Tumor growth rate for prognostic stratification and treatment optimization in unresectable hepatocellular carcinoma treated with TACE combined with systemic therapy

Abstract

Background

Tumor growth rate (TGR) is a dynamic biomarker for evaluating therapeutic response and prognosis in unresectable hepatocellular carcinoma (uHCC) with triple therapy, yet its clinical utility and role in guiding treatment optimization require further validation.

Methods

This study included 68 uHCC patients receiving transarterial chemoembolization (TACE) with systemic combination therapy. Propensity score matching (PSM) was applied to balance baseline confounders. Cubic spline models explored nonlinear associations between TGR and survival risk and found the ideal cut-off points. Kaplan-Meier (KM) analysis compared overall survival (OS) and progression-free survival (PFS) between TGR subgroups, Cox multivariate regression evaluated the independent prognostic value of TGR.

Results

Cox analysis confirmed TGR0 as an independent prognostic factor for OS (hazard ratio (HR) 1.035, 95% confidence interval (CI): 1.013–1.057, p = 0.020) and PFS (HR 1.033, 95%CI: 1.014–1.053, p = 0.001). After stratified for TGR0/TGR1 and matched, the low-TGR0 group showed significantly longer median OS (not reached vs. 13.09 months, p = 0.002) and PFS (12.11 vs. 4.38 months, p = 0.006) than the high-TGR0 group. In the subgroup analysis, after propensity score matching(PSM), the low-TGR1 group demonstrated a more favorable prognosis compared with high-TGR1(mOS: not reached vs. 16.97 months, p = 0.024; mOS1: not reached vs. 11.78 months, p = 0.022).

Conclusions

Dynamic TGR monitoring identifies heterogeneous therapeutic responses, guiding timely personalized treatment adjustments and prognostic stratification in uHCC.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12876-025-04596-2.

Introduction

Liver cancer continues to be a global health challenge, with high incidence and mortality in GLOBALCAN 2022 [1]. As governmental promotion of hepatitis B vaccinations and advances [2] in preventive medicine, the incidence of liver cancer ranking dropped to 4th, however the cancer death is occupied the 2th place in the nationwide statistics on cancer of the National Cancer Center in China [3]. There are a wide variety of primary liver cancers, but hepatocellular carcinoma accounts for more than 80% [4]. Treatment of uHCC using systemic therapy starting from the phase III SHARP trial shown a survival benefit of Sorafenib compared to placebo (10.7 months versus 7.9 months; HR 0.69, p < 0.001) [5]. Several first-line clinical trials ended with a fatality across almost 10 years until REFLECT trial showed Lenvatinib was non-inferior versus Sorafenib [6]. Since the phase III IMbrave150 trial reported that Atezolizumab plus Bevacizumab improved OS with a median OS 19.2 months, which suggest the era of immune checkpoint inhibitors (ICIs) combined with antiangiogenic drug [7].

TACE is an effective method in the treatment of middle and advanced liver cancer. It is widely recognized as the first-line treatment for patients with having intermediate-stage HCC according to the Barcelona Clinic Liver Cancer (BCLC) staging system [8].TACE combined with programmed death-(ligand)1 (PD-[L]1) inhibitors and molecular targeted therapies (MTT) has shown considerable promise in the treatment of HCC. A nationwide, retrospective cohort study involving 826 HCC patients demonstrated that those receiving the combination therapy had significantly improved progression-free survival (PFS) and overall survival (OS) compared to those undergoing TACE alone [9]. The conclusion was additionally verified by a multicenter, randomized, double-blind placebo-controlled Phase III trial (EMERALD-1), which showed that the median PFS for the combination group was 15 months, while it was only 8.2 months for the monotherapy group [10]. It is evident that triple treatment is the promising trend of unresectable HCC treatment.

TGR has emerged as a significant prognostic factor across various malignancies, providing insights into tumor behavior and treatment efficacy [11–14]. TGR was used as the evaluation criterion of the hyper progressive disease in the treatment by the immune checkpoint inhibitors-based combination therapy [15]. In previous study, our team demonstrated that higher TGR related to a worse prognosis in patients with huge HCC receiving TACE as an initial treatment [16]. However, the role that the TGR plays in the triple treatment and timing of switch to second-line treatment of uHCC remains unclear. This study now frame the current study explicitly as a validation and extension of our prior findings into this new, clinically relevant treatment paradigm.

Patients and methods

Study population and study variables

Conducted at Zhongshan Hospital (Xiamen Branch), Fudan University, this single-center retrospective cohort study focused on patients with uHCC treated with TACE combined with systemic therapy between April 2020 and December 2022. Patients who were still alive were censored as of 31 December 2023. Individuals were eligible if they satisfied the following: (I) 18 years or older; (II) unresectable HCC diagnosis; (III) TACE + systemic combination therapy (anti-angiogenic agents in combination with immune checkpoint inhibitors) as the first-line therapy; (IV) The classification of liver function as Child-Pugh (CP) class A or B and Eastern Cooperative Oncology Group performance status (ECOG) of 0 to 1. They were excluded if they had: (I) undergone previous TACE or systemic therapy; (II) the interval between triple treatment was more than 1 month; (III) Immunotherapy is less than three cycles; (IV) loss to Follow-up.

All eligible patients were treated with TACE and systemic combination therapy. Anti-angiogenic agents including Lenvatinib, Sorafenib, Apatinib or Bevacizumab. Immune checkpoint inhibitors include Pembrolizumab, Atezolizumab, Camrelizumab, Sintilimab or Tislelizumab. The use of medicines in accordance with the manufacturers^, recommended instructions. Individuals possessing at least one measurable lesion in any location, treated with a minimum of three doses of immunotherapy, and had their disease evaluated following the cycle.

According to globally recognized criteria, the diagnosis of HCC was confirmed through pathological or clinical methods [17]. Clinical examinations and tumor status assessments were conducted for all patients at baseline and during response evaluations throughout the follow-up, utilizing contrast-enhanced imaging. Data on Demographic characteristics, tumor characteristics, and baseline laboratory examinations were gathered for analysis. The albumin–bilirubin (ALBI) grade was determined using the formula of previous study [18] and the Neutrophil-to-Lymphocyte Ratio (NLR) was calculated as the absolute neutrophil count divided by the absolute lymphocyte count.

Procedure for TACE

Experienced physicians on our team conducted all TACE procedures with traditional interventional radiology techniques. Following a successful Seldinger puncture in the arterial cruralis, a 4-French/5-French catheter was selected to conduct arteriography of the hepatic arteries or ectopic vessel. Subsequently, a coaxial microcatheter was used for superselective catheterization of the hepatic artery’s tumor-feeding branch. Once angiography confirmed the catheter’s placement, a chemotherapy emulsion made up of lipiodol (3–10 ml) and epirubicin (dose range of 10–30 mg) was infused into the vessels feeding the tumor. Further embolization was achieved using a gelatin sponge or microsphere. Repeat TACE sessions were considered based on radiographic assessment and were typically indicated for one of the following reasons: (1) Incomplete embolization: Residual or recurrent viable tumor (enhancement on arterial phase) on follow-up imaging performed after the previous TACE. (2) New lesions: The appearance of new intrahepatic lesions. (3) Persistent/increased AFP: In patients with elevated alpha-fetoprotein (AFP), a lack of significant decline or a subsequent rise was also considered.

Response evaluation and TGR assessment

Overall survival (OS) was the primary endpoint of this study, defined as the time from treatment initiation to death from any cause. Patients who were alive at the last follow-up were censored on that date. Progression-free survival (PFS) was defined as the time from the time of the initial treatment until radiologic progression or death from any cause, whichever occurred first. Patients without an event were censored at the date of their last adequate tumor assessment. Deaths occurring before documented radiologic progression were considered events for the PFS endpoint, and was specified as a secondary endpoint. All analyses of PFS and other endpoints beyond OS were considered exploratory. The time of overall survival 1 (OS1) was measured from the date of the secondary TACE until death from any cause. Patients who were still alive at the end of the study or lost to follow-up were censored on the date of their last known contact. Treatment response was assessed by radiologic evaluations performed within a predefined 1 to 3-month window after treatment initiation. Tumor measurements were conducted according to Response Evaluation Criteria in Solid Tumors (RECIST) v1.1 [19]by two independent, board-certified radiologists blinded to clinical outcomes. TGR was determined using a previously described formula [20]: TGR0 = 100 × (exp (TG) − 1); TG0 (tumor growth) = (3 × log (D2/D1))/time (months), where D1 represents the tumor size at the first date when contrast-enhanced CT or enhanced MRI was conducted before the initial treatment; D2 refers to the tumor size on the date when the first imaging evaluation after the combined treatment. TGR1 = 100 × (exp (TG1) − 1); TG1 (tumor growth) = (3 × log (D3/D1))/time1 (months); D3 indicates the tumor size recorded on date3 when the imaging assessment was done before the secondary TACE during regular systemic treatment; time (months) = (date 2 − date 1)/30.4; time1 (months) = (date 3 − date 1)/30.4; Tumor size (D1, D2 and D3) was calculated based on the longest diameters of the largest tumor nodule exclusively.

Statistical analysis

Continuous variables were analyzed using the t-test or Mann-Whitney U test based on Shapiro-Wilk normality results; categorical variables were compared using χ²/Fisher’s exact test. To explore the TGR linear effect on OS of liver cancer, the TGR was modeled as a continuous variable and fitted in a Cox proportional hazard model using cubic restricted splines with knots at the 5th, 35th, 65th, and 95th percentiles of TGR. According to the results of the restricted cubic spline analysis, TGRs were dichotomized using HR = 1.0 as the cutoff. The Kaplan-Meier curve and the log-rank test were used to compare OS and PFS by TGR groups. Multivariable Cox proportional hazards regression models were used to test the associations between TGR and the prognosis. Variables with a significance level of p < 0.1 in the univariate analysis along with potential confounders were incorporated into the multivariate Cox proportional hazards regression model. Hazard Ratios (HRs) and corresponding 95% confidence intervals (CIs) were calculated. Further, propensity score matching (1:1 nearest neighbor, caliper = 0.2 SD) was used to control for potential confounders between TGR groups. The propensity score for each patient is estimated by using a binary logistic regression model upon potential confounding variables (age, sex, BCLC stage, tumor length, subsequent therapy and ALBI). The post-hoc power calculation was used to the effect of the small sample size. All statistical tests were two-tailed with α = 0.05 defining significance. Statistical analyses were performed using the R-4.3.3 software and PASS 2025 (version 25.0.2).

Results

Baseline characteristics of the study population

The study included 68 eligible patients with uHCC who received triple therapy(Fig. 1). For the entire population, the median follow-up durations were 25.76 months. The majority of patients were categorized as having BCLC stage C (77.9%), Child–Pugh class A (97.1%), and hepatitis B virus (HBV) infection (88.2%) and macrovascular tumor thrombosis (MVT, 61.8%). The tumor length was 92.38 ± 49.57 mm, and 66.2% patients with multiple tumor nodules. Vascular invasion was identified in most patients (72.1%), but only 27.9% had extrahepatic metastases (Table 1). 27.9% patients accepted subsequent therapy including radiotherapy, surgery, chemotherapy or thermal ablation.

Fig. 1.

Flowchart of study design

Table 1.

Univariable and multivariable Cox regression to identify the risk factors of overall survival and progression free survival

		OS						PFS
		Univariate analysis			Multivariate analysis			Univariate analysis			Multivariate analysis
Characteristics	Total	HR	95%CI	p value	HR	95%CI	p value	HR	95%CI	p value	HR	95%CI	p value
gender
male	60	Ref						Ref
female	8	0.802	0.243–2.645	0.717				0.533	0.192–1.479	0.227
Age	56.21 ± 11.12	0.999	0.967–1.032	0.957				0.988	0.962–1.016	0.403
LCS
Yes	42	Ref						Ref			Ref
No	26	0.664	0.312–1.414	0.288				0.605	0.338–1.083	0.091	0.701	0.387–1.269	0.241
Chlid-pugh
A	66	Ref						Ref		0.240
B	2	3.258	0.764–13.890	0.110				2.364	0.562–9.944
BCLC
A/B	1/14	Ref						Ref
C	53	2.613	0.911–7.496	0.074	2.284	0.657–7.933	0.194	1.198	0.614–2.336	0.597
HBV
Yes	60	Ref						Ref
No	8	0.629	0.191–2.073					0.470	0.169–1.309	0.149
HCV
Yes	1	0.048	0.000–22510.313.000.313	0.649				0.048	0.000–117.360.000.360	0.445
No	67	Ref						Ref
HBP
Yes	13	Ref						Ref
No	55	0.692	0.296–1.620	0.396				0.826	0.413–1.652	0.589
DM
Yes	9	Ref						Ref
No	59	1.267	0.440–3.648	0.661				1.080	0.486–2.402	0.850
Drink
Yes	9	Ref						Ref
No	59	0.456	0.186–1.121	0.870				0.609	0.285–1.298	0.199
Tumor length(mm)	85.650(59.250,124.625)	1.007	1.000–1.015.000.015	0.060	1.005	0.996–1.015	0.233	0.999	0.993–1.005	0.809
VI
Yes	49	Ref						Ref
No	19	0.496	0.203–1.214	0.125				0.947	0.519–1.726	0.859
PVTT
Yes	42	Ref						Ref
No	26	0.613	0.288–1.305	0.204				0.877	0.501–1.537	0.647
Distant metastasis
Yes	19	Ref						Ref
No	49	0.591	0.281–1.245	0.167				0.687	0.380–1.241	0.213
NOTN
Single	23	Ref						Ref
Multiple	45	0.759	0.349–1.651	0.486				0.595	0.32–1.108	0.102
AFP(ng/ml)
<400	34	Ref						Ref
≥ 400	34	1.704	0.833–3.482	0.144				0.966	0.561–1.664	0.901
Subsequent therapy
Yes	19	Ref			Ref			Ref
No	49	2.254	0.921–5.514	0.075	1.480	0.531–4.127	0.453	1.322	0.723–2.416	0.364
TGR0	−4.052(−14.641,6.904	1.030	1.012–1.047	0.001	1.035	1.013–1.057	0.020	1.036	1.017–1.055	<0.001	1.033	1.014–1.053	0.001
PLT	222.55 ± 113.94	1.000	0.996–1.003	0.826				0.999	0.996–1.002	0.421
Monocyte	0.57 ± 0.22	2.303	0.447–11.864	0.319				1.678	0.396–7.103	0.482
ALBI	−2.86 ± 0.41	3.016	1.233–7.380	0.016	2.215	0.813–6.033	0.120	1.523	0.752–3.082	0.135
NLR	2.87 ± 1.44	1.176	0.967–1.430	0.105				1.067	0.894–1.274	0.473
TNF	13.51 ± 17.18	0.990	0.961–1.021	0.537				0.979	0.947–1.013	0.228
IL-6	11.59 ± 12.03	0.990	0.960–1.021	0.530				1.004	0.983–1.025	0.711
IL-8	54.66 ± 54.25	1.004	0.998–1.010	0.206				1.003	0.997–1.008	0.303

Bolded p values are statistically significant. Provide median (IQR) for markedly skewed

Abbreviations: CI Confidence interval, HR Hazard ratio, OS Overall survival, PFS Progression Free Survival, Ref Reference, LCS Liver cirrhosis, ECOG Eastern Cooperative Oncology Group, VI Vascularinvasion, PVTT Portalveintumorthrombosis, NOTN Number of tumor nodule, AFP Alpha-fetoprotein, ALBI Albumin–bilirubin, NLR Neutrophil-to-lymphocyteratio, TNF Tumor necrosis factor

Indicators related to overall survival and progression-free survival

As shown on Table 1, according to the univariate analysis results and the clinical relevance, BCLC stages(HR 2.613, 95% CI: 0.911–7.496, p = 0.074), tumor length(HR 1.007, 95%CI: 1.000–1.015.000.015, p = 0.060), without subsequent therapy treatment(HR 2.254, 95%CI: 0.921–5.514, p = 0.075),ALBI (HR 3.016, 95%CI: 1.233–7.380, p = 0.016), and TGR0(HR 1.03, 95%CI: 1.012–1.047, p = 0.001),were included in the multivariate analysis for OS. After multivariate analysis, TGR0(HR 1.035, 95%CI: 1.013–1.057, p = 0.02), was found to be an independent predictor of prognosis. Likewise, TGR0(HR 1.033, 95%CI: 1.014–1.053, p = 0.001), remained strongly associated with progression-free survival of the enrolled population.

Optimal threshold value for TGR0 and TGR1

The Cox proportional hazard model with restricted cubic splines showed a linear association between TGR0/TGR1 and the overall survival of patients who met the enrollment criteria. The linear relationship between TGR0/TGR1 and overall survival in HCC patients was shown by restricted cubic splines in the Cox proportional hazard model (Fig. 2, p for linear trend < 0.05). The TGR0 and TGR1 cut-off values were optimally set at − 8.4% per month and 3% per month respectively. Based on this optimal threshold, HCC patients were categorized into a high-TGR0 group (TGR0 ≥ − 8.4% per month) and a low-TGR0 group (TGR < − 8.4% per month). Similarly, the enrolled patients were arranged in two sets depending on TGR1(high-TGR1 group (TGR1 ≥ 3% per month) and a low-TGR1 group (TGR1 < 3% per month)).

Fig. 2.

Association between TGR0 a The change of tumor size at the date of first assessment)/TGR1 b The change of tumor size at the date of assessment before the secondary TACE during regular systemic treatment) and OS. TGR was modeled as a continuous variable and fitted in a Cox proportional hazard model using cubic restricted splines with knots at the 5th, 35th, 65th, and 95th percentiles of TGR; Gray represents the 95% confidence interval

Evaluation of PFS and OS of two groups based on TGR0

The evaluation of PFS and OS was conducted by stratifying patients into two groups based on TGR0. A group characterized by a high TGR0, exhibited a median PFS of 4.38 months and a median OS of 14.97 months. In contrast, the other group with a low TGR0, demonstrated significantly better outcomes, with a median PFS of 12.11 months (p = 0.006) and the median OS was not attained(p <0.001). The HR for PFS between the two groups was 2.179 (95%CI: 1.232–3.855, p = 0.007), while the HR for OS was 4.419 (95%CI:1.805–10.820, p = 0.001), indicating that TGR0 is a strong prognostic factor for both PFS and OS.

Following 1:1 PSM, the survival disadvantage of the high-TGR0 group persisted. The median PFS remained shorter (4.38 months vs.14.61 months; HR = 2.647, 95% CI: 1.295–5.410; p = 0.008). Similarly, the median OS difference was maintained (13.09 months vs. not-reached; HR 4.742, 95% CI: 1.662–13.532; p = 0.004), and the 1-year OS rate was 46.0% vs. 82.0%, the 2-year OS rate was 27.0% vs. 76.0% (Fig. 3; Table 2). These findings suggest that TGR0 can effectively stratify patients into distinct risk groups.

Fig. 3.

Comparison of OS(a and c) and PFS(b and d) between low-TGR0 and high-TGR0 groups before(a and b) and after(c and d) propensity score matching (PSM). Kaplan-Meier curves demonstrate significantly improved OS and PFS in the low-TGR0 group compared to the high-TGR0 cohort, both in the original (pre-PSM, n = 68) and matched populations (post-PSM, n = 44)

Table 2.

Long-Term survival outcomes in TGR0 groups

	mOS(month)	mPFS(month)	1-year survival	2-year survival
Unmatched Cohort
low-TGR0	NR	12.11	86.00%	81.00%
high-TGR0	14.97	4.38	50.00%	33.00%
Matched Cohort
low-TGR0	NR	14.61	82.00%	76.00%
high-TGR0	13.09	4.38	46.00%	27.00%

Abbreviations: NR No reach

Evaluation of OS and OS1 of two groups based on TGR1

When evaluating the OS and OS1 by stratifying patients into two other groups based on TGR1, patients with high TGR1 (n = 15) exhibited significantly inferior OS compared to the low-TGR1 group (n = 30) (Fig. 4). The median OS was reduced by approximately 52.37% in the high-TGR1 cohort (16.97 months vs. 35.63 months; HR 2.912, 95%CI: 1.119–7.575; p = 0.028). This trend remained similar in OS1 when a second TACE was performed. The high-TGR1 group showed a median OS1 of 11.78 months versus 32.83 months in the low-TGR1 group (HR = 2.711, 95% CI: 1.036–7.094; p = 0.042). Post-PSM, the OS and OS1 difference remained statistically significant (mOS: 16.97 months vs. not-reached, HR = 5.032, 95%CI: 1.063–23.815, p = 0.042; mOS1: 11.78 months vs. not-reached, HR = 5.126, 95%CI: 1.083–24.262, p = 0.039), with consistent direction and magnitude of hazard ratios.

Fig. 4.

Kaplan-Meier survival analysis comparing overall survival (OS) and OS1 between low and high TGR1 groups before and after PSM (a) OS curves in the unmatched cohort (low TGR, n = 30 vs. high TGR, n = 15). (b) OS1 curves in the unmatched cohort (c) OS curves in the PSM-matched cohort (1:1 matched pairs, n = 14 per group). (d) OS1 curves in the PSM-matched cohort

Before PSM, the 1-year OS rate and 2-year OS rate of the high-TGR1 group was 67% and 38%, significantly less than that of the low-TGR1 group (83% and 77%). Similarly, the 1-year OS1 rate and 2-year OS1 rate favored the low-TGR1 group (80% vs. 46% and 65% vs. 36%). Post-matching analysis demonstrated that the low-TGR1 group maintained a significantly higher 1-year OS/OS1 rate (92% vs. 71% and 90% vs. 49%), with persistent differences in 2-year OS/OS1 rate (81% vs. 41% and 75% vs. 39%) when compared with high-TGR1 group (Table 3). Kaplan-Meier curves corroborated the consistent OS/OS1 trends before and after matching (Fig. 4). These findings suggest that TGR1 is an early signal for required reintervention.

Table 3.

Long-Term survival outcomes in TGR1 groups

	mOS(month)	1-year survival	2-year survival	mOS1(month)	1-year survival	2-year survival
Unmatched Cohort
low-TGR1	35.63	83.00%	77.00%	32.83	80.00%	65.00%
high-TGR1	16.97	67.00%	38.00%	11.78	46.00%	36.00%
Matched Cohort
low-TGR1	NR	92.00%	81.00%	NR	90.00%	75.00%
high-TGR1	16.97	71.00%	41.00%	11.78	49.00%	39.00%

Abbreviations: NR No reach

Evaluation of OS and PFS based on TACE -related procedure and complications

The median number of TACE procedures for the entire cohort was 4.04 ± 2.40 and 93.33% patients occurrence of adverse reactions related to TACE (Supplementary Table 1). The analyses of these variables, which no statistically significant relate to overall survival and progression-free survival, further support the idea that TGR serves as an independent biomarker of tumor biology.

Data on TACE-related complications have been analyzed and were included in the Supplementary Table 2. The most common complications were nausea or vomiting (82.22%) and hypohepatia (77.78%). Most of the complications were classified as grades 1 to 2 (82.22%) and no statistically significant correlation with OS (p = 0.211) or PFS (p = 0.767).

Discussion

In recent years, with the widespread application of TACE combined with systemic therapy in the treatment of uHCC [10, 21], accurate prediction of treatment response and patient prognosis has become increasingly important. The GREPONET-2 study demonstrated that TGR measured three months of treatment initiation could effectively predict PFS, highlighting its potential as a tool for early intervention and improved patient management [22]. The ability of TGR to predict PFS and OS has also has been demonstrated in other studies, making it a valuable biomarker in clinical oncology [23–25]. In our research, the identification of TGR as an independent predictor of efficacy and prognosis in uHCC patients receiving triple therapy underscores its critical role in tumor biology and therapeutic response. Elevated tumor growth rate demonstrated a significant correlation with unfavorable clinical outcomes in the cohort study. Multivariate analysis revealed that TGR0 ≥−8.4% independently predicted reduced overall survival. When compared to the results from our 2022 publicaiton, the remarkable consistency of the optimal TGR cut-off values (approximately − 8.6% per month) across two distinct patient populations (uHCC vs. huge HCC) and initial treatment eras (triple therapy vs. TACE). This negligible difference is a key strength that reinforces TGR’s robustness as a reproducible biomarker rather than being a weakness. This biological parameter may serve as an independent prognostic indicator requiring further validation in multicenter trials.

Our results corroborate emerging evidence on the temporal relationship between neoplastic proliferation dynamics and therapeutic efficacy. The recent study has revealed that accelerated tumor growth patterns, as quantified by volumetric expansion metrics, may serve as critical prognostic indicators across diverse malignancies. In the context of transarterial chemoembolization for large hepatocellular carcinoma, elevated TGR emerged as an autonomous prognostic factor for reduced survival duration (HR 2.06, 95% CI: 1.23–3.43) [16]. This biological parameter demonstrated superior discriminative capacity compared to conventional staging systems, potentially enabling stratification of candidates for multimodal therapeutic approaches and intensive monitoring protocols. Parallel observations in thoracic oncology further validate the biomarker potential of growth kinetic assessments. A bicenter analysis of immune checkpoint blockade recipients with advanced non-small cell lung cancer revealed that rapid early-phase tumor proliferation correlated significantly with diminished therapeutic response rates (3.3% versus 20.9%; P = 0.040) and shortened median survival (HR 2.93, 95% CI: 1.47–5.83; P = 0.002) [26]. These findings collectively propose that serial quantification of proliferative kinetics could enhance precision medicine paradigms through dynamic treatment adaptation and early identification of suboptimal responders.

Biomarkers play a significant role in determining the optimal timing for TACE procedures. A study demonstrated that the post-treatment serum IL-6 level, rather than pretreatment or dynamic changes of IL-6, serves as a practical marker for predicting tumor response [27]. This finding suggests that monitoring IL-6 levels after TACE can help in discriminating between patients who will benefit from repeated TACE and those who may require alternative therapeutic strategies. Another study highlights the prognostic value of preoperative MRI-derived 3D quantitative tumor arterial burden (TAB) in patients with HCC undergoing TACE. The study found that a high TAB was a strong independent predictor of TACE refractoriness and PFS [28]. IL-6 and TAB may serve as the biomarker for optimizing the timing of repeated TACE treatments according to the capacity in predicting clinical outcome of therapy. However, both of them absence of direct evidence in prolong survival time after repeated TACE. Our longitudinal analysis revealed that TGR1 exhibited a significant inverse correlation with post-procedural survival duration following the second TACE. Patients exceeding a TGR1 threshold of ≥ 3% experienced markedly shorter post-TACE survival, underscoring its stratification efficacy for interventional scheduling. These findings posit TGR1 dynamics as a quantitative decision-making tool to refine iterative TACE timing in HCC management, warranting prospective validation of its precision therapeutic utility.

Beyond its prognostic value, TGR may serve as a dynamic biomarker to guide clinical decision-making (supplementary Fig. 1). Based on our findings, we propose the following practical considerations. For patients with a higher TGR0, more intensive monitoring could be warranted to detect rapid progression sooner. Furthermore, these patients should be considered for upfront multimodal therapy, including combination with systemic agents or, where feasible, evaluation for locoregional alternatives such as radiation therapy or surgical resection. For patients with a high TGR1 (after initial TACE but prior to the second planned procedure), the treatment strategy should be re-evaluated. A TGR1 ≥ 3% per month suggests that the current therapeutic approach is insufficient. In such cases, proceeding with the planned second TACE without modification is likely to yield suboptimal outcomes. Instead, clinicians should consider: (1) switching or intensifying systemic therapy regimens; (2) combining TACE with other locoregional modalities like ablation or radiotherapy; or (3) for clearly refractory disease, foregoing repeat TACE altogether in favor of alternative treatment pathways. These proposals are hypothesis-generating and require validation in prospective studies designed specifically to test TGR-driven treatment algorithms.

This study has several inherent limitations due to its retrospective design and restricted sample size. First, the small cohort may reduce statistical power, increasing the risk of Type II errors and limiting the generalizability of findings to broader populations. We have explicitly acknowledged the implications of this limited sample size for interpreting our OS results in the study limitations section. To address this point, we have performed a post-hoc power calculation. The comparison between Group 1 (low-TGR) and Group 2 (high-TGR) for time-to-event outcomes was assessed using a two-sided logrank test with a Type I error rate (α) of 0.05. The calculation, performed with PASS 2025 (version 25.0.2), followed the method of Freedman (1982) [29], which relies on the number of events under the proportional hazards assumption. The parameters assumed an exponential distribution for time-to-event, an accrual time of 24 months, a follow-up time of 12 months, and no loss to follow-up. Based on the observed median survival times and sample size, the study demonstrated sufficient statistical power for progression-free survival (PFS, 94.42%) and for OS1 (71.19%) when groups were stratified by TGR0 and TGR1, respectively. Notably, the limited statistical power for OS directly compromises the reliability of OS-related hazard ratios (HRs) and p-values. Small cohorts reduce the ability to detect true effect sizes and inflate the variability of HR estimates, as the sample size is insufficient to stabilize the proportional hazards assumption - even with the Freedman (1982) method accounting for event counts. The power was limited for the analyses of overall survival (OS), with values of 30.33% for TGR0 and 39.87% for TGR1. Second, retrospective data collection introduces potential selection bias and incomplete documentation, which could affect outcome accuracy. PSM was usded to eminimize selection bias. Furthermore, we included a cohort of patients treated with immune checkpoint inhibitors and angiogenesis inhibitor, there was several types of agents used may lead to efficacy variations across patient subgroups. Treatment heterogeneity within the study population introduces additional confounding. Patients received diverse therapeutic regimens based on clinical decision-making, which may distort HRs and p-values by masking or amplifying TGR’s true effect on survival. Although heterogeneity existed in the types of agents, it is noteworthy that the core mechanisms of targeted-immunotherapy combinations may exhibit convergence. As a complementary approach, we have performed Kaplan-Meier analyses for the different systemic therapeutic drugs. The association between systemic regimens and OS/PFS remained no statistically significant (p > 0.05) within this more homogeneous subset, strongly supporting that the prognostic value was no associated with regimen mix (supplementary Table 3). Finally, the single-center design and the predominance of HBV-related HCC in our cohort as factors that limit the generalizability of our findings to western populations with different etiologies. The external validation in diverse cohorts should be a necessary next step. Future studies should stratify analyses by specific drug classes or biomarkers to elucidate their roles in modulating treatment responses, and validate the clinical benefits of standardized regimens through larger, prospective multi-center randomized controlled trials, witch validate the prognostic value of TGR and generate more robust HR estimates with reliable p-values.

Conclusion

TGR serves as a dynamic and quantitative biomarker in HCC combination therapy, demonstrating significant predictive value for both PFS and OS. Early TGR assessment enables the identification of patients with differential therapeutic responses, guiding timely treatment optimization and personalized strategies. These findings highlight the clinical utility of TGR as a prognostic tool and underscore its potential to refine therapeutic decision-making in uHCC management. Further prospective studies are warranted to validate its role in heterogeneous patient cohorts and multimodal treatment settings.

Authors’ contributions

Guobin Chen : Data curation; Formal analysis; Investigation; Methodology; Visualization; Writing original draft.Meixia Wang: Data analysis; Methodology; Resources; Visualization.Xing Huang: Formal analysis; Methodology; Writing review & editing; Data curation; Zhenzhen Zhang: Methodology; Data curation; Yanfang Wu: Resources; Data curation; Xiaochun Chen: Methodology; Data curation; Xinkun Guo: Conceptualization; Visualization; Susu Zheng: Methodology; Data curation; Tanghui Zheng: Methodology; Data curation; Hong Chen: Investigation; Data curation; Jingfang Wu: Formal analysis; Data curation; Boheng Zhang: Conceptualization; Methodology; Supervision; Writing – review & editing.

Funding

This investigation received financial support from Natural Science Foundation of Xiamen, China(3502Z20227386)and the medical and health guidance project of Xiamen city(3502Z20244ZD1107).

Data availability

The datasets generated and analyzed during this retrospective study are not publicly available due to institutional privacy policies. Anonymized data may be made available upon reasonable request to the corresponding author, subject to ethics committee approval. All materials are described within the manuscript.

Declarations

Ethics approval and consent to participate

This research received approval from the Institutional Review Board at Zhongshan Hospital (Xiamen Branch), Fudan University, and adhered to the ethical standards set by the institutional research committee and the latest Declaration of Helsinki(Approval No.B2023-035). Since the study was retrospective, obtaining patient consent was not needed. De-identified data are available from the corresponding author upon reasonable request, subject to a data use agreement.

Competing interests

The authors declare no competing interests.