model to adjust for potential center effects.
Based on the independent predictors identified by multivariable logistic regression, an individualized risk prediction nomogram was developed using the “rms” package in R software to estimate the probability of EPI occurrence for individual patients. In addition, an online risk calculator based on the Shiny framework was developed to facilitate clinical implementation of the model. Model performance was evaluated in terms of discrimination, calibration, and clinical utility. Discrimination was assessed using ROC curves and AUC values. Calibration was evaluated using calibration plots, calibration intercepts, calibration slopes, and Brier scores to assess the agreement between predicted probabilities and observed outcomes. Decision curve analysis (DCA) and clinical impact curves (CIC) were performed to evaluate the clinical net benefit and potential applicability of the model. The DeLong test was used to compare differences in AUC values between the INSEP model and individual predictors to assess the incremental predictive value of the combined model. All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant.
Results
Clinical characteristics of patients
According to the predefined inclusion and exclusion criteria, a total of 1,369 elderly HCC patients were included in this study. Among them, 988 patients from the First Hospital of Jilin University and the First Affiliated Hospital of Zhengzhou University were randomly assigned to the training dataset (n = 692) and the internal validation dataset (n = 296) at a ratio of 7:3 (Figure 1). In the training dataset, RCS analysis was performed to further explore the potential dose–response relationships between the SII, operative duration, and the risk of EPI. No significant nonlinear associations were observed between SII or operative duration and EPI risk, with P values for nonlinearity of 0.233 and 0.905, respectively (Supplementary Figures 1A, B).
Subsequently, ROC curve analysis was conducted to evaluate the predictive performance of SII and operative duration for EPI. Both variables demonstrated potential predictive value. The optimal cutoff values determined by ROC analysis were 597 for SII and 178 min for operative duration (Supplementary Figures 2A, B). Considering that the SII cutoff of 597 was close to 600, and that rounding to 600 had minimal impact on model discrimination while improving clinical applicability, the final cutoff value for SII was set at 600. Similarly, the cutoff value for operative duration was rounded to 180 min. These predefined cutoff values were subsequently used for EPI stratification and univariable and multivariable logistic regression analyses. As shown in Table 1, baseline characteristics were generally well balanced between the training and internal validation datasets, with no statistically significant differences observed among all variables (all P > 0.05).
Table 1.
Baseline characteristics of elderly patients with hepatocellular carcinoma in the training and internal validation datasets.
| Variables | Categories | Total (N = 988) | Training dataset (N = 692) | Internal validation dataset (N = 296) | P value |
|---|---|---|---|---|---|
| Demographic data | |||||
| Gender (%) | Male | 862 (87.25) | 599 (86.56) | 263 (88.85) | 0.376 |
| Female | 126 (12.75) | 93 (13.44) | 33 (11.15) | ||
| Age (%) | < 70 | 445 (45.04) | 319 (46.10) | 126 (42.57) | 0.341 |
| ≥70 | 543 (54.96) | 373 (53.90) | 170 (57.43) | ||
| BMI (%) | Under weight | 104 (10.53) | 74 (10.69) | 30 (10.14) | 0.564 |
| Normal weight | 532 (53.85) | 365 (52.75) | 167 (56.42) | ||
| Overweight/Obesity | 352 (35.63) | 253 (36.56) | 99 (33.45) | ||
| ECOG (%) | 0–1 | 720 (72.87) | 509 (73.55) | 211 (71.28) | 0.511 |
| 2–4 | 268 (27.13) | 183 (26.45) | 85 (28.72) | ||
| Smoke (%) | No | 611 (61.84) | 419 (60.55) | 192 (64.86) | 0.227 |
| Yes | 377 (38.16) | 273 (39.45) | 104 (35.14) | ||
| Drink (%) | No | 732 (74.09) | 509 (73.55) | 223 (75.34) | 0.612 |
| Yes | 256 (25.91) | 183 (26.45) | 73 (24.66) | ||
| DM (%) | No | 881 (89.17) | 616 (89.02) | 265 (89.53) | 0.901 |
| Yes | 107 (10.83) | 76 (10.98) | 31 (10.47) | ||
| History of liver disease (%) | No | 905 (91.60) | 641 (92.63) | 264 (89.19) | 0.097 |
| Yes | 83 (8.40) | 51 (7.37) | 32 (10.81) | ||
| Cirrhosis (%) | No | 271 (27.43) | 195 (28.18) | 76 (25.68) | 0.465 |
| Yes | 717 (72.57) | 497 (71.82) | 220 (74.32) | ||
| Child pugh grade (%) | No | 902 (91.30) | 629 (90.90) | 273 (92.23) | 0.579 |
| Yes | 86 (8.70) | 63 (9.10) | 23 (7.77) | ||
| BCLC stage (%) | 0 | 598 (60.53) | 428 (61.85) | 170 (57.43) | 0.386 |
| A | 167 (16.90) | 111 (16.04) | 56 (18.92) | ||
| B | 223 (22.57) | 153 (22.11) | 70 (23.65) | ||
| Laboratory data | |||||
| SII (%) | < 600 | 870 (88.06) | 610 (88.15) | 260 (87.84) | 0.975 |
| ≥600 | 118 (11.94) | 82 (11.85) | 36 (12.16) | ||
| CONUT (%) | < 3 | 773 (78.24) | 543 (78.47) | 230 (77.70) | 0.855 |
| ≥3 | 215 (21.76) | 149 (21.53) | 66 (22.30) | ||
| Preoperative AFP (%) | < 400 | 713 (72.17) | 511 (73.84) | 202 (68.24) | 0.085 |
| ≥400 | 275 (27.83) | 181 (26.16) | 94 (31.76) | ||
| Imaging data | |||||
| Tumor location (%) | Right lobe | 289 (29.25) | 207 (29.91) | 82 (27.70) | 0.543 |
| Left lobe | 575 (58.20) | 395 (57.08) | 180 (60.81) | ||
| Bilateral | 124 (12.55) | 90 (13.01) | 34 (11.49) | ||
| Maximum tumor diameter (%) | < 5 | 289 (29.25) | 207 (29.91) | 82 (27.70) | 0.5425 |
| 5–10 | 575 (58.20) | 395 (57.08) | 180 (60.81) | ||
| ≥10 | 124 (12.55) | 90 (13.01) | 34 (11.49) | ||
| Number of tumors (%) | Single | 834 (84.41) | 587 (84.83) | 247 (83.45) | 0.651 |
| Multiple | 154 (15.59) | 105 (15.17) | 49 (16.55) | ||
| Perioperative information | |||||
| Operative approach (%) | Open | 575 (58.20) | 385 (55.64) | 190 (64.19) | 0.015 |
| Laparoscopic | 413 (41.80) | 307 (44.36) | 106 (35.81) | ||
| Surgical duration (%) | < 180 | 703 (71.15) | 483 (69.80) | 220 (74.32) | 0.173 |
| ≥180 | 285 (28.85) | 209 (30.20) | 76 (25.68) | ||
| Intraoperative blood loss (%) | < 400 | 632 (63.97) | 440 (63.58) | 192 (64.86) | 0.755 |
| ≥400 | 356 (36.03) | 252 (36.42) | 104 (35.14) | ||
| Intraoperative blood transfusion (%) | No | 799 (80.87) | 558 (80.64) | 241 (81.42) | 0.842 |
| Yes | 189 (19.13) | 134 (19.36) | 55 (18.58) | ||
| Length of hospital stay [mean (SD)] | 11.523 (7.56) | 11.412 (6.72) | 11.784 (9.23) | 0.479 | |
| ICU admission postoperatively (%) | No | 907 (91.80) | 631 (91.18) | 276 (93.24) | 0.340 |
| Yes | 81 (8.20) | 61 (8.82) | 20 (6.76) | ||
| Minor complications (%) | No | 775 (78.44) | 544 (78.61) | 231 (78.04) | 0.907 |
| Yes | 213 (21.56) | 148 (21.39) | 65 (21.96) | ||
| Major complications (%) | No | 893 (90.38) | 620 (89.60) | 273 (92.23) | 0.242 |
| Yes | 95 (9.62) | 72 (10.40) | 23 (7.77) | ||
| EPI (%) | No | 839 (84.92) | 580 (83.82) | 259 (87.50) | 0.165 |
| Yes | 149 (15.08) | 112 (16.18) | 37 (12.50) | ||
BMI, body mass index; ECOG, eastern cooperative oncology group; DM, diabetes mellitus; BCLC, barcelona clinic liver cancer; SII, systemic immune-inflammation index; CONUT, controlling nutritional status; AFP, alpha-fetoprotein; ICU, intensive care unit admission; Minor complications, Clavien-Dindo Grade I-II; Major complications, Clavien-Dindo Grade III; EPI, early postoperative infection; CONUT score ≥3 indicates malnutrition.
Baseline characteristics were further compared between the model-development dataset and the independent external validation dataset (n = 381; Table 2). Significant differences were observed between the two datasets in several demographic, clinical, and perioperative variables, including sex, age, ECOG performance status, alcohol consumption, diabetes mellitus, cirrhosis, BCLC stage, CONUT score, maximum tumor diameter, surgical approach, operative duration, intraoperative blood loss, intraoperative blood transfusion, length of hospital stay, incidence of minor complications, incidence of major complications, and occurrence of EPI (all P < 0.05).
Table 2.
Comparison of baseline characteristics of elderly patients with hepatocellular carcinoma between the multicenter dataset and the independent validation dataset.
| Variables | Categories | Total (N = 1,369) | Multi-center dataset (N = 988) | Independent validation dataset (N = 381) | P value |
|---|---|---|---|---|---|
| Demographic data | |||||
| Gender (%) | Male | 1161 (84.81) | 862 (87.25) | 299 (78.48) | 0.001 |
| Female | 208 (15.19) | 126 (12.75) | 82 (21.52) | ||
| Age (%) | < 70 | 665 (48.58) | 445 (45.04) | 220 (57.74) | < 0.001 |
| ≥70 | 704 (51.42) | 543 (54.96) | 161 (42.26) | ||
| BMI (%) | Under weight | 160 (11.69) | 104 (10.53) | 56 (14.70) | 0.081 |
| Normal weight | 734 (53.62) | 532 (53.85) | 202 (53.02) | ||
| Overweight/Obesity | 475 (34.70) | 352 (35.63) | 123 (32.28) | ||
| ECOG (%) | 0–1 | 1037 (75.75) | 720 (72.87) | 317 (83.20) | 0.001 |
| 2–4 | 332 (24.25) | 268 (27.13) | 64 (16.80) | ||
| Smoke (%) | No | 826 (60.34) | 611 (61.84) | 215 (56.43) | 0.076 |
| Yes | 543 (39.66) | 377 (38.16) | 166 (43.57) | ||
| Drink (%) | No | 978 (71.44) | 732 (74.09) | 246 (64.57) | 0.001 |
| Yes | 391 (28.56) | 256 (25.91) | 135 (35.43) | ||
| DM (%) | No | 1198 (87.51) | 881 (89.17) | 317 (83.20) | 0.003 |
| Yes | 171 (12.49) | 107 (10.83) | 64 (16.80) | ||
| History of liver disease (%) | No | 1255 (91.67) | 905 (91.60) | 350 (91.86) | 0.961 |
| Yes | 114 (8.33) | 83 (8.40) | 31 (8.14) | ||
| Cirrhosis (%) | No | 447 (32.65) | 271 (27.43) | 176 (46.19) | < 0.001 |
| Yes | 922 (67.35) | 717 (72.57) | 205 (53.81) | ||
| Child pugh grade (%) | No | 1258 (91.89) | 902 (91.30) | 356 (93.44) | 0.234 |
| Yes | 111 (8.11) | 86 (8.70) | 25 (6.56) | ||
| BCLC stage (%) | 0 | 854 (62.38) | 598 (60.53) | 256 (67.19) | 0.001 |
| A | 198 (14.46) | 167 (16.90) | 31 (8.14) | ||
| B | 317 (23.16) | 223 (22.57) | 94 (24.67) | ||
| Laboratory data | |||||
| SII (%) | < 600 | 1195 (87.29) | 870 (88.06) | 325 (85.30) | 0.200 |
| ≥600 | 174 (12.71) | 118 (11.94) | 56 (14.70) | ||
| CONUT (%) | < 3 | 1097 (80.13) | 773 (78.24) | 324 (85.04) | 0.006 |
| ≥3 | 272 (19.87) | 215 (21.76) | 57 (14.96) | ||
| Preoperative AFP (%) | < 400 | 993 (72.53) | 713 (72.17) | 280 (73.49) | 0.671 |
| ≥400 | 376 (27.47) | 275 (27.83) | 101 (26.51) | ||
| Imaging data | |||||
| Tumor location (%) | Right lobe | 392 (28.63) | 289 (29.25) | 103 (27.03) | 0.265 |
| Left lobe | 814 (59.46) | 575 (58.20) | 239 (62.73) | ||
| Bilateral | 163 (11.91) | 124 (12.55) | 39 (10.24) | ||
| Maximum tumor diameter (%) | < 5 | 489 (35.72) | 365 (36.94) | 124 (32.55) | < 0.001 |
| 5–10 | 639 (46.68) | 521 (52.73) | 118 (30.97) | ||
| ≥10 | 241 (17.60) | 102 (10.32) | 139 (36.48) | ||
| Number of tumors (%) | Single | 1149 (83.93) | 834 (84.41) | 315 (82.68) | 0.483 |
| Multiple | 220 (16.07) | 154 (15.59) | 66 (17.32) | ||
| Perioperative information | |||||
| Operative approach (%) | Open | 837 (61.14) | 575 (58.20) | 262 (68.77) | 0.001 |
| Laparoscopic | 532 (38.86) | 413 (41.80) | 119 (31.23) | ||
| Surgical duration (%) | < 180 | 877 (64.06) | 703 (71.15) | 174 (45.67) | < 0.001 |
| ≥180 | 492 (35.94) | 285 (28.85) | 207 (54.33) | ||
| Intraoperative blood loss (%) | < 400 | 834 (60.92) | 632 (63.97) | 202 (53.02) | 0.003 |
| ≥400 | 535 (39.08) | 356 (36.03) | 179 (46.98) | ||
| Intraoperative blood transfusion (%) | No | 1046 (76.41) | 799 (80.87) | 247 (64.83) | < 0.001 |
| Yes | 323 (23.59) | 189 (19.13) | 134 (35.17) | ||
| Length of hospital stay [mean (SD)] | 12.023 (7.596) | 11.523 (7.564) | 13.318 (7.534) | 0.001 | |
| ICU admission postoperatively (%) | No | 1261 (92.11) | 907 (91.80) | 354 (92.91) | 0.564 |
| Yes | 108 (7.89) | 81 (8.20) | 27 (7.09) | ||
| Minor complications (%) | No | 949 (69.32) | 775 (78.44) | 174 (45.67) | < 0.001 |
| Yes | 420 (30.68) | 213 (21.56) | 207 (54.33) | ||
| Major complications (%) | No | 1089 (79.55) | 893 (90.38) | 196 (51.44) | < 0.001 |
| Yes | 280 (20.45) | 95 (9.62) | 185 (48.56) | ||
| EPI (%) | No | 1187 (86.71) | 839 (84.92) | 348 (91.34) | 0.002 |
| Yes | 182 (13.29) | 149 (15.08) | 33 (8.66) | ||
BMI, body mass index; ECOG, eastern cooperative oncology group; DM, diabetes mellitus; BCLC, barcelona clinic liver cancer; SII, systemic immune-inflammation index; CONUT, controlling nutritional status; AFP, alpha-fetoprotein; ICU, intensive care unit admission; minor complications, Clavien-Dindo Grade I-II; major complications, Clavien-Dindo Grade III; EPI, early postoperative infection.
Comparison of baseline characteristics and clinical outcomes between the EPI and non-EPI groups
As shown in Supplementary Figure 3A, pulmonary infection was the most common type of EPI among elderly HCC patients, accounting for 45% of all infection events, followed by SSI (15%) and intra-abdominal infection (12%). Other types of infection included other infections (11%), urinary tract infection (10%), and sepsis (7%).
Regarding the distribution of causative pathogens, Escherichia coli was the most frequently identified pathogen (33%), followed by Klebsiella pneumoniae (17%) and Pseudomonas aeruginosa (12%). Other detected pathogens included Acinetobacter baumannii (7%), Enterobacter species (4%), and other bacterial species (2%) (Supplementary Figure 3B). According to the occurrence of EPI, patients were further classified into the EPI group (n = 182) and the non-EPI group (n = 1,187), and baseline characteristics were compared between the two groups (Table 3). No statistically significant differences were observed between the two groups in terms of sex, age, BMI, smoking history, alcohol consumption, diabetes mellitus, history of liver disease, Child–Pugh grade, tumor location, or tumor number (all P > 0.05).
Table 3.
Baseline characteristics between elderly patients with hepatocellular carcinoma with and without EPI.
| Variables | Categories | Total (N = 1,369) | Non-EPI (N = 1187) | EPI (N = 182) | P value |
|---|---|---|---|---|---|
| Demographic data | |||||
| Gender (%) | Male | 1161 (84.81) | 1009 (85.00) | 152 (83.52) | 0.682 |
| Female | 208 (15.19) | 178 (15.00) | 30 (16.48) | ||
| Age (%) | < 70 | 665 (48.58) | 574 (48.36) | 91 (50.00) | 0.739 |
| ≥70 | 704 (51.42) | 613 (51.64) | 91 (50.00) | ||
| BMI (%) | Under weight | 160 (11.69) | 137 (11.54) | 23 (12.64) | 0.176 |
| Normal weight | 734 (53.62) | 627 (52.82) | 107 (58.79) | ||
| Overweight/Obesity | 475 (34.70) | 423 (35.64) | 52 (28.57) | ||
| ECOG (%) | 0–1 | 1037 (75.75) | 928 (78.18) | 109 (59.89) | < 0.001 |
| 2–4 | 332 (24.25) | 259 (21.82) | 73 (40.11) | ||
| Smoke (%) | No | 826 (60.34) | 712 (59.98) | 114 (62.64) | 0.548 |
| Yes | 543 (39.66) | 475 (40.02) | 68 (37.36) | ||
| Drink (%) | No | 978 (71.44) | 840 (70.77) | 138 (75.82) | 0.187 |
| Yes | 391 (28.56) | 347 (29.23) | 44 (24.18) | ||
| DM (%) | No | 1198 (87.51) | 1044 (87.95) | 154 (84.62) | 0.251 |
| Yes | 171 (12.49) | 143 (12.05) | 28 (15.38) | ||
| History of liver disease (%) | No | 1255 (91.67) | 1087 (91.58) | 168 (92.31) | 0.850 |
| Yes | 114 (8.33) | 100 (8.42) | 14 (7.69) | ||
| Cirrhosis (%) | No | 447 (32.65) | 402 (33.87) | 45 (24.73) | 0.018 |
| Yes | 922 (67.35) | 785 (66.13) | 137 (75.27) | ||
| Child pugh grade (%) | No | 1258 (91.89) | 1093 (92.08) | 165 (90.66) | 0.611 |
| Yes | 111 (8.11) | 94 (7.92) | 17 (9.34) | ||
| BCLC stage (%) | 0 | 854 (62.38) | 756 (63.69) | 98 (53.85) | 0.020 |
| A | 198 (14.46) | 170 (14.32) | 28 (15.38) | ||
| B | 317 (23.16) | 261 (21.99) | 56 (30.77) | ||
| Laboratory data | |||||
| SII (%) | < 600 | 1195 (87.29) | 1076 (90.65) | 119 (65.38) | < 0.001 |
| ≥600 | 174 (12.71) | 111 (9.35) | 63 (34.62) | ||
| CONUT (%) | < 3 | 1097 (80.13) | 1017 (85.68) | 80 (43.96) | < 0.001 |
| ≥3 | 272 (19.87) | 170 (14.32) | 102 (56.04) | ||
| Preoperative AFP (%) | < 400 | 993 (72.53) | 896 (75.48) | 97 (53.30) | < 0.001 |
| ≥400 | 376 (27.47) | 291 (24.52) | 85 (46.70) | ||
| Imaging data | |||||
| Tumor location (%) | Right lobe | 392 (28.63) | 348 (29.32) | 44 (24.18) | 0.113 |
| Left lobe | 814 (59.46) | 693 (58.38) | 121 (66.48) | ||
| Bilateral | 163 (11.91) | 146 (12.30) | 17 (9.34) | ||
| Maximum tumor diameter (%) | < 5 | 489 (35.72) | 411 (34.63) | 78 (42.86) | 0.016 |
| 5–10 | 639 (46.68) | 555 (46.76) | 84 (46.15) | ||
| ≥10 | 241 (17.60) | 221 (18.62) | 20 (10.99) | ||
| Number of tumors (%) | Single | 1149 (83.93) | 997 (83.99) | 152 (83.52) | 0.956 |
| Multiple | 220 (16.07) | 190 (16.01) | 30 (16.48) | ||
| Perioperative information | |||||
| Operative approach (%) | Open | 837 (61.14) | 749 (63.10) | 88 (48.35) | 0.001 |
| Laparoscopic | 532 (38.86) | 438 (36.90) | 94 (51.65) | ||
| Surgical duration (%) | < 180 | 877 (64.06) | 791 (66.64) | 86 (47.25) | < 0.001 |
| ≥180 | 492 (35.94) | 396 (33.36) | 96 (52.75) | ||
| Intraoperative blood loss (%) | < 400 | 834 (60.92) | 741 (62.43) | 93 (51.10) | 0.005 |
| ≥400 | 535 (39.08) | 446 (37.57) | 89 (48.90) | ||
| Intraoperative blood transfusion (%) | No | 1046 (76.41) | 925 (77.93) | 121 (66.48) | 0.001 |
| Yes | 323 (23.59) | 262 (22.07) | 61 (33.52) | ||
| Length of hospital stay [mean (SD)] | 12.023 (7.596) | 11.107 (5.359) | 17.995 (14.370) | < 0.001 | |
| ICU admission postoperatively (%) | No | 1261 (92.11) | 1134 (95.53) | 127 (69.78) | < 0.001 |
| Yes | 108 (7.89) | 53 (4.47) | 55 (30.22) | ||
| Minor complications (%) | No | 949 (69.32) | 851 (71.69) | 98 (53.85) | < 0.001 |
| Yes | 420 (30.68) | 336 (28.31) | 84 (46.15) | ||
| Major complications (%) | No | 1089 (79.55) | 997 (83.99) | 92 (50.55) | < 0.001 |
| Yes | 280 (20.45) | 190 (16.01) | 90 (49.45) | ||
BMI, body mass index; ECOG, eastern cooperative oncology group; DM, diabetes mellitus; BCLC, barcelona clinic liver cancer; SII, systemic immune-inflammation index; CONUT, controlling nutritional status; AFP, alpha-fetoprotein; ICU, intensive care unit admission; minor complications, Clavien-Dindo Grade I-II; major complications, Clavien-Dindo Grade III; EPI, early postoperative infection.
However, significant differences were observed in several clinical characteristics and perioperative variables. Compared with the non-EPI group, patients in the EPI group had poorer ECOG performance status (2–4 points), a higher proportion of cirrhosis, and more advanced BCLC stage (all P < 0.05). In addition, the proportions of patients with SII ≥600, CONUT score ≥3, and preoperative AFP ≥400 were significantly higher in the EPI group than in the non-EPI group (all P < 0.001). Regarding tumor characteristics, maximum tumor diameter differed significantly between the two groups (P = 0.016).
For perioperative factors, patients in the EPI group had a higher proportion of open surgery, longer operative duration (≥180 min), greater intraoperative blood loss (≥400 mL), and a higher rate of intraoperative blood transfusion compared with those in the non-EPI group (all P < 0.05). Regarding clinical outcomes (Supplementary Table 2), significant differences were also observed between the EPI and non-EPI groups. Compared with patients without EPI, those with EPI experienced a significantly longer hospital stay (17.99 ± 14.37 days vs. 11.11 ± 5.36 days) and a higher rate of postoperative intensive care unit (ICU) admission (30.22% vs. 4.47%). Furthermore, the incidences of minor complications (46.15% vs. 28.31%) and major complications (49.45% vs. 16.01%) were significantly higher in the EPI group than in the non-EPI group (all P < 0.05).
Identification of risk factors for EPI
To evaluate correlations and potential collinearity among candidate predictors, a heatmap was first generated to visualize pairwise correlations among all candidate variables and explore the underlying correlation patterns (Figure 2). Subsequently, multicollinearity was quantitatively assessed using the variance inflation factor (VIF). All candidate variables showed relatively low VIF values (range, 1.00–1.47), indicating no significant multicollinearity among the predictors (Supplementary Figure 4).
Figure 2.
Correlation heatmap of variables.
LASSO regression analysis was subsequently performed to identify important predictors associated with EPI (Figure 3). Candidate variables are presented in Supplementary Table 3. Ten-fold cross-validation was applied to determine the optimal penalty parameter, and the λ value corresponding to the minimum mean cross-validation error (λ.min = 0.00682) was selected as the criterion for variable selection. At this parameter level, LASSO regression identified 13 candidate predictors, including ECOG performance status, drinking history, Child–Pugh grade, BCLC stage, SII, CONUT score, AFP level, tumor location, maximum tumor diameter, number of tumors, operative approach, operative duration, and intraoperative blood loss (Supplementary Table 4). These variables were subsequently entered into multivariable logistic regression analysis to identify independent predictors of EPI and develop the INSEP prediction model.
Figure 3.
The selection process of the optimum value of the parameter λ in the Lasso regression model by cross-validation method; (A) the LASSO path plot; (B) the cross-validation plot.
Considering the multicenter design of this study, potential differences in baseline characteristics and perioperative management strategies among participating centers may introduce center effects. Therefore, the impact of center effects on model parameter stability was further evaluated. In the training dataset, study center (the First Hospital of Jilin University and the First Affiliated Hospital of Zhengzhou University) was forced into the multivariable logistic regression model as an adjustment covariate. Compared with the model without center adjustment, changes in the odds ratios (ORs) of all predictors were less than 10%, and the statistical significance of predictors remained largely unchanged (Supplementary Table 5). Moreover, the study center itself was not significantly associated with EPI risk (OR = 1.12, 95% CI: 0.69–1.82, P = 0.642). These findings suggested that center effects had no substantial influence on model parameter estimation.
Based on the multivariable logistic regression analysis after adjustment for center effects (Table 4), ECOG performance status, SII, CONUT score, surgical approach, and operative duration were identified as independent risk factors for EPI.
Table 4.
Risk factors for EPI in elderly HCC patients: multivariable logistic regression analysis adjusted for center effects.
| Variables | OR comparison | UV OR (95% CI) | UV P | MV OR (95% CI) | MV P |
|---|---|---|---|---|---|
| Demographic data | |||||
| ECOG (%) | 0–1 | Reference | < 0.001 | Reference | < 0.001 |
| 2–4 | 2.59 (1.70–3.94) | 4.96 (2.73–9.00) | |||
| Drink (%) | No | Reference | 0.123 | ||
| Yes | 1.21 (1.03–3.15) | ||||
| Child pugh grade (%) | A | Reference | 0.668 | ||
| B | 1.85 (1.41–2.78) | ||||
| BCLC stage (%) | 0 | Reference | < 0.001 | ||
| A | 1.78 (1.03–3.10) | 0.038 | |||
| B | 2.39 (1.50–3.81) | < 0.001 | |||
| Laboratory data | |||||
| SII (%) | < 600 | Reference | < 0.001 | Reference | 0.023 |
| ≥600 | 7.12 (4.33–11.71) | 2.73 (1.43–5.22) | |||
| CONUT (%) | < 3 | Reference | < 0.001 | Reference | < 0.001 |
| ≥3 | 8.57 (4.74–14.58) | 7.84 (3.46–15.34) | |||
| Preoperative AFP (%) | < 400 | Reference | < 0.001 | ||
| ≥400 | 2.89 (1.90–4.41) | ||||
| Imaging data | |||||
| Tumor location (%) | Right lobe | Reference | 0.009 | ||
| Left lobe | 1.64 (1.02–2.64) | 0.041 | |||
| Bilateral | 0.56 (0.23–1.34) | 0.195 | |||
| Maximum tumor diameter (CM) (%) | < 5 | Reference | 0.190 | ||
| 5–10 | 1.99 (1.64–2.52) | 0.969 | |||
| ≥10 | 1.45 (1.19–2.09) | 0.079 | |||
| Number of tumors (%) | Single | Reference | 0.775 | ||
| Multiple | 1.92 (1.52–2.64) | ||||
| Perioperative information | |||||
| Operative approach (%) | Open | Reference | 0.019 | Reference | 0.025 |
| Laparoscopic | 0.62 (0.28–0.74) | 0.83 (0.48–0.91) | |||
| Surgical duration (%) | < 180 | Reference | < 0.001 | Reference | < 0.001 |
| ≥180 | 3.05 (2.02–4.62) | 7.44 (4.02–13.78) | |||
| Intraoperative blood loss (%) | < 400 | Reference | < 0.001 | ||
| ≥400 | 2.78 (1.84–4.20) | ||||
ECOG, eastern cooperative oncology group; BCLC, barcelona clinic liver cancer; SII, systemic immune-inflammation index; CONUT, controlling nutritional status; AFP, alpha-fetoprotein; EPI, early postoperative infection; OR, odds ratio. All ORs presented in this table were adjusted for study site in the multivariable logistic regression model.
Specifically, compared with patients with an ECOG performance status of 0–1, those with ECOG scores of 2–4 had a significantly increased risk of EPI (OR = 4.96, 95% CI: 2.73–9.00, P < 0.001). Patients with SII ≥600 had a higher risk of EPI than those with SII < 600 (OR = 2.73, 95% CI: 1.43–5.22, P = 0.023). Similarly, a CONUT score ≥3 was associated with a markedly increased risk of EPI (OR = 7.84, 95% CI: 3.46–15.34, P < 0.001).
Furthermore, compared with open surgery, laparoscopic surgery was associated with a lower risk of EPI (OR = 0.83, 95% CI: 0.48–0.91, P = 0.025). Longer operative duration (≥180 min) was significantly associated with an increased risk of EPI, with patients undergoing prolonged surgery having approximately a 7.44-fold higher risk compared with those with operative duration < 180 min (OR = 7.44, 95% CI: 4.02–13.78, P < 0.001).
Development and validation of the INSEP model
Based on the independent risk factors identified above, an individualized risk prediction model was developed to accurately estimate the probability of EPI occurrence in elderly patients with HCC (Figure 4). To facilitate the integration of multiple predictors and provide individualized risk assessment, an INSEP nomogram was subsequently constructed.
Figure 4.
INSEP model predicting the occurrence of EPI in patients with elderly HCC patients; EPI, early postoperative infection.
The nomogram demonstrated excellent discriminatory performance, with AUC values of 0.870, 0.803, and 0.798 in the training dataset, internal validation dataset, and external validation dataset, respectively (Figures 5A–C). DeLong's test showed that the AUC values of the nomogram were significantly higher than those of each individual predictor in all datasets (all P < 0.05), indicating superior discrimination of the combined prediction model compared with any single predictor alone (Figures 5D–F).
Figure 5.
ROC curves of the INSEP model in the training dataset (A), internal validation dataset (B), and external validation dataset (C); DeLong test comparisons of the INSEP model versus other predictive models in the training dataset (D), internal validation dataset (E), and external validation dataset (F). ROC, receiver operating characteristic; AUC, area under the curve. *represents p < 0.05, **represents p < 0.01, ***represents p < 0.001. DeLong test was used to compare AUCs between models.
Calibration curves demonstrated good agreement between the predicted probabilities and the observed outcomes, with minimal overall deviation, indicating satisfactory calibration and reliability of the model. Specifically, points located above the diagonal reference line indicate that the model overestimated the risk, whereas points below the line indicate risk underestimation. In all validation datasets, the calibration curves closely approximated the ideal diagonal line, further supporting the robustness and potential clinical applicability of the INSEP nomogram (Figures 6A–C).
Figure 6.
Calibration curves of the INSEP model for predicting EPI in the training dataset (A), internal validation dataset (B), and external validation dataset (C); EPI, early postoperative infection.
Clinical utility and net benefit analysis
DCA and CIC were performed to evaluate the clinical utility of the INSEP prediction model and quantify its potential net benefit. As shown in Figures 7A–C, within a clinically relevant range of threshold probabilities, the INSEP model achieved greater net benefit compared with the “treat-all” strategy, the “treat-none” strategy, and individual predictors.
Figure 7.
DCA and CIC of the INSEP model for predicting EPI. The net clinical benefit of the model was evaluated across different threshold probabilities. (A–C) Show the DCA results for the training dataset, internal validation dataset, and external validation dataset, respectively, while (D–F) present the corresponding CICs. DCA, Decision curve analysis; CIC, clinical impact curves; EPI, early postoperative infection.
The clinical impact curves further demonstrated that the INSEP model maintained a favorable ability to identify patients at high risk of EPI across a wide range of risk thresholds (Figures 7D–F). These findings suggest that the INSEP model provides substantial advantages in supporting clinical decision-making and improving the accuracy of individualized risk stratification.
Furthermore, these results indicate that a prediction model integrating multiple prognostic factors provides superior predictive performance compared with single-variable approaches and may serve as a practical decision-support tool for assisting clinicians in identifying elderly HCC patients who are at increased risk of EPI.
Deployment of a web-based clinical decision support system
To facilitate the clinical implementation of the INSEP prediction model, we further deployed the model as an online web-based risk calculator, enabling clinicians to rapidly estimate the risk of EPI in elderly HCC patients based on individual clinical characteristics and laboratory parameters. The online calculator incorporates five independent risk factors, including surgical approach, operative duration, CONUT score, SII, and ECOG performance status (Figure 8). This web-based risk calculator is publicly accessible at: https://shao1234567.shinyapps.io/epi-risk-predictor/.
Figure 8.
Deployment of a web-based clinical decision support system based on the INSEP model.
Clinical scenario demonstration and user experience
To intuitively demonstrate the practical value of the INSEP online decision-support system for EPI risk assessment, two representative cases from the external validation dataset were selected for illustration (Supplementary Table 6).
In the low-risk case, the model predicted an EPI probability of 3.3%, and the patient did not develop any postoperative infectious events. In contrast, in the high-risk case, the predicted infection probability was 97.0%, and the patient developed pneumonia on postoperative day 5, which was consistent with the model prediction. Screenshots of the online calculator interface and corresponding risk prediction results are presented in Supplementary Figure 5.
Furthermore, three independent hepatobiliary surgeons, with 6, 8, and 10 years of clinical experience, respectively, were invited to evaluate the usability of the INSEP online calculator. All three physicians considered the tool easy to operate, with an estimated assessment time of approximately 25–60 s per patient. They reported that the interface was intuitive and that the risk prediction outputs were clear and easy to interpret. The physicians indicated that the risk stratification results could facilitate preoperative patient–physician communication and support perioperative clinical decision-making. No system errors or usability barriers were identified during the testing process. One physician suggested that incorporating specific infection types (e.g., pneumonia and surgical site infection) into the model output might further enhance its clinical utility (Supplementary Table 7). Collectively, these findings support the INSEP online calculator as a practical individualized risk stratification tool for assessing the risk of early postoperative infection among elderly HCC patients undergoing curative hepatectomy.
Discussion
In this study, we developed and externally validated an inflammation–nutrition–surgery-based predictive model for EPI in elderly HCC patients undergoing curative hepatectomy. Using a large multicenter dataset and a robust variable selection strategy based on LASSO and multivariable logistic regression analyses, we identified several independent predictors of EPI, including ECOG ≥2, SII ≥600, CONUT score ≥3, laparoscopic vs. open surgical approach, and surgical duration ≥180 min. Patients with these high-risk features exhibited a significantly higher incidence of postoperative EPI, which was further associated with adverse clinical outcomes, including prolonged hospital stay, increased ICU admission rates, and greater severity of postoperative complications.
Our findings support and extend previous evidence demonstrating a close association between poor preoperative nutritional status, a heightened systemic inflammatory state, and an increased risk of adverse postoperative outcomes in patients with malignancies (24–27). A recent multicenter retrospective study involving 210 patients with gallbladder cancer undergoing curative resection reported that an elevated systemic SII was independently associated with worse short-term outcomes and poorer long-term prognosis (28). Moreover, a combined model incorporating SII and hemoglobin achieved better predictive performance than either marker alone, highlighting the synergistic prognostic value of inflammatory and nutritional status (24, 29). Consistent with these findings, our study demonstrated that elevated preoperative SII and a higher CONUT score were independently associated with an increased risk of EPI in elderly HCC patients, supporting the importance of integrating inflammatory and nutritional parameters for perioperative risk assessment.
Mechanistically, an elevated SII reflects enhanced neutrophil- and platelet-mediated inflammatory activity together with impaired lymphocyte-mediated immune function. Excessive activation of the innate immune response may promote tissue injury and immune dysregulation, whereas lymphopenia compromises host antimicrobial defense, thereby increasing susceptibility to postoperative infections (25). Meanwhile, poor nutritional status, as reflected by a high CONUT score, may further aggravate immunosuppression by impairing immune cell proliferation, protein synthesis, and tissue repair. Consequently, patients with concurrent systemic inflammation and malnutrition may develop a state characterized by “high inflammation, impaired immunity, and poor nutritional reserve,” which substantially increases the risk of postoperative infectious complications following major surgery (26, 27, 30).
Notably, these associations may be particularly pronounced in elderly patients because aging is accompanied by two hallmark biological processes, namely immunosenescence and inflammaging (31–33). With advancing age, immune function progressively declines, while chronic low-grade inflammation persists, rendering elderly individuals more susceptible to exaggerated inflammatory responses and immune dysregulation following surgical stress (34). In addition, elderly patients often have reduced physiological reserve and a greater burden of comorbidities, further limiting their ability to recover from surgical trauma and increasing their vulnerability to postoperative infectious complications (28). Therefore, inflammatory and nutritional biomarkers may capture not only conventional inflammatory and nutritional status but also age-related physiological vulnerability in elderly HCC patients.
In addition, this study also found that a higher ECOG performance status score (2–4) was significantly associated with the occurrence of EPI. As an important indicator reflecting overall functional status and physiological reserve, the ECOG score has been widely used in prognostic assessment of cancer patients (35–37). A higher ECOG score generally indicates limited physical activity, poorer baseline condition, and reduced tolerance to surgical stress (38, 39).
Furthermore, patients with poor performance status are often accompanied by impaired immune function and malnutrition. Impaired immune function weakens the host's ability to eliminate pathogens, while malnutrition may compromise immune cell proliferation and functional recovery, thereby increasing the risk of postoperative infection. In addition, reduced perioperative mobility in patients with higher ECOG scores may further contribute to an increased incidence of pulmonary infections and other complications, which is consistent with the findings observed in this study (40, 41). Therefore, the ECOG score not only serves as a simple tool for assessing baseline patient status, but may also be combined with inflammatory and nutritional indicators to enable early identification of patients at high risk of postoperative infection.
In this study, we found that laparoscopic surgery was associated with a lower risk of EPI compared with open surgery, while prolonged surgical duration (>3 h) was significantly associated with an increased risk of EPI in elderly patients with HCC. These findings highlight the important role of operative approach and surgical burden in the development of perioperative infections. Minimally invasive approaches, such as laparoscopic surgery, are generally associated with reduced surgical trauma, shorter tissue exposure time, and a more controlled intraoperative environment. These advantages may lead to a less pronounced systemic inflammatory response and better preservation of postoperative immune function (42, 43). In addition, laparoscopic procedures are typically associated with less intraoperative blood loss and less extensive tissue manipulation, which may further reduce the risk of postoperative infectious complications (44–46). In contrast, open surgery often involves greater surgical trauma, prolonged exposure of tissues, and more complex operative procedures, all of which may contribute to increased susceptibility to postoperative infection.
On the other hand, prolonged surgical duration is often indicative of increased procedural complexity and a higher perioperative stress burden. Extended operative time may increase the risk of intraoperative contamination and sustain systemic inflammatory activation, thereby exacerbating postoperative immunosuppression (47–49). Furthermore, prolonged anesthesia exposure and ischemia–reperfusion injury may impair immune recovery and delay postoperative healing (50, 51). These effects may be particularly pronounced in elderly patients, who often exhibit reduced physiological reserve and immunosenescence. Taken together, our findings suggest that optimizing surgical strategies—such as prioritizing minimally invasive approaches when feasible and improving operative efficiency to reduce surgical duration—may represent important avenues for reducing the risk of postoperative infection in elderly HCC patients.
Furthermore, the web-based INSEP calculator demonstrated good clinical applicability in representative clinical scenarios. By providing rapid and individualized risk estimation, the tool may facilitate preoperative risk stratification, perioperative management, and clinical decision-making for elderly HCC patients. Compared with conventional prediction models, the INSEP calculator combines favorable predictive performance with ease of use, supporting its potential for routine clinical implementation. Nevertheless, further prospective multicenter studies are warranted to evaluate its real-world effectiveness and clinical impact.
Several limitations of this study should be acknowledged. First, this was a retrospective study. Although a relatively large sample size and multicenter validation were employed, selection bias and information bias could not be completely eliminated. Prospective studies are therefore required to further validate the reliability of the model. Second, the study population was derived predominantly from Chinese patients. Although the INSEP model demonstrated good discriminative ability in the independent external validation dataset (AUC = 0.798), its applicability to other ethnic populations and healthcare systems requires further validation using larger international multicenter datasets. Third, the model was primarily developed using preoperative inflammatory and nutritional indicators together with perioperative variables. More detailed intraoperative parameters, such as real-time physiological changes and anesthesia-related factors, as well as dynamic postoperative indicators, were not included. The absence of these variables may limit the model's ability to fully characterize the mechanisms and dynamic evolution of EPI. Future studies integrating multidimensional clinical information may further improve prediction accuracy. Fourth, several continuous variables were converted into categorical variables according to optimal cutoff values derived from ROC curve analysis to improve model interpretability and facilitate clinical application. Although restricted cubic spline analysis demonstrated no significant nonlinear association between SII or operative duration and EPI risk, categorization of continuous variables inevitably results in some loss of information. In addition, the ROC-derived optimal cutoff values for SII (597) and operative duration (178 min) were rounded to 600 and 180 min, respectively, to enhance clinical practicality. The stability of these cutoff values should be further confirmed in larger prospective datasets. Fifth, baseline characteristics differed between the model development dataset and the external validation dataset, reflecting the inherent heterogeneity in patient populations, perioperative management strategies, and surgical practices across different medical centers. To minimize the potential influence of center effects, study center was included as an adjustment variable in the multivariable logistic regression analysis. The results showed that study center was not independently associated with EPI and that adjustment for center effects resulted in only minimal changes in the estimated effects of the predictors. Moreover, despite the observed inter-center heterogeneity, the INSEP model maintained stable predictive performance in the external validation dataset, suggesting satisfactory generalizability beyond specific institutional settings. Sixth, large-scale prospective evidence evaluating the real-world implementation of the INSEP model is still lacking. Future multicenter prospective studies are warranted to further assess its robustness, generalizability, and clinical impact on perioperative management and patient outcomes across diverse clinical settings. Finally, although independent hepatobiliary surgeons provided favorable feedback regarding the usability and clinical practicality of the INSEP online calculator, one physician suggested that the current model predicts only the overall risk of EPI without distinguishing specific infection types. Future studies incorporating more detailed infection phenotypes and developing multitask prediction models may further enhance its value for infection-specific risk prediction and clinical decision support.
5. Conclusion
This study developed and validated a comprehensive predictive model (INSEP) based on inflammatory, nutritional, and surgical factors for estimating the risk of early postoperative infection (EPI) in elderly patients undergoing curative hepatectomy for hepatocellular carcinoma (HCC). The model demonstrated stable and excellent discriminative ability and calibration performance across multicenter datasets, outperforming conventional single biomarkers. The INSEP model provides a simple and reliable tool for preoperative risk stratification, enabling clinicians to identify high-risk patients and implement individualized perioperative management strategies, thereby potentially improving patient outcomes. Future prospective studies are still required to further validate its clinical utility.
Acknowledgments
The authors thank Dr. Zhi-Chao Huang, Dr. Run-Jiang Guo, and Dr. Chang-Cheng Zhao for their independent testing and valuable feedback on the web-based clinical decision support tool.
Funding Statement
The author(s) declared that financial support was not received for this work and/or its publication.
Edited by: Rodrigo Barderas, Carlos III Health Institute (ISCIII), Spain
Reviewed by: Zhangyou Guo, Yunnan Cancer Hospital, China
Zhipeng Jin, Liaoning Cancer Hospital and Institute, China
Abbreviations: HCC, hepatocellular carcinoma; EPI, early postoperative infection; INSEP, inflammation-nutrition-surgery-based model; SII, systemic immune-inflammation index; CONUT, controlling nutritional status score; AFP, alpha-fetoprotein; BCLC, Barcelona Clinic Liver Cancer staging system; ASA, American Society of Anesthesiologists; BMI, body mass index; ROC, receiver operating characteristic; AUC, area under the curve; CI, confidence interval; OR, odds ratio; DCA, decision curve analysis; CIC, clinical impact curve; ICU, intensive care unit; LOS, length of stay.
Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
Ethics statement
The studies involving humans were approved by the Ethics Committee of the First Hospital of Jilin University. The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation was not required from the participants or the participants' legal guardians/next of kin in accordance with the national legislation and institutional requirements.
Author contributions
G-ZS: Data curation, Investigation, Writing – original draft, Writing – review & editing. D-DH: Conceptualization, Formal analysis, Investigation, Writing – original draft. F-XL: Conceptualization, Formal analysis, Methodology, Project administration, Writing – review & editing. D-SZ: Conceptualization, Methodology, Project administration, Writing – review & editing. Q-ZW: Data curation, Investigation, Project administration, Writing – review & editing. G-TF: Conceptualization, Methodology, Project administration, Writing – review & editing. H-TZ: Investigation, Methodology, Project administration, Visualization, Writing – review & editing. S-FS: Data curation, Investigation, Validation, Writing – review & editing. WQ: Data curation, Funding acquisition, Methodology, Project administration, Visualization, Writing – original draft. Y-GC: Formal analysis, Funding acquisition, Methodology, Project administration, Visualization, Writing – review & editing.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that Generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2026.1883480/full#supplementary-material
Restricted cubic spline analysis of the associations between SII index, surgical duration, and EPI. RCS, restricted cubic spline; SII, systemic immune-inflammation index; EPI, early postoperative infection; OR, odds ratio; CI, confidence interval.
ROC curve analysis of SII index and surgical duration for predicting EPI. ROC, Receiver operating characteristic; SII, systemic immune-inflammation index; EPI early postoperative complications.
Common types of EPI (A) and their associated pathogens (B); EPI, early postoperative infections.
Assessment of multicollinearity among candidate variables using variance inflation factor (VIF).
Demonstration of the INSEP online calculator input and output interfaces for a low-riskcase (Left) and a high-risk case (Right).
Definitions of postoperative infections.
Postoperative clinical outcomes stratified by EPI.
Candidate variables included in LASSO regression analysis.
Variables retained by LASSO regression at λ.min.
Comparison of odds ratios for EPI predictors before and after adjustment for center effects.
Clinical characteristics and outcomes of two representative cases from the external validation dataset.
Usability feedback from three independent hepatobiliary surgeons on the INSEP online calculator.
References
- 1.Hwang SY, Danpanichkul P, Agopian V, Mehta N, Parikh ND, Abou-Alfa GK, et al. Hepatocellular carcinoma: updates on epidemiology, surveillance, diagnosis and treatment. Clin Mol Hepatol. (2025) 31:S228–54. doi: 10.3350/cmh.2024.0824 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Mauro E, de Castro T, Zeitlhoefler M, Sung MW, Villanueva A, Mazzaferro V, et al. Hepatocellular carcinoma: epidemiology, diagnosis and treatment. JHEP Rep. (2025) 7:101571. doi: 10.1016/j.jhepr.2025.101571 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Magyar C, Rajendran L, Li Z, Vogel A, O'Kane GM, Chan ACY, et al. Precision surgery for hepatocellular carcinoma. Lancet Gastroenterol Hepatol. (2025) 10:350–68. doi: 10.1016/S2468-1253(24)00434-5 [DOI] [PubMed] [Google Scholar]
- 4.Malik AK, Geh D, Jeffry Evans TR, Chow P, Mann DA, White SA. Improving surgical treatments for hepatocellular carcinoma. Nat Rev Gastroenterol Hepatol. (2026) 23:208–26. doi: 10.1038/s41575-025-01143-y [DOI] [PubMed] [Google Scholar]
- 5.Knitter S, Noltsch AR, Pesthy S, Beierle AS, Krenzien F, Schöning W, et al. Postoperative infectious complications are associated with diminished overall survival of patients undergoing resection for colorectal liver metastases. Eur J Surg Oncol. (2023) 49:106920. doi: 10.1016/j.ejso.2023.04.019 [DOI] [PubMed] [Google Scholar]
- 6.Asai Y, Yamaguchi J, Mizuno T, Onoe S, Watanabe N, Igami T, et al. Impact of preoperative muscle mass and quality on surgical outcomes in patients undergoing major hepatectomy for perihilar cholangiocarcinoma. J Hepatobiliary Pancreat Sci. (2023) 30:202–11. doi: 10.1002/jhbp.1220 [DOI] [PubMed] [Google Scholar]
- 7.Wei T, Zhang XF, Bagante F, Ratti F, Marques HP, Silva S, et al. Postoperative infectious complications worsen long-term survival after curative-intent resection for hepatocellular carcinoma. Ann Surg Oncol. (2022) 29:315–24. doi: 10.1245/s10434-021-10565-2 [DOI] [PubMed] [Google Scholar]
- 8.She S, Shi J, Zhu J, Yang F, Yu J, Dai K. Impact of inflammation and the immune system on hepatocellular carcinoma recurrence after hepatectomy. Cancer Med. (2024) 13:e7018. doi: 10.1002/cam4.7018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Zhang S, Tang Z. Prognostic and clinicopathological significance of systemic inflammation response index in patients with hepatocellular carcinoma: a systematic review and meta-analysis. Front Immunol. (2024) 15:1291840. doi: 10.3389/fimmu.2024.1291840 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Du Q, Zhao Y, Yang J, Yang Y. Visit-to-visit variability of inflammation-immunity indices and prognosis in hepatocellular carcinoma. BMC Gastroenterol. (2025) 26:70. doi: 10.1186/s12876-025-04506-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Sun Y, Liang M, Wang X, Dong W, Wu Z, Sun F, et al. Inflammatory and nutritional markers predict response and prognosis of patients with locally advanced gastric cancer receiving neoadjuvant immunochemotherapy. BMC Gastroenterol. (2025) 25:478. doi: 10.1186/s12876-025-03874-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Megyesfalvi E, Ghimessy A, Bauer J, Pipek O, Saghi K, Gellert A, et al. Diagnostic and prognostic relevance of inflammatory markers in surgically treated thymic epithelial tumors: an international multicenter study. Lung Cancer. (2025) 200:108111. doi: 10.1016/j.lungcan.2025.108111 [DOI] [PubMed] [Google Scholar]
- 13.Bektaş O, Bektaş K. Prognostic value of systemic inflammatory markers in predicting neoadjuvant platinum response and platinum resistance in advanced epithelial ovarian cancer: a peritoneal cancer index-based analysis. Gynecol Oncol. (2025) 199:147–51. doi: 10.1016/j.ygyno.2025.07.001 [DOI] [PubMed] [Google Scholar]
- 14.Zhu P, Liao W, Zhang WG, Chen L, Shu C, Zhang ZW, et al. A prospective study using propensity score matching to compare long-term survival outcomes after robotic-assisted, laparoscopic, or open liver resection for patients with BCLC stage 0-A hepatocellular carcinoma. Ann Surg. (2023) 277:e103–11. doi: 10.1097/SLA.0000000000005380 [DOI] [PubMed] [Google Scholar]
- 15.Tsai KY, Chen HA, Wang WY, Huang MT. Long-term and short-term surgical outcomes of laparoscopic vs. open liver resection for hepatocellular carcinoma: might laparoscopic approach be better in early HCC. Surg Endosc. (2019) 33:1131–9. doi: 10.1007/s00464-018-6372-0 [DOI] [PubMed] [Google Scholar]
- 16.Wu HL, Liu HY, Liu WC, Hou MC, Tai YH. A. predictive model incorporating inflammation markers for high-grade surgical complications following liver resection for hepatocellular carcinoma. J Chin Med Assoc. (2022) 85:845–52. doi: 10.1097/JCMA.0000000000000713 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Bao M, Zhu Q, Aji T, Wei S, Tuergan T, Ha X, et al. Development of models to predict postoperative complications for hepatitis B virus-related hepatocellular carcinoma. Front Oncol. (2021) 11:717826. doi: 10.3389/fonc.2021.717826 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Shen Y, Hu YL, Xu JH, Zhu S, Cai L, Wu YF, et al. Incidence, risk factors, outcomes, and prediction model of surgical site infection after hepatectomy for hepatocellular carcinoma: a multicenter cohort study. Eur J Surg Oncol. (2025) 51:109486. doi: 10.1016/j.ejso.2024.109486 [DOI] [PubMed] [Google Scholar]
- 19.Liu X, Qiu Z, Ndhlovu E, Wan Y., Sun H., Wang S., et al. Establishing and externally validating a hemoglobin, albumin, lymphocyte, and platelet (HALP) score-based nomogram for predicting early recurrence in BCLC stage 0/A hepatocellular carcinoma patients after radical liver resection: a multi-center study. J Hepatocell Carcinoma. (2024) 11:1127–41. doi: 10.2147/JHC.S465670 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Loman BR, Luo M, Baggs GE, Mitchell DC, Nelson JL, Ziegler TR, et al. Specialized high-protein oral nutrition supplement improves home nutrient intake of malnourished older adults without decreasing usual food intake. JPEN J Parenter Enteral Nutr. (2019) 43:794–802. doi: 10.1002/jpen.1467 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Jacenko S, Blough S, Grant G, Tohme R, McFarland, et al. Lessons learnt from the applying the centers for disease control and prevention (CDC) evaluation framework to the measles incident management system response, USA, 2020–2021. BMJ Glob Health. (2023) 8:e011861. doi: 10.1136/bmjgh-2023-011861 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Shehab N, Alschuler L. McILvenna S, Gonzaga Z, Laing A, deRoode D, et al. The national healthcare safety network's digital quality measures: CDC's automated measures for surveillance of patient safety. J Am Med Inform Assoc. (2024) 31:1199–205. doi: 10.1093/jamia/ocae064 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Abbassi F, Pfister M, Lucas KL, Domenghino A, Puhan MA, Clavien PA. Milestones in surgical complication reporting: clavien-dindo classification 20 years and comprehensive complication index 10 years. Ann Surg. (2024) 280:763–71. doi: 10.1097/SLA.0000000000006471 [DOI] [PubMed] [Google Scholar]
- 24.Hu B, Yang XR. Xu Y, Sun YF, Sun C, Guo W, et al. Systemic immune-inflammation index predicts prognosis of patients after curative resection for hepatocellular carcinoma. Clin Cancer Res. (2014) 20:6212–22. doi: 10.1158/1078-0432.CCR-14-0442 [DOI] [PubMed] [Google Scholar]
- 25.González García E, Arroyo T, Galván M, Becerra MJ, Gallego M, Mauro I, et al. Association between nutritional status and early postoperative infection risk in kidney transplant patients. Nutrients. (2025) 17:1935. doi: 10.3390/nu17111935 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Sowerbutts AM, Burden S, Sremanakova J, French C, Knight SR, Harrison EM. Preoperative nutrition therapy in people undergoing gastrointestinal surgery. Cochrane Database Syst Rev. (2024) 4:CD008879. doi: 10.1002/14651858.CD008879.pub3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Zhao W, Liu L. Zhang J, Shi L, Zhang C, Ma Y, et al. Preoperative geriatric nutritional risk index predicts prognosis and postoperative complications in elderly patients undergoing non-cardiac surgery: a retrospective cohort study. BMC Geriatr. (2025) 25:101. doi: 10.1186/s12877-024-05667-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Liu F, Yin P, Jiao B, Shi Z, Qiao F, Xu J. Detecting the preoperative peripheral blood systemic immune-inflammation index (SII) as a tool for early diagnosis and prognosis of gallbladder cancer. BMC Immunol. (2025) 26:7. doi: 10.1186/s12865-025-00683-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Xie R, Xiao M. Li L, Ma N, Liu M, Huang X, et al. Association between SII and hepatic steatosis and liver fibrosis: a population-based study. Front Immunol. (2022) 13:925690. doi: 10.3389/fimmu.2022.925690 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Boubekki A, Myhre JN, Luppino LT, Mikalsen KO, Revhaug A, Jenssen R. clinically relevant features for predicting the severity of surgical site infections. IEEE J Biomed Health Inform. (2022) 26:1794–801. doi: 10.1109/JBHI.2021.3121038 [DOI] [PubMed] [Google Scholar]
- 31.He W, Yao C, Wang K, Duan Z, Wang S, Xie L. Single-cell landscape of immunological responses in elderly patients with sepsis. Immun Ageing. (2024) 21:40. doi: 10.1186/s12979-024-00446-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Ziogas DC, Theocharopoulos C, Aravantinou K, Boukouris AE, Stefanou D, Anastasopoulou A, et al. Clinical benefit of immune checkpoint inhibitors in elderly cancer patients: current evidence from immunosenescence pathophysiology to clinical trial results. Crit Rev Oncol Hematol. (2025) 208:104635. doi: 10.1016/j.critrevonc.2025.104635 [DOI] [PubMed] [Google Scholar]
- 33.Ajoolabady A, Pratico D, Tang D, Zhou S, Franceschi C, Ren J. Immunosenescence and inflammaging: Mechanisms and role in diseases. Ageing Res Rev. (2024) 101:102540. doi: 10.1016/j.arr.2024.102540 [DOI] [PubMed] [Google Scholar]
- 34.Kaminski HJ, Kusner LL, Cutter GR, Le Panse R, Wright CD, Perry Y, et al. Does surgical removal of the thymus have deleterious consequences. Neurology. (2024) 102:e209482. doi: 10.1212/WNL.0000000000209482 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Nomura M, Yamaguchi T. Chin K, Hato S, Kato K, Baba E, et al. Phase II trial of adjuvant S-1 following neoadjuvant chemotherapy and surgery in patients with locally advanced esophageal squamous cell carcinoma: the PIECE trial. Ann Surg Oncol. (2025) 32:302–11. doi: 10.1245/s10434-024-16325-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Kolligs F, Arnold D. Golfieri R, Pech M, Peynircioglu B, Pfammatter T, et al. Factors impacting survival after transarterial radioembolization in patients with hepatocellular carcinoma: results from the prospective CIRT study. JHEP Rep. (2023) 5:100633. doi: 10.1016/j.jhepr.2022.100633 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Ma D, Liu M, Zhai X, Li X, Jin B, Liu Y. Development and validation of prognostic risk prediction models for hepatocellular carcinoma patients treated with immune checkpoint inhibitors based on a systematic review and meta-analysis of 47 cohorts. Front Immunol. (2023) 14:1215745. doi: 10.3389/fimmu.2023.1215745 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Xu B, Lu D. Liu K, Lv W, Xiao J, Zhang X, et al. Efficacy and prognostic factors of regorafenib in the treatment of BCLC Stage C hepatocellular carcinoma after failure of the first-line therapy. Drug Des Devel Ther. (2023) 17:507–18. doi: 10.2147/DDDT.S400533 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Endo K, Kakisaka K. Abe T, Yusa K, Nakaya I, Watanabe T, et al. Grip strength complements performance status in assessing general condition in patients with unresectable hepatocellular carcinoma treated with atezolizumab and bevacizumab. Hepatol Res. (2024) 54:1158–73. doi: 10.1111/hepr.14084 [DOI] [PubMed] [Google Scholar]
- 40.Matsuura S. Patient-related factors in advanced lung cancer: a review of ECOG-PS, malnutrition, and sarcopenia. Respir Investig. (2025) 63:1250–61. doi: 10.1016/j.resinv.2025.10.005 [DOI] [PubMed] [Google Scholar]
- 41.Bhargavan R, Philip FA, Km JK, Augustine P, Thomas S. Comparison of modified frailty index, clinical frailty scale, ECOG score, and ASA PS score in predicting postoperative outcomes in cancer surgery: a prospective study. Indian J Surg Oncol. (2024) 15:938–45. doi: 10.1007/s13193-024-01995-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Fall S, Souche R, Bardol T, Fabre JM, Borie F. Laparoscopic versus open pancreaticoduodenectomy for pancreatic or periampullary tumors: a multicenter propensity score-matched comparative study. Surg Endosc. (2025) 39:3037–48. doi: 10.1007/s00464-025-11677-6 [DOI] [PubMed] [Google Scholar]
- 43.Xu S, Xu Y. Wang S, Chu Q, Zhang H, Gong W, et al. Comparison of short- and long-term outcomes between laparoscopic and open pancreaticoduodenectomy in overweight patients: a propensity score-matched study. Surg Endosc. (2025) 39:881–90. doi: 10.1007/s00464-024-11418-1 [DOI] [PubMed] [Google Scholar]
- 44.Aulicino M, Ferracci F. Abatini C, Lodoli C, Di Giorgio A, Attalla El Halabieh M, et al. Short-term outcomes of laparoscopic versus open cytoreductive surgery: a propensity score matching analysis. Surg Endosc. (2025) 39:6430–9. doi: 10.1007/s00464-025-12026-3 [DOI] [PubMed] [Google Scholar]
- 45.Mueller TC, Kehl V. Dimpel R, Blankenstein C, Egert-Schwender S, Strudthoff J, et al. Intraoperative wound irrigation for the prevention of surgical site infection after laparotomy: a randomized clinical trial by CHIR-net. JAMA Surg. (2024) 159:484−92. doi: 10.1001/jamasurg.2023.7985 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Zaman S, Mohamedahmed A, Abdelrahman W, Abdalla HE, Wuheb AA, Issa MT, et al. Minimally invasive surgery for inflammatory bowel disease: a systematic review and meta-analysis of robotic versus laparoscopic surgical techniques. J Crohns Colitis. (2024) 18:1342–55. doi: 10.1093/ecco-jcc/jjae037 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.de Angelis P, Tan KS, Chudgar NP, Dycoco J, Adusumilli PS, Bains MS, et al. Operative time is associated with postoperative complications after pulmonary lobectomy. Ann Surg. (2023) 278:e1259-e1266. doi: 10.1097/SLA.0000000000005696 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Jacobs RC, Valukas CS, Visa MA, Logan CD, Feinglass JM, Lung KC, et al. Association of operative time and approach on postoperative complications for esophagectomy. Surgery. (2024) 176:1106–14. doi: 10.1016/j.surg.2024.06.021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Kuemmerli C, Sijberden JP, Cipriani F, Osei-Bordom D, Aghayan D, Lanari J, et al. Is prolonged operative time associated with postoperative complications in liver surgery? An international multicentre cohort study of 5424 patients. Surg Endosc. (2024) 38:7118–7130. doi: 10.1007/s00464-024-11276-x [DOI] [PubMed] [Google Scholar]
- 50.Rivas E, Palermo J. Cekmecelioglu BT, Li Y, Yang D, Saha AK, et al. Postoperative oxygen saturation and surgical-site infection after major non-cardiac surgery: a retrospective analysis. J Clin Anesth. (2025) 106:111947. doi: 10.1016/j.jclinane.2025.111947 [DOI] [PubMed] [Google Scholar]
- 51.Feng Y, Sun JF, Wei HC, Cao Y, Yao L, Du BX. Correlation between anesthesia methods and adverse short-term postoperative outcomes depending on frailty: a prospective cohort study. Clin Interv Aging. (2024) 19:613–26. doi: 10.2147/CIA.S448898 [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
Restricted cubic spline analysis of the associations between SII index, surgical duration, and EPI. RCS, restricted cubic spline; SII, systemic immune-inflammation index; EPI, early postoperative infection; OR, odds ratio; CI, confidence interval.
ROC curve analysis of SII index and surgical duration for predicting EPI. ROC, Receiver operating characteristic; SII, systemic immune-inflammation index; EPI early postoperative complications.
Common types of EPI (A) and their associated pathogens (B); EPI, early postoperative infections.
Assessment of multicollinearity among candidate variables using variance inflation factor (VIF).
Demonstration of the INSEP online calculator input and output interfaces for a low-riskcase (Left) and a high-risk case (Right).
Definitions of postoperative infections.
Postoperative clinical outcomes stratified by EPI.
Candidate variables included in LASSO regression analysis.
Variables retained by LASSO regression at λ.min.
Comparison of odds ratios for EPI predictors before and after adjustment for center effects.
Clinical characteristics and outcomes of two representative cases from the external validation dataset.
Usability feedback from three independent hepatobiliary surgeons on the INSEP online calculator.
Data Availability Statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.