We collected clinical and follow-up data of elderly patients > 65 years of age who underwent radical hepatectomy for HCC at West China Hospital of Sichuan University (Chengdu, Sichuan Province, China) from January 2015 to September 2018. The preoperative diagnosis of HCC was performed according to the criteria of the American Association for the Study of Liver Diseases (AASLD) (12). The absolute contraindications for liver resection are American Society of Anesthesiologists (ASA) grade >= III, ascites, extrahepatic metastasis, unresectable large vessel tumor invasion, or future residual liver <40%-50% (13). The inclusion criteria were as follows:(1) Patients aged >= 65 years (male or female); (2) Primary hepatocellular carcinoma was confirmed by postoperative pathological examination; (3) The years of diagnosis were 2015–2018; (4) Without the absolute contraindications for liver resection; (5) no preoperative anticancer treatments. The exclusion criteria were as follows: (1) Patients aged < 65 years; (2) Intrahepatic cholangiocarcinoma or mixed-type HCC was confirmed by postoperative pathological examination; (3) Patients have absolute contraindications for liver resection; (4) Severe dysfunction of vital organs; (5) History of any other malignancy.

For the included patients, we collected the basic information of the patients at the time of admission in the medical records, including age, sex, overweight (BMI>24), viral hepatitis B, liver cirrhosis, abdominal surgery history, hypertension, diabetes, chronic obstructive pulmonary disease (COPD), tumor size (cm), tumor numbers (single or multiply), tumor location (VII/VIII/IVa or not), alpha-fetoprotein (AFP, ng/L), indocyanine green retention rate at 15 minutes (ICGR15, %), TB (mmol/L), ALB (g/L), aspartate aminotransferase (AST, IU/L), alanine aminotransferase (ALT, IU/L), platelet (PLT, IU/L), white blood count (WBC, 10⁹/L).

In addition, we also collected the data of patients undergoing surgical treatment, postoperative pathological data, and postoperative complications in the medical records. The information includes ASA(I~IV), surgical type(laparoscopic or open), major resection, blood loss(ml), intraoperative transfusion, total pringle time(min), margin distance(mm), cancer embolus, microvascular invasion, capsular invasion, microsatellite nodules, poor differentiation, necrosis (0~4), fibrosis (0~4), overall complications, major complications, liver-specific complications, liver failure, hemorrhage, ascites, biliary leakage, general complications, respiratory complications, respiratory infection, wound infection, pleural effusion, atelectasis respiratory insufficiency, Clavien-Dindo Grade(I~IV) and hospital stay.

All patients were followed up every 1 month for the first 3 months after discharge and every 3 months thereafter. The median follow-up time was 34.5 months. The primary endpoints of the study were overall survival (OS) and disease-free survival (DFS), and the secondary endpoint was complication rate. DFS was defined from the end of surgery to death or recurrence; OS was defined from the end of surgery to death.

All patients were randomly assigned into training cohort and validation cohort in a ratio of 3:1. For descriptive statistics of patient clinical characteristics data, the statistical description median [interquartile range (IQR)] was used for continuous variables, and frequency (%) was used for categorical variables. Univariate Cox regression was used to screen the risk factors related to prognosis. Prognostic variables were further selected by least absolute shrinkage and selection operator regression model (LASSO) and multivariate Cox regression to identify the predictors of overall survival (OS) and disease-free survival (DFS). Proportional hazards assumption was used to assessed the Cox regression models. These indicators were then used to construct a predictive nomogram.

A series of validation methods were used to validate the accuracy and discrimination of the nomogram, including AUC, calibration curve and consistency index (C-index). In order to explore clinical application value of the model, decision analysis curve (DCA) was used to calculate net benefit under different thresholds. At the same time, we divided all patients into low-risk and high-risk groups based on each patient’s nomogram score. Log-rank test and Kaplan-Meier (K-M) curve were used to compare survival differences among patients in different groups.

All data analyses were performed using R 4.1.1. and SPSS 25.0. All tests were two-sided, and a P<0.05 was considered statistically significant.

A total of 188 elderly HCC patients who underwent hepatectomy were included in this study, as shown in Figure 1 . Of these patients, males accounted for 77.66% and the median age was 69 (66,72) years. Among them, 40.43% of the patients were overweight (BMI>24). 39.36% and 41.49% of the patients had a history of hepatitis B and liver cirrhosis respectively. 29.26% of the patients had abdominal surgery history. In terms of common chronic diseases in the elderly, hypertension accounted for 34.57%, diabetes accounted for 17.55%, and COPD accounted for only 3.19% ( Table 1 ).

The median tumor size of patients was 5.65cm (4.00,7.00). The tumors of 72.87% of the patients were single lesions, and the tumors of 51.06% of the patients were located in the posterosuperior segments (VII/VIII/IVa) ( Table 1 ). Part of the blood routine, blood biochemical, and serum tumor marker data of the patients are shown in Table 1 . All data were collected at the time of admission. Table 2 shows the surgical, pathological, and postoperative information of 188 patients.

In the training cohort, 46 patients died, while in the validation cohort, 15 patients died. The 1-year and 3-year OS of the training cohort was 83.7% and 72.3% of that of the validation cohort, respectively, and the OS of the validation cohort was 85.1% and 74.5%, respectively.

In order to initially determine the factors related to the OS and DFS of the elderly patients, univariate Cox regression analysis was performed. The results showed that ALB, viral hepatitis B, TB, PLT, cancer embolus, capsular invasion, major resection, microvascular invasion, blood loss, and overweight (BMI>24) were significantly related to the OS of the patients ( Table 3 ). ALB, major resection, capsular invasion, TB, pleural effusion, microvascular invasion, and PLT were significantly related to the DFS of the patients ( Table 3 ).

The LASSO regression analysis was used to reduce high-dimensional data (14). The features with a P value <0.2 in the univariate Cox regression analysis of OS and DFS were included in the LASSO. The cross-validation was used to determine the most appropriate λ value as optimal parameters ( Figure 2 ). The results for OS yield a λ value of 12, and the results for DFS yield a λ value of 13. Indicating that 12 features associated with OS and 13 features associated with DFS included in LASSO were important.

Significant variables output by LASSO regression analysis were included in multivariate Cox regression analysis. The results of multivariate Cox regression analysis showed that ALB, cancer embolus, blood loss, viral hepatitis B, TB, microvascular invasion, overweight (BMI>24), and major resection were independent risk factors for OS in elderly patients ( Table 4 ). Major resection, ALB, microvascular invasion, laparoscopic surgery, blood loss, TB, and pleural effusion were independent risk factors for DFS in elderly patients ( Table 4 ). The proportional hazards assumption was evaluated and found reasonable for each variable ( Table S1 ).

The following factors were used to construct the nomogram ( Figures 3 , 4 ). 1-, 2-, and 3-year OS and DFS in elderly patients were predicted by the nomogram. We used ROC curves to verify the predictive value of these independent prognostic factors. As shown in Figures 3 and 4 , the area under the curve (AUC value) of the eight independent prognostic factors for the prediction of OS was 0.827. The AUC value of 6 independent prognostic factors for the prediction of DFS was 0.739. In the validation cohort, the AUC values were 0.798 and 0.694, respectively. This suggests that these factors have good predictive value for OS and DFS. Subsequently, calibration curves were depicted, and the results showed that the predicted values of the prediction models were generally consistent with the actual observed values ( Figures 3 , 4 ).

The predictive value of the constructed nomogram was compared with the 8th edition American Joint Committee on Cancer (AJCC) T-staging system in terms of clinical practicability. The results are shown in Figure 5 . In the training cohort, the C-index of the nomogram for OS and DFS was 0.825 and 0.699, respectively, which was significantly higher than that of the T-staging (OS: 0.590; DFS: 0.537). Similarly, in the validation cohort, the C-index of the nomogram for OS (0.914) and DFS (0.761) was also significantly higher than that of the T-staging (OS: 0.613; DFS: 0.621). Additionally, DCA suggested that the nomogram had better predictive power than the T-staging ( Figure 6 ). Overall, the nomogram models exhibited superior performance compared to the T-staging.

All patients were assigned to the high-risk group or the low-risk group based on their nomogram scores. In both the training and validation cohorts, patients in the low-risk group exhibited significantly higher survival rates and lower recurrence rates than those in the high-risk group ( Figure 7 ).

At present, the prognosis prediction model of malignant tumor patients has been widely used. In recent years, the nomogram related to HCC has been gradually developed, which has played a very active role in the diagnosis, treatment, and prognosis prediction of HCC. For example, Li et al. constructed a nomogram to predict the risk of liver nodule malignant transformation into HCC, which is helpful for the early diagnosis of HCC (15). Lin et al. established a simplified model to help guide decisions about prophylactic transarterial chemoembolization after hepatectomy for patients with HCC (16). Wang et al. developed a nomogram to predict recurrence in patients with early-stage HCC (17). However, the prognosis of HCC patients is closely related to age. In addition, the incidence of HCC is highest in the elderly population, whose postoperative prognostic factors are different from other age groups. For example, the underlying health conditions of elderly patients and the higher incidence of postoperative complications will affect the prognosis of patients. Currently, there is still a lack of predictive models for recurrence and survival after liver resection in elderly patients, so we constructed a nomogram to predict the prognosis of such elderly patients.

Unlike earlier prognostic indicators such as BCLC and Child–Pugh grade (18), the nomogram we developed calculates the survival rate and recurrence rate for each patient rather than simply categorizing patients into different risk groups. This approach reduces the impact of heterogeneity, thereby aiding clinicians in making individualized treatment decisions for elderly HCC patients and establishing a foundation for managing high-risk patients in clinical practice.

Our study collected the clinical information of elderly HCC patients over 65 years old who underwent liver resection in our hospital. After univariate analysis, LASSO regression analysis, and multivariate Cox analysis, it was found that ALB, cancer embolus, blood loss, viral hepatitis B, TB, microvascular invasion, overweight (BMI>24), and major resection were closely related to the OS of this group of people. Major resection, ALB, microvascular invasion, laparoscopic surgery, blood loss, TB, and pleural effusion are closely related to the DFS of this group. Among them, ALB, blood loss, TB, microvascular invasion, and major resection are related to both DFS and OS. Previous studies have shown that these factors are associated with the prognosis of HCC.

Overall, the clinical features we used to construct the nomogram after statistical screening were reported in previous studies. It is suggested that these factors may be related to the OS and DFS of patients. However, in terms of these clinical features, no studies have systematically analyzed them in elderly HCC patients undergoing hepatectomy and determined their impact on prognosis. We initially confirmed the validity of the nomogram for OS and DFS prediction in elderly HCC patients who underwent hepatectomy, and found that it exhibited superior predictive power compared to T-staging, as suggested by DCA. This nomogram can effectively identify patients at high risk of death and recurrence. Therefore, the model we developed is an intuitive clinical tool with good predictive performance, assisting physicians in making rational treatment decisions for elderly HCC patients.

However, this study still has some limitations. First of all, we constructed the nomogram based on the clinical data of West China Hospital of Sichuan University. This does not necessarily represent other countries and regions. Secondly, as a retrospective analysis, this study inevitably has information bias and selection bias, which may affect the conclusion to a certain extent. Third, the nomogram we constructed has only been internally verified, and more clinical data and multi-center studies are still needed for external verification to further prove the effectiveness of the nomogram.