BPG is committed to discovery and dissemination of knowledge
Systematic Reviews Open Access
Copyright: ©Author(s) 2026. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution-NonCommercial (CC BY-NC 4.0) license. No commercial re-use. See permissions. Published by Baishideng Publishing Group Inc.
World J Gastrointest Pharmacol Ther. Sep 5, 2026; 17(3): 120377
Published online Sep 5, 2026. doi: 10.4292/wjgpt.120377
Survival in unresectable hepatocellular carcinoma: Comparison of four treatments based on patients reconstructed from Kaplan–Meier curves of randomized trials
Vera Damuzzo, Department of Pharmacy, Vittorio Veneto Hospital, AULSS2 Marca Trevigiana, Vittorio Veneto 31029, Italy
Stefano Vecchia, Department of Pharmacy, AUSL Piacenza Guglielmo da Saliceto Hospital, Piacenza 29121, Italy
Lorenzo Gasperoni, Department of Pharmaceutical, USL Toscana Centro, Prato 59100, Italy
Luna Del Bono, Department of Pharmacy, University of Pisa, Pisa 56126, Italy
Andrea Ossato, UOC Territorial Pharmaceutical Service, Azienda ULSS 8 Berica, Vicenza 36100, Italy
Elena Orlandi, Department of Oncology-Haematology, Piacenza General Hospital, Piacenza 29191, Italy
Andrea Messori, Osservatorio Innovazione Section, HTA Regione Toscana, Firenze 50139, Italy
ORCID number: Vera Damuzzo (0000-0002-3685-6789); Stefano Vecchia (0000-0003-0578-0870); Lorenzo Gasperoni (0000-0001-6427-0809); Luna Del Bono (0009-0007-8094-1677); Andrea Ossato (0000-0001-8984-4733); Elena Orlandi (0000-0002-2559-7558); Andrea Messori (0000-0002-5829-107X).
Co-first authors: Vera Damuzzo and Stefano Vecchia.
Author contributions: Damuzzo V and Vecchia S contributed to conceptualization, data curation, formal analysis, investigation, methodology, software, supervision, validation, writing-original draft, writing-review and editing, both authors have made crucial and indispensable contributions towards the completion of the project and thus qualified as the co-first authors of the paper; Gasperoni L, Del Bono L, Ossato A, Orlandi E, Messori A contributed to conceptualization, data curation, formal analysis, investigation, methodology, supervision, validation, writing-original draft, writing-review and editing.
AI contribution statement: AI tools (specifically ChatGPT) were used solely for linguistic refinement and formatting assistance. No AI tool was involved in the generation of research data, interpretation of results, or formulation of conclusions. All AI-generated outputs were critically reviewed and revised by the authors.
Conflict-of-interest statement: The authors declare no conflict of interests.
PRISMA 2009 Checklist statement: The authors have read the PRISMA 2009 Checklist, and the manuscript was prepared and revised according to the PRISMA 2009 Checklist.
Corresponding author: Andrea Messori, PharmD, Osservatorio Innovazione Section, HTA Regione Toscana, Via Alderotti 26/N, Firenze 50139, Italy. andrea.messori@osservatorioinnovazione.net
Received: February 27, 2026
Revised: April 30, 2026
Accepted: May 22, 2026
Published online: September 5, 2026
Processing time: 188 Days and 15.2 Hours

Abstract
BACKGROUND

The therapeutic landscape for unresectable hepatocellular carcinoma (HCC) has evolved with the advent of combinations of immune checkpoint inhibitors (ICI) and tyrosine kinase inhibitors (TKI). However, the absence of direct comparisons in clinical trials makes it challenging to determine their relative efficacy.

AIM

To explore the comparison between ICI and TKI in combination therapy and determine the relative efficacy.

METHODS

We conducted a PubMed database search from inception to January 2026, selecting all first-line regimens supported by a randomized controlled trial (RCT). Overall survival (OS) was the endpoint. A series of indirect comparisons was performed across the selected regimens. To conduct our analysis, individual patient data (IPD) were reconstructed from Kaplan-Meier curves. Survival was assessed by the Cox univariate model. Hazard ratio (HR) and restricted mean survival time (RMST) were estimated. Heterogeneity between the sorafenib and lenvatinib control arms was assessed in two separate analyses.

RESULTS

Based on our PubMed search, eight RCTs were included. Most combinations compared with sorafenib significantly improved OS, with camrelizumab plus rivoceranib showing the most favourable HR. This was followed by atezolizumab plus bevacizumab, tremelimumab plus durvalumab, toripalimab plus bevacizumab and anlotinib plus penpulimab. Cabozantinib plus atezolizumab did not demonstrate a clear benefit. Camrelizumab plus rivoceranib was superior to tremelimumab plus durvalumab (HR = 0.78; 95%CI: 0.61-0.99) and anlotinib plus penpulimab (HR = 0.73; 95%CI: 0.57-0.93). Compared with lenvatinib, nivolumab plus ipilimumab and pembrolizumab plus lenvatinib yielded modest improvements, with no significant difference between the two. RMST analyses suggested that nivolumab plus ipilimumab, lenvatinib and pembrolizumab plus lenvatinib produced similar results to those of camrelizumab plus rivoceranib. The choice of comparator meaningfully shapes the estimated treatment effect.

CONCLUSION

Our study compared the efficacy of the main first-line treatments recommended for unresectable HCC. While we were able to rank the magnitude of efficacy across these treatments, the intrinsic limitations of our methods (primarily the reconstruction of IPD and the indirect nature of the treatment comparisons) should be borne in mind.

Key Words: Hepatocellular carcinoma; Immune checkpoint inhibitors; Tyrosine kinase inhibitors; Indirect comparison; Overall survival

Core Tip: In the absence of head-to-head trials for first-line unresectable hepatocellular carcinoma, we reconstructed individual patient data from eight phase III trials to indirectly compare immune checkpoint inhibitor and tyrosine kinase inhibitor combinations. Most regimens outperformed sorafenib, with camrelizumab plus rivoceranib emerging as the most effective option, demonstrating superiority over tremelimumab plus durvalumab and anlotinib plus penpulimab. Restricted mean survival time analyses revealed that nivolumab plus ipilimumab, lenvatinib, and pembrolizumab plus lenvatinib achieved results similar to those of camrelizumab plus rivoceranib. These findings provide a practical framework for treatment positioning and underscore the critical influence of comparator on estimated efficacy.



INTRODUCTION

Sorafenib has long been the standard of care for patients with unresectable hepatocellular carcinoma (HCC)[1]. However, the therapeutic armamentarium for HCC has substantially expanded because of the development of immune checkpoint inhibitors (ICIs) and multitarget tyrosine kinase inhibitors (TKIs), which are often used in combination regimens[2-10]. The first therapeutic advantage was demonstrated for lenvatinib, which demonstrated noninferiority to sorafenib in terms of overall survival (OS) and showed advantages in progression-free survival (PFS) and objective response rate[2]. The IMbrave150 trial established atezolizumab + bevacizumab as a new reference standard superseding sorafenib and demonstrating superior OS and PFS together with a favorable toxicity profile[3]. The HIMALAYA[4] and CheckMate 9DW[5] trials evaluated dual immune checkpoint blockade with tremelimumab + durvalumab administered according to the STRIDE regimen and ipilimumab + nivolumab, respectively. These combinations showed promising OS outcomes and notably prolonged response durations. Additional phase III trials have explored combinations of TKIs/antibodies targeting VEGF or multiple growth factors involved in angiogenesis with anti–PD-1/PD-L1 antibodies, with notable examples including cabozantinib + atezolizumab in COSMIC-312, camrelizumab + rivoceranib in CARES-310, toripalimab + bevacizumab in HEPATORCH, and lenvatinib + pembrolizumab in LEAP-002[6-9]. Additional evidence was provided by the APOLLO trial, in which case penpulimab (a novel high-affinity ICI) was combined with anlotinib (a multitarget TKI)[10].

In this rapidly evolving therapeutic landscape, sorafenib has been progressively superseded by several therapeutic options as the first-line standard of care. This development has triggered the evolution of treatment standards and the choice of control arms in pivotal clinical trials, with agents such as lenvatinib being increasingly used as reference comparators[1]. The abovementioned studies underscore the central role of ICI and TKI combinations in first-line HCC treatment and highlight the absence of direct head-to-head comparisons among the leading regimens and the consequent uncertainty regarding their relative therapeutic value.

To close this evidence gap, we herein apply the IPDfromKM algorithm, an artificial intelligence-assisted method that reconstructs individual patient data (IPD) from published Kaplan-Meier (KM) curves, extracting information from time-to-event curves to create a database representing individual patients (including follow-up time and event status) and thereby enabling indirect comparisons in the absence of head-to-head trials[11]. By combining this reconstruction strategy with the careful assessment of between-trial heterogeneity and consistency assumptions, our analysis aims to position the main combinatorial regimens for first-line HCC treatment within a coherent comparative framework and generate clinically interpretable estimates of their relative efficacy.

MATERIALS AND METHODS

We systematically searched PubMed to identify randomized controlled trials (RCTs) pertinent to this analysis, with the final search conducted on January 28, 2026. The search strategy was based on the following combination of terms: [(“hepatocellular carcinoma” OR HCC OR “hepatocellular cancer”) AND (“first line” OR “first-line” OR unresectable OR advanced) AND (lenvatinib OR sorafenib)]. Study selection followed PRISMA recommendations (Figure 1)[12].

Figure 1
Figure 1 PRISMA flowchart of the process of trial selection. RCT: Randomized clinical trial; HCC: Hepatocellular carcinoma; ICI: Immune checkpoint inhibitor.

Each eligible RCT was phase III, was performed in patients with previously untreated unresectable HCC, reported OS, and presented outcomes with KM survival curves. When multiple reports from the same study were identified, the most up-to-date and comprehensive publication was included to prevent data duplication.

Eight trials providing evidence for first-line combination regimens included in indirect comparisons were identified (Table 1 and Supplementary Table 1)[3-10]. As these trials were based on long-term follow-ups, OS data were analyzed using a validated approach based on the abovementioned IPDfromKM method[11], which enables the simulation of comparative treatment analyses even when original patient-level data are unavailable. Herein, these comparative analyses involved indirect cross-treatment comparisons based on KM curves generated from reconstructed patient data.

Table 1 Summary of the main clinical characteristics of patients included in the analysis, n (%).
Ref.
Treatments under comparison
Regions
Patient number
Number of events OS
HR (reported in the original trial)
Geographical region (Asia)
Etiology (HBV-HCV positivity)
Cheng et al[3], 2022, IMbrave150Atezolizumab + bevacizumab vs sorafenib17 countries and regions336; 165179; 990.66 (95%CI: 0.52-0.85)133 (57); 68 (41)236 (70); 112 (68)
Sangro et al[4], 2024, HIMALAYATremelimumab + durvalumab vs sorafenibAsia (except Japan), 83%393; 389291; 3160.78 (95%CI: 0.67-0.92)41 (39.8); 23 (35.9)63 (61.1); 42 (65.6)
Yau et al[6], 2024, COSMIC-312Cabozantinib + atezolizumab vs sorafenibAsia, 29%; other, 71%432; 217262; 1210.98 (95%CI: 0.78-1.24)127 (29); 63 (29)263 (60); 131 (60)
Qin et al[7], 2025, CARES-310Camrelizumab + rivoceranib vs sorafenibAsia, 83%272; 271111; 1510.64 (95%CI: 0.49-0.80)225 (83); 224 (83)230 (84); 226 (84)
Shi et al[8], 2025, HEPATORCHToripalimab + bevacizumab vs sorafenibChina, 96%162; 164104; 1260.76 (95%CI: 0.58-0.99)NR293 (90); 15 (5)
Zhou et al[10], 2025, APOLLOAnlotinib + penpulimab vs sorafenibChina, 100%433; 216217; 1220.69 (95%CI: 0.55-0.87)433 (100); 216 (100)381 (88); 188 (87)
Llovet et al[9], 2023, LEAP-002Lenvatinib + pembrolizumab vs lenvatinibAsia (excluding Japan), 31%; Japan and Western regions, 69%395; 399252; 2820.84 (95%CI: 0.71-1.00)172 (44); 173 (43)247 (63); 237 (59)
Yau et al[5], 2025, CheckMate 9DW1Nivolumab + ipilimumab vs sorafenib/lenvatinibEurope or North America, 43%; Asia, 42%335; 333194; 2280.79 (95%CI: 0.65-0.96)140 (42); 152 (46)211 (63); 214 (65)

In brief, IPD were derived from KM survival curves presented in both the experimental and control groups of the included RCTs. Curve digitization was accomplished through WebPlotDigitizer (version 4.7, retrieved January 29, 2026; distance = 20, Δx = Δy = 15). The resulting coordinate pairs, combined with participant enrolment figures and event counts, were processed using the IPDfromKM software package (version 1.2.3.0; most recent update March 22, 2022; retrieved January 29, 2026). This process produced a synthetic dataset containing time-to-event information (measured from study entry to the final observation point) and patient status classifications (surviving/censored or deceased/progressed). The above procedure was used to generate synthetic individual-level data for each trial arm. Digitization and data reconstruction were independently executed by two separate investigators, and the findings were subsequently cross-verified to reduce human errors.

Following the generation of these individual-level datasets, we conducted indirect treatment comparisons between experimental combination regimens and conventional therapy, applying statistical methodologies identical to those utilized in investigations involving actual patient populations. To validate our method, we performed heterogeneity analysis, comparing the control arms of the included RCTs with one another. Heterogeneity was assessed using the likelihood ratio and Wald’s tests, along with the corresponding p-values. The statistics on the reconstructed patients used the same tests as studies based on real patients, namely the univariate Cox model with hazard ratios (HRs) and 95%CIs. Restricted mean survival times (RMSTs) were calculated using the R-platform (version 4.3.2). KM curves were truncated at 28 months for RMST analysis, which corresponds to the minimum follow-up time available across the included RCTs.

Relevant RCTs were divided into two subgroups, namely those with sorafenib and lenvatinib as comparators. The CheckMate 9DW trial was the only exception, as both lenvatinib and sorafenib could be administered to the control group. However, as these trials generated two separate network geometries with no in-between connection, we performed two separate IPDfromKM analyses based on sorafenib or lenvatinib as a comparator.

RESULTS

Preliminary analysis aimed to assess the comparability of the included studies revealed a high degree of heterogeneity among TKI controls (likelihood ratio test statistic = 31.49 on seven degrees of freedom, P = 5 × 10−5) (Supplementary Figure 1). This difference was due to lenvatinib being superior to sorafenib (HR: 0.77, 95%CI: 0.69-0.86), i.e., an indirect comparison of all the studies planned for analysis could not be performed because of the lack of a common comparator. Consequently, we presented the results of the indirect comparison of RCT efficacy by dividing studies into two groups, namely those with (1) Sorafenib; and (2) Lenvatinib as common comparators. The CheckMate 9DW trial involved both TKIs as comparators; however, as 75% of the patients in the control group received lenvatinib, this study was included in the second group. Given these premises, the heterogeneity within the two groups remained within the specifications necessary to ensure comparability (Supplementary Figure 2). The first group included the CARES-310, HIMALAYA, IMbrave150, COSMIC-312, APOLLO, and HEPATORCH trials. Heterogeneity within this group was minimal (likelihood ratio test statistic = 8.24 on five degrees of freedom, P = 0.1). The second group included the LEAP-002 and CheckMate 9DW trials, and the heterogeneity of the control groups remained low (likelihood ratio test statistic = 0.01 on one degree of freedom, P = 0.9) despite the inclusion of patients treated with sorafenib (15% for CheckMate 9DW).

In our main analysis, the OS benefit of each of the eight combination regimens was compared with that of a standard TKI (sorafenib or lenvatinib, with all respective control groups pooled together) and, more importantly, with those of the other combination regimens.

The KM curves for OS in the first group are compared in Figure 2. Five combination regimens superior to sorafenib were identified, ranked in order of decreasing efficacy as camrelizumab + rivoceranib (HR: 0.61, 95%CI: 0.50-0.75), atezolizumab + bevacizumab (HR: 0.75, 95%CI: 0.64-0.89), tremelimumab + durvalumab (HR: 0.78, 95%CI: 0.68-0.90), toripalimab + bevacizumab (HR: 0.78, 95%CI: 0.63-0.97), and anlotinib + penpulimab (HR: 0.84, 95%CI: 0.73-0.96). However, this was not the case for atezolizumab + cabozantinib (HR: 0.90, 95%CI: 0.78-1.03). These findings are consistent with those reported in the original trials. The median survival advantage ranged from 22.2 months (95%CI: 20.4-27.2) for camrelizumab + rivoceranib to 16.3 months (95%CI: 15.5-19.8) for anlotinib + penpulimab, exceeding the median OS of sorafenib controls (14.4 months, 95%CI: 13.6-15.7).

Figure 2
Figure 2 Comparison of overall survival benefit in studies using sorafenib as control arm. In red, the 6 sorafenib control arms pooled together; in gold, camrelizumab + rivoceranib[7]; in light green, tremelimumab + durvalumab[4]; in light blue, cabozantinib + atezolizumab[6]; in dark green, atezolizumab + bevacizumab[3]; in purple, toripalimab + bevacizumab[8]; in pink, anlotinib + penpulimab[10]. Endpoint: Overall survival; time in months.

In addition, we compared the efficacies of the abovementioned five combination regimens with each other, revealing that camrelizumab + rivoceranib was superior to tremelimumab + durvalumab (HR: 0.78, 95%CI: 0.61-0.99) and anlotinib + penpulimab (HR: 0.73, 95%CI: 0.57-0.93), with no significant difference observed between the other combinations.

In the second group, the approximation made because 15% of control patients in the CheckMate 9DW trial received sorafenib[5] was thought to have a negligible effect. The analysis of this second group (Figure 3) showed that pembrolizumab + lenvatinib achieved a median survival of 22.8 months with a survival advantage over lenvatinib at limits of statistical significance (HR: 0.84; 95%CI: 0.71-0.98); nivolumab + ipilimumab achieved a median survival of 25.5 months with HR of 0.79 (95%CI: 0.67-0.94). The head-to-head indirect comparison of the two treatment regimens (pembrolizumab + lenvatinib vs nivolumab + ipilimumab) showed no significant difference in efficacy (HR: 1.05, 95%CI: 0.83-1.33).

Figure 3
Figure 3 Comparison of overall survival benefit in studies using lenvatinib as control arm. In orange, the two lenvatinib control arms pooled together; in green, pembrolizumab + lenvatinib[9]; in blue, nivolumab + ipilimumab[5]. Time in months.

When OS was estimated using the RMST method (truncation time = 27 months), camrelizumab + rivoceranib presented an RMST of 19.64 months (95%CI: 18.42-20.87), which was 3.8 months longer than that of the sorafenib control. A more modest advantage was observed compared with the other combinations, except for toripalimab + bevacizumab, which had a notably shorter RMST of 16.88 months (95%CI: 15.93-17.82). Conversely, nivolumab + ipilimumab, lenvatinib, and pembrolizumab + lenvatinib had RMSTs similar to that of camrelizumab + rivoceranib. These results are detailed in Table 2.

Table 2 Overall survival estimates of first line treatments for advanced hepatocellular carcinoma compared to sorafenib and to camrelizumab plus rivoceranib are reported as restricted mean survival time (expressed in months) with 95%CI in the overall population at 28 months of follow-up.
Arm
RMST (months)
Lower 95%CI
Upper 95%CI
Difference in RMST compared to sorafenib/lenvatinib (months)
Difference in RMST compared to camrelizumab + rivoceranib (months)
Sorafenib (pooled)15.8715.3316.400-3.78
Camrelizumab + rivoceranib19.6518.4220.873.780
Tremelimumab + durvalumab17.1816.1718.201.32-2.46
Atezolizumab + bevacizumab18.2217.1119.322.35-1.42
Toripalimab + bevacizumab16.8815.9317.821.01-2.77
Anlotinib + penpulimab17.9416.3719.512.07-1.71
Lenvatinib (pooled)18.8618.2119.583.01-0.77
Pembrolizumab + lenvatinib19.6318.7320.543.77-0.01
Nivolumab + ipilimumab19.1418.0320.243.27-0.51
DISCUSSION

Our indirect comparison of first-line systemic therapies for unresectable HCC demonstrates that most ICI + TKI combination regimens are significantly superior to sorafenib in terms of OS, with camrelizumab + rivoceranib consistently ranking as the most efficaciouvs HR and RMST analyses. The survival hierarchy observed in the sorafenib-based group aligns with the findings of individual trials, lending credibility to the reconstructed IPD approach. Although camrelizumab + rivoceranib demonstrated significant superiority over tremelimumab + durvalumab and anlotinib + penpulimab in indirect comparisons, differences among the remaining regimens were nonsignificant, suggesting broadly comparable efficacy.

A key methodological limitation of our analysis is the lack of a single consistent comparator across all included RCTs. Although TKIs were consistently used as controls, they were not equivalent. Our preliminary heterogeneity analysis demonstrated a significant difference between the two comparators, reflecting the established survival advantage of lenvatinib over sorafenib[2]. This finding implies a violation of the fundamental assumptions underlying indirect treatment comparisons, namely the principles of transitivity and consistency, which require relative treatment effects to be estimated against a common and comparable reference (the so-called anchored comparisons). Consequently, performing a single unified indirect comparison across all trials would have been methodologically inappropriate and prone to bias.

To mitigate this limitation, we conducted analysis in two comparison networks, categorized according to the TKI comparator (sorafenib or lenvatinib) used in the original trials. This approach ensured internal consistency within each network and allowed indirect comparisons to be conducted only among studies sharing a sufficiently comparable reference treatment. The CheckMate 9DW trial, which included both TKIs in the control arm, was assigned to the lenvatinib network because most patients (85%) received lenvatinib. Within each network, residual heterogeneity remained low, supporting the validity of this stratified analytical strategy.

Within this framework, OS results expressed as HRs should be interpreted separately for each comparator-defined group, as the apparent magnitudes and temporal patterns of treatment effects are inherently influenced by the efficacy of the reference therapy. Conversely, OS estimation using RMST is less dependent on the treatment effect of the control arm, although cross-treatment comparisons can only be qualitative when this indicator is used.

Our analysis of OS based on reconstructed IPD yielded the following main results. Most combination regimens demonstrated an OS benefit compared with TKI monotherapy, particularly vs sorafenib. An important exception was atezolizumab + cabozantinib, which did not show any OS advantage over sorafenib in the COSMIC-312 trial[6]. This finding may reflect the discordance between PFS and OS observed in this trial, which is possibly related to its dual primary endpoint design and the early disease control effect of cabozantinib. In addition, subgroup analyses suggested heterogeneous treatment effects across patient populations, which may have diluted any OS benefit in the intention-to-treat population. Pembrolizumab + lenvatinib also failed to meet its co-primary endpoints of superiority over lenvatinib in the LEAP-002 trial. These findings are consistent with the literature, supporting an overall advantage of combination strategies over TKI monotherapy, albeit with nonuniform outcomes across different regimens. In our indirect comparison, camrelizumab + rivoceranib was associated with a greater HR advantage than tremelimumab + durvalumab and anlotinib + penpulimab and showed a numerical RMST advantage over atezolizumab + bevacizumab and toripalimab + bevacizumab.

In the qualitative comparison of the two study groups, combination regimens using sorafenib as the control arm generally showed HRs of 0.6-0.8, with OS benefits clearly separated from the corresponding control curves. As expected, in the second group with lenvatinib as a more effective comparator, the benefit was more limited, as observed in the LEAP-002 trial, or emerged later, as in the CheckMate 9DW trial-a pattern consistent with the delayed treatment effect typically observed with immunotherapy, which may contribute to the observed HR advantage over time. The main advantage of indirect comparisons using the IPDfromKM method over classic network meta-analysis is that the former leverages reconstructed IPD, thereby accounting for the follow-up duration, whereas the latter relies on simple binary comparisons derived from HR against a common comparator. Although the IPDfromKM method allows for cross-treatment comparisons when information from real patients is unavailable, it has certain limitations. In particular, multivariate analyses cannot be performed unless separate KM curves are reported in the published articles for each covariate to be evaluated. Therefore, subgroup analyses such as hepatitis B virus (HBV) vs non-HBV or Asian vs. non-Asian populations were not conducted because separate KM curves for these subgroups were not reported in the original RCTs. Finally, given the univariate (or unadjusted) nature of the HRs estimated by the IPDfromKM method, our results should be interpreted as descriptive rather than causal.

In terms of RMST, camrelizumab + rivoceranib achieved OS values comparable with those of immunotherapy combinations and lenvatinib monotherapy. As the CARES-310 trial was not as mature as other comparable RCTs, the minimum common truncation time adopted in this analysis was 27-28 months. Although RMST enables the qualitative comparisons of survival benefits within a predefined time window, the survival advantages of different approaches may be more fully realized beyond the selected time horizon. Indeed, although applying a common truncation time is methodologically necessary for temporal comparability across trials, it limits the full capture of long-term survival benefits, particularly for immunotherapy-based strategies such as the STRIDE regimen, which exhibits the late separation of OS curves and durable advantages beyond disease progression[13]. Moreover, the clinical effects of immunotherapy may persist across subsequent lines of systemic therapy, as reflected by the prolonged time to first subsequent therapy or death and preserved OS even after the introduction of additional treatments[14]. These long-term dynamics may therefore be only partially represented in truncated RMST analyses.

Several considerations regarding population heterogeneity across the included trials should be acknowledged. In particular, the CARES-310 trial enrolled a markedly Asia- and HBV-enriched population, with 83% of patients recruited in Asia and nearly three quarters presenting HBV-related HCC. In contrast, HIMALAYA is a global study with a substantially lower proportion of HBV-related disease in the overall population despite the availability of a dedicated Asian subgroup analysis. The APOLLO trial was conducted exclusively in China, reflecting a population with a distinct geographic and etiological background. However, most other pivotal trials included in this analysis are global studies with a substantial proportion of Asian patients, typically 30%-60% (e.g., IMbrave150, HIMALAYA, CheckMate 9DW, and LEAP-002). Therefore, the overall dataset reflects predominantly mixed populations rather than a clear contrast between Asian and non-Asian cohorts, although CARES-310 represents a more pronounced enrichment. These between-trial differences may represent unmeasured sources of heterogeneity and potential effect modification in indirect comparisons, particularly for long-term OS outcomes. This aspect is especially relevant when the common comparator is sorafenib, a drug that is now largely de-emphasized in first-line clinical practice and associated with a high toxicity burden. Importantly, definitive evidence indicating that Asian or HBV-positive patients derive a systematically greater relative benefit from immunotherapy compared with TKIs is lacking[15]; rather, HBV-related disease has been associated with prognostic differences across studies, which may variably influence survival outcomes depending on the clinical context[16]. Therefore, etiological and geographic heterogeneity should be interpreted as a source of baseline prognostic imbalance rather than as a validated predictive factor for differential treatment benefit. Finally, new experimental treatments are being developed for HCC[16-19], although none of them were administered to patients enrolled in the included trials.

CONCLUSION

Our indirect comparison of first-line treatments for unresectable HCC suggests that most immunotherapy + TKI combination regimens provide an OS advantage over TKI monotherapy, particularly that based on sorafenib. By contrast, clinically relevant heterogeneity exists across individual strategies. When lenvatinib is used as the control arm, the benefit magnitude and timing are more modest or delayed, which highlights the impact of comparator efficacy on the interpretation of treatment effects. RMST-based analyses indicate that average survival gains within a fixed time horizon may be similar across different regimens; however, short- to mid-term benefits should not be interpreted as evidence of long-term clinical superiority. Overall, these findings provide a critical interpretation of first-line combination therapies for HCC, accounting for study design, comparator choice, and follow-up duration.

References
1.  Amin N, Anwar J, Sulaiman A, Naumova NN, Anwar N. Hepatocellular Carcinoma: A Comprehensive Review. Diseases. 2025;13:207.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 27]  [Cited by in RCA: 31]  [Article Influence: 31.0]  [Reference Citation Analysis (1)]
2.  Kudo M, Finn RS, Qin S, Han KH, Ikeda K, Piscaglia F, Baron A, Park JW, Han G, Jassem J, Blanc JF, Vogel A, Komov D, Evans TRJ, Lopez C, Dutcus C, Guo M, Saito K, Kraljevic S, Tamai T, Ren M, Cheng AL. Lenvatinib versus sorafenib in first-line treatment of patients with unresectable hepatocellular carcinoma: a randomised phase 3 non-inferiority trial. Lancet. 2018;391:1163-1173.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 4538]  [Cited by in RCA: 4347]  [Article Influence: 543.4]  [Reference Citation Analysis (13)]
3.  Cheng AL, Qin S, Ikeda M, Galle PR, Ducreux M, Kim TY, Lim HY, Kudo M, Breder V, Merle P, Kaseb AO, Li D, Verret W, Ma N, Nicholas A, Wang Y, Li L, Zhu AX, Finn RS. Updated efficacy and safety data from IMbrave150: Atezolizumab plus bevacizumab vs. sorafenib for unresectable hepatocellular carcinoma. J Hepatol. 2022;76:862-873.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 1370]  [Cited by in RCA: 1381]  [Article Influence: 345.3]  [Reference Citation Analysis (9)]
4.  Sangro B, Chan SL, Kelley RK, Lau G, Kudo M, Sukeepaisarnjaroen W, Yarchoan M, De Toni EN, Furuse J, Kang YK, Galle PR, Rimassa L, Heurgué A, Tam VC, Van Dao T, Thungappa SC, Breder V, Ostapenko Y, Reig M, Makowsky M, Paskow MJ, Gupta C, Kurland JF, Negro A, Abou-Alfa GK; HIMALAYA investigators. Four-year overall survival update from the phase III HIMALAYA study of tremelimumab plus durvalumab in unresectable hepatocellular carcinoma. Ann Oncol. 2024;35:448-457.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 298]  [Cited by in RCA: 266]  [Article Influence: 133.0]  [Reference Citation Analysis (3)]
5.  Yau T, Galle PR, Decaens T, Sangro B, Qin S, da Fonseca LG, Karachiwala H, Blanc JF, Park JW, Gane E, Pinter M, Peña AM, Ikeda M, Tai D, Santoro A, Pizarro G, Chiu CF, Schenker M, He A, Chon HJ, Wojcik-Tomaszewska J, Verset G, Wang QQ, Stromko C, Neely J, Singh P, Jimenez Exposito MJ, Kudo M; CheckMate 9DW investigators. Nivolumab plus ipilimumab versus lenvatinib or sorafenib as first-line treatment for unresectable hepatocellular carcinoma (CheckMate 9DW): an open-label, randomised, phase 3 trial. Lancet. 2025;405:1851-1864.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 207]  [Cited by in RCA: 262]  [Article Influence: 262.0]  [Reference Citation Analysis (0)]
6.  Yau T, Kaseb A, Cheng AL, Qin S, Zhu AX, Chan SL, Melkadze T, Sukeepaisarnjaroen W, Breder V, Verset G, Gane E, Borbath I, Rangel JDG, Ryoo BY, Makharadze T, Merle P, Benzaghou F, Milwee S, Wang Z, Curran D, Kelley RK, Rimassa L. Cabozantinib plus atezolizumab versus sorafenib for advanced hepatocellular carcinoma (COSMIC-312): final results of a randomised phase 3 study. Lancet Gastroenterol Hepatol. 2024;9:310-322.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 76]  [Cited by in RCA: 77]  [Article Influence: 38.5]  [Reference Citation Analysis (1)]
7.  Qin S, Gu S, Chan SL, Bai Y, Ren Z, Lin X, Chen Z, Jia W, Jin Y, Guo Y, Hu X, Meng Z, Liang J, Xiong J, Ren H, Yang F, Li W, Chen Y, Zeng Y, Xu L, Yuan X, Li D, Sultanbaev A, Pazgan-Simon M, Pisetska M, Melisi D, Ponomarenko D, Du J, Shi W, Cheng AL, Kaseb A, Vogel A; CARES-310 Study Group. Camrelizumab plus rivoceranib versus sorafenib as first-line therapy for unresectable hepatocellular carcinoma (CARES-310): final analysis of a randomised, open-label, international, phase 3 study. Lancet Oncol. 2025;26:1598-1611.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 17]  [Reference Citation Analysis (0)]
8.  Shi Y, Han G, Zhou J, Shi X, Jia W, Cheng Y, Jin Y, Hua X, Wen T, Wu J, Gu S, Bai Y, Wang X, Zhang T, Chen Z, Zhang B, Huang M, Liu H, Mao Y, Zhou L, Wang R, Shan Y, Zhang W, Song T, Guo Y, Zhou F, Shao B, Zhang M, Liang B, Zheng J, Zhang G, Shen J, Su W, Zhang F, He Y, Hu S, Liu R, Zhang C, Shen S, Zeng H, Wang TE, Guo W, Shen Y, Chen Y, Li Y, Samol J, Hu H, Zhang W, Du C, Li E, Liu C, Pin CS, Li X, Xu H, Huang JF, Hao C, Lv J, Wang W, Xu Q, Bai A, Zhang X, Liu B, Jin C, Fan J. Toripalimab plus bevacizumab versus sorafenib as first-line treatment for advanced hepatocellular carcinoma (HEPATORCH): a randomised, open-label, phase 3 trial. Lancet Gastroenterol Hepatol. 2025;10:658-670.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 35]  [Cited by in RCA: 40]  [Article Influence: 40.0]  [Reference Citation Analysis (0)]
9.  Llovet JM, Kudo M, Merle P, Meyer T, Qin S, Ikeda M, Xu R, Edeline J, Ryoo BY, Ren Z, Masi G, Kwiatkowski M, Lim HY, Kim JH, Breder V, Kumada H, Cheng AL, Galle PR, Kaneko S, Wang A, Mody K, Dutcus C, Dubrovsky L, Siegel AB, Finn RS; LEAP-002 Investigators. Lenvatinib plus pembrolizumab versus lenvatinib plus placebo for advanced hepatocellular carcinoma (LEAP-002): a randomised, double-blind, phase 3 trial. Lancet Oncol. 2023;24:1399-1410.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 382]  [Cited by in RCA: 368]  [Article Influence: 122.7]  [Reference Citation Analysis (2)]
10.  Zhou J, Bai L, Luo J, Bai Y, Pan Y, Yang X, Gao Y, Shi R, Zhang W, Zheng J, Hua X, Xu A, Hu S, Zhang F, Yang X, Da M, Wang R, Ma J, Jia W, Quan D, Peng C, Yang W, Yin G, Qi Y, Zhang G, Du X, Mao X, Meng Z, Jiao S, Fan J; APOLLO Study Group. Anlotinib plus penpulimab versus sorafenib in the first-line treatment of unresectable hepatocellular carcinoma (APOLLO): a randomised, controlled, phase 3 trial. Lancet Oncol. 2025;26:719-731.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 38]  [Cited by in RCA: 42]  [Article Influence: 42.0]  [Reference Citation Analysis (0)]
11.  Liu N, Zhou Y, Lee JJ. IPDfromKM: reconstruct individual patient data from published Kaplan-Meier survival curves. BMC Med Res Methodol. 2021;21:111.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 615]  [Cited by in RCA: 665]  [Article Influence: 133.0]  [Reference Citation Analysis (4)]
12.  Shamseer L, Moher D, Clarke M, Ghersi D, Liberati A, Petticrew M, Shekelle P, Stewart LA; PRISMA-P Group. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015: elaboration and explanation. BMJ. 2015;350:g7647.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 9935]  [Cited by in RCA: 8818]  [Article Influence: 801.6]  [Reference Citation Analysis (4)]
13.  Rimassa L, Chan SL, Sangro B, Lau G, Kudo M, Reig M, Breder V, Ryu MH, Ostapenko Y, Sukeepaisarnjaroen W, Varela M, Tougeron D, Crysler OV, Bouattour M, Van Dao T, Tam VC, Faccio A, Furuse J, Jeng LB, Kang YK, Kelley RK, Paskow MJ, Ran D, Xynos I, Kurland JF, Negro A, Abou-Alfa GK. Five-year overall survival update from the HIMALAYA study of tremelimumab plus durvalumab in unresectable HCC. J Hepatol. 2025;83:899-908.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 78]  [Cited by in RCA: 113]  [Article Influence: 113.0]  [Reference Citation Analysis (0)]
14.  van Bömmel F, Abou-Alfa GK, Chan SL, Sangro B, Lau G, Kudo M, Breder VV, Varela M, Crysler OV, Bouattour M, Dao TV, Faccio A, Furuse J, Kang YK, Jeng L, Kelley RK, Nakamura H, Schmidt J, Ali S, Rimassa L. Subsequent anticancer therapy analysis of the Phase 3 HIMALAYA study of tremelimumab plus durvalumab in unresectable hepatocellular carcinoma. Z Gastroenterol. 2025;63:e551-e552.  [PubMed]  [DOI]  [Full Text]
15.  Shao YY, Akhmetzhanov AR. Demographics and Immunotherapy Efficacy for Advanced Hepatocellular Carcinoma: A Systematic Review and Meta-Analysis of Phase III Clinical Trials. Liver Cancer. 2025;14:302-310.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 3]  [Cited by in RCA: 10]  [Article Influence: 10.0]  [Reference Citation Analysis (0)]
16.  Cranford HM, Jones PD, Wong RJ, Liu Q, Kobetz EN, Reis IM, Koru-Sengul T, Pinheiro PS. Hepatocellular Carcinoma Etiology Drives Survival Outcomes: A Population-Based Analysis. Cancer Epidemiol Biomarkers Prev. 2024;33:1717-1726.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 20]  [Cited by in RCA: 17]  [Article Influence: 8.5]  [Reference Citation Analysis (0)]
17.  Seda P, Abbey R, Peng L, Ezeaku I, Laroia ST, Aziz H. A review of established and new developments in local therapies for liver cancer. ESMO Gastrointest Oncol. 2026;11:100283.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 1]  [Cited by in RCA: 1]  [Article Influence: 1.0]  [Reference Citation Analysis (0)]
18.  Chan LL, Chok KSH, Ip AYH, Poon DMC, Chan SL. Recent advancements in radiotherapy for hepatocellular carcinoma. ESMO Gastrointest Oncol. 2026;11:100308.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 1]  [Reference Citation Analysis (0)]
19.  Cui Y, Du Z, Xu J, Gu L, Du J, Wang M, Yang T. Surgery-centered integrated strategies for personalized hepatocellular carcinoma care. Cancer Biol Med. 2026;j.issn.2095-3941.2026.0045.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 1]  [Reference Citation Analysis (0)]
Footnotes

Peer review: Externally peer reviewed.

Peer-review model: Single blind

Specialty type: Gastroenterology and hepatology

Country of origin: Italy

Peer-review report’s classification

Scientific quality: Grade B, Grade B, Grade D

Novelty: Grade B, Grade C, Grade D

Creativity or innovation: Grade B, Grade C, Grade C

Scientific significance: Grade B, Grade C, Grade C

P-Reviewer: Denisov A, Additional Professor, MD, PhD, Researcher, Spain; Nakamura K, Associate Professor, MD, PhD, Japan; Shi Q, PhD, China S-Editor: Liu H L-Editor: A P-Editor: Wang WB

Write to the Help Desk