Balloon-occluded versus conventional transarterial chemoembolization for the treatment of early to intermediate stage hepatocellular carcinoma: a meta-analysis and trial sequential analysis

Abstract

Background

We performed a meta-analysis and trial sequential analysis (TSA) to compare the therapeutic efficacy and adverse events (AEs) of balloon-occluded transarterial chemoembolization (B-TACE) with conventional transarterial chemoembolization (cTACE) in treating early-to-intermediate-stage hepatocellular carcinoma (HCC).

Methods

We systematically searched PubMed, Web of Science, Cochrane Library, Embase, China National Knowledge Infrastructure (CNKI), and Wanfang databases for studies comparing B-TACE and cTACE in the treatment of HCC. The outcomes included the complete response (CR) rate, objective response rate (ORR), lipiodol consumption, and adverse events (AEs). Depending on the heterogeneity assessment, either a fixed-effects or random-effects model was utilized, followed by a meta-analysis using Review Manager 5.3. Additionally, a TSA was conducted to assess the adequacy of the sample size.

Results

Five studies including a total of 1166 patients were analyzed. This meta-analysis revealed that compared with cTACE, B-TACE significantly improved the CR rate (risk ratio [RR] = 1.21, 95% confidence interval [CI] 1.04–1.42, p = 0.02) and the ORR (RR = 1.23, 95% CI 1.09–1.38, p = 0.0006). These findings were validated using TSA, which did not require a larger information size. The TSA results indicated that B-TACE consumed more lipiodol than cTACE, potentially leading to more satisfactory embolization efficacy. In terms of AEs, only post-embolization syndrome was found to occur more frequently in patients treated with B-TACE than in those treated with cTACE (RR = 1.30, 95% CI 1.01–1.68, p = 0.04). However, the TSA suggested that additional cases are necessary to confirm this difference.

Conclusions

B-TACE consumed more lipiodol and demonstrated superior effects on the CR rate and ORR compared to cTACE in the treatment of HCC. Importantly, this improvement in efficacy did not correspond to a significant increase in AEs. Based on these findings, it is recommended that well-designed, large-scale randomized controlled trials be conducted to further validate and expand upon these results.

Trial registration

This study was registered in the international prospective register of systematic reviews PROSPERO (registration No: CRD42023489055).

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-025-14435-y.

Introduction

Globally, hepatocellular carcinoma (HCC) ranks third in cancer-related deaths and sixth in cancer diagnoses [1]. The Barcelona Clinic Liver Cancer (BCLC) staging system is a prevalent framework for HCC care, offering both tumor classification and treatment guidance [2]. Transarterial chemoembolization (TACE) is presently a guideline-recommended global standard treatment for patients with unresectable intermediate-stage HCC as per the BCLC system. TACE is also considered in early-stage HCC patients for whom recommended treatments, such as surgical resection or ablation therapy, are not feasible or have failed [3]. TACE can be performed using two techniques: conventional TACE (cTACE), which uses a mixed emulsion of lipiodol and chemo-agents delivered via tumor-feeding arteries, and drug-eluting beads TACE (DEB-TACE), also known as drug-eluting microsphere TACE (DEM-TACE), in which chemo-agents are loaded outside the body and controlled released in tumor.

According to recent studies, the treatment efficacy of TACE has greatly increased because of the introduction of novel techniques such as super-selective catheterization, navigation software, and cone-beam computed tomography (CBCT) [4, 5]. However, TACE has a significant limitation in the high rate of tumor recurrence and refractoriness. According to a recent report, the 3-year recurrence rate after TACE can reach 48% [6]. Patients who achieved complete response (CR) after the initial TACE treatment had better survival outcomes than those who achieved CR after at least two TACE sessions (70.2 vs. 40.6 months) [7]. Moreover, repeated TACE may damage liver function if TACE does not achieve a radiological CR after a single session [8]. To enhance the CR rate of single-session TACE, balloon-occluded TACE (B-TACE) was introduced. In B-TACE, a balloon microcatheter is inflated to temporarily occlude tumor-feeding arteries and compressively deliver lipiodol emulsion without reflux. The putative advantage of B-TACE arises from the hemodynamic alterations induced by balloon inflation. In B-TACE, the occlusive action of the proximal arteries lowers the balloon-occluded arterial stump pressure (BOASP). With a BOASP decreased to ≤ 64 mm Hg, the drug-embolic mixture, whether lipiodol or DEB, can be forcefully injected into the tumor arteries, leading to a higher complete necrosis rate of the tumor. Furthermore, various types of balloon microcatheters may also influence therapeutic outcomes. A recent study indicated that the use of double-balloon microcatheters can more effectively reduce the hepatic arterial blood pressure, which may potentially lead to better embolic outcomes [9]. While conventional indications for B-TACE remain undefined, several studies indicate that tumor responses are superior in patients receiving B-TACE compared to those undergoing cTACE methods. A case–control study [10] found that B-TACE had better oncological outcomes and longer recurrence-free time than DEM-TACE. However, due to the lack of large-sample randomized controlled trials (RCTs), it is not possible to make a reliable conclusion on whether B-TACE is superior to cTACE. Additionally, whether B-TACE is associated with a higher rate of adverse events (AEs) than cTACE is still debated. This meta-analysis sought to find all existing trials to compare the efficacy and outcomes of B-TACE with cTACE. To ensure the reliability of our conclusions, we performed a trial sequential analysis (TSA), which integrates the repeated testing of the significance of accumulating data with the estimation of the required information size (RIS). This method helps correct the increased risk of random errors and biases often present in meta-analyses. Additionally, it clarifies whether the sample size of the included studies is sufficient to draw definitive conclusions.

Materials and methods

This meta-analysis was conducted in accordance with the principles of the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement [11]. This study was registered in the international prospective register of systematic reviews PROSPERO (registration No: CRD42023489055).

Search strategy

From the first accessible date to April 26, 2025, we searched PubMed, Web of Science, Embase, the Cochrane Library, China National Knowledge Infrastructure (CNKI), and Wanfang databases extensively for relevant studies. The search included the following terms: “liver cancer” or “hepatic cancer” or “HCC” or “hepatic neoplasm” or “liver neoplasm” or “hepatocellular cancer,” and “balloon-occluded transcatheter arterial chemoembolization” or “balloon occluded” or “B-TACE.” The search strategy is detailed in Table S1 (see Additional file). The language of this study was not restricted, and only RCTs and cohort studies were eligible for inclusion. To ensure comprehensive coverage, we also examined the reference lists of relevant studies to identify any additional research. The literature review was conducted independently by two investigators.

Eligibility criteria

Study participants were selected by three independent investigators, with any disagreements resolved through discussion. Studies meeting the criteria as follows were incorporated into the analysis: (1) according to the EASL guidelines, patients who were diagnosed with HCC; (2) treatment modalities included cTACE and DEM-TACE with or without balloon occlusion; (3) patients with potentially unresectable HCC of BCLC stage 0 to B and stage I to II HCC according to the TNM staging system of the Liver Cancer Study Group of Japan [12]; and (4) patients classified as Child–Pugh class A or B, or those who were in Child–Pugh class C with minimal ascites that were managed with diuretics. The exclusion criteria were: (1) case–control studies, reviews, meeting abstracts, studies without control groups, expert opinions, case reports, and letters; (2) irrelevant studies or studies lacked sufficient data; (3) studies with repeated data or the same population in different publications from the same author group; (4) patients with portal vein invasion or extrahepatic metastasis; (5) patients who had undergone prior systemic therapies; and (6) individuals exhibiting a platelet count below 50,000/μL or a bilirubin level exceeding 3 mg/dL. Figure 1 shows the flowchart of the study inclusion procedure.

Fig. 1.

Flow chart of trials selection process

B-TACE technical procedures

The B-TACE procedure was carried out with a 2.0, 2.8, or 3.0 Fr microcatheter with an occlusion balloon at the tip. The micro-balloon catheter was inserted through a standard 4.0 or 5.0 Fr angiography catheter via the coaxial technique. It was thereafter advanced as near as feasible to the target vessel and situated in the segmental or subsegmental branch of the artery. Inflation of the micro-balloon was carried out using a contrast media/saline solution, and the balloon was inflated to the diameter of the target vessel to achieve flow occlusion. The infusion of lipiodol emulsion or drug-eluting microspheres was then performed under micro-balloon occlusion until sufficient filling of the cancer nodule or overflow into the intrahepatic collateral pathway was observed.

Data extraction and quality assessment

An investigator used the predetermined data extraction form to collect data from the included full-text publications; two other investigators checked the extraction to ensure accuracy. Any disagreements were resolved through discussions until a consensus was reached. The following data were collected: authors'names, publication year, study location (country or region), study design, treatment type, number of patients, average patient age, sex of participants, and stages of HCC (as classified by the BCLC staging system or the TNM staging system of the Liver Cancer Study Group of Japan), Child–Pugh classes, Eastern Cooperative Oncology Group Performance Status (ECOG-PS), CR rate, objective response rate (ORR), dose of lipiodol consumed in lipiodol TACE, and AEs (including clinical events and biological events). The ORR was defined as the rate of patients reaching a CR or partial response (PR), as evaluated using the Response Evaluation Criteria in Cancer of the Liver (RECICL) [13] or the modified Response Evaluation Criteria in Solid Tumors (mRECIST) [14]. The amount of lipiodol consumed in lipiodol TACE was recorded from the beginning of the lipiodol emulsion injection until the end of the injection process. The injection process ended when the lipiodol emulsion flowed into the peripheral portal vein branches of the normal liver or the dense accumulation of lipiodol in the tumor was confirmed. As the evaluation methods for necrotic changes in patients with CR were similar between RECICL and mRECIST [13], the outcome of the CR rate was extracted together. However, the evaluation method of PR differed significantly between these two criteria, and the outcome of the ORR was extracted separately. For studies using propensity score matching (PSM), data were extracted both before and after matching respectively. The tumor response, including the CR rate and ORRs evaluated using different standards, was defined as the primary outcome. The secondary outcomes were the dose of lipiodol consumed in TACE and AEs.

The Newcastle–Ottawa Scale (NOS) was employed to evaluate the methodological quality of the cohort studies incorporated in this meta-analysis. Each study was scored based on the following criteria: representativeness of the exposed cohort, selection of the non-exposed cohort, ascertainment of exposure, absence of the outcome at baseline, comparability of cohorts (based on design or analysis), outcome assessment, length of follow-up, and adequacy of follow-up. Two investigators independently performed the quality assessment using this scale. The NOS scores for each cohort study ranged from 0 to 9 points. To improve the reliability of our analysis, studies that scored > 6 points were included.

Statistical analysis

The meta-analysis was conducted using Review Manager 5.3 (The Nordic Cochrane Center, The Cochrane Collaboration, Copenhagen, Denmark) and Stata SE 15 (STATA Corp., College Station, TX, USA). As all included studies were cohort studies, we calculated risk ratios (RR) with 95% confidence intervals (CIs) for dichotomous outcomes. Since the units for continuous outcomes were identical, we computed mean differences (MD) for these outcomes. The RR was defined as the ratio of CR rate, ORR, or AEs in the B-TACE group to that in the cTACE group. For interpretation, an observed RR > 1 indicated that events (CR rate, ORR, or AEs) were more frequent in the B-TACE group compared to the cTACE group. Similarly, an observed MD > 0 suggested that the B-TACE group consumed more lipiodol than the cTACE group. Statistical heterogeneity among the studies included in each meta-analysis was assessed using the Q test and I² test, with heterogeneity defined as I² > 50% [15]. When there was moderate heterogeneity among the included studies (I² < 50%), a fixed-effects model was utilized. Conversely, a random-effects model was employed when there was significant heterogeneity (I² > 50%). All p-values were two-tailed, with p < 0.05 being statistically significant. All forest plots were generated using GraphPad Prism 9 (GraphPad Software, San Diego, CA, USA) in accordance with the findings from Review Manager 5.3.

TSA Performance

The accumulated Z-value is the statistic derived from dividing the logarithm of the pooled intervention effect by its standard error. The anticipated intervention effect was strongly supported when the Z-curve crossed the trial sequential monitoring boundaries, entered the futility area, or crossed the RIS line. Our TSA was designed to maintain an overall type I error risk of 5% and 80% power for calculating the RIS. We calculated the proportion of control events from the low-risk cTACE cohort of trials incorporated in our meta-analysis. For the continuous outcomes, the empirical MD and variance were applied to calculate the RIS. Due to the unclear relative risk reduction (RRR), we selected the RRR estimated from studies classified as having a low risk of bias using TSA. To compare B-TACE and cTACE across different aspects, two-sided tests were utilized. Based on the findings of the heterogeneity evaluation in our meta-analysis, we utilized either a fixed-effects or random-effects model. Due to the limited number of participants, the random-effects models were applied using the Biggerstaff-Tweedie (BT) method. TSA 0.9.5.10 Beta (http://www.ctu.dk/tsa/) was applied to perform TSA.

Results

Literature search results

The initial search identified 594 studies, of which five cohort studies [16–20] were included. The characteristics of studies included in our search results were listed in Table 1. The study enrolled 1166 patients, comprising 262 treated with B-TACE and 904 treated with cTACE. Among them, patients in the study by Arai et al. [16] were staged using the TNM staging system. Thirty-eight patients had stage I HCC (39.2%), and fifty-nine had stage II HCC (60.8%). For liver function evaluation, 927 (79.5%) patients were classified into Child–Pugh class A, and 232 (19.9%) patients were classified into Child–Pugh class B. However, seven patients with Child–Pugh C liver function were included in the cTACE group before PSM in the trial by Golfieri et al. [19], as long as the ascites were minor and could be treated with diuretics.

Table 1.

Studies included in meta-analysis

Study	Publication year	Country/Region	Study type	Groups	Number of cases	Sex (M/F)	Age (years)	BCLC stage	TNM stage (I/II)	Child–Pugh class (A/B)	Tumor size (cm)
Arai et al. [16]	2015	Japan	Retrospective cohort study	B-TACE	49	33/16	71.9 (62–84)	N.A	16/33	36/13	2.9 (0.8–7.3)
				cTACE	48	34/14	69.9 (54–91)	N.A	22/26	37/11	2.45 (1.4–9.0)
Chu et al. [20]	2023	Korea	PSM study	B-TACE	44	31/13	65.4 ± 10.6	0-B	N.A	39/5	3.7 ± 1.5
				cTACE	523	387/136	63.2 ± 10.2	0-B	N.A	454/69	3.2 ± 1.8
Golfieri et al. [19]	2021	Europe	PSM study	B-TACE	91	75/16	68 (40–91)	0-B	N.A	67/24	3.6 (0.9–15.9)
				cTACE	234	177/57	65 (39–86)	0-B	N.A	156/71	3.3 (1.0–15.0)
Irie et al. [17]	2016	Japan	Retrospective cohort study	B-TACE	28	22/6	72.5 ± 9.0	0-B	N.A	17/11	3.92 ± 2.29
				cTACE	49	35/14	71.8 ± 8.8	0-B	N.A	39/10	4.03 ± 2.71
Maruyama et al. [18]	2016	Japan	Retrospective cohort study	B-TACE	50	N.A	N.A	B	N.A	43/7	3.2 ± 2.8
				cTACE	50	N.A	N.A	B	N.A	39/11	2.8 ± 1.7

M Male, F Female, BCLC Barcelona Clinic Liver Cancer, PSM Propensity score matching, B-TACE Balloon-occluded TACE, cTACE Conventional TACE, N.A. Not available

Two of the studies [19, 20] conducted PSM. After the matching, as is presented in Table 2, 246 patients were included, with 123 undergoing B-TACE treatment and 123 undergoing cTACE treatment. In total, 186 (75.6%) patients had a Child–Pugh class A liver function, while 60 (24.4%) were classified as Child–Pugh class B. The seven patients with Child–Pugh class C HCC, as mentioned above, were all excluded through PSM.

Table 2.

Studies included in the meta-analysis after PSM

Study	Groups	Number of cases	Sex (M/F)	Age (years)	Child–Pugh class (A/B)	Tumor size (cm)
Chu et al. [20]	B-TACE	32	23/9	65.4 ± 10.6	28/4	3.4 ± 1.2
	cTACE	32	22/10	63.2 ± 10.2	25/7	3.6 ± 2.4
Golfieri et al. [19]	B-TACE	91	75/16	68 (40–91)	67/24	3.6 (0.9–15.9)
	cTACE	91	67/24	66 (44–86)	66/25	3.4 (1.0–15.0)

M Male, F Female, B-TACE Balloon-occluded TACE, cTACE Conventional TACE

Risk of bias

The risk of bias assessment results for the included studies are presented in Table 3. Two of the studies [17, 18] included patients who had undergone ablation therapy before the TACE treatment; we considered that the outcomes of interest may have been presented before the start of their studies, and they scored 7 points. Although the study by Golfieri et al. [19] performed PSM, the control for additional factors was not satisfactory, mainly because patients’ characteristics, such as BCLC stage, did not completely meet those mentioned in their eligibility criteria, thus scoring 8 points on the NOS.

Table 3.

Methodological quality assessment of cohort studies: the Newcastle–Ottawa Scale

Study	Representativeness of the exposed cohort	Selection of the non-exposed cohort	Selection of exposure	Demonstration that outcome of interest was not present at start of study	Comparability of cohorts on the basis of the design or analysis	Assessment of outcome	Sufficient follow-up	Adequacy of follow-up of cohorts
Arai et al., 2015	★	★	★	★	★✰	★	★	★
Chu et al., 2023	★	★	★	★	★★	★	★	★
Golfieri et al., 2021	★	★	★	★	★✰	★	★	★
Irie et al., 2016	★	★	★	✰	★✰	★	★	★
Maruyama et al. 2016	★	★	★	✰	★✰	★	★	★

★A point is given for meeting the corresponding criterion

✰No points

Complete response rate

One of the studies compared CR rates evaluated using the mRECIST criteria before PSM, and another two studies compared CR rates evaluated using the RECICL criteria. Because no significant heterogeneity was observed across the studies (p = 0.36, I² = 2%), a fixed-effects model was applied. The CR rates of B-TACE evaluated using both mRECIST and RECICL were significantly higher than those of cTACE (63.1% vs. 53.2%, RR = 1.21, 95% CI 1.04–1.42, p = 0.02; Fig. 2a). The cumulative Z curve exceeded the monitoring boundary, further indicating that the CR rate in the B-TACE group was significantly higher than that in the cTACE group. (Fig. 2b).

Fig. 2.

a Comparison of the complete response (CR) rates before PSM b TSA of the CR rates before PSM. The red region on the right side of the horizontal axis indicates the area of futility. The red lines on the left represent the trial sequential boundaries for benefit or harm, whereas the flat green lines denote the conventional boundaries for benefit or harm. The RIS of 645 patients was determined using an event proportion of 52.0% in the control arm. B-TACE Balloon-occluded TACE, cTACE conventional TACE, CR Complete response, CI confidence interval, RIS required information size

Two studies reported the CR evaluated using the mRECIST criteria after PSM. A fixed-effects model was used because no significant heterogeneity was observed (p = 0.24, I² = 26%). Compared with cTACE, B-TACE significantly improved the CR rate (68.3% vs. 52.8%, RR = 1.29, 95% CI 1.06–1.58, p = 0.01; Fig. 3a). The cumulative Z curve reached neither the monitoring boundary nor the RIS line, which means further studies are necessary for reliable results (Fig. 3b).

Fig. 3.

a Comparison of the complete response (CR) rates after PSM b TSA of the CR rates after PSM. The red region on the right side of the horizontal axis indicates the area of futility. The red lines on the left represent the trial sequential boundaries for benefit or harm, whereas the flat green lines denote the conventional boundaries for benefit or harm. The RIS of 676 patients was determined using an event proportion of 43.0% in the control arm. B-TACE Balloon-occluded TACE, cTACE conventional TACE, CR Complete response, CI confidence interval, RIS required information size

Objective response rate

In two studies, the ORR values evaluated using mRECIST were compared after PSM. A random-effects model was used due to the significant heterogeneity observed among the studies. (p = 0.10, I² = 63%). No evidence of a difference in ORR was found between the two groups (92.7% vs. 94.3%, RR = 1.00, 95% CI 0.90–1.12, p = 0.95; Fig. 4a). The random-effects model (BT) was used in the TSA of the ORR evaluated using mRECIST. According to the TSA results, boundary RIS was ignored due to there being insufficient information, and 143,711 patients were deemed necessary for accurate conclusions (Fig. 4b).

Fig. 4.

a Comparison of the objective response rates (ORRs) evaluated using mRECIST b TSA of the ORR was performed using mRECIST. The flat green lines denote the conventional boundaries for benefit or harm. The RIS of 143,711 patients was determined using an event proportion of 94.0% in the control arm. B-TACE Balloon-occluded TACE, cTACE conventional TACE, ORR Objective response rate, CI confidence interval, RIS required information size

Two studies reported the ORR evaluated using RECICL. A fixed-effects model was used because no significant heterogeneity was observed in the ORR evaluated using RECICL (p = 0.30, I² = 9%). The ORR of patients treated with B-TACE evaluated using RECICL was significantly higher than that of the cTACE group (96.1% vs. 79.4%, RR = 1.23, 95% CI 1.09–1.38, p = 0.0006; Fig. 5a). According to the TSA calculations, the Z curve exceeded both the monitoring boundary and the RIS, further indicating that the ORRs in the B-TACE group were higher than those in the cTACE group (Fig. 5b).

Fig. 5.

a Comparison of the objective response rates (ORRs) evaluated using RECICL b TSA of the ORR evaluated by RECICL. The red region on the right side of the horizontal axis indicates the area of futility. The red lines on the left represent the trial sequential boundaries for benefit or harm, whereas the flat green lines denote the conventional boundaries for benefit or harm. The RIS of 117 patients was determined using an event proportion of 79.0% in the control arm. B-TACE Balloon-occluded TACE, cTACE conventional TACE, ORR Objective response rate, CI confidence interval, RIS required information size

Lipiodol consumption

Two included studies reported the dose of lipiodol consumed during TACE. A random-effects model was applied because the heterogeneity was significant (p = 0.008, I² = 86%). Meta-analysis indicated that no significant difference in the lipiodol dose was found between the B-TACE and cTACE procedures (4.7 vs. 2.8 mL, MD = 1.94 mL, 95% CI −0.22–4.09, p = 0.08; Fig. 6a). A random effects model (BT) was applied to the TSA of the lipiodol dose. Notably, the cumulative Z curve reached not only the monitoring boundary but also the RIS line, which indicated that the lipiodol dose was higher in B-TACE than in cTACE (Fig. 6b).

Fig. 6.

a Comparison of the lipiodol consumption b TSA of the lipiodol consumption. We performed TSA to compare the lipiodol consumption between the B-TACE and cTACE groups. The red region on the right side of the horizontal axis indicates the area of futility. The red lines on the left represent the trial sequential boundaries for benefit or harm, whereas the flat green lines denote the conventional boundaries for benefit or harm. The RIS of 154 patients was calculated using the empirical MD and variance. B-TACE Balloon-occluded TACE, cTACE conventional TACE, CI confidence interval, RIS required information size

Adverse events

AEs were categorized into clinical events, including post-embolization syndrome, fever, nausea or vomiting, abdominal pain, and liver abscess, as well as blood biochemical indicators such as elevated ALT and total bilirubin. The AEs were compared between the B-TACE and cTACE groups. Overall, patients treated with B-TACE had an increased risk of post-embolization syndrome (any grade, 52.8% vs. 40.7%, RR = 1.30, 95% CI 1.01–1.68, p = 0.04; Table 4) and nausea or vomiting (any grade, 18.4% vs. 9.0%, RR = 2.05, 95% CI 1.20–3.50, p = 0.009; Table 4). However, these two AEs ≥ grade 3 were not reported in the included studies. The risk of other AEs (any grade) did not differ significantly between B-TACE and cTACE (p > 0.05) (Table 4). Regarding severe AEs, 16 patients in B-TACE group and 7 patients in cTACE group exhibited elevated ALT (≥ grade 3). In comparison to cTACE, B-TACE did not elevate the occurrence of severe AEs (≥ grade 3) (elevated ALT, p = 0.06; Fig. S1a, see Additional file). The serum ALT levels returned to the baseline levels in all patients either spontaneously or after treatments.

Table 4.

The comparation of adverse events (Any grade)

Study or subgroup subtotal (95% CI)	B-TACE (%)	cTACE (%)	Weight (%)	Risk ratio M–H, fixed, 95% CI	P
Clinical events
Post-embolization syndrome	52.8	40.7	31.2	1.30 [1.01,1.68]	0.04
Fever	34.7	31.7	37.5	1.09 [0.85,1.39]	0.48
Nausea or vomiting	18.4	9.0	10.6	2.05 [1.20,3.50]	0.009
Abdominal pain	23.7	16.9	20.0	1.39 [0.93,2.08]	0.10
Liver abscess	3.5	0	0.6	6.00 [0.73,49.12]	0.09
Biological events
Elevation of ALT	51.6	49.2	71.0	1.04 [0.95,1.14]	0.34
Elevation of total-bilirubin	42.4	38.8	29.0	1.09 [0.78,1.52]	0.61

B-TACE Balloon-occluded TACE, cTACE Conventional TACE, CI Confidence interval

The cumulative Z curve for post-embolization syndrome only crossed the conventional boundary, not the trial sequential monitoring boundary, and a sample size of 2,046 patients was required to draw reliable conclusions (Fig. S2a, see Additional file). For nausea or vomiting, the Z curve only crossed the RIS line but not the monitoring boundary, which indicated that no evidence of a difference was found in the ratio of nausea or vomiting between the B-TACE and cTACE groups (Fig. S2c, see Additional file). For abdominal pain and grade 3 elevation of ALT, the cumulative Z curve reached neither the α-spending monitoring boundary nor the conventional boundary, indicating no significant differences were observed between B-TACE and cTACE treatment, and 1,131 and 408 patients were required to reach the RIS, respectively (Figs. S2 d and S1b, see Additional file). For fever and elevation of ALT and total bilirubin, the boundary TSA was ignored because of the lack of sufficient information (Fig. S2b, e, f, see Additional file). For liver abscess, TSA could not be performed because the incidence of liver abscess was 0 in the cTACE groups of these two studies.

Discussion

Our meta-analysis found that compared to cTACE, B-TACE improved the CR rate and ORR evaluated using RECICL, and TSA confirmed these results. Thus, B-TACE produced a more pronounced tumor response than cTACE. The possible reasons for this are as follows: with a blocking effect of the proximal arteries, the BOASP is effectively decreased. Subsequently, a more intensively administered drug-embolic mixture can be forcefully delivered not only into the tumor vessels but also into the arterioportal micro-anastomoses at the tumor's periphery, resulting in an enhanced therapeutic effect compared with cTACE [21]. As is well known, meta-analyses are regarded as the highest level of evidence. However, they are not without biases and errors. A higher risk of random errors in meta-analyses may arise due to small sample sizes and repeated significance testing of included studies [22]. As a result, we used TSA to create trial sequential monitoring boundaries and estimate the RIS by adjusting the conventional significant levels [23]. In addition, the TSA applied an α-spending function and β-spending function to adjust for type I and type II errors. These functions yielded statistical significance thresholds and futility boundaries [24].

The pooled CR rate was higher in B-TACE group than in cTACE group according to our meta-analysis, whereas no significant heterogeneity was observed (I² = 2%). In the subgroup analysis by Golfieri et al. [19], only patients with an HCC tumor diameter of 30–50 mm showed a significant benefit on the CR rate in the B-TACE group (48.9% for cTACE; 71.7% for B-TACE; p = 0.033). In addition, the TSA result indicates that additional research is necessary to attain a more dependable conclusion on CR rates assessed by mRECIST after PSM.

The ORR evaluated using mRECIST in a study by Golfieri et al. [19] revealed a slight difference between the B-TACE and cTACE groups (94.5% for cTACE; 90.1% for B-TACE; p = 0.405). In our meta-analysis, we found no evidence of a difference in the ORR evaluated using mRECIST between B-TACE and cTACE, with significant heterogeneity (I² = 63%). The ORR evaluated using mRECIST has been successfully increased to close to 100% in recent years using super-selective TACE and CBCT, which aimed to improve the efficacy of TACE and reduce the occurrence of AEs [4]. We considered that the ORR of TACE evaluated using mRECIST was too high to make a significant difference between different TACE procedures. The TSA results also showed that more studies are necessary to make a reliable conclusion. However, our meta-analysis found that the ORR of B-TACE evaluated using RECICL was significantly higher than that of cTACE, which was confirmed using TSA. Thus, we believe that different standards for tumor treatment effect evaluation played a part. Developed in 1993, RECICL was commonly used in Japan, and it incorporated RECIST1.1 and mRECIST as response evaluation criteria in systemic therapy, particularly for HCC [13]. By comparing mRECIST to RECICL, we found that the evaluation method of CR was similar between the two standards. However, the evaluation of PR was stricter in RECICL than in mRECIST. Therefore, the evaluation of the ORR with RECICL was stricter than that with mRECIST, and the ORR evaluated using RECICL could not be as high as that evaluated using mRECIST. As a consequence, a significant ORR difference between the two TACE procedures could be easier to observe to some extent. Apart from this, B-TACE was performed much earlier in Japan than in other countries in the world, with more clinical experiences, which may have resulted in a more pronounced outcome.

No significant difference in the lipiodol dose consumed during TACE was found in our meta-analysis. However, TSA revealed that B-TACE consumed more lipiodol than cTACE. We believe that the use of various random-effects model methods might have contributed to the aforementioned differences in results. The DerSimonian-Laird method (DL) for random-effects models is commonly used in meta-analyses. However, as demonstrated in recent studies [25, 26], for random-effects models, as implemented in TSA, is more reliable for studies with a small sample size and high heterogeneity, where errors may arise when using the DL method. The BT method was deemed more appropriate than the DL method as the I² of the lipiodol consumption reached 86%. Therefore, we support the TSA findings and suggest that B-TACE uses more lipiodol than other cTACE methods. The lipiodol dose in TACE is determined by blood flow and tumor size and should correspond to the tumor size to achieve a good curative effect. Necrosis of tumor cells can be achieved only when lipiodol has obstructed all supply vessels and the drug flows completely into the tumor [27]. The injection of lipiodol often stops at the onset of portal branch depiction. In clinical experience, many cTACE procedures do not achieve a satisfactory embolization effect because portal branch depiction appeared in a relatively early stage when the lipiodol dose consumed was insufficient to kill all tumor cells. In addition, lipiodol infused into the liver parenchyma was often washed out through arterioportal communication to the portal venous system during TACE using a conventional microcatheter [9]. By blocking the hepatic arterial flow, forceful embolization of both the accompanying portal veins and the hepatic artery was achieved through the peribiliary arterial plexus by B-TACE, with compact accumulation of the embolic materials [20]. As a result, we considered that with more consumed lipiodol, B-TACE may achieve a more satisfactory embolization effect and thus a better tumor response. At the same time, liver cirrhosis caused by an overdose of drugs and lipiodol can be effectively prevented by blocking the way in which the drug flows back and disperses into the normal liver parenchyma using a balloon occlusion microcatheter.

The primary advantage of B-TACE lies in its hemodynamic modulation effect. However, hepatic artery anatomical variations may influence blood flow dynamics and consequently affect B-TACE outcomes. Hepatic artery stenosis and extrahepatic collateral blood supply may have resulted in restricted tumor uptake of lipiodol emulsion. And thus, in the study by Maruyama et al. [18], no statistically significant difference of the lipiodol emulsion ratio was shown between the B-TACE and cTACE groups at the level of lobar and segmental. A recent study indicates that anatomical abnormalities, including convoluted arterial anatomy, stenosis at the origin of the celiac artery, and arteriosclerosis, may lead to an unstable backup of the parent catheter, which restricts the advancement of the micro-balloon catheter. The balloon anchor technique can be utilized to deal with the abovementioned problem [28].

According to a single-center study by Kim et al. [29], 90% of the included patients who underwent B-TACE experienced post-embolization syndrome, which may result in prolonged hospital stay. Our meta-analysis showed that the rates of post-embolization syndrome and vomiting or nausea were not as high as previously reported, but were significantly higher in the B-TACE group compared to the cTACE group. This result may be attributed to the higher drug infusion and absorption both within the tumor and the peritumoral area, which could have contributed to a better tumor response, as previously demonstrated. The TSA of post-embolization syndrome indicated that more studies are necessary to confirm the results. However, the TSA of nausea and vomiting made a firm conclusion that no evidence of a difference between the two groups was found as the cumulated Z curve reached the RIS line, and overestimation by meta-analysis was successfully controlled. A study by Arai et al. [16] demonstrated that more patients in the B-TACE group had increased serum ALT levels than those in the cTACE group. However, no significant difference in the elevation of ALT was found between the two groups in our meta-analysis. Moreover, no evidence of a difference was found between the B-TACE and cTACE groups regarding other AEs, including clinical and biological events. Further studies are required to confirm these results. Apart from this, no significant difference was observed between the B-TACE and cTACE regarding AEs ≥ grade 3.

In the study by Golfieri et al. [19], both balloon-occluded lipiodol TACE and balloon-occluded DEM-TACE were included in the B-TACE group, whereas other studies only included balloon-occluded cTACE in the B-TACE group. Because only two studies compared the ORRs evaluated with mRECIST with high heterogeneity, a sensitivity analysis could not be conducted, and we could not determine whether the introduction of balloon-occluded DEM-TACE contributed to the heterogeneity. However, a recent study revealed no evidence of a difference in response rates between lipiodol-based B-TACE and B-TACE with drug-eluting microspheres [21].

Our meta-analysis has several limitations. First, given the lack of RCTs comparing B-TACE and cTACE, the included studies were all retrospective cohort studies, which may have resulted in selection bias. In addition, because the number of studies comparing B-TACE and cTACE was relatively small and only two or three studies could be pooled in our meta-analysis in many aspects, sensitivity analyses of significant heterogeneity and publication bias evaluation could not be performed. Aside from that, the survival outcomes and subgroup analyses were insufficient to be pooled in our meta-analysis. Apart from this, the majority of the included studies were single-center investigations, which may lead to an overestimation of evidence quality. According to the results of the TSA and the abovementioned limitations, future trials should address the pressing necessity for large-sample RCTs. Additionally, survival outcomes and various aspects of subgroup analyses need to be further explored. Researchers should focus on recurrence rates, and trials comparing B-TACE with cTACE in patients with HCC of different tumor sizes are warranted.

Conclusion

Our meta-analysis and TSA both demonstrated that B-TACE resulted in a better tumor response compared to cTACE. With more consumed lipiodol, a more satisfactory embolization effect may be achieved with B-TACE. However, due to the limited sample size, a definitive conclusion could not be drawn regarding whether there is a significant difference in the incidence of AEs between the B-TACE and cTACE groups, except for nausea and vomiting. Further high-quality, large-sample RCTs comparing cTACE and B-TACE are necessary to provide more substantial evidence.

Abbreviations

HCC

Hepatocellular carcinoma

BCLC

Barcelona Clinic Liver Cancer

TACE

Transarterial chemoembolization

cTACE

Conventional transarterial chemoembolization

DEB-TACE

Drug-eluting beads transarterial chemoembolization

DEM-TACE

Drug-eluting microsphere transarterial chemoembolization

CBCT

Cone-beam computed tomography

CR

Complete response

B-TACE

Balloon-occluded TACE

RCTs

Randomized controlled trials

AEs

Adverse events

TSA

Trial sequential analysis

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-analyses

ECOG-PS

Eastern Cooperative Oncology Group Performance Status

ORR

Objective response rate

PR

Partial response

RECICL

Response Evaluation Criteria in Cancer of the Liver

mRECIST

Modified Response Evaluation Criteria in Solid Tumors

PSM

Propensity score matching

NOS

Newcastle-Ottawa Scale

RR

Risk ratios

MD

Mean difference

BT

Biggerstaff-Tweedie

BOASP

Balloon-occluded arterial stump pressure

DL

DerSimonian-Laird

Authors’ contributions

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Jiaxi Liu, Weiwei Wang, Huan Zhai, Yanan Ma and Qiwen You. The first draft of the manuscript was written by Jiaxi Liu and Haibo Shao and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This study was partially supported by National Natural Science Foundation of China (No.82072037, 82372070, 82111530100); Liaoning Medical-Engineering Cross Fund (2022-YGJC-53).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.