Overall, a cohort of 115 uHCC patients underwent triple therapy from April 2021 to December 2022 at our institution were included in this retrospective study. All patients were deemed unresectable through multidisciplinary evaluation. The following inclusion criteria were used: (1) no previously received SBRT, immunotherapy or targeted therapy, (2) Barcelona Clinic Liver Cancer (BCLC) stage B or C, (3) Eastern Cooperative Oncology Group (ECOG) performance status score of 0 or 1, (4) Child-Pugh classification A and B7, and at least 1 measurable intrahepatic lesion. The exclusion criteria were without imaging evaluation or incomplete data on tumor markers, combination of PD-1 inhibitors less than 2 cycles and no elevation of both AFP and DCP at baseline. In accordance with ethical standards, this study strictly adhered to the principles stated in the Declaration of Helsinki. Ethical approval was obtained from the Ethics Committee of the PLA General Hospital’s Fifth Medical Centre. Informed consent was unnecessary for participation due to the retrospective nature.

After the implantation of 3-5 localization markers, patients underwent SBRT using a CyberKnife VSI image-guided robotic stereotactic radiosurgery system. Following CT simulation to identify the treatment site, an oncologist contoured the gross tumor volume (GTV) and delineated the organs at risk (OARS). The GTV included the tumor and portal vein tumor thrombus (PVTT) if it met the dose-volume limits of the critical organs. Otherwise, just the PVTT is described as the GTV (28). The planning target volume (PTV) extends the GTV by 3-5 mm and avoided the OARs. The prescribed doses were 45 to 55 Gy/5 to 10 fractions. Acceptable doses of OARs were derived from the American Association of Physicists in Medicine (AAPM) TG-101 report (29).

All patients were initiated ICIs in combination with targeted drugs within one week after the completion of the last radiotherapy session. The specific regimens included Sintilimab + Bevacizumab, Sintilimab + Lenvatinib, Tislelizumab + Lenvatinib, or Toripalimab + Lenvatinib. The prescribed dosages were as follows: Sintilimab at 200mg every three weeks, Tislelizumab at 200mg every three weeks, Toripalimab at 240mg every three weeks, and the recommended initial dose of lenvatinib for is 12 mg/day for individuals weighing more than 60 kg, and 8 mg/day for those weighing 60 kg or less. Bevacizumab was administered at 15 mg/kg of body weight every three weeks.

We systematically gathered and analyzed baseline clinical parameters and follow-up data for applicable individuals, encompassing age, gender, etiology, ALBI classification (30), BCLC stage, blood count, transaminase, AFP, DCP, the occurrence of macrovascular invasion and extrahepatic metastases. Baseline AFP and DCP levels were assessed within one week prior to the start of triple therapy. High AFP was defined >10 ng/ml and high DCP was defined >40 mAU/ml according to our institution’s reference range. To monitor AFP and DCP responses, data were obtained 6-8 weeks after the triple therapy. Radiological response was evaluated using Magnetic Resonance Imaging (MRI) per modified Response Evaluation Criteria in Solid Tumors (mRECIST) (26, 27), with tumor assessments conducted every 6-8 weeks by three expert radiologists. ORR was determined on the basis of optimal treatment response, with complete remission (CR) or partial response (PR) lasting≧1 month. Patients received triple therapy until the onset of intolerable adverse events or disease progression. PFS was defined as the duration between the initiation of treatment and the occurrence of disease progression, death, or the final follow-up. OS was the time between the commencement of therapy and death or the last follow-up.

Continuous data were analyzed using independent T-tests and reported as the median and interquartile spacing. Numbers and percentages were used to characterize the categorical data, and chi-square or Fisher exact tests were applied to compare the results. ROC curves were then plotted between the radiographic response and the magnitude of tumor marker decline at 6-8 weeks post-therapy. The biomaker cut-offs were chosen by identifying the points on the ROC curve that maximized the sum of sensitivity and specificity in differentiating patients by best confirmed response. To find acceptable objective response (CR+PR) predictors, binary logistic regression analysis was employed. The Kaplan-Meier approach produced survival curves, and the log-rank test was used to assess group differences. Cox regression models were used to identify PFS and OS predictors. In the multivariate analysis, factors with p < 0.1 in the univariate analysis were included, and p < 0.05 was deemed statistically significant. Software called SPSS and Graphpad Prism were used for all statistical analyses.

A total of 20 of the 115 patients were excluded due to no imaging evaluation (n=8) or missing tumor marker data (n=5), or no elevated baseline AFP and DCP levels (n=7) ( Figure 1 ). The baseline characteristics of the remaining individuals were shown in Table 1 . Patients were predominantly over 50 years of age (70, 73.7%) and male (88, 92.6%). Furthermore, a significant proportion of patients had a history of chronic hepatitis B infection (89, 93.7%). Vascular invasion was found in 75 (78.9%) patients and extrahepatic metastases were present in 18 (19.0%) patients. According to the BCLC staging, 9 patients (9.5%) were classified as stage B, and 86 patients (90.5%) were classified as stage C. In terms of Child-Pugh classification, 73 patients (76.8%) were classified as class A, while 22 patients (23.2%) fell into class B. Regarding the ALBI classification, 18 patients (19.0%) were assigned to class 1, and 77 patients (81.0%) to class 2. The median baseline AFP was 604.5 ng/ml, while the median baseline DCP was 1396.0 mAU/ml.

A total of 88 individuals exhibited elevated baseline AFP levels, 64 (72.7%) of them were older than 50, and 83 (94.3%) were infected with HBV. Of this subgroup, 70(79.5%) had vascular invasion, and 17 (19.3%) had extrahepatic metastases, and the median baseline AFP value was 948.5 ng/ml ( Table 1 ).

After undergoing triple therapy for 6–8 weeks, analysis based on the ROC curve revealed that a 51.5% decrease in AFP was the optimal threshold, corresponding to the maximum Youden index value of 0.414 ( Supplementary Table S1 ). Consequently, a decrease in AFP of 50% was established as the cut-off value for defining AFP response ( Supplementary Table S1 ). In this group, a total of 55 patients (62.5%) exhibited AFP response ( Table 1 ). As depicted in the waterfall diagram in Figure 2A , most of the patients who demonstrated AFP response also achieved objective response. As shown in Table 2 , the multivariate analysis confirmed that AFP response independently influenced objective response (OR 5.50, 95% CI 2.04-14.83; p=0.001).

The high AFP cohort had an overall response rate (ORR) of 62.5%, including 11 cases (12.5%) achieving CR and 44 cases (50.0%) achieving PR, respectively. In the AFP response subgroup, 9 cases (16.4%) achieved CR, 33 cases (60.0%) achieved PR, resulting in a favorable ORR of 76.4%. In contrast, the ORR of the AFP non-response subgroup was significantly lower at 39.4%, with a statistically significant difference (p=0.001) ( Supplementary Table S2 ).

A total of 91 individuals presented with elevated baseline DCP levels, of whom 66 (72.5%) were over 50 years of age, and the majority were infected with HBV (85, 93.4%). Among this cohort, vascular invasion was observed in 71 patients (78.0%), 17 patients (18.6%) experienced extrahepatic metastases, and the median baseline DCP value was 2386.0 mAU/ml ( Table 1 ).

According to ROC curve analysis, a 70% reduction in DCP levels after 6–8 weeks of triple therapy was identified as the optimal cut-off threshold ( Supplementary Table S1 ). In this cohort, a total of 42 patients (46.1%) achieved DCP response ( Table 1 ). Most patients with a DCP response also obtained objective response, as shown in Figure 2B . The multivariate analysis further demonstrated that DCP response independently influenced objective response (OR 7.99; 95% CI 2.82-22.60; p < 0.001), as shown in Table 3 .

The high DCP cohort achieved an ORR of 62.6%, with CR observed in 12 cases (13.2%) and PR in 45 cases (49.5%). In the DCP response subgroup, 10 cases (23.8%) achieved CR, and 26 cases (61.9%) achieved PR, resulting in an increased ORR of 85.7%. In contrast, the ORR of 42.9% in the AFP non-response subgroup was lower with a significant difference between the two groups (p=0.001) ( Supplementary Table S2 ).

The median PFS in high AFP cohort was 11.7 months (95% CI 9.0-14.4 months). Notably, the AFP response subgroup exhibited a longer median PFS compared to those without response (13.7 months vs. 6.2 months, p=0.013), as illustrated in Figure 3A . In the multivariate analysis, the absence of extrahepatic metastasis emerged as a favorable and meaningful factor influencing PFS (HR 0.40, 95% CI 0.21-0.75; p=0.004), and achieving AFP response was correlated with increased PFS (HR 0.36; 95% CI 0.20-0.62; p<0.001) ( Table 4 ).

The median PFS in the high DCP cohort was 11.2 months (95% CI 8.8-13.5 months), and the DCP response subgroup exhibited a longer median PFS than the DCP non-response subgroup (15.6 months vs. 9.3 months, p=0.001), as shown in Figure 3B . In the multivariate analysis, the absence of extrahepatic metastasis was a favorable factor influencing PFS (HR 0.51, 95% CI 0.29-0.93; p=0.027), and achieving DCP response was correlated with increased PFS (HR 0.44, 95% CI 0.26-0.74; p=0.002) ( Table 5 ).

Patients who achieved AFP or DCP response exhibited more favorable OS compared to the non-response subgroup (not reached vs. 21.9 months, p=0.012; not reached vs. 20.6 months, p=0.007, respectively) ( Figure 4 ). The multivariate analysis revealed that AFP or DCP response was associated with prolonged OS (HR 0.47, 95% CI 0.22-0.99; HR 0.35, 95% CI 0.14-0.86; p=0.047, p=0.022, respectively) ( Tables 4 , 5 ).