Transarterial chemoembolization outcomes for advanced hepatocellular carcinoma vs sorafenib and best supportive care in a sub-Saharan African cohort

Abstract

BACKGROUND

Sub-Saharan Africa (SSA) and Southeast Asia account for 80% of hepatocellular carcinoma (HCC) cases globally. Public healthcare systems in low- and middle-income countries often face significant economic constraints, resulting in limited treatment options. The objectives of this study were to identify factors associated with poor outcomes in patients with Barcelona Clinic Liver Cancer (BCLC) stage C and D undergoing transarterial chemoembolization (TACE) and to compare their outcomes to patients treated tyrosine kinase inhibitors (TKIs) or best supportive care (BSC) only.

AIM

To assess clinical outcomes and identify predictive factors that may facilitate the broader implementation of TACE in patients with advanced HCC within resource-constrained settings such as SSA.

METHODS

A single-center, retrospective cohort study was conducted to investigate the risk factors associated with the outcome of TACE in patients with BCLC stage C and D using univariate and multivariate regression analysis. Frequency matching was used to ensure comparable distributions of confounding factors across patients treated with TACE, TKIs, or BSC. Survival analysis was performed to compare outcomes among the matched groups.

RESULTS

Patients with BCLC stage C and D presenting with elevated gamma-glutamyl transferase levels or elevated aspartate aminotransferase levels or portal vein infiltration were identified as high-risk and demonstrated poor response to TACE treatment. In contrast, patients with BCLC stage C disease who lacked these high-risk features showed significantly longer overall survival when treated with TACE compared to those who received BSC or TKIs.

CONCLUSION

Gamma-glutamyl transferase levels, aspartate aminotransferase levels, and portal vein infiltration are critical risk factors to consider when determining treatment strategies for HCC patients in SSA. Patients without these factors can derive significant benefits from TACE as an alternative to BSC or TKIs.

Key Words: Hepatocellular carcinoma; Transarterial chemoembolization; Tyrosine kinase inhibitor; Best supportive care; Outcomes

Core Tip: Hepatocellular carcinoma is one of the most prevalent cancers on the African continent. In low-income countries, transarterial chemoembolization (TACE) is frequently offered as a life-prolonging treatment option, even for late-stage hepatocellular carcinoma patients. This single-center retrospective cohort study evaluated the outcomes of TACE in Barcelona Clinic Liver Cancer (BCLC) stage C and D patients. Furthermore, outcomes were compared in three matched cohorts’ patients with BCLC stage C and D treated with TACE, patients who received tyrosine kinase inhibitors, and patients who were treated with best supportive care. Furthermore, the overall survival of patients with BCLC stage C disease undergoing TACE was significantly better compared to tyrosine kinase inhibitors and best supportive care. Finally, portal vein tumor infiltration, high aspartate aminotransferase and gamma-glutamyl transferase levels were significant risk factors for poor outcomes.

INTRODUCTION

GLOBOCAN reports hepatocellular carcinoma (HCC) as the 6^th most common and 3^rd deadliest cancer worldwide[1]. Approximately 80% of new HCC patients are reported in low- and middle-income countries, specifically Southeast Asia and sub-Saharan Africa (SSA)[2,3]. In SSA, the high incidence is associated with the high prevalence of chronic hepatitis B virus (HBV) infection[4-6]. The true impact of HCC in SSA may be underestimated due to inadequate reporting and incomplete data collection[7]. In high-income countries, HCC typically occurs in older patients with liver cirrhosis. In these countries, around 40% of patients are treated with curative intent, with five-year overall survival (OS) rates of above 70%[8,9]. This is in contrast to SSA, where the reported median age at HCC diagnosis ranges from 28 years to 54 years, often occurring in non-cirrhotic patients[10]. In addition, the disease more frequently occurs in males, and patients present more often with symptomatic disease, with large tumor burdens, metastatic disease, and tumor-related complications, which result in a high fatality ratio[10].

The treatment of HCC is guided by guidelines from several international and regional associations and organizations. The Barcelona Clinic Liver Cancer (BCLC) guidelines form the basis of Western guidelines but are not widely included in Asian guidelines[11,12]. Considering tumor burden, liver function/dysfunction, and physical performance status, patients are classified into five different stages: Very early stage (0), early stage (A), intermediate stage (B), advanced stage (C), and terminal stage (D). Curative-intended treatments (liver resection, ablation, and liver transplantation) are reserved for patients with stage 0 and stage A disease. Liver transplantation can also be considered in stage B patients who are downstaged to within transplant criteria with transarterial chemoembolization (TACE). The expected OS of patients treated with curative intent within guidelines exceeds 5 years. TACE is reserved for selected patients with stage B disease (well-defined nodules and preserved portal flow where selective access to the arterial supply of the tumor is technically possible) with an expected OS exceeding 2.5 years. Systemic therapies, like tyrosine kinase inhibitors (TKIs), are recommended for stage B patients with diffuse, infiltrative, extensive bilobar disease and stage C patients[13]. Tyrosine kinases play a central role in tumorigenesis and cancer progression and have therefore become key targets for drugs. TKIs block the phosphorylation of tyrosine residues on kinase substrates, thereby preventing activation of downstream signaling pathways[14]. The benefit of sorafenib, the first TKI registered for use in HCC was demonstrated in the Sharp randomized controlled trial, which showed a survival benefit of 2.8 months in the sorafenib-treated cohort, compared to placebo[15]. With novel combination systemic therapies such as atezolizumab-bevacizumab and tremelimumab-durvalumab, OS rates beyond 2 years can be expected[9,16,17]. Patients with stage D disease receive best supportive care (BSC) with expected demise before 3 months.

Global disparities exist in the treatment and outcome of HCC. In high-income countries, around 40% of patients are diagnosed sufficiently early to qualify for curative-intended therapies, resulting in five-year OS rates surpassing 70%[10,18,19]. In a recent systematic review that included all population-based studies describing the management and outcomes of HCC in SSA, most patients (84%) received only BSC, while only 4% received curative-intended treatments, 2% received TACE, and 8% received some form of systemic therapy[10]. These dismal figures are firstly related to the advanced stage at presentation, but also to the unavailability of curative treatment options, TACE, and systemic therapies[20]. Due to the lack of systemic therapies for HCC in the public service in South Africa, the use of TACE was extended outside the BCLC criteria in our institution. Although the use of TACE outside guidelines has been reported, there are no published reports on the use and outcomes of TACE for advanced HCC in SSA where the extended use is driven by the lack of systemic therapies[21]. In this retrospective study, we evaluated the outcomes in patients with BCLC stage C and D disease who underwent TACE and compared them to those treated with TKIs or BSC. Additionally, we investigated factors associated with poor prognosis among patients treated with TACE to inform risk-stratified treatment selection in resource-limited settings.

MATERIALS AND METHODS

Patient recruitment and data collection

The institutional review board of the Faculty of Health Sciences at the University of Cape Town granted ethical approval for this study. Data were extracted from two prospectively maintained HCC registries at Groote Schuur Hospital, Cape Town, South Africa. The first, a paper-based registry, included HCC patients treated from January 1, 1990 to December 31, 2016, while the second captured patients from January 1, 2017 using a faculty-secure REDCap (Research Electronic Data Capture) platform hosted by the University of Cape Town.

Patients with BCLC stage C and D HCC who were treated with TACE as definitive primary treatment between January 1, 2005 and August 31, 2024 at Groote Schuur Hospital comprised the study cohort. Age less than 18 years, pregnancy, treatment with bland transarterial embolization, patients who, in addition to TACE, had systemic therapies, and patients who underwent TACE as a bridge to surgery or transplantation and were transplanted or resected were exclusion criteria. Patients with BCLC stage B disease who underwent TACE as definitive management, patients who received TKIs, and patients who were treated with BSC only during the same time period comprised the comparative cohorts. The study flow chart is reported in Supplementary Figure 1. We adopted a retrospective cohort design, as a prospective approach was not feasible in this setting due to resource constraints and the need to capture real-world clinical practice in SSA. Although this design entails inherited limitations, such as potential selection bias and incomplete data, these were minimized through strict inclusion criteria, standardized definitions and rigorous data verification.

TACE

In line with the BCLC guidelines, TACE was considered as definitive palliative treatment in patients with BCLC intermediate stage B disease, but also for patients with BCLC stage C and D disease with Eastern Cooperative Oncology Group Performance Status (ECOG-PS) less than 4. Extrahepatic metastases and macrovascular invasion (MVI) were not regarded as absolute contraindications, provided patients’ life expectancy was assessed as three months or more. TACE was performed using the Seldinger technique, via a common femoral artery access. After delineating the liver arterial supply and tumor vascularization. Super-selective cannulation of segmental and subsegmental hepatic artery branches was performed with a microcatheter (Progreat 2.7 Frcoaxi Corporation, Japan). The selected branches feeding the tumor were embolized with a doxorubicin/Lipiodol emulsion (Guerbet, Paris, France) or doxorubicin-loaded drug-eluting beads (HepaSpheres, Meritmedical, UT, United States). The TACE procedure was deemed complete when blood flow stasis or near standstill (sluggish) flow was confirmed on control angiography. At two-month follow-up, a contrasted computed tomography scan of the chest and the abdomen was performed. Patients with disease progression (new extrahepatic lesions, increase in size and/or number of liver lesions) or worsening ECOG-PS were referred for BSC. Patients with disease stability were offered TACE again three months or more after their first session. OS was calculated from the date of the TACE procedure.

TKI treatment

Sorafenib was only available from January 1, 2009 to December 31, 2017 at Groote Schuur Hospital as part of a donation program. Patients with irresectable HCC and reasonable ECOG-PS who were expected to survive at least three months were offered sorafenib. Extrahepatic metastases, MVI, and Child-Turcotte-Pugh (CTP) B and C disease were not regarded as contraindications at our center. The standard dosage was 400 mg orally daily, and a cycle was equal to four weeks of treatment. Patients were re-evaluated clinically and biochemically after 28 days. Metoclopramide (10 mg orally every eight hours) and loperamide (2 mg orally as required) were used to treat nausea/vomiting and diarrhea, respectively. Patients who developed hand-foot syndrome were managed with moisturizers and pain medication, while hypertension was managed by initiating or optimizing antihypertensive treatment. Dose reduction was applied when necessary. In patients with hematological toxicity as evidenced by an absolute neutrophil count of less than 1 × 10⁹/L, platelet count of less than 60 × 10⁹/L, or life-threatening bleeding, sorafenib was stopped. Renal impairment was managed with electrolyte monitoring and intravenous fluids. OS was calculated from the start date of the sorafenib treatment.

BSC

Patients who did not qualify for sorafenib or TACE were offered BSC only. The treatment regimen included oral morphine sulfate, metoclopramide as an antiemetic, and lactulose to prevent constipation. A palliative care nurse performed home visits as required, and selected patients were referred for hospice care. OS was calculated from the start date of the treatment regimen.

Statistical analysis

Statistical analysis and graphical illustrations of the data were generated in the R (version 4.3.3) and R studio (version 2023.12.1) software. Categorical variables were presented as n (%); Fisher’s exact test was used to assess associations between independent groups, while the Jonckheere-Terpstra test was applied to evaluate trends across ordered groups. Continuous variables were examined as medians and compared using the Wilcoxon rank sum test for two groups, the Kruskal-Wallis test for multiple groups, and Spearman’s correlation test to evaluate trends across ordered groups. The survival curves were generated using the Kaplan-Meier method and compared with the Wald test. For patients with BCLC stage C disease, cohorts were frequency-matched in a 1:1:2 ratio (TACE/sorafenib/BSC) to enable comparison between treatment groups. For patients with BCLC stage D disease, those treated with TACE were compared to BSC using a 1:5 frequency-matching ratio (TACE:BSC). The frequency matching was based on the presence of jaundice, human immunodeficiency virus infection, chronic HBV infection, ECOG-PS, alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT), model of end-stage liver disease (MELD)-sodium (Na) score, lung metastases, CTP grade, gender, age, and BCLC stage using a function implemented into the R package KODAMA[22].

Factors associated with OS of patients (with BCLC stage C or D) undergoing TACE were assessed using univariate and multivariate Cox proportional hazards regression models. To address the influence of extreme values and potential non-linearity in continuous variables, laboratory parameters including GGT, aspartate aminotransferase (AST), alanine aminotransferase, ALP, bilirubin, and international normalized ratio were categorized using the 90^th and 10^th percentiles, as fixed clinical cut-offs may not adequately reflect the distribution and variability within this cohort. Additional analyses using quartiles, quintiles, and clinically defined cut-offs were performed to confirm the robustness of associations across different categorization schemes. Only variables with univariate significance were entered into a multivariate Cox model. Variables with the highest P value were removed from the model using backward stepwise elimination until all variables in the model remained statistically significant. Statistical significance was based on a two-sided test at 5% significance level. To assess the stability and robustness of the multivariate Cox model, we calculated variance inflation factors[23] for all included covariates to rule out multicollinearity. Bootstrap resampling[24] with 1000 iterations was performed to evaluate the consistency of model coefficients across resamples, and percentile-based confidence intervals (CIs) were derived. The model’s discrimination was assessed using Harrell’s concordance index (c-index)[25]. Restricted mean survival time (RMST) was estimated using the R package survRM2, with the truncation time set to the 80^th percentile of the observed follow-up distribution. RMST was chosen instead of relying solely on median OS because several groups exhibit heavy right-censoring, which prevents accurate estimation of the 95%CI; RMST provides a robust and interpretable summary of average OS time under these conditions.

RESULTS

Patient baseline data

During the study period, 569 patients with HCC were treated at Groote Schuur Hospital, of whom two (0.4%), 31 (5.4%), 35 (6.2%), 323 (56.8%), and 178 patients (31.3%) presented with BCLC stage 0, A, B, C, and D disease, respectively. The median age was 49 [interquartile range (IQR): 38-63] years and the majority (n = 427; 75.0%) were male: Over half (n = 329; 57.8%) had chronic HBV infection, of which 5 (0.9%) had co-infection with chronic hepatitis C virus (HCV) and 68 (12.0%) had co-infection with human immunodeficiency virus. The median HBV viral load was 28308 (IQR: 547-1060260) IU/mL. Excessive alcohol intake was reported by 180 (31.6%) patients and 133 (23.4%) were smokers. According to the CTP classification, 263 (46.5%) patients were CTP-A, 208 (36.8%) were CTP-B, and 94 (16.6%) were CTP-C, while the median MELD-Na score was 12 (IQR: 9-12). Out of 71 (12.5%) patients receiving TACE, 17 (23.9%), 49 (69.0%), and 5 (7.0%) were classified as BCLC stage B, C, and D disease, respectively. Demographic and clinical features of these patients are shown in Supplementary Table 1. No statistically significant differences were observed in the prevalence of clinical signs or comorbidities between patients in the different BCLC stages at presentation.

A total of 501 (88.0%) patients had advanced or terminal disease (BCLC stage C or D disease). Of these, the majority (n = 380, 75.8%) received BSC only, while 54 (10.8%) were treated with TACE and 40 (8.0%) received sorafenib, as a definitive treatment. Demographic and clinical features of these patients are shown in Table 1. Patients treated with TACE were significantly older than those who received sorafenib (P value < 0.0001) and BSC (P value = 0.0003). As expected, patients receiving BSC only were characterized by a worse disease according to CTP grade and BCLC criteria. Patients with jaundice and ascites were more likely to be treated with BSC only. Also, the incidence of hepatomegaly varies among the patient treatment groups, as it can limit the eligibility for TACE.

Table 1 Demographic and clinical features of patients with Barcelona Clinic Liver Cancer stage C or D disease treated with transarterial chemoembolization, sorafenib, or best supportive care, n (%)/median (interquartile range).

Feature	TACE (n = 54)	Sorafenib (n = 40)	BSC (n = 381)	P value
Demographics
Gender				0.3740
Female	16 (29.6)	7 (17.5)	101 (26.5)
Male	38 (70.4)	33 (82.5)	280 (73.5)
Age (years)	54.5 (44, 69)	42.5 (36, 57.25)	48 (37.75, 61)	0.0009
Symptoms
Pain	41 (75.9)	32 (80.0)	297 (78.0)	0.8960
Weight loss	33 (61.1)	22 (55.0)	247 (64.8)	0.4190
GI bleeding	0 (0.0)	1 (2.5)	28 (7.3)	0.0440
Fever	2 (3.7)	2 (5.0)	23 (6.0)	0.9310
Signs on presentation
BMI (kg/m2)	21.65 (19.05, 26.25)	22.6 (20.8, 24.2)	22.1 (20.1, 25.5)	0.8860
Jaundice	4 (7.4)	1 (2.5)	81 (21.3)	0.0005
Ascites	9 (16.7)	9 (22.5)	144 (37.8)	0.0019
Palpable mass	23 (42.6)	15 (37.5)	119 (31.2)	0.2050
Hepatomegaly	17 (31.5)	28 (70.0)	204 (53.5)	0.0006
Comorbidities
HIV	7 (13.0)	5 (12.5)	58 (15.2)	0.9130
HBV	32 (59.3)	26 (65.0)	222 (58.3)	0.7540
HCV	6 (11.1)	1 (2.5)	14 (3.7)	0.0495
Obesity	1 (1.9)	0 (0.0)	6 (1.6)	1.0000
Habits
Smoker	11 (20.4)	13 (32.5)	86 (22.6)	0.3500
Drinker	20 (37.0)	13 (32.5)	115 (30.2)	0.5480
Staging/scoring
CTP grade				< 0.0001
A	36 (66.7)	31 (77.5)	124 (32.9)
B	15 (27.8)	8 (20.0)	163 (43.2)
C	3 (5.6)	1 (2.5)	90 (23.9)
MELD-Na score	10.5 (8, 12)	9 (7.75, 13)	14 (10, 20)	< 0.0001
BCLC stage				0.0005
C	49 (90.7)	38 (95.0)	213 (55.9)
D	5 (9.3)	2 (5.0)	168 (44.1)
ECOG-PS				0.0005
0	2 (3.8)	0 (0.0)	11 (3.0)
1	37 (69.8)	22 (55.0)	103 (27.7)
2	13 (24.5)	16 (40.0)	117 (31.5)
3	1 (1.9)	2 (5.0)	113 (30.4)
4	0 (0.0)	0 (0.0)	28 (7.5)
HBV VL (× 103)	1 (0, 455)	3 (1, 3983)	93 (2, 1278)	0.0806
WBC (cells/nL)	6.65 (5.092, 8.97)	7.95 (5.815, 11.788)	8.4 (6.09, 11.45)	0.0024
Platelet count (U/mL)	228.5 (178.75, 334.25)	316 (192, 488.5)	263 (169.5, 396)	0.0376
INR	1.18 (1.11, 1.37)	1.19 (1.118, 1.355)	1.31 (1.16, 1.52)	0.0005
Sodium (mEq/L)	137 (134.25, 139)	138.5 (136.75, 140)	135 (132, 138)	< 0.0001
Urea (mmol/L)	4.5 (3.7, 6.05)	4.5 (3.675, 5.45)	4.9 (3.5, 7.1)	0.3500
Creatinine (mmol/L)	71 (58, 84.75)	70.5 (61, 80)	68.5 (55, 84)	0.9070
Albumin (g/L)	38 (32, 41)	36 (35, 42)	31 (26, 36)	< 0.0001
T-bilirubin (mmol/L)	15 (9, 21.75)	10.5 (8, 15.5)	27 (14, 53)	< 0.0001
C-bilirubin (mmol/L)	8 (5, 14)	5 (4, 7.25)	15 (8, 42)	< 0.0001
ALT (U/L)	51 (26.25, 82.75)	44 (24.5, 71.5)	58 (34, 97.5)	0.0521
AST (U/L)	83 (51, 186)	71 (39.75, 190)	149 (80.5, 253.5)	< 0.0001
ALP (U/L)	166 (135.5, 295.75)	225 (139.75, 326.5)	282 (175, 479.5)	< 0.0001
GGT (U/L)	257 (117.25, 500.75)	337.5 (176.25, 418.5)	339 (174.25, 564.75)	0.1660
CEA (U/L)	3.1 (2, 4.3)	1.9 (1.1, 3.55)	2.75 (1.9, 5.575)	0.3960
CA19.9 (U/L)	67 (16, 164.5)	56 (31.3, 237.05)	74 (17.5, 180.5)	0.7900
AFP (mg/L)	1.5 (0.1, 20.1)	6.7 (0.1, 83.7)	5.8 (0.2, 101.0)	0.1220
Imaging
Cirrhosis present				0.2500
Unknown/not recorded	1 (1.9)	5 (12.5)	38 (10.1)
No	23 (43.4)	17 (42.5)	166 (44.0)
Yes	29 (54.7)	18 (45.0)	173 (45.9)
Degree of differentiation				0.0016
Not assessed, not reported	1 (16.7)	16 (94.1)	61 (81.3)
Poorly	0 (0.0)	1 (5.9)	6 (8.0)
Moderately	2 (33.3)	0 (0.0)	5 (6.7)
Well	3 (50.0)	0 (0.0)	3 (4.0)
PVI	20 (37.7)	9 (22.5)	151 (39.8)	0.1060
PVI with occlusion	19 (35.8)	7 (17.5)	145 (38.3)	0.0290
Focal				0.0170
Multi-focal	31 (59.6)	18 (52.9)	230 (72.8)
Single-focal	21 (40.4)	16 (47.1)	86 (27.2)
Lung metastasis	4 (7.4)	3 (7.5)	85 (22.3)	0.0033

GI: Gastrointestinal; BMI: Body mass index; HIV: Human immunodeficiency virus; HBV: Hepatitis B virus; HCV: Hepatitis C virus; CTP: Child-Turcotte-Pugh; MELD-Na: Model of end-stage liver disease-sodium; BCLC: Barcelona Clinic Liver Cancer; ECOG-PS: Eastern Cooperative Oncology Group Performance Status; VL: Viral load; WBC: White cell count; INR: International normalized ratio; C-bilirubin: Conjugated bilirubin; T-bilirubin: Total bilirubin; ALT: Alanine transaminase; AST: Aspartate aminotransferase; ALP: Alkaline phosphatase; GGT: Gamma-glutamyl transferase; CEA: Carcinoembryonic antigen; CA19.9: Carbohydrate antigen 19-9; AFP: Alpha-fetoprotein; PVI: Portal vein infiltration; TACE: Transarterial chemoembolization; BSC: Best supportive care.

Figure 1 shows the Kaplan-Meier plot of OS for the 71 patients who underwent TACE, stratified by BCLC stages B, C, and D. Although statistically significant differences among the survival curves were not reached based on the Wald test (P value = 0.155), Cox hazard analysis shows a progressive incremental hazard ratio (HR) with the higher BCLC stages. The HR between patients with BCLC stage C and stage B disease was 1.35 with a 95%CI of 0.72-2.53. A higher HR of 2.17 (95%CI: 0.77-6.16) was reported for BCLC-D compared to patients with BCLC stages B disease.

Figure 1 Kaplan-Meier plot of the overall survival for the 71 patients who underwent transarterial chemoembolization as per Barcelona Clinic Liver Cancer stage B, C, and D disease. BCLC: Barcelona Clinic Liver Cancer.

Factors impacting survival in patients with BCLC stage C and D disease treated with TACE

Univariate and multivariate analyses were performed to determine which pretreatment features were significantly linked to OS in patients with BCLC stage C and D disease treated with TACE (Figure 2). We did not observe any statistically significant differences in OS between patients with BCLC stage C and D disease. This may be due to the limited number of patients with BCLC stage D (n = 5).

Figure 2 Univariate and multivariate analysis of hazard ratios for patients treated with transarterial chemoembolization using forest plot representation. a: The hazard ratio was calculated by comparing patients in the lowest 10% to those in the highest 90% of the distribution, serving as the reference group; b: The hazard ratio was calculated by comparing patients in the highest 10% to those in the lower 90% of the distribution, serving as the reference group. CI: Confidence interval; HIV: Human immunodeficiency virus; HBV: Hepatitis B virus; HCV: Hepatitis C virus; VL: Viral load; CTP: Child-Turcotte-Pugh; MELD-Na: Model of end-stage liver disease-sodium; BCLC: Barcelona Clinic Liver Cancer; ECOG-PS: Eastern Cooperative Oncology Group Performance Status; PVI: Portal vein infiltration; WBC: White cell count; CEA: Carcinoembryonic antigen; CA19.9: Carbohydrate antigen 19-9; AFP: Alpha-fetoprotein; INR: International normalized ratio; ALT: Alanine transaminase; AST: Aspartate aminotransferase; ALP: Alkaline phosphatase; GGT: Gamma-glutamyl transferase; C-bilirubin: Conjugated bilirubin; T-bilirubin: Total bilirubin.

A palpable mass, portal vein infiltration (PVI), PVI with occlusion, alpha-fetoprotein, lung metastasis, platelet count, ALP, and GGT levels were associated with a shorter OS. Interestingly, HCV infection was associated with a longer survival. Upon further review of our cohort, we observed that HCV-positive patients had distinct baseline characteristics compared to HCV-negative individuals. Specifically, 5 out of 6 HCV-positive patients (83.3%) were female, compared to only 22.9% in the HCV-negative group (P value = 0.0067). Their median age was also higher (73 years; IQR 68.7-78.7 vs 61 years; IQR 52-70, P value = 0.0023), and they had better preserved liver function, as indicated by a lower MELD-Na score (median 8; IQR 7-9.7 vs 11; IQR 8-13, P value = 0.049).

Analysis using alternative categorizations, including quartiles, quintiles, and clinical thresholds, produced similar trends, reinforcing the validity of the observed associations. These results are presented in Supplementary Table 2. Applying multivariate analysis to capture variables independently linked to OS, only GGT and AST levels and the presence of PVI were significantly associated with higher HRs in patients treated with TACE. The multivariate model showed no evidence of multicollinearity, with all variance inflation factors below 1.3. The final model achieved a c-index of 0.748 with a standard error of 0.035, indicating good discriminative performance. Bootstrap analysis confirmed the robustness of all retained predictors, with CIs consistent with the original model estimates. Although the CI for GGT was wide, this was also observed in the univariate analysis and likely reflects the small sample size and skewed distribution of GGT values.

Among the 54 patients who underwent TACE, 20 (37.7%) had PVI. Only 5 (9.3%) patients showed a GGT value above 800 U/L, and 4 (7.4%) patients showed an AST value above 250 U/L. The demographic and clinical features differences between patients with a survival time equal to or shorter than 10 months and longer than 10 months are shown in Supplementary Table 3. Weight loss, chronic HCV infection, a higher platelet count, higher AST and ALP levels, PVI, and PVI with occlusion were significantly associated with a survival ≤ 10 months. To further elucidate the impact of the identified independent factors predicting outcome for TACE treatment, patients were categorized into two distinct groups based on their risk profiles. The low-risk group consisted of patients with GGT levels below 800 U/L, AST levels below 250 U/L, and no PVI. In contrast, the high-risk group included patients with elevated GGT levels exceeding 800 U/L, AST levels exceeding 250 U/L, or PVI. A graphical representation of the TACE treatment for each patient is reported in Supplementary Figure 2.

Survival analysis

To compare the OS among the different treatments, a frequency matching procedure was implemented to select a group of patients with clinical features independently for patients with BCLC stage C and D disease. The demographic and clinical differences between patients receiving different treatments are presented in Supplementary Tables 4 and 5 for those with BCLC stage C and D disease, respectively.

The survival analysis of matched patients with BCLC stage C disease (Figure 3A) showed the benefit of TACE treatment over BSC with an HR of 0.41 (95%CI: 0.26-0.65; P value = 0.0001) and over sorafenib treatment with an HR of 0.44 (95%CI: 0.28-0.70; P value = 0.0004). The comparison between sorafenib and BSC approached statistical significance, with an HR of 0.68 (95%CI: 0.45-1.03; P value = 0.0715). Among high-risk patients, sorafenib was associated with a trend toward better survival compared to BSC, though the difference was not statistically significant (P value = 0.1490). The results may indicate that patients in this group (with high levels of GGT or high levels of AST or PVI) could benefit from sorafenib (Figure 3B). However, no patients treated with sorafenib in this group survived longer than one year; while the one-year survival probability for patients treated with TACE was 13.3% (95%CI: 3.7%-48.4%), showing higher probabilities of survival compared to 4.3% (95%CI: 0.6%-28.9%) in patients receiving BSC. In the low-risk group, only TACE showed a substantial benefit compared to the treatment with sorafenib and BSC (Figure 3C) with an HR of 0.32 (95%CI: 0.16-0.61; P value = 0.0006) and of 0.33 (95%CI: 0.17-0.65; P value = 0.0012), respectively. Patients treated with TACE showed a one-year survival probability of 54.0% (95%CI: 35.5%-82.1%), which was significantly higher than patients treated with sorafenib (19.1%; 95%CI: 7.9%-46.2%) or BSC (10.8%; 95%CI: 3.9%-30.5%). Median OS and RMST values are reported in Supplementary Table 6.

Figure 3 The survival rates. A-C: Across all matched patients, in risk-categorized groups, high-risk group and low-risk group with Barcelona Clinic Liver Cancer (BCLC) stage C disease; D-F: Across all matched patients, in risk-categorized groups, high-risk group and low-risk group with BCLC stage D disease. BCLC: Barcelona Clinic Liver Cancer; BSC: Best supportive care; TACE: Transarterial chemoembolization; GGT: Gamma-glutamyl transferase; AST: Aspartate aminotransferase.

Among patients with BCLC stage D disease, the survival comparison was limited to those treated with TACE vs BSC, given that only two patients in this subgroup received sorafenib. The survival analysis of matched patients (Figure 3D) showed a no clear difference between the TACE treatment over BSC with an HR 0.53 (95%CI: 0.18-1.56; P value = 0.2490), although the one-year survival probability for patients treated with TACE was 20.0% (95%CI: 3.5%-100.0%), showing higher probabilities of survival compared to 4.6% (95%CI: 0.7%-31.7%) in patients receiving BSC. The analyses of the high-risk (Figure 3E) and low-risk (Figure 3F) groups are limited by the small sample size.

Among all patients with BCLC stage C and D disease, 12.5% (5 of 40) of patients treated with sorafenib had to interrupt the treatment or had a dose reduction, mainly due to side effects related to the gastrointestinal tract (diarrhea, gastrointestinal bleeding, and gastroenteritis). Adverse events arising from TACE and sorafenib are depicted in Supplementary Table 7. There was only one death resulting from a cerebrovascular accident.

DISCUSSION

This study provides important insight into the outcomes of patients with HCC BCLC stage C and D disease treated with TACE outside the BCLC criteria. Among matched patients with BCLC stage C disease, those who underwent TACE demonstrated a significant survival benefit compared to those treated with TKIs or BSC. This finding is particularly relevant in resource-limited settings where systemic therapies are often unavailable or unaffordable. Furthermore, GGT, AST, and the presence of PVI were independent predictors of poor OS post-TACE outside the BCLC criteria. Patients with GGT levels < 800 U/L, AST levels < 250 U/L, and no PVI had a one-year survival probability of 54.0% following TACE, compared to 19.1% for sorafenib and 10.8% for BSC. In contrast, patients with a GGT > 800 U/L, AST > 250 U/L, or PVI had a one-year survival probability of only 13.3% with TACE. Our observations regarding GGT and AST are in line with findings in an Ethiopian study which reported that elevated GGT (≥ 200 U/L) and AST levels (≥ 60 U/L) are independent predictors of OS following TACE[26].

The results of our study align with and expand the existing literature on TACE for advanced HCC. While TACE is traditionally reserved for BCLC stage B disease, several studies, mostly from Asia, have explored its use in stage C and D patients. In general, the use of TACE is more extensive in Asian countries with TACE recommended at stages of the applied guidelines that would translate into stage C disease according to the BCLC guidelines[27-29]. An analysis of the few reports on the use of TACE in SSA showed that it was widely used in patients with advanced and even terminal diseases[30-32]. In the Ethiopian study, 88.1% of treated patients had BCLC C or D disease[26].

The identification of GGT as a predictor of poor TACE outcomes is consistent with previous studies, which show that elevated GGT levels, apart from being associated with poor liver function and worse outcomes in cirrhotic patients with HCC, also may impact survival due to associated aggressive tumor behavior[33]. GGT plays a central role in glutathione metabolism and redox regulation, influencing antioxidant defense, cellular proliferation, and apoptosis. In hepatocarcinogenesis, increased GGT expression provides a growth advantage by helping cells maintain elevated glutathione levels and enhancing resistance to pro-oxidant anticancer therapies[34,35]. Experimental studies have shown that overexpression of GGT enhances the invasive capacity of cancer cells and may even play a role in the development of anticancer drug resistance[36,37]. Additionally, our identification of elevated AST levels as an independent prognostic factor is supported by existing literature linking higher AST levels to impaired liver function and reduced OS in HCC patients[38-40]. AST is a key metabolic enzyme involved in amino acid metabolism and energy production through the replenishment of Krebs cycle intermediates. Elevated serum AST levels reflect hepatocyte injury, necrosis, or inflammation and are commonly observed in viral hepatitis, cirrhosis, alcohol-related liver disease, and drug-induced hepatotoxicity[41]. Similarly, our findings of worse outcomes following TACE in patients with PVI are in line with several studies that have reported PVI as a strong negative prognostic factor in HCC[6,7].

Lung metastasis is a recognized adverse prognostic factor in HCC and has been widely reported as such in previous studies[42-45]. In our cohort, however, lung metastasis did not retain independent significance in the multivariate analysis. This may be explained by its frequent co-occurrence with PVI, which itself remained a strong and independent predictor of poor survival. The overlapping presence of these features suggests a possible causal or collinear relationship, in which lung metastasis may act as a surrogate marker of aggressive disease driven by vascular invasion. Therefore, while lung metastasis was associated with worse survival in univariate analysis, its prognostic impact was not independent when adjusted for PVI. This finding highlights the complex interplay between systemic spread and vascular involvement in advanced HCC and supports the prioritization of PVI as a multivariate prognostic factor in our analysis. The observation of better survival in patients with HCV-related HCC, who more often presented with localized disease, eligible for TACE, compared to HBV-related patients, who more often present with multifocal disease, is in line with previous results. In a Korean nationwide study[46], poorer OS and younger patients in HBV-related HCC compared to HCV-related HCC was reported. Emerging evidence suggests that HCV-related HCC may exhibit a distinct tumor biology[47].

The low rate of severe adverse events in our study is consistent with the literature, which reports a mortality rate of < 1% for TACE[8]. In contrast, sorafenib is associated with significant toxicity, particularly in patients with poor liver function, as seen in our cohort. The results of the initial two randomized controlled trials, which showed median OS rates of 10.7 months and 6.5 months, could not be duplicated in the assessment of real-world data from low- and middle-income countries[15,48]. In an Ethiopian study, patients who received Sorafenib had a median survival of only 106 days[30]. In line with this, our study did not show a significant survival benefit for sorafenib compared to BSC. In subsequent randomized trials, with lenvatinib as first-line treatment, regorafenib as second-line treatment, cabozantinib as second- or subsequent-line treatment, and in two trials with ramucirumab as second-line treatment the median OS ranged from 8.5 months to 13.6 months[49-53]. Combination therapy with atezolizumab plus bevacizumab as first-line treatment resulted in 6- and 12-month OS rates of 84.8% and 72.2%, respectively[54].

In this study, we observed the potential of extending the eligibility criteria of TACE treatment in African patients with advanced-stage HCC. In the most recent version of the guidelines, survival of more than 30 months is expected for patients within the guidelines (BCLC B stage)[17]. Extending criteria for TACE that would result in a median OS of 10 months or more if the procedure can be performed with acceptable morbidity and mortality rates would be ethically defensible, especially in regions where systemic therapies are not readily available.

There are several limitations to our study. The retrospective nature introduces potential biases, including selection bias and incomplete data. The decision to perform TACE was not randomized but based on multidisciplinary clinical judgment, considering hepatic reserve, performance status, tumor burden, portal vein thrombosis, and the presence of extrahepatic disease. Consequently, patients selected for TACE generally had more favorable clinical profiles, whereas patients with severe liver dysfunction, extensive disease, or poor functional status were more frequently allocated to BSC, potentially confounding survival comparisons. To minimize these imbalances, a frequency matching procedure was applied, which substantially reduced baseline differences caused by treatment selection. While this approach improves comparability and strengthens the validity of our findings, we acknowledge that residual confounding cannot be completely excluded due to the non-randomized nature of the study. Secondly, the small sample size of patients undergoing TACE (n = 54) and sorafenib (n = 40) limits the statistical power of subgroup analyses. Thirdly, the study does not provide data on other long-term outcome parameters, such as progression-free survival (PFS) or quality of life, which are critical for evaluating the overall impact of TACE. However, in our setting, parameters such as PFS (defined as time from treatment to radiologic progression or death) may not reflect meaningful clinical benefit, particularly if: (1) The tumor doesn’t shrink but remains stable; (2) The patient dies from liver dysfunction or other causes unrelated to tumor growth; or (3) Progression happens radiologically, but the patient’s quality of life or liver function hasn’t declined. Given the advanced stage of disease in our cohort (BCLC stage C and D disease) and the high early mortality rate (over 50% of patients died within six months), PFS would not provide meaningful additional information. Most deaths occurred before radiologic progression could be assessed and were primarily due to liver dysfunction rather than tumor growth. The retrospective design of this study restricts the ability to establish causality and may affect the generalizability of the findings. Therefore, the three proposed outcome indicators (GGT > 800 U/L, AST > 250 U/L, or PVI) should be validated in future prospective cohorts or real-world registries to confirm their applicability across broader clinical settings. This study presents real-world practice in a resource-limited setting, providing valuable insights into the challenges and outcomes of managing advanced HCC in SSA, making a compelling case for the extension of indications for TACE to compensate for the lack of systemic therapies. Further studies, preferably larger prospective controlled trials, are needed to validate the findings of this study and establish standardized criteria for the use of TACE in advanced HCC. Further research into novel biomarkers could improve risk stratification and patient selection for TACE.

CONCLUSION

In conclusion, this study contributes to the growing body of evidence supporting the use of TACE in more advanced HCC stages by providing comparative effectiveness data against alternative treatments within the specific SSA context. We showed that TACE as a palliative intervention can provide a survival benefit for selected patients with advanced HCC. This is of particular relevance in resource-limited settings, where systemic therapies are often unavailable. The identification of elevated GGT and AST levels and the presence of PVI as predictors of poor outcomes following TACE in this patient group provides a foundation for developing a potential risk-stratification model to guide off-guideline TACE use in this population. Future prospective multicenter studies are needed to validate these results to enable refinements of patient selection strategies that could culminate in generally applicable guidance for off-guideline TACE use.