We conducted a retrospective analysis of 230 HCC patients who received combination therapy with PD-1 inhibitors and TKIs at the Department of Biological Injury, Fifth Medical Center of PLA General Hospital between June 2020 and August 2024. Among these patients, we further selected those with baseline TBIL levels >3×ULN and PTA ≥40% for subsequent analysis (n=108). Hyperbilirubinemia is defined as bilirubin levels exceeding the ULN (ULN = 17.1 μmol/L) (15, 16). These patients were not recommended for treatment with ICIs, according to the prescribing information for PD-1 inhibitors and the NCCN clinical practice guidelines (12, 17, 18). These patients were refractory to standard interventions, including PTCD and BA therapy. Prior to initiating combination therapy with ICIs and TKIs, these patients were managed solely with conservative supportive care. Given the absence of alternative therapeutic options and after thorough multidisciplinary team evaluation, we proceeded with PD-1 plus TKIs combination therapy in this patient cohort. Comprehensive safety monitoring and efficacy assessments were implemented. The patient inclusion criteria were as follows: age ≥18 years; histologically/imaging confirmed advanced HCC or intrahepatic cholangiocarcinoma (ICC); patients who received at least one cycle of ICIs combined with TKIs; TBIL level at baseline >3× ULN (51 μmol/L); PTA at baseline >40%; and at least one measurable target lesion based on modified Response Evaluation Criteria in Solid Tumors (mRECIST) (19, 20). The exclusion criteria were as follows: combination of other primary tumors; severe coagulation disorders; combination of severe cardiac and renal insufficiency; and other diseases such as diabetes and hypertension that were under control. The study was approved by the Ethics Committee of The Fifth Medical Center of the Chinese PLA General Hospital (IRB No. KY-2023-7-46-1), with informed consent waived due to its retrospective nature.

The PD-1 inhibitors: sintilimab (Innovent Biologics, China), camrelizumab (Hengrui Pharma, China), and tislelizumab (BeiGene, China) were prescribed at a fixed dose of 200 mg every three weeks, while toripalimab (Top Alliance, China) was administered at 240 mg. The PD-1/CTLA-4 bispecific inhibitor, cadonilimab (Akesobio, China), were prescribed at 240 mg every three weeks. Sorafenib (Bayer Schering Pharma, Germany) was administered daily at 400 mg, lenvatinib (Eisai, Japan) at 8 to 12 mg/day according to body weight, and bevacizumab (Qilu Pharmaceutical, China) at 7.5 mg/kg every three weeks. The choice of agents was based on physician discretion, patient tolerance, and individual financial considerations of patients. Treatment options for patients who met the criteria for transcatheter arterial chemoembolization (TACE), PTCD or BA treatment were decided together by the patient and the physician before or after systemic combination therapies. Among the enrolled patients, 65 patients were identified with confirmed history of HBV infection and continued to receive NAs during the treatment to control viral replication. NAs including entecavir (ETV) (0.5 mg/day), tenofovir (TDF) (300 mg/day), and adefovir dipivoxil (ADV) (10 mg/day) were administered during immunotherapy.

Baseline radiological responses were documented using computed tomography (CT) or magnetic resonance imaging (MRI), with subsequent assessments performed every three months. Tumor responses were evaluated by mRECIST (19, 20). Complete response (CR) was defined as the disappearance of arterial enhancement in all target lesions based on the sum of baseline diameters; partial response (PR) was designated when the sum of target lesion diameters decreased by at least 30%, while an increase of at least 20% in the sum of diameters was defined as progressive disease (PD). Stable disease (SD) was characterized by no change in the sum of diameters between PR and PD. AEs were recorded according to the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0 (17, 21). Patients who experience AEs associated with ICIs or TKIs therapy may undergo dose reduction, treatment suspension, discontinuation, or a switch to alternative agents depending on the severity of toxicity. Immunosuppression was administered according to the severity of the AEs.

The cutoff date for follow-up was August 31, 2024, with data extracted from patient medical records. The measurements of the study were similar to those of a previous study (22). Baseline information comprised patient demographic details, such as the Child–Pugh stage, Barcelona Clinic Liver Cancer (BCLC) stage, type of combination therapy, TACE, PTCD and BA treatment. Data collection encompassed baseline, one-week, and three-month routine blood tests, hepatic functional assessments, lymphocyte subset characterization, along with quantification of C-reactive protein (CRP) and interleukin-6 (IL-6). Serum CRP levels and hepatic functional assessments were quantified employing the automated biochemical analyzer (Beckman Coulter AU5400, CA). Serum IL-6 levels were evaluated utilizing the Roche Cobas 8000 system (Roche Diagnostics GmbH, Germany). The lymphocyte subsets were measured from whole blood, and the following subsets were identified: total lymphocytes (CD45+), CD8+ T cells (CD45+CD3+CD8+), CD4+ T cells (CD45+CD3+CD4+), NK cells (CD45+CD3-CD16+CD56+), and B cells (CD45+CD3-CD19+). The cells were stained with the following human antibodies: anti-CD3-FITC, anti-CD45-PerCP, anti-CD8-PE, anti-CD4-APC, anti-CD16-PE, anti-CD56-PE, and anti-CD19-APC. All antibodies were sourced from BD Biosciences (Franklin Lakes, NJ). For the gating strategy, we first identified lymphocytes based on CD45 expression and light scatter properties. Subsequently, T cells were gated as CD3+ cells and further subdivided into CD4+ and CD8+ subsets. NK cells were identified as CD3-CD16+CD56+, and B cells as CD3-CD19+. Routine blood parameters were analyzed using the automated hematology analyzer (SYSMEX HN-2000 series, Japan).

All statistical analyses were conducted using the SPSS Statistics version 27.0 (IBM, USA). Baseline data and adverse events were summarized using descriptive statistics. Continuous variables were presented as mean ± standard deviation(SD) or median (interquartile ranges[IQR]), while categorical variables were presented as counts and percentages. Student t test and nonparametric test were used to compare continuous variables. Categorical variables were compared using Fisher’s exact test or the χ² test. The cutoff date for survival analysis was August 31, 2024. OS and progression-free survival (PFS) were assessed using Kaplan–Meier curves with comparisons facilitated by the log-rank test. Cox survival analysis (Backward-Wald) was used to analyze hazard ratios (HRs) and 95% confidence intervals (CIs), identifying independent risk factors associated with OS. Variables with P <0.1 in univariate analysis, as well as those deemed necessary for adjustment by the researchers, were included in the subsequent multivariate analysis. Data were visualized using GraphPad Prism version 9.4 (GraphPad Software, USA). Statistical significance was set at P <0.05.

In our cohort, 108 participants were included, comprising patients with HCC (n=75, 69.4%) and patients with ICC (n=33, 30.6%) ( Supplementary Figure S1 ). Among the patients receiving ICI treatment, 61 patients were treated with sintilimab, 28 patients with cadonilimab, and 19 patients with other PD-1 inhibitors. A total of 86 participants received combination therapy with lenvatinib, and 22 patients received sorafenib, bevacizumab or rigotinib in combination. Among the 38 patients who underwent PTCD treatment during the combination therapy, TBIL levels could not be effectively corrected. Notably, 79% of these patients received PTCD treatment prior to immunotherapy. A total of 71 patients had a history of chronic viral hepatitis (hepatitis B or C) and 65 of them received antiviral interventions. Twenty-five percent of patients underwent TACE within one week prior to immunotherapy. Patients with severe TACE-related AEs (e.g., fever, hemorrhage) were excluded. Among the 108 patients, the median baseline TBIL was 105.50 μmol/L, and the median baseline PTA was 68.60%. The median MELD score was 11.66. Further details are listed in Table 1 .

In this study, the administration of combination therapy demonstrated overall good tolerability. Among the patients, 51.9% (n=56) experienced at least one treatment-related adverse events (TRAEs) of any grade, and 15.7% of patients (n=17) reported grade 3 (G3) or grade 4 (G4) TRAEs. The most common TRAEs reported were fever (n=18, 16.7%), increased TBIL (n=12, 11.1%), rash (n=8, 7.4%), pruritus (n=8, 7.4%) and fatigue (n=7, 6.5%). The most frequent G3/4 TRAEs primarily manifested as increased TBIL and rashes. All patients who experienced severe adverse events (AEs) showed improvement following treatment with corticosteroids. The overview of AEs occurrences in the patient population is provided in Table 2 . During treatment, sintilimab was permanently discontinued in two patients due to AEs. One patient who experienced AEs suspended sintilimab. Subsequently, this patient was rechallenged with sintilimab after improvement in AEs and did not suffer G3/4 AEs. One patient discontinued camrelizumab due to cutaneous hemangioma formation. Three patients reduced the dose of TKIs due to AEs; two patients discontinued TKIs because of complications while five others switched to regorafenib due to tumor progression. Overall, the incidence and nature of AEs appeared to be generally consistent with those in our previous study cohort (22).

To evaluate the patient’s tolerance to the combination therapy, we assessed baseline characteristics and tracked early changes in key clinical parameters ( Figure 1A ). Compared with baseline levels, significant improvements in liver inflammation-related markers were observed after 12 weeks of treatment, including the levels of alanine aminotransferase (ALT) (P <0.001), Aspartate aminotransferase (AST) (P =0.008), gamma-glutamyltransferase (γGT) (P =0.003), and TBIL (P<0.001). γGT levels significantly decreased within one week (P <0.001). The MELD score did not show a significant increase during the treatment period (P =0.742), indicating that the combined therapy did not exacerbate liver burden, and overall liver function remained stable, with early improvements noted in cholestasis-related markers. It is noteworthy that serum albumin (ALB) levels significantly decreased at 12 weeks (P =0.021), potentially due to inadequate nutritional intake following patient discharge. Coagulation functions such as PTA and INR did not show significant differences compared to baseline levels after 12 weeks of treatment (P =0.197, P =0.240); however, an improving trend was observed at 12 weeks ( Figure 1B ). Furthermore, compared with the established baseline level, the level of IL-6 showed a significant increase at one week (P < 0.001), but there was no significant difference at 12 weeks relative to baseline level (P =0.938). The trend in level of CRP paralleled that of IL-6 ( Figure 1C ).

Tumor responses were assessed in all patients using the mRECIST criteria, as shown in Table 3 . Among the 67 patients who could be evaluated for tumor response, one patient achieved CR, seven patients achieved PR, 33 patients had SD, and 26 patients had PD. Notably, 10 patients were lost to follow-up during the study. The ORR was 11.9%, while the disease control rate (DCR) reached 61.2%. We analyzed patient survival with a median follow-up of 9.60 months (95% CI: 6.68–12.52 months), in which 53 patients died and ten patients were lost to follow-up. The mOS was 5.03 months (95% CI: 1.36–8.70 months) ( Figure 2A ). Excluding patients lost to follow-up, 60 patients experienced either tumor progression or death, with a median progression-free survival (median PFS) of 3.63 months (95% CI: 1.98–5.29 months) ( Figure 2B ).

The application of PD-1 inhibitors in cancer patients can reactivate functions of immune cells, particularly T cells, restoring their immune function to exert antitumor effects (23). Studies have demonstrated that the early response and proliferation of peripheral blood PD-1⁺ CD8⁺ T cells can predict the response to anti-PD-1 therapy in solid tumors (24). In this study, we also conducted an initial assessment of the impact of PD-1 therapy on immune activation in this specific population by analyzing changes in the distribution of lymphocyte subpopulations in patients with HCC combined with hyperbilirubinemia. E ( Figure 3A ). When patients were grouped according to tumor response and changes in lymphocyte subpopulation counts were evaluated among those with different treatment outcomes ( Figures 3B–G ), we observed that both the counts of T cells (P =0.047), CD4⁺ T (P =0.020), and CD8⁺ T (P =0.030) in patients with PD showed progressive decline in T-cell counts. In contrast, these counts remained stable during the 12 weeks of treatment in patients with DC ( Figures 3B, C, G ). The change in NK cell counts between the two groups mirrored the trend observed for T cells, though the difference did not reach statistical significance ( Figure 3E ). These findings suggest that lymphocyte dynamics during PD-1 therapy may correlate with treatment outcomes, warranting further investigation.

After 12 weeks of combined PD-1 inhibitors and TKIs treatment, we observed a significant improvement in TBIL levels. However, a slight increase in TBIL was noted within the first week (P=0.026) ( Figure 1A ). This transient rise may be attributed to the progressive worsening of liver function in some patients with compromised hepatic function. Therefore, we conducted a Cox regression analysis to identify prognostic risk factors. To minimize potential bias from the choice of different TKI treatments, we included only patients treated with lenvatinib. After adjusting for age, sex, and other relevant covariates, the multivariate Cox survival analysis revealed that MELD score >18 (P =0.016, HR=3.982 [95% CI: 1.292–12.273]) and decrease in TBIL level within the first week(P =0.001, HR=0.243 [95% CI: 0.102–0.575]) were significant risk factors impacting OS, as shown in Table 4 .

We stratified patients according to the risk factors to assess differences in survival outcomes. Patients who experienced a decrease in TBIL levels within the first week showed a prolonged mOS (not reached [NR] vs 3.3months [95% CI: 2.24–4.36 months], P =0.013) and mPFS (7.03 months [95% CI: 0.59–13.48 months] vs 2.77 months [95% CI: 1.60–3.93 months], P =0.010) compared to those of patients with TBIL levels consistently elevated during the same period ( Figures 4A, B ).In contrast, patients with a MELD score ≤18 did not show a significantly longer mOS compared to those with a MELD score >18(5.03 months vs 5.52 months, P =0.509), as well as no significant difference in mPFS (4.43 months vs 3.53 months, P =0.820) ( Supplementary Figures S2A, B ).

To evaluate the impact of different levels of TBIL on PD-1 treatment, we also included a subset of patients with TBIL levels between ULN and 3×ULN and compared the survival outcomes between the two groups. Baseline characteristics are shown in Supplementary Table S1 . We found that patients with TBIL levels > 3×ULN did not exhibit significant differences in OS or PFS compared to those with lower levels ( Supplementary Figures S2C, D ). The findings suggest that patients with hyperbilirubinemia for whom PD-1 is not recommended by guidelines or drug descriptions have a favorable safety profile when treated with PD-1 plus TKIs combination therapy. This particular group can achieve survival benefits similar to other patients recommended for treatment.

We also grouped the influencing factors to compare the differences in immune cell counts. However, no statistically significant difference was found between the two groups in terms of MELD score ( Supplementary Figure S3 ) or early decline in TBIL ( Supplementary Figure S4 ).