The data of 370 patients pathologically diagnosed with HBV-HCC who received curative hepatectomy as the first line of treatment were retrospectively collected at the Guangdong Provincial People’s Hospital from January 2013 to December 2019. The exclusion criteria included the following: 1) patients younger than 18 years; 2) patients without the negative surgical margin; 3) patients who underwent neoadjuvant downstage therapy; 4) patients with recurrent HCC; 5) patients with preoperative incomplete data on blood indices; and 6) patients with a follow-up period of less than 6 months. This study was approved by the ethics committee of Guangdong Provincial People’s Hospital (KYZ202132501) and performed following the Declaration of Helsinki.

Preoperative blood indices and clinicopathological data within 7 days before surgery were collected retrospectively from the electronic medical record system. Contrast-enhanced CT or contrast-enhanced MRI combined with serum alpha-fetoprotein (AFP) was performed during the first month of the first-year follow-up and every 3 months after that to detect early recurrence. Following this, the patients were recommended to undergo imaging examinations every 6 months for 1 year after surgery. Additionally, imaging examinations were performed if patients had chief complaints such as abdominal pain.

Moreover, a multidisciplinary treatment discussion was conducted for every patient before hepatectomy. As the commonly used adjuvant management, all patients were recommended to receive adjuvant TACE 1–2 months after surgery. The highly selective conventional TACE was used for this adjuvant management.

Recurrence-free survival (RFS) was defined as the time from curative hepatectomy to tumor recurrence at any site. The overall survival (OS) was defined as the time from the curative hepatectomy to death, due to any cause, or last contact.

Continuous variables were presented as the mean and SD, whereas categorical variables were expressed as frequency and percentage. Univariate Cox regression analysis was performed to estimate the indices associated with RFS, and stepwise multivariate Cox regression analysis was performed to build a blood index signature (BIS). The following formula was used to calculate the BIS: H0 * Exp[Ʃ(βi × xi)]. Furthermore, the univariate and multivariate Cox regression analyses were used to assess the independence of novel signature in evaluating RFS based on clinicopathological variables and construct a predictive nomogram. The area under the receiver operating characteristic (ROC) curve was used to evaluate the predictive performance. Calibration curves were used to compare the association between the actual outcomes and predicted probabilities. The best cutoff value for risk stratification was calculated using the X-tile software (Yale School of Medicine, USA). Additionally, the Kaplan–Meier survival curves with log-rang text and subgroup Cox regression analyses were plotted to compare differences in RFS or OS. Decision curve analysis (DCA) was used to evaluate the clinical applicability of the nomogram.

Statistical analyses in this study were performed using the R software version 4.0.5. Unless otherwise stipulated, a two-tailed p-value <0.05 was considered statistically significant.

The clinicopathological characteristics of the study’s cohort are summarized in Table 1 . The patients were predominantly male, with a solitary tumor, without capsule and microvascular invasions (MVIs), and had well-reserved liver function (Child–Pugh class A). The average age of the cohort is 54.22 years. A total of 141 (38.1%) patients underwent laparoscopic surgery, and 126 (34.1%) received adjuvant TACE. The average tumor size is 5.44 cm. A total of 164 (44.3%) patients with Ki67 ≥ 20% and 172 (46.5%) patients with grade III/IV tumor differentiation were observed. Notably, 171 (46.2%) patients showed negative AFP (AFP < 20 ng/ml) result.

Univariate Cox analyses were performed to filter blood indices associated with the RFS of HBV-HCC. Blood indices significantly associated with RFS were further analyzed using stepwise multivariate COX regression. Subsequent stepwise elimination of the blood indices resulted in APRI and SIRI as significant predictors for the RFS of HBV-HCC ( Table 2 ). The following formula was used to calculate the BIS: 0.447 * Exp(0.340552 * APRI + 0.447854 * SIRI).

Univariate Cox regression analysis was used to identify the predicted values of RFS in BIS and clinicopathological parameters. Furthermore, the association between RFS and Ki67, size, tumor number, AFP, MVI, and BIS score was analyzed using ( Figure 1A ) multivariate regression analyses, wherein Ki67, size, MVI, and BIS were found to be independent prognostic predictors of RFS ( Figure 1B ). Moreover, the BIS achieved the highest AUC value as compared with the other clinicopathological parameters ( Figure 1C ). Following this, BIS levels in different clinicopathological subgroups were assessed. The results indicated BIS levels were significantly higher in patients with AFP positive values, grade III/IV, MVI, and tumor size ≥5 cm (p < 0.001) ( Figures 2A–F ).

In multivariate Cox regression analysis, tumor number with p < 0.1 and hazard ratio (HR) = 1.416 (95% CI = 0.947−2.119) was screened for constructing the model owing to its greater clinical significance. A nomogram was constructed based on Ki67, tumor size, tumor number, MVI, and BIS ( Figure 3A ), with each clinicopathological characteristic corresponding to a specific point. The sum of the values was represented on the total points axis, whereas the probabilities of 0.5-, 1-, and 2-year RFS were represented on the corresponding axis.

The AUC values of ROC of the nomogram were 0.778, 0.752, and 0.706 in 0.5-, 1-, and 2-year RFS prediction, respectively ( Figure 3B ). The calibration plots for the probabilities of 0.5-, 1-, and 2-year RFS demonstrated positive concurrences between the nomogram predictions and actual observations ( Figures 3C–E ). The study cohort was divided into high and low recurrence risk groups according to an optimal cutoff value (28.7) determined using the X-tile software. As shown in Figures 4A, B , patients in the low recurrence risk group showed better RFS (p = 5.766e−13) and OS (p = 2.998e−15) than those in the high recurrence risk group. The nomogram showed a better net benefit than that of MVI and tumor size within a wide range of threshold probability ( Figure 4C ).

Cox regression analysis was used to determine the efficiency of adjuvant TACE in different subgroups. Patients in the low recurrence risk group, who received adjuvant TACE, showed no improvement in RFS (HR = 1.165, 95% CI = 0.786−1.727, p = 0.447) and OS (HR = 1.109, 95% CI = 0.532−2.310, p = 0.780). However, patients in high recurrence risk group showed improvement in OS (HR = 0.527, 95% CI = 0.281−0.991, p = 0.047) but not in RFS (HR = 0.862, 95% CI = 0.525−1.414, p = 0.556) ( Figure 4D ). Furthermore, the Kaplan–Meier analysis was performed to evaluate the RFS and OS of patients with or without adjuvant TACE in high recurrence and low recurrence risk groups. In the low recurrence risk group, patients who underwent adjuvant TACE showed no benefits regarding RFS (p = 0.4456) and OS (p = 0.7842) ( Figures 5A, B ). Although the patients in the high recurrence risk group who received adjuvant TACE showed no benefits regarding RFS (p = 0.5727), the OS (p = 0.04516) of those who underwent TACE was more beneficial than of those without adjuvant TACE ( Figures 5C, D ). Considering the short-term recurrence rate in the high recurrence risk group, the 3-month recurrence rate was significantly lower in patients who underwent adjuvant TACE (p = 0.02557) ( Figure 5E ). Although the 6-month RFS showed no significant difference, a better and beneficial RFS trend was observed in patients who underwent adjuvant TACE ( Figure 5F ).