Introduction

Hepatocellular carcinoma (HCC) constitutes 80%−90% of primary liver cancers and is the third leading cause of cancer-related deaths worldwide1. However, HCC remains a challenging disease to treat, particularly in advanced stages where patients are often ineligible for curative surgery and therapeutic options are limited2. Immune checkpoint inhibitors (ICIs) combined with anti-angiogenic targeted therapy represent first-line treatment for advanced liver cancer (typically unresectable advanced HCC) and hold promise for conversion therapy, potentially offering long-term survival benefits for some patients with intermediate to advanced stages. Nevertheless, the clinical response rate to programmed death-1(PD-1) antibodies is only approximately 20%3. The objective response rate for HCC patients treated with the combination of atezolizumab and bevacizumab is 33.2%4, indicating that a substantial proportion of patients derive no benefit from immunotherapy and may even experience severe immune-related adverse events (irAEs).

Diminished immune function may increase the risk of tumorigenesis and progression5. Immunosenescence refers to the progressive decline of immune function, leading to reduced clearance of abnormal cells, promotion of tumor immune escape, thereby increasing the risk of cancer development and progression, and is associated with treatment resistance and recurrence6. Studies in melanoma and non-small cell lung cancer (NSCLC) have suggested that patients with high proportions of both CD8⁺ terminally differentiated effector memory T cells (TEMRA) and senescent CD8⁺ T cells CD8⁺ T cells may have a poorer prognosis7,8. In T-cell differentiation, the surface expression of CD28 and CD57 serves as a key indicator of immunosenescence. CD28 is a critical co-stimulatory molecule for T-cell activation, while CD57 is a recognized marker of terminal differentiation. The CD28⁻CD57⁺ phenotype on CD8⁺ T cells characterizes a senescent subset with reduced proliferative capacity and altered effector functions, which has been associated with poor outcomes in several cancers6.

This study analyzed changes in the levels of senescent CD8⁺ T cells in the peripheral blood of patients with newly diagnosed hepatocellular carcinoma (NDHCC) before initiation and after completing 3 to 4 cycles of therapy with atezolizumab plus bevacizumab, assessed their correlation with treatment response, and evaluated their predictive value for treatment efficacy.

Materials and methods

Study subjects

Thirty-two patients with newly diagnosed HCC who were initially treated with atezolizumab plus bevacizumab at our hospital between January 2022 and December 2024 were enrolled.

Inclusion Criteria: HCC patients scheduled for atezolizumab plus bevacizumab therapy; no prior surgery, intervention, immunotherapy, or targeted therapy; measurable lesions; survival expectancy > 6 months; complete pre- and post-treatment clinical data including contrast-enhanced computed tomography (CT) or magnetic resonance imaging (MRI). The assessment of survival expectancy (> 6 months) was determined by the treating physicians based on the patient’s overall clinical condition, liver function reserve (Child-Pugh score), and tumor burden at diagnosis.

Exclusion Criteria: Other malignancies; comorbidities (e.g., diabetes, autoimmune diseases); recent use (within 2 weeks) of immunosuppressants/immunomodulators; active infections (excluding patients with controlled chronic viral hepatitis), such as active bacterial or fungal infections; inflammation, hematological diseases, or severe organ dysfunction.

Efficacy evaluation

Treatment response was evaluated according to mRECIST criteria9 based on contrast-enhanced CT or MRI. The scans were reviewed by two experienced radiologists who were blinded to the flow cytometry results. Complete response (CR) was defined as the disappearance of any intratumoral arterial enhancement in all target lesions. Partial response (PR) was defined as at least a 30% decrease in the sum of the diameters of enhancing target lesions, taking as reference the baseline sum of the diameters. Progressive disease (PD) was defined as at least a 20% increase in the sum of the diameters of enhancing target lesions, relative to the smallest sum observed during the study, or the appearance of one or more new lesions. Stable disease (SD) was defined as cases that did not qualify for PR, CR, or PD. Patients achieving PR or better were classified as responders. Patients with SD, PD, or those who did not achieve PR were classified as non-responders.

Sample collection and analysis

Baseline samples were collected prior to the first treatment cycle, while follow-up samples were obtained after approximately 3–4 cycles, immediately before the infusion at the time of the first radiological response assessment. At each time point, 2 mL of fresh peripheral blood was drawn from each patient into EDTA-containing vacuum tubes for phenotypic characterization. Analysis was completed within 2 h of collection. The following monoclonal antibody cocktail was used to identify T lymphocyte surface phenotypes: 20 µL CD45-PerCP (25 µg/mL, Clone 2D1, Cat. No. 664934), 20 µL CD57-FITC (12.5 µg/mL, Clone HNK-1, Cat. No. 663497), 20 µL CD28-PE (25 µg/mL, Clone L293, Cat. No. 662797), 5 µL CD8-APC (50 µg/mL, Clone SK1, Cat. No. 663524), and 5 µL CD3-APC-H7 (100 µg/mL, Clone SK7, Cat. No. 663490; all antibodies from BD Biosciences, San Jose, CA). Whole blood was stained directly without prior peripheral blood mononuclear cell isolation. The antibody mixture was incubated with 50 µL of the sample at room temperature (20–25 °C) in the dark for 20 min. Subsequently, red blood cells were lysed using 1.5 mL of BD FACS Lysing Solution for 8 min, followed by centrifugation at 500 × g for 5 min. Cells were washed once with phosphate-buffered saline and immunophenotyping was performed using an 8-color multiparameter flow cytometer (FACS Canto II, BD Bioscience). During analysis, single-cell gating was performed using FSC-A vs. FSC-H to exclude doublets and aggregates. Subsequently, the lymphocyte population was gated, followed by gating on the CD3⁺CD8⁺ T-cell population. The proportion of CD28⁻CD57⁺ cells was analyzed within this CD3⁺CD8⁺ T-cell population and is presented as a percentage of CD3⁺CD8⁺ T cells. (Fig. 1). A viability dye was not included in the flow cytometry panel. However, to minimize the inclusion of dead cells and debris, we rigorously gated on the lymphocyte population based on forward and side scatter properties, and analysis was performed promptly (< 2 h) after staining to ensure high cell viability.

Fig. 1
figure 1

Representative flow cytometry gating strategy and analysis of peripheral blood. (A1-A4) Gating sequence for a patient with a high baseline proportion of CD28⁻CD57⁺CD8⁺ T cells. (B1-B4) Gating sequence for a patient with a low baseline proportion of CD28⁻CD57⁺CD8⁺ T cells. ① (A1, B1): Single cells were gated from all acquired events using FSC-A versus FSC-H to exclude doublets and aggregates. ② (A2, B2): Lymphocytes were identified and gated from the single-cell population based on CD45 expression and side scatter (SSC-A) properties. ③ (A3, B3): CD3⁺CD8⁺ T cells were gated from the lymphocyte population. ④ (A4, B4): The expression of CD28 and CD57 was analyzed on the CD3⁺CD8⁺ T-cell population (P3). The proportion of CD28⁻CD57⁺ cells (quadrant Q1-UL) is indicated and represents the senescent T-cell subset of interest. FSC, forward scatter; SSC, side scatter.

Full size image

Treatment regimen

Patients were intravenously administered 1200 mg of atezolizumab combined with 15 mg/kg of bevacizumab every three weeks. The treatment continued until disease progression, intolerable toxicity occurred, or the patient withdrew from the treatment. All finally enrolled patients completed one course of treatment and underwent the first scheduled efficacy assessment after 3 to 4 cycles of therapy. Crucially, no patient discontinued treatment due to adverse events prior to this first assessment. Since the focus of this biomarker study was not on toxicity, a detailed analysis of adverse events was not conducted.

Data collection

Baseline clinical data were collected, including gender, age, alcohol consumption history, number of tumors, maximum tumor diameter, history of hepatitis B, presence of cirrhosis, presence of portal vein tumor thrombus, Child-Pugh liver function classification, alpha-fetoprotein (AFP) level, disease stage (based on the Barcelona Clinic Liver Cancer staging system10, and presence of extrahepatic metastases.

Statistical analysis

The Shapiro-Wilk test was used to assess data normality. Normally distributed continuous data are presented as mean ± standard deviation (mean ± SD) and compared using independent samples t-tests. Non-normally distributed continuous data are presented as median (interquartile range) [M (Q1, Q3)] and compared using the Mann-Whitney U test. Categorical variables are expressed as frequency (percentage) and compared using the chi-square test (χ² test) or Fisher’s exact test. Receiver operating characteristic (ROC) curve analysis was employed to evaluate the predictive performance of the CD28⁻CD57⁺CD8⁺ T-cell proportion and to determine the optimal diagnostic cutoff value. Binary logistic regression analysis was used to identify factors influencing early treatment response. To assess the associations between the baseline CD28⁻CD57⁺CD8⁺ T-cell proportion and clinical characteristics, Spearman’s rank correlation analysis was used for continuous variables (age, maximum tumor diameter, and AFP level). The Mann-Whitney U test was employed to compare the CD28⁻CD57⁺CD8⁺ T-cell proportion between categorical variable groups (number of tumors [< 3 vs. ≥3] and BCLC stage [B vs. C]). All statistical analyses were two-sided, and a P value < 0.05 was considered statistically significant. The figures were drawn using GraphPad Prism 10 (GraphPad Software, Inc.).

Results

Patient characteristics

A total of 32 NDHCC patients treated with the “T + A” regimen were enrolled in this study. The cohort included 23 males and 9 females, with a mean age of 62.14 ± 11.60 years. The majority were male (71.9%) and aged ≥ 60 years old (62.5%). The vast majority of patients had hepatitis B virus infection (90.6%), among whom 37.5% had liver cirrhosis. Regarding tumor characteristics, 75.0% of patients had fewer than 3 tumors, and 68.8% of patients had a maximum tumor diameter ≥ 10 cm. According to the BCLC staging, 62.5% of patients were in stage C; all patients had Child-Pugh liver function of class A or B (each accounting for 50%). In addition, 68.8% of patients had an AFP level ≥ 400 ng/mL. Table 1. Following the first radiological evaluation conducted after 3 to 4 cycles of therapy according to mRECIST criteria, 14 patients (43.8%) were classified as responders (achieving PR or better), and 18 patients (56.2%) were classified as non-responders (SD or PD). A summary of treatment cycles and a detailed breakdown of objective responses are provided in Supplementary Table S1.

Table 1 Baseline demographic and disease characteristics of the study cohort (N = 32).
Full size table

Changes in CD28⁻CD57⁺CD8⁺ T-cell proportions following ‘T + A’ therapy

Before treatment, the median proportion of the CD28⁻CD57⁺ subset within CD3⁺CD8⁺ T cells in the entire cohort of HCC patients was 26.45% (17.68, 43.23). After treatment, it significantly decreased to 20.15% (13.68, 30.28), a difference that was statistically significant (p = 0.039).

Association between baseline CD28⁻CD57⁺CD8⁺ T-cell proportion and treatment response

Analysis of the association between baseline CD28⁻CD57⁺CD8⁺ T-cell levels and treatment response was conducted. The baseline proportion of CD28⁻CD57⁺CD8⁺ T cells was significantly lower in responders compared to non-responders (21.45% vs. 33.90%, p = 0.0269). Notably, the dynamics of this cell population after treatment differed between the groups. In the responder group, the cell proportion decreased from 21.45% to 17.15%, a change that was not statistically significant (p = 0.334). In contrast, the non-responder group showed a statistically significant reduction from 33.90% to 25.30% (p = 0.0283).

Comparison of baseline characteristics between responders and non-responders

The baseline clinicopathological characteristics of the two groups were compared. Baseline characteristics were comparable between responders and non-responders (Table 2). The mean ages of the two groups were similar (61.2 ± 10.8) years in the responder group vs. (62.9 ± 12.3) years in the non - responder group), and the gender distributions were similar (proportion of males: 78.6% vs. 66.7%). Regarding disease characteristics, there were no significant statistical differences between the two groups in the proportion of liver cirrhosis (42.9% vs. 33.3%), number of tumors (proportion of multiple tumors: 28.6% vs. 22.2%), maximum tumor diameter (proportion of ≥ 10 cm: 78.6% vs. 61.1%), extrahepatic metastasis (14.3% vs. 27.8%), and BCLC stage (proportion of stage C: 57.1% vs. 66.7%) (all p > 0.05). In addition, the distributions of baseline AFP levels in the two groups were also basically balanced (proportion of AFP ≥ 400 ng/mL: 64.3% vs. 77.8%, p = 0.308).

Table 2 Comparison of baseline characteristics between responder and Non-Responder Groups.
Full size table

Associations between baseline CD28⁻CD57⁺CD8⁺ T-cell proportion and clinical characteristics

The association between the baseline CD28⁻CD57⁺CD8⁺ T-cell proportion and key clinical variables was further investigated. No significant correlation was found with patient age (Spearman’s r = 0.187, p = 0.306). However, a statistically significant positive correlation was observed with the maximum tumor diameter (Spearman’s r = 0.430, p = 0.014). No significant correlations were identified with AFP level (Spearman’s r = −0.089, p = 0.626). Furthermore, the baseline proportion of CD28⁻CD57⁺CD8⁺ T cells did not differ significantly between groups stratified by the number of tumors (Mann-Whitney U = 92, p = 0.875) or BCLC stage (Mann-Whitney U = 107.5, p = 0.494). Table 3.

Table 3 Associations between baseline CD28⁻CD57⁺CD8⁺ T-cell proportion and clinical characteristics.
Full size table

Predictive value of baseline CD28⁻CD57⁺CD8⁺ T-cell proportion for initial treatment response in NDHCC patients

To evaluate the predictive value of the baseline CD28⁻CD57⁺CD8⁺ T-cell proportion for treatment response in NDHCC patients and determine its optimal cutoff value, ROC curve analysis was performed using treatment non-response as the state variable. The results showed that the area under the curve (AUC) for predicting treatment response was 0.730 (95% CI: 0.541–0.919, p = 0.028). The sensitivity was 66.70%, and the specificity was 85.70% at the optimal cutoff value of 29.05% (Fig. 2A). Based on this cutoff (29.05%), the 32 NDHCC patients were divided into two groups: Group A (≥ 29.05%) and Group B (< 29.05%). In Group A, 85.71% of patients were non-responders, whereas in Group B, only 33.33% were non-responders (Fig. 2B). This difference was statistically significant (p = 0.005).

Fig. 2
figure 2

Predictive value of CD28⁻CD57⁺CD8⁺ T cell proportion in early treatment response among NDHCC patients. (2 A: ROC curve of CD28⁻CD57⁺CD8⁺ T cell proportion for predicting early treatment response in NDHCC patients; 2B: Comparison of early treatment outcomes between NDHCC patient groups stratified by the threshold.).

Full size image

Factors influencing early treatment response based on logistic regression analysis

Binary logistic regression was performed to identify factors associated with early treatment response. The model included the following baseline variables: age, viral hepatitis, cirrhosis, significant alcohol history, AFP level, maximum tumor diameter, extrahepatic metastasis, and BCLC stage. The results indicated that the CD28⁻CD57⁺CD8⁺ T-cell proportion (p = 0.020) and BCLC stage (p = 0.034) were statistically significant independent factors influencing early treatment response. The maximum tumor diameter was also a significant predictor (p = 0.035), although the estimate was highly unstable. In contrast, viral hepatitis (p = 0.302) and significant alcohol history (p = 0.192) were not identified as significant predictors in this model. The complete results are presented in Table 4.

Table 4 Binary logistic regression analysis of factors influencing early treatment response in NDHCC Patients.
Full size table

Association between treatment cycle number and T-cell dynamics

Two analyses were conducted to assess potential confounding by the number of treatment cycles prior to initial response assessment on senescent T-cell changes. First, cycle numbers were comparable between responders and non-responders (median: 3.5 (3, 4) vs. 4 (3, 4); Mann-Whitney U = 112, p = 0.721). Next, no significant correlation was found between cycle number and the change in CD28⁻CD57⁺CD8⁺ T-cell levels across the cohort (Spearman’s r = 0.068, p = 0.711). Thus, treatment exposure difference is unlikely to account for the intergroup immunophenotypic disparity.

Discussion

ICIs, particularly those targeting the PD-1/programmed Death-Ligand 1(PD-L1) pathway, have revolutionized cancer treatment and are increasingly used across various stages of therapy, improving survival outcomes for some patients with solid tumors. However, standard immune checkpoint blockade demonstrates limited efficacy in most HCC patients11,12,13. Many patients fail to benefit from immunotherapy and may even experience severe immune-related adverse events. Immunosenescence is considered a key factor underlying this differential response14. Studies indicate that immunosenescence alters immune function and the tumor microenvironment, reducing the response to ICI treatment and increasing the risk of immune-related adverse events15. A retrospective analysis of 332 patients with nasopharyngeal carcinoma revealed that CD28⁻CD57⁺CD8⁺ T cell levels were significantly elevated compared to healthy controls, and long-term follow-up confirmed this subset as an independent predictor of poor radiotherapy prognosis6. Zhang et al.16, using a spontaneous mouse hepatocellular carcinoma model, demonstrated that reversing T cell senescence to expand the functional CD8⁺ T cell pool could enhance the efficacy of anti-PD-1 therapy. Furthermore, elevated levels of senescent CD8⁺ T cells have been observed in various cancers, including lung, breast, colorectal cancer, and melanoma, further supporting the close association between immunosenescence and tumor progression17,18,19,20. These findings not only underscore the critical role of immunosenescence in disease progression but also highlight the clinical importance of assessing immunosenescence status for predicting ICI response and improving the efficiency of cancer treatment. Consequently, peripheral blood-based immune function assessment is gaining increasing attention in clinical practice. Accurate evaluation and potential reversal of immunosenescence may help identify patients most likely to benefit from immunotherapy, thereby improving overall treatment outcomes. Based on this rationale, this study aimed to prospectively validate whether the baseline peripheral blood proportion of CD28⁻CD57⁺CD8⁺ T cells could serve as an effective biomarker for predicting early treatment response in patients with hepatocellular carcinoma receiving first-line “T + A” therapy.

This study found that the proportion of CD28⁻CD57⁺CD8⁺ T cells decreased in NDHCC patients after “T + A” treatment, suggesting that the therapy may partially reverse or alleviate systemic immunosenescence. More importantly, the baseline proportion of these cells was significantly higher in non-responders than in responders. These statistical differences support the concept that the senescent CD28⁻CD57⁺CD8⁺ T-cell phenotype is associated with poorer efficacy of the “T + A” regimen. This finding aligns with studies in rectal cancer, melanoma, and NSCLC8,20,21, which reported that high proportions of immunosenescent and immunoactivated CD8⁺ T cells correlate with negative outcomes such as cancer recurrence, progression, or death.

An important finding of this study is that the baseline level of senescent CD28⁻CD57⁺CD8⁺ T cells was not correlated with patient age. This provides evidence contrasting with the conventional view of immunosenescence as a solely age-related process and suggests that in hepatocellular carcinoma (HCC), the accumulation of these cells may be an important driver by the tumor itself—a phenomenon termed “tumor-induced immunosenescence”22. The significant positive correlation observed in this study between the proportion of senescent T cells and the maximum tumor diameter strongly supports this concept. A larger tumor burden likely fosters a more immunosuppressive and senescence-inducing microenvironment, potentially through mechanisms such as sustained antigen exposure23 and the secretion of pro-inflammatory cytokines24.

The biological and clinical implications of CD57⁺ CD8⁺ T cells appear to be profoundly context-dependent. For instance, in acute myocarditis, a specialized population of CD57⁺CD8⁺ T cells with potent cytotoxic and migratory capabilities has been identified, and their differentiation is driven by Interleukin-18 signaling24. This demonstrates that in certain inflammatory contexts, CD57⁺ CD8⁺ T cells can represent a highly active effector population. However, within the oncological setting, the accumulation of senescent CD8⁺ T cells, often marked by the CD28⁻CD57⁺ phenotype, is increasingly recognized as a hallmark of tumor-induced immune dysfunction and a significant barrier to successful immunotherapy22,25. This concept is strongly supported by clinical studies across various malignancies. For example, in advanced NSCLC, a defined “senescent immune phenotype” based on CD28⁻CD57⁺KLRG1⁺ cells was an independent predictor of inferior outcomes following PD-1/PD-L1 inhibitor therapy8. The convergence of our findings in HCC with those in NSCLC solidifies the role of the CD28⁻CD57⁺CD8⁺ T-cell subset as a cross-canceral marker of a T-cell state committed to therapeutic resistance. Additionally, the lack of association with other clinical features, such as AFP level or BCLC stage, further underscores the potential of the CD28⁻CD57⁺CD8⁺ T-cell proportion as a unique and independent biomarker.

To evaluate the predictive performance of this marker, we constructed an ROC curve. The results showed that the AUC for the baseline CD28⁻CD57⁺CD8⁺ T-cell proportion in predicting treatment non-response was 0.730 (95% CI: 0.541–0.919). The optimal cutoff value determined by the Youden Index was 29.05%. At this threshold, the specificity was as high as 85.70%, meaning that when a patient’s baseline level exceeds this value, the probability of being correctly identified as a non-responder is high. The sensitivity was 66.70%, indicating a moderate ability to identify potential non-responders. The high specificity is particularly valuable in clinical practice. It could help identify patients unlikely to respond to the ‘T + A’ regimen before treatment initiation. This would spare them unnecessary costs and risks of immune-related adverse events, facilitating more precise, personalized treatment decisions.

Binary logistic regression analysis further confirmed that the baseline CD28⁻CD57⁺CD8⁺ T-cell proportion is an independent factor influencing early treatment response (OR = 1.135, 95% CI: 1.020–1.263, p = 0.020). An OR greater than 1 indicates that this proportion is a risk factor; for each unit increase in the cell proportion, the risk of treatment non-response increases by 13.5%. This finding suggests that the predictive value of this senescent T-cell proportion was more stable than that of some traditional tumor burden indicators in our model, which is consistent with the findings of Ferrara et al.8 regarding NSCLC. It is worth noting that BCLC stage (p = 0.034) and maximum tumor diameter (p = 0.035) also retained statistical significance in the comprehensive model. However, their odds ratios exhibited extremely wide confidence intervals (e.g., OR = 83.923 for BCLC stage; OR = 0.000 for max tumor diameter), likely reflecting instability due to the limited sample size and potential collinearity. Therefore, these particular point estimates should be interpreted with extreme caution. This analysis suggests that the baseline CD28⁻CD57⁺CD8⁺ T-cell proportion, as a quantifiable immune micro-level indicator, may possess independent predictive value for early treatment failure in “T + A” therapy, and appeared to demonstrate greater statistical stability than some traditional macro-level tumor burden indicators in our cohort. This conclusion, however, awaits verification in larger-scale studies.

The potential biological mechanism likely lies in the multiple functional defects of CD28⁻CD57⁺CD8⁺ T cells as terminally differentiated senescent cells. Given that the efficacy of atezolizumab—which targets PD-L1 on tumor and/or immune cells—is contingent upon a pre-existing, functionally competent T-cell population, the predominance of a senescent CD8⁺ T-cell subset may fundamentally undermine this prerequisite. On one hand, they undergo metabolic reprogramming, featuring mitochondrial dysfunction, reactive oxygen species accumulation, and decreased proteasome activity, directly weakening their cytotoxicity and proliferative capacity26. On the other hand, the senescence-associated secretory phenotype of these cells alters, secreting large amounts of pro-inflammatory factors (e.g., tumor necrosis factor-α, interferon-γ) and matrix-remodeling enzymes, shaping an immunosuppressive tumor microenvironment that promotes tumor progression and immune escape26. Concurrently, reduced Interleukin-2 production by senescent T cells may facilitate the stability and functional maintenance of Treg cells, further suppressing anti-tumor immunity8, collectively leading to ICI treatment failure.

Notably, the objective response rate observed in our study (43.8%) was numerically higher than the 33.2% reported in the IMbrave1504. This discrepancy could be attributed to our relatively small sample size, patient selection differences (all were newly diagnosed and treatment-naive in our study), or the single-center nature of the study. Furthermore, our study assessed early response after 3 to 4 cycles of therapy, which might capture initial responders who could later progress. Despite this variation in the overall response rate, our central finding regarding the predictive value of baseline CD28⁻CD57⁺CD8⁺ T-cell proportion remained consistent and statistically significant. Furthermore, we confirmed that the number of treatment cycles received before the first assessment was comparable between responders and non-responders and was not correlated with the observed immunophenotypic changes, supporting the notion that the baseline CD28⁻CD57⁺CD8⁺ T-cell proportion itself may be a predictive biomarker, independent of initial treatment exposure intensity.

The pursuit of reliable biomarkers to predict outcomes to “T + A” therapy in HCC remains a critical and unmet clinical need. Several candidate biomarkers have been investigated based on their established roles in tumor immunity. Tumor mutational burden, which reflects tumor neoantigen load, is an established pan-cancer predictor of response to certain immunotherapies27. Its predictive utility in hepatocellular carcinoma (HCC), however, is constrained by generally low baseline levels and an inconsistent correlation with clinical outcomes28,29,30. Similarly, PD-L1 expression has a strong rationale as the direct target of atezolizumab. However, its predictive value in HCC is confounded by tumor heterogeneity, dynamic expression, and lack of standardized assays, leading to conflicting results across trials31,32. Microsatellite instability-high status, resulting from deficient mismatch repair, is a potent biomarker for immunotherapy in multiple solid tumors due to its high immunogenicity33. However, it is exceedingly rare in HCC34, drastically limiting its applicability in this cancer type. The density of CD8+ tumor-infiltrating lymphocytes directly quantifies a pre-existing anti-tumor immune response at the tumor site and is a favorable prognostic factor35. Nonetheless, its assessment requires invasive tissue biopsies, rendering it unsuitable for dynamic monitoring and widespread clinical application. A study investigating gut microbiota dynamics in HCC patients during anti-PD-1 treatment suggested that the gut microbiota can modulate the efficacy of anti-PD-1 immunotherapy to some extent, providing a foundation for further research36. However, clinical translation is challenged by the system’s complexity and the absence of standardized, actionable protocols. Although AFP is the most widely used serological marker in HCC and its on-treatment decline correlates with improved outcomes, it primarily serves as an early response indicator and does not inform pre-therapeutic decisions37.

The collective limitations of tissue-based, complex, or static serological biomarkers have spurred interest in readily accessible and dynamic alternatives. In this context, peripheral blood-based immune profiling presents a compelling alternative, offering advantages of easy accessibility, non-invasiveness, and suitability for repeated assessment. Against this backdrop, we explored the role of peripheral blood immune subsets and identified that the baseline proportion of CD28⁻CD57⁺CD8⁺ T cells is associated with early non-response. This immune metric represents a candidate that leverages the practical advantages of liquid biopsy, potentially offering an immune-centric metric for pre-treatment stratification.

This study has several limitations. Its single-center design and small sample size limit statistical power, increase the risk of Type II errors, and require validation in larger, multi-center cohorts. The absence of an independent validation cohort and mature long-term survival data necessitates cautious interpretation of our findings. Furthermore, the observational nature of this study and the lack of functional assays prevent us from establishing a causal link between the senescent phenotype and treatment resistance. Methodologically, the flow cytometry analysis did not include a viability dye, and the results are based on relative cell proportion rather than absolute counts. Future studies addressing these limitations are warranted to confirm the clinical utility of this biomarker.

Conclusion

In conclusion, our findings suggest that the baseline peripheral blood CD28⁻CD57⁺CD8⁺ T-cell proportion is a promising biomarker for predicting early response to “T + A” combination therapy in HCC. This indicator is non-invasive and readily measurable, which could aid in identifying patients at higher risk of initial treatment resistance. These results support the need for further validation of this biomarker to assess its potential for guiding individualized treatment strategies.