Benefit of perioperative radiotherapy for hepatocellular carcinoma: a quality-based systematic review and meta-analysis

Abstract

Introduction:

Although surgery is the standard curative modality for hepatocellular carcinoma, more than two-thirds experience intrahepatic recurrence. Since no standard perioperative treatment has been established, the authors performed a meta-analysis to evaluate the benefits of perioperative radiotherapy (RT).

Methods:

The PubMed, MEDLINE, EMBASE, and Cochrane Library were searched until May 2023. Randomized or propensity-matched studies evaluating at least five major clinical factors investigating benefit of perioperative RT, were included. The main effect measure were the pooled odds ratios (OR) regarding the benefit of perioperative RT using 2-year overall survival (OS) and 1-year disease-free survival (DFS) data.

Results:

Seven studies (five randomized and two propensity-matched studies) involving 815 patients were included. The pooled ORs for 1-year DFS and 2-year OS were 0.359 (95% CI: 0.246–0.523) and 0.371 (95% CI: 0.293–0.576), respectively, favoring perioperative RT, with very low heterogeneity. In the subgroup analyses, the benefits of OS and DFS were consistent between the two subgroups [portal vein thrombosis (PVT) and narrow resection margin (RM) groups]. In the PVT subgroup, the pooled OS rates at both 1-year and 2-year (75.6 vs. 36.9%, P<0.001; 25.6 vs. 9.9%, P=0.004) and DFS rates at both 1-year and 2-year (25.2 vs. 10.3%, P=0.194; 11.9 vs. 3.0%, P=0.022) were higher in the perioperative RT group. In the narrow RM subgroup, the surgery and RT groups showed higher pooled OS rates for both 1-year and 2-year (97.3 vs. 91.9%, P=0.042; 90.4 vs. 78.7%, P=0.051) and DFS (88.1 vs. 72.6%, P<0.001; 70.1 vs. 51.7%, P<0.001). Grade 5 toxicity was not reported, and three studies reported grade ≥3 or higher liver function test abnormalities, ranging from 4.8–19.2%.

Conclusion:

The present study supports the oncological benefits of perioperative RT, for cases with high-risk of recurrence. Oncologic outcomes between subgroups differed according to clinical indications.

Introduction

Highlights

• The standard curative treatment for hepatocellular carcinoma is surgery, but about two-thirds experience intrahepatic recurrence.

• Unlike other cancers, the effectiveness of perioperative treatment of hepatocellular carcinoma has not been established.

• This study tried to verify the effectiveness of perioperative radiotherapy by pooled analyzing randomized studies and well-designed propensity matching studies.

• Perioperative radiotherapy was effective in terms of overall survival (odds ratio 0.371) and disease-free survival (odds ratio 0.359), and heterogeneity between studies was low.

• When analyzed by dividing into portal vein thrombosis and narrow resection margin groups according to clinical indication, there was a significant difference in survival rate between groups, but the effectiveness of perioperative radiotherapy was effective in both groups.

Surgical resection is the standard curative modality for the treatment of hepatocellular carcinoma (HCC), with a 5-year survival rate of 56–70%^1–3. However, since intrahepatic recurrence within 5 years is as high as 67–79%^4–6, the necessity of perioperative adjuvant treatment to reduce recurrence has been suggested. Nevertheless, a meta-analysis of randomized trials investigating various adjuvant modalities, including chemotherapy, internal radiation, and heparanase inhibitors, reported that their benefits were limited, and possible toxicities are concerned^7,8. So far, the effectiveness of perioperative therapeutic modalities has not been recognized in major treatment guidelines for HCC^9,10.

Radiotherapy (RT) has been used as a perioperative adjuvant modality for the treatment of various malignancies. Of note, recent computed tomography (CT)-based planning can reduce recurrence while minimizing liver function damage by selectively irradiating high-risk areas in the liver parenchyma. Considering that postoperative recurrence is related to anatomic factors such as narrow margins or vascular invasion^6,11, focal irradiation, including a high-risk area, has a rationale for lowering intrahepatic recurrences. Recently, several randomized studies have reported the benefits of perioperative RT. Wei et al.¹² conducted a randomized study in which neoadjuvant irradiation was administered to patients with HCC and portal vein thrombosis (PVT), and both disease-free survival (DFS) and overall survival (OS) significantly improved (1-year OS: 75.2 vs. 43.1%; 1-year DFS: 33 vs. 14.9%). Shi et al.¹³ performed a randomized study on postoperative stereotactic body radiation therapy in surgical cases with narrow margins and reported improved OS and DFS (3-year OS: 89.5 vs. 69.4%; 3-year DFS: 65.8 vs. 36.87%).

Although several studies investigating the benefit of perioperative RT have been published, they included a small number of patients, and phase 3 studies, which can provide clear clinical indications, are lacking. Therefore, we intended to suggest the therapeutic benefits and clinical indications of perioperative RT by performing a meta-analysis and qualitatively selecting randomized or controlled studies with well-considered comparability.

Method

Searching process and data collection

This study was designed to evaluate the oncological benefit of perioperative RT in HCC. The hypothetical PICO question was: ‘Could perioperative RT (either neo-RT or adjuvant RT) provide additional oncologic benefit as compared to surgery alone?’ This study has been reported according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)¹⁴ (Supplemental Digital Content 1, http://links.lww.com/JS9/B355, Supplemental Digital Content 2, http://links.lww.com/JS9/B356), Assessing the Methodological Quality of Systematic Reviews (AMSTAR) guidelines¹⁵ (Supplemental Digital Content 3, http://links.lww.com/JS9/B357), and the Cochrane handbook version 6.2 in methodologic regard. Studies fulfilling the following inclusion criteria were included in the present study: 1) clinical studies comparing perioperative RT and surgery with surgery alone; 2) randomized studies or propensity score-matched studies evaluating major clinical factors including at least liver function (Child-Pugh criteria), tumor size, multiplicity, vessel invasion, and alpha-fetoprotein (AFP) level; 3) inclusion of 10 or more patients in both arms; and 4) provision of at least one measure of OS or DFS. We systematically searched the PubMed, MEDLINE, EMBASE, and Cochrane Library databases, as recommended by the Cochrane Handbook¹⁶, for studies published until 7 May 2023. The reference lists of the searched studies were also reviewed. We used the following detailed keywords to find studies related to perioperative radiotherapy and HCC, the main subjects of this study, and survival data, the main outcome: (‘adjuvant radiotherapy’ OR ‘adjuvant radiation therapy’ OR ‘adjuvant rt’ OR ‘postoperative radiotherapy’ OR ‘postoperative radiation’ OR ‘postoperative rt’ OR ‘neoadjuvant radiotherapy’ OR ‘neoadjuvant radiation’ OR ‘neoadjuvant rt’ OR ‘preoperative radiotherapy’ OR ‘preoperative radiation’ OR ‘preoperative rt’) AND (hcc OR hepatocellular) AND survival. The full search strategy utilizing the internal functions of the literature search are provided in Supplement Note 1 (Supplemental Digital Content 4, http://links.lww.com/JS9/B358). Language restrictions were not imposed. Conference abstracts that met the inclusion criteria were included. Studies from the same institution with larger numbers of patients were included. Among studies with similar numbers of patients, more recent studies were included.

The primary endpoints were OS and DFS rates, and grade ≥3 complications were investigated as secondary endpoints. Referring to a meta-analysis including 44 surgical series, 1-year and 3-year OS ranged 73–100% and 61–95%, and 1-year and 3-year DFS ranged 46–95% and 20–72%, respectively¹⁷. Furthermore, perioperative RT is commonly considered for cases with a high-risk of recurrence. Therefore, we mainly analyzed the short-term outcomes (e.g. rates at 1-year and 2-year) considering the level of OS and DFS that can determine the efficacy of the modality and the clinical situation in which perioperative RT is applied. A prestandardized form was used for data collection, which included 1) general information including authors, year of publication, affiliation, study design, and patient recruiting period; 2) clinical information including indication of perioperative RT [e.g. PVT or narrow resection margin (RM)], number of patients, viral cause, extrahepatic metastases, main PVT, OS, and DFS rates at 1-years and 2 years, and rates and profiles of grade ≥3 complication. In the absence of numerical data on the OS and progression-free survival (PFS) rates, data were retrieved from descriptive graphs. Above processes were performed by two independent researchers, and any disagreements were resolved through discussion and re-evaluation of the literature.

Risk of bias and quality of study evaluation

In clinical practice, perioperative RT is attempted when there is a high possibility of recurrence, such as a narrow RM or the presence of vascular invasion. However, the indications for perioperative RT are not internationally standardized and may differ among institutions. Therefore, to prove the benefits of perioperative RT through a meta-analysis, only randomized or nonrandomized studies with reliable comparability should be included. Therefore, this study included only randomized and intentionally matched studies (i.e. propensity score-matching studies) that considered at least five major clinical indicators (Child-Pugh criteria, tumor size, multiplicity, vessel invasion, and AFP level). As our study included both randomized and nonrandomized series, we used the Newcastle–Ottawa scale for quantitative quality analysis¹⁸. A score of 8–9 was regarded as high-quality, 5–7 as medium quality, and less than 5 as low quality. Nonrandomized studies regarded as having low quality were excluded based on agreement between authors, referencing the Cochrane handbook chapter, which states that only observational studies with moderate to low risk of bias should be included in pooled analyses¹⁹. For randomized studies, the Jadad scoring²⁰ was performed for referential evaluation of randomized analysis methods.

Statistics

The main effect measure was the pooled odds ratio (OR) regarding the benefit of perioperative RT using OS and PFS data. Considering that we only included randomized and intentionally matched studies with reliable comparability, a fixed-effects model was used as the main effect measure. Since the data availability and clinical significance were based on the literature from a preliminary search^{12,13,21–23}, the pooled OR was calculated from the 2-year OS and 1-year DFS rates. We also obtained the pooled percentile rates of OS and PFS at 1-year and 2 years for the clinical studies. The mixed-effects model, which uses a random effects model combining studies within the subgroups and a fixed-effects model combining studies between subgroups, was used. Subgroup analyses were performed with regard to clinical indications because OS and PFS differed according to these indications. The clinical indications were categorized as 1) PVT and 2) narrow RM (±microvascular invasion), of which the former group showed inferior survival results in the preliminary literature review. Therefore, subgroup analyses were performed for both groups. Sensitivity analyses were performed, of the main effect measure, according to the study design (randomized or propensity-matched studies) and the setting of perioperative RT (neoadjuvant or adjuvant). For the interpretation of results with different study designs, we referred to the stepwise hierarchical pooled analysis for synergistic interpretation by Shin and Rim²⁴ (i.e. the ascending pattern of effect size, statistical significance, and validity in accordance with the higher level of study design strengthens the hypothesis, whereas the descending pattern weakens).

Heterogeneity among studies was assessed using the Cochran Q test²⁵ and I² statistics²⁶; studies with I² statistics of 25, 50, and 75% were regarded as low, moderate, and high, respectively. Publication bias was quantitatively assessed using visual funnel plot analysis and Egger’s test²⁷. If the two-tailed P-value of Egger’s test was <0.1 with visual asymmetry in the funnel plot analyses, Duval and Tweedie’s trim-and-fill method²⁸ was performed, yielding estimates that corrected for publication bias. Above statistical analyses were performed using Comprehensive Meta-Analysis version 4 (Biostat Inc.).

We also performed trial sequential analysis (TSA) to account for the risk of random errors. TSA is a method that assess the information size (the total sample sizes from all incorporated trials) for a meta-analysis with the statistical significance threshold, and quantifies the statistical reliabilty in the cumulative meta-analysis^29,30. We set the possibilities of type I error up to 5% and of type II error up to 20% (80% power). We calculated the information size required to demonstrate or reject an anticipated intervention effect of a -30% relative risk reduction (30% increase of DFS and OS). We also heterogeneity corrected the required information size assuming 30% diversity.

Protocol registration

This study is registered in PROSPERO (CRD42023437315).

Results

Study selection and characteristics

A total of 994 studies were identified in the initial database search. After excluding studies with irrelevant formats (e.g. reviews, letters, editorials, case reports, trial protocols, and laboratory studies), the abstracts of 212 studies were screened. A full-text review was performed for 29 studies that remained after abstract screening, and finally, seven studies^{12,13,21–23,31,32} involving 815 patients who fulfilled all inclusion criteria were included in the present study. The study inclusion process is summarized in Figure 1, and the full search terms are provided in Supplement Note 1 (Supplemental Digital Content 4, http://links.lww.com/JS9/B358).

Figure 1.

The study inclusion process.

Among the seven studies, five were randomized, and two were propensity-matched. The earliest study recruited patients from 2007 to 2012, and the latest from 2019 to 2022. No studies reported significant conflicts of interests or supported by commercial funds. The main clinical indications for perioperative RT were PVT in two studies and narrow RM in the other five studies. Among the patients included in the studies, 86–95% had hepatitis B virus (HBV) as a causative agent, and 87–100% had reserved liver function classified as Child-Pugh class A. None of the studies included patients with extrahepatic metastasis. General information about the included studies is summarized in Table 1, and detailed information, including all clinical information, is provided in Supplement Table 1 (Supplemental Digital Content 5, http://links.lww.com/JS9/B359).

Table 1.

Characteristics of included studies of surgery group.

References	Group	Patient recruiting period	Study design	n	Indication	RT scheme	HBV (%)	CPC A (%)	Macrovasvular inv (%)	Main PVT (%)	Tumor size	O-S 1#year	O-S 2#year	p	D-FS 1#year	D-FS 2#year	P
Sun et al.,21	Surgery	2013–2016	RCT	26	PVT	IMRT 50Gy/25F to cutting bed, main PV branch and trunk	92.3	100	100	26.9	≥5 cm 84.6%	26.9	11.5	0.005	3.8	0	0.001
	Surgery and adjRT			26			92.3	100	100	24.0	≥5 cm 96.2%	76.9	19.2		15.3	7.7
Wei, et al.12	Surgery	2016–2017	RCT	82	PVT	3DCRT 18Gy/6F to tumor and PVT	92.7	97.6	100	37.8	≥10 cm 51.2%	43.1	9.4	<0.001	14.9	3.3	0.009
	Surgery and NART			82			91.5	96.3	100	50.0	≥10 cm 39%	75.2	27.4		33	13.3
Yu et al.,22	Surgery	2007–2012	RCT	61	RM <1 cm, central tumor only	3DCRT 60Gy/30F to cutting bed with a 1cm margin	86.9	100	0	0.0	Mean 5.6 cm	89.6	74.6	0.480	72.4	45.5	0.060
	Surgery and adjRT			58			87.9	100	0	0.0	Mean 4.7 cm	96.2	80.4		78.1	63.1
Shi et al.,13	Surgery	2015–2016	RCT	38	RM <1 mm or R1, mVI	SBRT 35Gy/5F to cutting bed with a 1–3 cm margin	94.7	100	0	0.0	Mean 4.9 cm	100.0	89.2	0.053	76.3	55.2	0.005
	Surgery and adjRT			38			94.7	100	0	0.0	Mean 4.9 cm	100.0	96.8		92.1	70.9
Kuanget al., (abst)32	Surgery	2019–2022	RCT	59	RM ≤1 cm, pathologic mVI	IMRT 50Gy		100	0	0.0					75.9		0.036
	Surgery and adjRT			59				100	0	0.0					90.1
Gou et al.,23	Surgery	2011–2020	PSM	78	RM ≤1 cm, pathologic mVI	M50Gy/25F to cutting bed with a 1cm margin	91	87.2	0	0.0	>5 cm 87.2%	96.0	69.3	0.028	74.5	49.9	0.011
	Surgery and adjRT			78			91	92.3	0	0.0	>5 cm 87.2%	97.1	89.3		90.1	68.7
Long et al.,31	Surgery	2008–2016	PSM	65	RM ≤1 cm, mVI 16.9% (surgery) 9.2% (surgery and RT)	50-60Gy/25-30F to cutting bed with a 1 cm margin	87.7	100	1.5	0.0	>5 cm 33.8%	86.2	84	0.045	66.2	57.6	0.001
	Surgery and adjRT			65			87.7	100	0	0.0	>5 cm 24.6%	98.5	94.8		90.8	78.4

3DCRT, 3-dimensional conformal radiotherapy; CPC, Child-Pugh class; DFS, disease-free survival; EHM, extrahepatic metastasis; HBV, hepatitis B virus; IMRT, intensity-modulated radiotherapy; mVI, microvascular invasion; NOS, Newcastle–Ottawa scale; OS, overall survival; PSM, propensity score-matching; PVT, portal vein thrombosis; RCT, randomized controlled trial; RT, radiotherapy; SBRT, stereotactic body radiotherapy.

Capital M at the heading denotes median value (e.g. M50Gy, median 50 Gy); Main PVT denotes Vp4 or bilateral invasion.

Quality assessment

This study included patients with narrow clinical ranges who underwent surgery and perioperative RT. Both the exposed and control groups were from the same institution, and the data were retrieved from the hospital records of tertiary hospitals. We did not find points for score loss in the selection category. No studies reported follow-up losses, which could have induced a significant bias. One conference abstract study did not report overall survival outcomes, losing one point in the outcome category. Because we included only randomized studies and intentionally matched studies that considered at least five major clinical indicators, all studies pointed to two in the comparability category. All studies were regarded as high-quality; therefore, all seven studies were included in the pooled analyses. A detailed score sheet is provided in Supplement Table 2 (Supplemental Digital Content 5, http://links.lww.com/JS9/B359). Interpretation of quality assessment and adequacy of study design for randomized trials are described in Supplement Table 3 (Supplemental Digital Content 5, http://links.lww.com/JS9/B359).

Clinical results

Regarding DFS, six out of seven studies reported a significant statistical benefit of perioperative RT, whereas one study reported borderline significance (P=0.06); regarding OS, four out of six studies reported a significant statistical benefit whereas one study reported borderline significance (P=0.053) and one study showed nonsignificance (P=0.480)²². In pooled analyses, the pooled ORs for 1-year DFS and 2-year OS were 0.359 (95% CI: 0.246–0.523) and 0.371 (95% CI: 0.293–0.576), in favor of perioperative RT, with very low heterogeneity (P=0.598, I²=~0.0%; P=0.603, I²=~0.0%; respectively). In the PVT subgroup, corresponding ORs were 0.338 (95% CI: 0.164–0.697) and 0.326 (95% CI: 0.151–0.703), respectively, with very low heterogeneity, and in the narrow RM subgroup, corresponding ORs were 0.366 (95% CI: 0.236–0.570) and 0.396 (0.232–0.675), respectively, with very low heterogeneity, all in favor of perioperative RT. Publication bias was not observed in the visual funnel plot analyses (Supplement Fig. 1, Supplemental Digital Content 4, http://links.lww.com/JS9/B358) or quantitative Egger’s test. The above results are summarized in Figure 2 as forest plots, and the data are presented in Table 2.

Figure 2.

Forest plots of pooled analyses of (A) 1-year disease-free survival and (B) 2-year overall survival.

Table 2.

Pooled results of main outcomes.

Subject /Subgroups	No of studies	No. of patients	Heterogeneity (p, I2)	Effects model	Pooled effects (OR, 95% CI)	Egger’s P
1-year DFS
All studies	7	815	0.598, ~0.0%	Fixed	0.359 (0.246–0.523)	0.378
PVT	2	216	0.692, ~0%	Fixed	0.338 (0.164–0.697)
Narrow RM	5	599	0.355, 9.0%	Fixed	0.366 (0.236–0.570)
2-year OS
All studies	6	697	0.603, ~0.0%	Fixed	0.371 (0.239–0.576)	0.9
PVT	2	216	0.450, ~0%	Fixed	0.326 (0.151–0.703)
Narrow RM	4	481	0.408, ~0%	Fixed	0.396 (0.232–0.675)

DFS, disease-free survival; OR, odds ratio; OS, overall survival; PVT, portal vein thrombosis; RM, resection margin.

Sensitivity analyses were performed according to the study design. Regarding 1-year DFS, pooled ORs were 0.419 (95% CI: 0.265–0.663, P<0.001, heterogeneity P=0.610, I²=~0.0%) and 0.258 (95% CI: 0.133–0.501, P<0.001, heterogeneity P=0.480, I²=~0.0%), respectively, among randomized and propensity-matched studies. Regarding 2-year OS, pooled ORs were 0.444 (95% CI: 0.255–0.773, P=0.004, heterogeneity P=0.463, I²=~0.0%) and 0.276 (95% CI: 0.135–0.565, P<0.001, heterogeneity P=0.937, I²=~0.0%), respectively, among randomized and propensity-matched studies. In both analyses, including randomized and propensity-matched studies respectively, effect size was valid and heterogeneity was low, in similar context with overall analysis. Sensitivity analyses were performed excluding the study by Wei et al.¹² because it was the only study in the neoadjuvant setting. The pooled OR was 0.360 (95% CI: 0.233–0.555, heterogeneity P=0.469, I²=~0.0%) regarding 1-year DFS and 0.409 (95% CI: 0.247–0.678, heterogeneity P=0.550, I²=~0.0%) for 2-year OS, both of which were similar to the results of the overall pooled analyses.

In the PVT subgroup, the surgery and RT groups showed higher pooled OS rates at both 1-year and 2-year (pooled percentiles: 75.6 vs. 36.9%, P<0.001 at 1-year; 25.6 vs. 9.9%, P=0.004 at 2 years). Regarding DFS, the surgery and RT groups showed higher pooled rates at 2-year (25.2 vs. 10.3%, P=0.194 at 1-year; 11.9 vs. 3.0%, P=0.022 at 2-year). In the narrow RM subgroup, the surgery and RT groups showed higher pooled OS rates at 1-year and 2 years (97.3 vs. 91.9%, P=0.042 at 1-year; 90.4 vs. 78.7%, P=0.051 at 2-year). Regarding DFS, the surgery and RT groups showed higher pooled rates for both years (88.1 vs. 72.6%, P<0.001 at 1-year; 70.1 vs. 51.7%, P<0.001 at 2-year). The heterogeneity was moderate or less than moderate in the pooled analyses. Detailed results are shown in Supplement Table 4 (Supplemental Digital Content 5, http://links.lww.com/JS9/B359) and Figure 3, and as forest plots in Supplement Fig. 2 (Supplemental Digital Content 4, http://links.lww.com/JS9/B358).

Figure 3.

Pooled temporal percentile of subgroups of (A) overall survival and (B) disease-free survival. RM, resection margin; RT, radiotherapy; PVT, portal vein thrombosis.

Grade 5 toxicity was not reported, and grade 4 toxicity was reported in only one study (two cases of liver function test (LFT) abnormalities in the study by Sun et al.²¹). Three studies reported no grade 3 toxicities. Three studies reported grade 3 or higher LFT abnormalities, ranging from 4.8–19.2%, which was the most commonly reported type of toxicity. The detailed types and incidences of grade 3 or higher toxicities among the included studies are shown in Supplement Table 1 (Supplemental Digital Content 5, http://links.lww.com/JS9/B359).

TSA

We performed TSA for analyses of pooled analyses of 1-year DFS and 2-year OS, which are main analyses of the study. In both analyses, cumulative z-score constantly increase and acrossed into O’Brien-Fleming boundary, exceeding both z-score of 1.96 (correlate to P=0.05) and required information size (event size), suggesting the conclusion of original analyses are reliable. Relevant figures and interpretations are shown in Supplement Fig. 3 (Supplemental Digital Content 4, http://links.lww.com/JS9/B358).

Discussion

In the present meta-analysis, the benefit of perioperative RT was significant for both DFS and OS, with very low heterogeneity. The benefits were consistent in the overall and subgroup analyses, including in the PVT and narrow RM subgroups. The pooled percentiles of DFS and OS were also significantly different between the perioperative RT and surgery alone groups. Notably, although differences in oncologic outcomes due to perioperative RT existed in both subgroups, the prognosis of the PVT and the narrow RM subgroups was different, although they were treated with a similar treatment decision of primary surgery and RT. Toxicities due to RT exist; however, mortality was not noted, and only one case of grade 4 toxicity was observed among 815 patients. Although damage to liver function should be carefully considered, the present study supports the beneficial use of perioperative RT for the treatment of HCC.

The causes and characteristics of HCC vary by region, leading to differences in treatment approaches³³. For instance, the Barcelona Clinic Liver Cancer (BCLC) system and European guidelines recommend surgery for BCLC A or 0, while recent Japanese and Korean guidelines^34,35, advocate surgery for selected cases with portal invasion. A recent meta-analysis encompassing observational studies, revealed that surgery achieved more favorable survival compared to the palliative RT subgroup³⁶. Advances in surgical techniques are prompting attempts at resections in previously challenging-to-access areas. Additionally, there is a trend towards attempting surgery for cases of BCLC intermediate stage or higher, including those with multiple tumors^37,38. While broadening the indications is expected to improve oncologic outcomes, it also introduces an elevated risk of recurrence.

Regarding the recurrence of HCC after surgery, the clinical factors affecting recurrence differed according to the time of recurrence. Early recurrence within 2 years of surgery is mainly related to tumor factors such as vascular invasion and narrow RM, and late recurrence is known to be related to underlying liver disease-related factors such as viral load and degree of cirrhosis^5,6,11 Sun et al.²¹ investigated patients with HCC presenting with PVT, the cutting bed, and the main trunk and branch of the portal vein, which were targeted by RT. In a study by Wei et al.¹² the tumor as well as the filling defect in the portal vein (PVT) on CT were targeted by RT. Postoperative recurrence near the resection margin is particularly common in patients with narrow resection margins. In a randomized study by Shi et al.³⁹, 29.5% of 84 patients with a margin of less than 1 cm experienced recurrence near the cutting bed, but none in the control group with a margin of greater than 1 cm experienced it. Accordingly, most of the studies included in this review set the clinical target volume by securing a margin of ~1 cm on the cutting bed. All studies in the present meta-analysis used CT-based treatment (3-dimensional conformal or intensity-modulated RT), and the ability to selectively irradiate high-risk areas, such as around the resection margins and vessels, might contribute to the effectiveness of perioperative RT. Therefore, we recommend a narrow margin and PVT status as possible indications for perioperative RT, based on the efficacy of the meta-analysis and therapeutic rationale.

Investigating recurrence patterns after RT in randomized or well-designed comparative studies is helpful for establishing specific treatment strategies. Among the studies included in this meta-analysis, in the study by Long et al.³¹, early intrahepatic recurrence rates within 18 months were 15.3 and 32.3% in the perioperative RT and surgery alone groups, respectively, whereas the recurrence rates after 18 months were 32.3 and 27.7%, respectively. In a study by Shi et al.³⁹, the recurrence rates in the local segment in the perioperative RT and surgery groups were 10.5 and 31.6%, respectively; whereas, in the distal segment the corresponding rates were 31.6 and 42.1%, respectively. Therefore, perioperative RT appears to reduce early recurrence associated with tumor factors mentioned above, and it is assumed that microvascular spread or local progression can be reduced by reducing subclinical disease in the target area. It is necessary to investigate the patterns of tumor recurrence in more randomized or well-designed comparative studies to establish RT targets and clinical indications.

Because the liver volume significantly decreases after surgery, liver function damage due to perioperative RT may be of concern. Among the studies included in this meta-analysis, grade ≥3 liver function damage due to perioperative RT was reported less than 10%, in most of the study. However, in a study by Sun et al.²¹, grade ≥3 liver function impairment was reported in 19.2% of patients. Of note, in this study, the RT target included not only the margined cutting bed but also the portal trunk and the main branch that remained. Although Wei et al.¹² included patients with PVT, RT targeted both the PVT and the tumor involved; serious damage to liver function was low (4.7%), probably because the portal vein in the remaining liver was not targeted and the dose of RT was relatively low. The strategy of targeting ~1 cm margin on the cutting bed in the postoperative setting seems safe, considering the results of the studies included in this meta-analysis. Therefore, we suggest targeting perioperative RT to the cutting bed or viable tumor; however, prophylactic treatment of the remaining liver portal vein should be discussed cautiously.

Our study suggests following future perspectives. Although various perioperative modalities have been tested, no treatment has been recognized as standard in major guidelines^9,34,37. Nevertheless, some modalities have reported effective results. In a recent meta-analysis, the application of postoperative transarterial chemoembolization (TACE) significantly improved both OS and DFS, in multinodular tumors and microvascular invasive cases⁴⁰. Several PD-L1 inhibitors, such as nivolumab (NCT03383458), pembrolizumab (NCT03867084), durvalumab (NCT03847428), atezolizumab & bevacizumab (NCT04102098) are being tested in the perioperative setting. Combination treatment of TACE and RT significantly improved survival compared to TACE alone, and particularly improved long-term survival⁴¹. Combining RT with TACE allows for the treatment of incompletely treated lesions in hypovascularized areas, and the lipiodolization of TACE enhances treatment efficiency enabling precise targeting of RT⁴². RT increases the efficacy of immunotherapy by releasing antigens in the microenvironment and increasing the expression of immune receptors⁴³. In a recent study by Wang et al.⁴⁴, the combination of atezolizumab/bevacizumab and RT achieved a response rate of 76.6%⁴⁴. There are still limited studies in the perioperative setting that combine RT with other modalities. Our study aims to provide information on the perioperative efficacy of RT, serving as a cornerstone for further research on perioperative treatment combinations.

This study has several limitations. All studies included in this review were conducted on Chinese populations. Therefore, a vast majority of the included patients had HBV-related HCC. Information regarding the benefits of perioperative RT for HCC associated with hepatitis C virus (HCV) or nonviral chronic hepatitis may require further investigation. Most studies included in the present meta-analysis involved RT in the postoperative setting; however, only one neoadjuvant RT study was included. However, the heterogeneity between studies in the pooled OR analyses, which is the main outcome, was very low, and similar results to the original study were obtained in the sensitivity analysis, excluding the neoadjuvant RT study. Therefore, we suggest the benefits of perioperative RT. Although the hazard ratio is a useful indicator in survival analysis, considering the time progression, few studies provided available data for pooled analysis. Therefore, we calculated pooled ORs based on data at specific time points, as the best available option, along with the pooled temporal percentile analyses which could be easily referenced by clinicians. This study investigated a subject similar to that in the systematic review conducted by Wang et al.⁷. However, only a few well-designed comparative studies have been conducted. Their study included three randomized, one propensity-matched, and four retrospective studies that compared surgery alone with surgery and perioperative RT. This meta-analysis included five randomized controlled studies and two high-quality propensity-matched studies investigating the benefits of perioperative RT, therefore providing more reliable synthesized evidence. In addition, our study reported temporal pooled survival percentiles of subgroups classified according to clinical indications, which are helpful for clinical decision-making.

Conclusion

The present study supports the oncological benefit of perioperative RT and surgery compared to surgery alone, for cases with high-risk of recurrence. Selective analyses, including high-quality comparative series and randomized studies, provide reliable evidence. Although perioperative RT was shown to increase prognosis in both subgroups, oncologic outcomes between subgroups differed according to clinical indication. As studies to date have been limited to HBV-related HCC populations, the role of perioperative RT in patients with HCC of various etiologies is warranted.

Ethical approval

This study is meta-analysis of published literature without using any personal information. The authors are accountable for all aspects of this work, and ensure that questions related to the accuracy or integrity of any part of the study are appropriately investigated and resolved.

Consent

This study is meta-analysis of published literature without using any personal information.

Sources of funding

This study was supported by a Korea University Grant (K2125791). The research grant supported only methodological aspects, including statistical analysis and linguistic correction, and did not affect the major contents, including the results and conclusions.

Author contribution

R.C.: conceived and designed the analysis, performed the analysis, and wrote the paper; R.C. and P.S.: collected the data and contributed data or analysis tools; R.C., P.S., and Y.W.S.: supervision and editing.

Conflicts of interest disclosure

Conflicts of interest relevant to this article was not reported.

Research registration unique identifying number (UIN)

This study is registered in PROSPERO (CRD42023437315).

Guarantor

Chai Hong Rim.

Data sharing statement

Data are available within the article or its supplementary materials.

Supplementary Material

js9-110-1206-s001.docx ^{(31.6KB, docx)}

js9-110-1206-s002.jpg ^{(144.4KB, jpg)}

js9-110-1206-s003.pdf ^{(325.5KB, pdf)}

js9-110-1206-s004.docx ^{(676.3KB, docx)}

js9-110-1206-s005.xlsx ^{(100.2KB, xlsx)}