Regorafenib use in patients with unresectable hepatocellular carcinoma: analyses of subgroups of special interest in the observational REFINE study.

Introduction
Comprising 90% of primary liver tumors, hepatocellular carcinoma (HCC) represents a significant public health issue worldwide.1 However, liver cancer remains among the least favorable cancer prognoses. Improved 5-year survival trends in recent years may reflect the impact of the improved treatment landscape for HCC, as well as improvements in preventive and curative treatment options for HCC-causing infections, such as hepatitis B and hepatitis C.2, 3, 4 Nevertheless, estimates predict that the global burden of HCC will rise significantly over the next 20 years, forecasting a 56% increase in deaths.5 This is largely due to the growing aging population, as well as increasing rates of obesity and diabetes.5
The treatment landscape for HCC has evolved significantly over recent decades. The oral tyrosine kinase inhibitor (TKI) sorafenib was approved in 2007 as a first-line systemic treatment of advanced HCC.6 Since 2020, immunotherapy treatments alone or in combination with other agents (e.g. atezolizumab + bevacizumab and durvalumab + tremelimumab) have been approved in the first- and second-line settings, becoming new standards of care.7 In 2017, the oral multikinase inhibitor regorafenib was the first agent approved as a second-line treatment of patients with unresectable HCC (uHCC) who were previously treated with sorafenib.8 This approval was based on the findings of the phase III RESORCE trial in which regorafenib demonstrated improved overall survival (OS) and a manageable safety profile versus placebo.9 However, as in all other phase III randomized controlled trials (RCTs), RESORCE had strict eligibility criteria, which excluded patients with certain characteristics, such as prior liver transplantation, Child–Pugh (CP) B/C liver function, Eastern Cooperative Oncology Group performance status (ECOG PS) 2-4, and varices with an increased risk of bleeding. Therefore, there is a need for further assessment of the safety and effectiveness of regorafenib in a broad, real-world setting.
The prospective, observational, international, REFINE study was designed to evaluate regorafenib in a broad, real-world patient population (N = 1005), including patients with CP–B liver function, patients with prior orthotopic liver transplantation, patients who were unable to tolerate sorafenib, and those who had received prior immunotherapy.10 The primary analysis results of REFINE support the safety and efficacy findings for regorafenib that were reported in RESORCE, in a more clinically diverse population.9 The median OS in REFINE [13.2 months, 95% confidence interval (CI) 11.6-14.8 months] was numerically similar to that in RESORCE (10.6 months, 95% CI 9.1-12.1 months) and the safety profile of regorafenib was consistent, with no unexpected safety signals.10
Here, several key post hoc subgroup analyses of the REFINE study are presented to further evaluate the real-world safety and effectiveness of regorafenib in patient populations that were either underrepresented or excluded from RESORCE and other phase III RCTs. Patients were analyzed according to liver function (a well-defined prognostic factor in HCC11), Barcelona Clinic Liver Cancer (BCLC) stage, and prior treatment, including transplantation.

Methods

Study design and participants
REFINE (NCT03289273) was an international, prospective, multicenter, observational study conducted between September 2017 and January 2022. Details of the complete study design, including the full eligibility criteria, have been previously published.10 Briefly, eligible patients were adults with a confirmed diagnosis of uHCC who initiated regorafenib, according to a physician’s decision, before study enrollment.
Patients were stratified in subgroups of interest by prior liver function [CP–B(7-9), CP–B7] and prior treatments, including patients who had received prior locoregional therapy with transarterial chemoembolization (TACE) or transarterial radioembolization (TARE) and in a post hoc analysis, those who underwent prior orthotopic liver transplantation. Patients were also analyzed by BCLC stage (B and C). Patients may have been treated outside of guideline recommendations before entering the study.

Objectives and assessments
The primary objective was to evaluate the real-world safety of regorafenib, including the incidence and severity of treatment-emergent adverse events (TEAEs) and dose modifications due to TEAEs. Dose modifications included dose interruptions and dose reductions.
The secondary objective of this subgroup analysis was to assess the effectiveness of regorafenib in the various subgroups through the analysis of OS and duration of treatment (DoT). OS was defined as the time from the start of regorafenib treatment to the date of death due to any cause. DoT was defined as the time between regorafenib initiation to the day of permanent discontinuation of regorafenib (including death).

Data analyses and statistical considerations
Data relating to safety and effectiveness were evaluated in all patients who received at least one dose of regorafenib and were analyzed according to subgroup [CP–B(7-9) or CP–B7 liver function, BCLC stage B/C disease, prior orthotopic liver transplant, prior TACE (defined as any history of prior TACE), prior TARE]. Adverse events (AEs) were summarized according to the Medical Dictionary for Regulatory Activities v25.0, and severity was categorized by the National Cancer Institute Common Terminology Criteria for Adverse Events v4.03.
OS was estimated using the unadjusted Kaplan–Meier method. Patients were observed from the start of regorafenib treatment until the date of premature discontinuation, withdrawal of consent, or death or until the end of the study. P values were calculated using the log-rank test. P values and 95% CIs are reported without any adjustments for confounding factors and multiple testing.
The study did not aim to confirm or reject predefined hypotheses. All variables were analyzed descriptively, with no adjustment for confounders: categorical variables by frequency tables and continuous variables by sample statistics.

Results

Patient disease characteristics and demographics
A total of 1005 patients were analyzed in the REFINE study (Tables 1A and B). Of these, 123 patients (12%) had CP–B(7-9) liver function and 84 (8%) were classified as having CP–B7. Most patients [625 (62%)] had BCLC stage C disease, and 133 patients (13%) had BCLC stage B disease. A small minority of patients [25 (2%)] had received prior liver transplantation, 584 patients (58%) had received prior TACE, and 28 (3%) had received prior TARE.

Safety

Child–Pugh B(7-9)/B7 liver function subgroups
Across CP–B(7-9) and CP–B7 subgroups and the overall REFINE population, the incidence of drug-related grade ≥3 [CP(7-9) 28%; CP–B7 27%; overall population 27%] and serious drug-related TEAEs [CP(7-9) 11%; CP–B7 12%; overall population 9%] was largely consistent, whereas drug-related TEAEs leading to permanent regorafenib discontinuation were more common in the CP–B(7-9) and CP–B7 subgroups (28% and 25%, respectively, versus 16%) (Table 2A and Supplementary Table S1A, available at https://doi.org/10.1016/j.esmogo.2025.100292). In these subgroups, treatment-emergent hand-foot skin reaction (HFSR) due to any cause or related to regorafenib occurred less frequently than in the overall REFINE population (Supplementary Tables S2A and S3A, available at https://doi.org/10.1016/j.esmogo.2025.100292).

BCLC stage B/C subgroups
In the BCLC stage B subgroup, a lower proportion of patients experienced serious TEAEs (25%) compared with the overall REFINE population (37%), whereas similar incidences were observed in patients with BCLC stage C (37%) and the overall REFINE population (Table 2B). Furthermore, the incidence of drug-related grade ≥3 and serious TEAEs was generally similar across BCLC stage B (27% and 5%, respectively) and C (27% and 9%, respectively) subgroups and the overall REFINE population (27% and 9%, respectively; Table 2B). TEAEs leading to dose modification, dose reduction, and permanent discontinuation of regorafenib occurred at similar rates in BCLC–B and BCLC–C patients (Supplementary Table S1B, available at https://doi.org/10.1016/j.esmogo.2025.100292). TEAEs that were among the most common in the REFINE study occurred with similar frequencies in patients with BCLC–B and BCLC–C (Supplementary Table S2B, available at https://doi.org/10.1016/j.esmogo.2025.100292).

Prior treatment subgroups

Prior orthotopic liver transplantation
In patients who underwent prior liver transplantation, the incidence of grade ≥3 TEAEs was higher than in the no prior liver transplantation group (72% versus 53%), and both drug-related TEAEs (84% versus 74%) and grade ≥3 drug-related TEAEs (48% versus 26%) were more common (Table 2C). In comparison with the no prior liver transplantation group, a higher proportion of patients with prior liver transplantation had any-cause and drug-related HFSR (Supplementary Tables S2C and S3C, available at https://doi.org/10.1016/j.esmogo.2025.100292).

Prior TACE or TARE
Across prior TACE and TARE subgroups, the incidence of grade ≥3 TEAEs was similar to that in the no prior TACE (50% and 58%, respectively) and no prior TARE (54% and 54%, respectively) subgroups (Table 2C). In the prior TARE group, the incidence of drug-related dose modification (46% versus 37%) and dose reduction (32% versus 23%) was higher than in the no prior TARE group (Supplementary Table S1C, available at https://doi.org/10.1016/j.esmogo.2025.100292). In the prior TACE subgroup, any-cause and drug-related HFSR occurred more frequently than in the no prior TACE subgroup (Supplementary Tables S2C and S3C, available at https://doi.org/10.1016/j.esmogo.2025.100292).

Effectiveness

Child–Pugh B(7-9)/B7 liver function subgroups
Median OS was shorter in patients with CP–B(7-9) liver function [n = 123, 6.3 months (nominal 95% CI 4.9-7.8 months)] and CP–B7 liver function [n = 84, 6.7 months (nominal 95% CI 5.1-8.7 months)] than in the overall REFINE population [13.2 (nominal 95% CI 11.6-14.8 months)10; Table 3A and Figure 1A]. The median DoT was similar between CP–B(7-9) and CP–B7 patients (2.3 months versus 2.5 months, respectively), and both were numerically shorter than in the overall REFINE population (3.7 months10; Table 3A).

BCLC stage B/C subgroups
Median OS was longer in the BCLC–B subgroup (n = 133) than in the BCLC–C subgroup [n = 625, 19.3 months (nominal 95% CI 16.3-26.0 months) versus 12.3 months (nominal 95% CI 10.7-14.6 months); Table 3B and Figure 1B] and the overall REFINE population [13.2 months (nominal 95% CI 11.6-14.8 months10)]. The median DoT was similar between BCLC–B and BCLC–C subgroups (3.9 versus 3.8 months); both were numerically slightly longer than the median DoT in the overall REFINE population (3.7 months10; Table 3B).12

Prior treatment subgroups

Prior orthotopic liver transplantation
In patients with prior orthotopic liver transplantation (n = 25), the median OS was numerically longer than in the overall REFINE and no prior transplantation populations [n = 980, 15.8 months (nominal 95% CI 10.1-25.4 months) compared with 13.2 months (nominal 95% CI 11.6-14.8 months)10 and 12.8 months (nominal 95% CI 11.4-14.6 months), respectively; Table 3C and Figure 1C]. Furthermore, the median DoT was numerically longer in these patients than in the overall REFINE population (5.5 months versus 3.7 months10; Table 3C).

Prior TACE or TARE
Median OS was 14.4 months (nominal 95% CI 12.1-16.2 months) in patients who received prior TACE (n = 584) and 12.0 months (95% CI 10.2-14.1 months) in those who received no prior TACE (n = 421). In patients who received prior TACE, the median OS was similar to that seen in the overall REFINE population [13.2 months (nominal 95% CI 11.6-14.8 months)10; Table 3C and Figure 1D], whereas in patients who received no prior TACE, it was numerically shorter. The median OS was 17.5 months (nominal 95% CI 8.8-24.5 months) in patients who received prior TARE (n = 28) and 12.9 months (nominal 95% CI 11.4-14.6 months) in patients who received no prior TARE (n = 977; Table 3C and Figure 1E). The median OS in patients who received prior TARE was numerically longer than in the overall REFINE population; however, a small sample of patients were included in this subgroup. The median DoT was similar between subgroups with prior TACE (3.8 months) and prior TARE (3.2 months) and the overall REFINE population (3.7 months10; Table 3C).

Discussion
The phase III RESORCE trial, which led to the approval of regorafenib in patients with uHCC in the second-line setting following progression on sorafenib, demonstrated the safety profile and efficacy of regorafenib in a select population of patients with uHCC.9 In common with other phase III RCTs, the trial population lacked the diversity that exists in everyday clinical practice. The prospective, observational, real-world REFINE study confirmed the safety and effectiveness of regorafenib in a clinically diverse population of patients with uHCC, including those who were either underrepresented or excluded from the phase III RESORCE trial.10 This subgroup analysis was focused on four specific underrepresented patient subpopulations: patients with CP–B(7-9) liver function, those with BCLC stage B or C disease, and those who underwent prior orthotopic liver transplantation.

Child–Pugh B(7-9)/B7 liver function subgroups
Child–Pugh B(7-9) liver function is an established negative prognostic factor in HCC, with poor OS documented in numerous clinical studies.12, 13, 14, 15 Although many patients with advanced HCC have Child–Pugh B(7-9) (impaired) liver function, these patients are typically excluded from phase III RCTs, including RESORCE.9,16,17 This underrepresentation has resulted in an absence of efficacy and safety data to support treatment recommendations in these patients. In this subgroup analysis, and as expected based on the known prognostic significance of Child–Pugh status, the median OS in CP–B(7-9) and CP–B7 patients was shorter than that in the overall REFINE population (6.3 months and 6.7 months compared with 13.2 months, respectively). Real-world studies of other TKIs, including cabozantinib, lenvatinib, and sorafenib, have reported varied median OS findings, ranging between 2.5 and 13.8 months, in the second- or later-line in CP–B(7-9) patients.12,13,18, 19, 20, 21 In several real-world studies, the combination of the immune checkpoint inhibitor (ICI) atezolizumab with bevacizumab has also shown median OS findings comparable with those observed with regorafenib in this study, ranging from 6.4 to 8.3 months in pretreated and treatment-naïve patients with CP–B(7-9).15,21, 22, 23 Furthermore, the ICI nivolumab has shown a median OS of 7.4 months in a phase I/II study in patients with CP–B(7-9) who had previously received treatment with sorafenib.24 Compared with the overall REFINE population, regorafenib had a similar safety profile in patients with CP–B(7-9) and CP–B7 liver function, with similar rates of TEAEs. Nevertheless, a higher incidence of serious TEAEs and TEAEs leading to permanent discontinuation of regorafenib treatment was observed in patients with CP–B(7-9). Effectiveness and safety data were unavailable for the subgroups of patients with CP–B(8/9) liver status; even so, REFINE provides additional real-world evidence as a large, international, observational study that may support the use of regorafenib in patients with CP–B7 after careful clinical assessment. According to BCLC guidelines, factors including albumin-bilirubin (ALBI) grade (for compensated disease) should be considered when assessing liver function and prognosis.25 In the REFINE study, patients with ALBI grade 1 had more favorable survival outcomes than those with ALBI grade 2 or 3, with a median OS of 20.2 months (nominal 95% CI 17.4-23.3 months) compared with 10.2 months (nominal 95% CI 8.7-11.4 months) and 3.5 months (nominal 95% CI 2.4-6.3 months), respectively.10 Therefore, ALBI grade may be an important factor to consider alongside Child–Pugh liver function class when assessing eligibility for regorafenib treatment in patients with advanced HCC. In the present analysis, patients in the CP–B subgroup were distributed across all ALBI grades; 8 (7%) were ALBI grade 1, 82 (67%) were ALBI grade 2, and 24 (20%) were ALBI grade 3. Further studies are needed to evaluate prognostic models in patients with CP–B liver function to more fully identify those who may benefit from regorafenib.

BCLC stage B/C subgroups
The effectiveness of regorafenib in this subgroup analysis supports its use in patients with BCLC–B and those with BCLC–C disease, with a longer median OS reported in patients with BCLC–B than the overall REFINE population and a consistent median OS in patients with BCLC–C.10 As expected, patients with BCLC–B had a longer median OS than those with BCLC–C, likely owing to the more favorable prognostic features that are typically present in these patients. In addition, the median OS for both subgroups (19.3 months and 12.3 months, respectively) was numerically higher than in the regorafenib cohort of RESORCE (10.6 months),9 which also included patients with BCLC stage B and C.9 Similar safety findings were observed in these subgroups and the overall REFINE cohort; however, fewer grade ≥3 TEAEs and serious TEAEs were observed in BCLC stage B patients.10 The findings of this subgroup analysis also corroborate the safety profile reported in RESORCE,9 with slightly lower rates of TEAEs (BCLC–B 93%, BCLC–C 92%, RESORCE 100%) and fewer drug-related TEAEs (BCLC–B 82%, BCLC–C 75%, RESORCE 93%) reported, compared with RESORCE. As most patients with BCLC–B did not receive prior locoregional therapy with TACE or TARE in this analysis, these findings may suggest that regorafenib may be a suitable systemic therapy option in patients with BCLC–B who are no longer candidates for locoregional therapy. However, it is unknown as to whether these patients were refractory to TACE or TARE, and as such these findings should be interpreted with caution.

Prior treatment subgroups
Real-world studies are essential to understanding the impact of prior treatments on the effectiveness and safety of later-line therapies. Following the results of the phase III RESORCE study, regorafenib was approved for use in patients who had received prior sorafenib.6,8,9 As RESORCE aimed to clearly define a patient population for the second-line use of regorafenib, the inclusion criteria specified that only patients who were able to tolerate sorafenib were eligible for enrollment. Accordingly, the safety profile of regorafenib was unknown in patients who were unable to tolerate sorafenib. In REFINE, there were no such limitations, and patients who discontinued prior sorafenib due to an AE were eligible for inclusion. Overall, regorafenib demonstrated similar effectiveness in these patients, with a median OS (11.1 months) similar to that reported in RESORCE (10.6 months). Moreover, a safety profile in sorafenib-intolerant patients consistent with that seen in the overall REFINE population was observed.10 These findings subsequently led the European Medicines Agency to update the product information (Summary of Product Characteristics, Section 4.4) for regorafenib by removing disease-specific precautions related to the use of regorafenib in patients with HCC who were unable to tolerate sorafenib.26 Furthermore, the National Comprehensive Cancer Network® (NCCN®) subsequently removed a footnote in their Clinical Practice Guidelines in Oncology regarding regorafenib data being reflected in patients who previously tolerated sorafenib only (footnote h, page 27/167, NCCN Clinical Practice Guidelines in Oncology – Hepatobiliary Cancers versions before 2024; HCC-I page 27/111 in NCCN Clinical Practice Guidelines in Oncology – Hepatocellular Carcinoma Version 1, 2025), and the 2025 guidelines now include regorafenib following progression on first-line systemic therapy.27 Further studies that analyze patients by prior treatment may ultimately lead to updates to clinical practice guidelines that may support treatment decision-making in these patient populations. For example, the REFINE-IO study (NCT06117891) in patients with unresectable HCC is designed to assess treatment outcomes according to systemic therapies received following first-line ICI combinations, to gain insights on optimal treatment sequencing following immunotherapy.

Prior orthotopic liver transplantation
Although a liver transplant is recommended in patients with early-stage HCC, patients who have undergone prior liver transplantation are typically excluded from clinical trials, such as RESORCE.9,28 Thus, there is a lack of evidence to support systemic treatment decisions in these patients. Immunotherapies, which are the first-line standard of care in advanced HCC, are not typically recommended in these patients due to the risk of transplant-related complications as a result of rejection.29 Nevertheless, the multikinase inhibitor sorafenib has previously demonstrated significant survival benefit in these patients.30 In the REFINE study, most patients (92%) with prior liver transplantation had received prior sorafenib. In these patients, the effectiveness and safety of regorafenib was consistent with that seen in the overall REFINE population. The median OS was numerically higher than in the overall REFINE population and in those with no prior liver transplantation (15.8 months versus 13.2 months and 12.8 months, respectively), and the incidence and severity of TEAEs were generally consistent, although a higher proportion of patients experienced grade ≥3 drug-related TEAEs.10 However, due to the small numbers of patients who had received a prior liver transplant, interpretation of these findings is limited. These findings also support the efficacy and safety of regorafenib that was reported in RESORCE.9 A numerically higher median OS was reported in this subgroup (15.8 months compared with 10.6 months in RESORCE), and a lower incidence of drug-related any-grade (84% compared with 93%, respectively) and grade ≥3 TEAEs (48% compared with 50%, respectively) was observed. Furthermore, these results corroborate those of retrospective studies that report that use of regorafenib following first-line sorafenib treatment may offer a survival benefit in patients with HCC recurrence after liver transplant.31, 32, 33 These findings are also similar to the retrospective study of the TKI lenvatinib in patients who experience HCC recurrence following liver transplant, which revealed a median OS of 14.5 months.13 Overall, the results of this subgroup analysis support the use of regorafenib following sorafenib treatment in patients with HCC recurrence after liver transplantation.

Prior TACE/TARE
Minimally invasive locoregional therapies, including TACE and TARE, are typically used in the first-line setting in patients with intermediate-stage HCC (BCLC stage B)34; TACE has been the standard of care for >2 decades, with TARE becoming more widely adopted. Therefore, evaluating regorafenib in patients who have previously received these therapies provides a true representation of patients in real-world clinical practice. In this subgroup analysis of REFINE, the efficacy findings in patients with prior TACE and TARE support those reported in the overall REFINE population.10 In patients who received prior TACE, the median OS was numerically similar to that in the overall REFINE population and numerically longer than in patients who did not receive prior TARE. The safety profile of regorafenib was generally consistent with the overall population and patients who did not receive prior TACE, with similar incidences of any-grade and grade ≥3 drug-related TEAEs. These findings further corroborate the safety profile of regorafenib observed in RESORCE, which also included patients who were treated with prior TACE.9 In patients who had received prior TARE, there appeared to be a trend toward a longer median OS compared with the overall REFINE population, and a higher incidence of any-grade and grade ≥3 drug-related TEAEs; however, as only 25 patients were included in this subgroup, the sample size is not sufficient to draw conclusions.10
As previously detailed, the REFINE study is limited by its observational design, and as such, the findings may have been subject to biases.10 Another limitation of this post hoc analysis is that the comparisons presented are descriptive/numerical in nature, rather than based on statistical analyses. Moreover, as this analysis did not adjust for baseline differences in subgroups, any differences observed in safety and effectiveness may be influenced by selection bias and prognostic imbalance.

Conclusions
This subgroup analysis of the real-world REFINE study confirmed the safety profile of regorafenib in underrepresented populations of patients with advanced HCC, including in those with impaired liver function [CP–B(7-9)] and those with BCLC stage B/C disease. The safety profile of regorafenib was also confirmed in patients who had received prior locoregional therapy with TACE. This analysis also provides further insight into the safety profile of regorafenib in patients treated with prior TARE and in those with prior liver transplantation, although these sample sizes were small. CP–B(7-9) liver function is a well-recognized prognostic factor of poor outcomes and is commonly observed in patients with advanced HCC, with some patients eligible for subsequent systemic therapy; therefore, this subgroup is of particular importance. As expected, the median OS of these patients was shorter than in the overall REFINE population; however, it was generally longer than that observed in phase III and real-world studies of other systemic therapies in patients with CP–B(7-9) liver function and advanced HCC.10,14,18, 19, 20 Based on the effectiveness and safety profile of regorafenib in the CP–B7 subgroup compared with the overall CP–B(7–9) group, the findings suggest that regorafenib could be considered for the treatment of patients with CP–B7 liver function after careful clinical assessment. Further studies are required to evaluate the use of regorafenib in patients with CP–B8/9 liver function. The overall results of this subgroup analysis support the safety profile and efficacy of regorafenib reported in the phase III RESORCE trial and provide additional real-world evidence that may help guide treatment decisions in these specific patient populations.