Preoperative transarterial chemoembolization (TACE) + liver resection versus upfront liver resection for large hepatocellular carcinoma (≥5 cm): a systematic review and meta-analysis

Abstract Background Hepatocellular carcinoma (HCC) accounts for majority of primary liver cancer. Use of preoperative neoadjuvant transarterial chemoembolization (PN-TACE) may result in tumor shrinkage and improve resectability. This study aims to summarize the outcomes of PN-TACE versus upfront liver resection (Up-LR) in large HCC (≥5 cm). Methods PubMed, Embase, The Cochrane Library, and Scopus were systematically searched till September 2022 for studies comparing PN-TACE versus Up-LR. The primary study outcomes were overall survival (OS), disease-free survival (DFS), and recurrence. Our secondary outcomes were postoperative morbidity and mortality. Results There were 12 studies with 15 data sets including 3960 patients (PN-TACE n = 2447, Up-LR n = 1513). Majority (89.5%, n = 1250/1397) of patients had Child’s A liver cirrhosis. Incidence of Child’s B cirrhosis was higher in PN-TACE compared to Up-LR (Odds ratio (OR) 1.69, 95% CI: 1.18, 2.41, p = 0.004). Pooled hazard ratio (HR) for OS showed no significant difference between PN-TACE and Up-LR (HR 0.87, 95% CI: 0.64, 1.18, p = 0.37), but DFS was superior in PN-TACE (HR 0.79, 95% CI: 0.63, 0.99, p = 0.04). Subgroup analysis based on study design failed to show any significant effect in randomized controlled trials (n = 2/15 data sets). However, operating time (mean difference (MD) 31.94 min, 95% CI: 2.42, 61.45, p = 0.03) and blood loss (MD 190.93 ml, 95% CI: 10.22, 317.65, p = 0.04) were higher in PN-TACE. Intrahepatic and extrahepatic recurrence, post-operative morbidity and in-hospital mortality were comparable between PN-TACE and Up-LR. Conclusion In retrospective studies, PN-TACE resulted in superior DFS compared to Up-LR. However, this may be confounded by selection bias.


Introduction
Primary liver cancer is the sixth most common diagnosed cancer and is the third leading cause of mortality worldwide [1].Hepatocellular carcinoma (HCC) accounts for the majority of primary liver cancer [1].Management of HCC requires a multipronged approach, taking into consideration tumor characteristics (e.g.number, size and location of the tumor, proximity to vascular structures and presence of portal vein thrombi), liver function, associated patient co-morbidities, available resources, and Eastern Cooperative Oncology Group (ECOG) performance status [2,3].
In the past decade, about 22 guidelines have been published based on several classification systems [4].Commonly used guidelines such as the Barcelona Clinic Liver Cancer (BCLC) guidelines, the Hong Kong Liver Cancer (HKLC) staging system and the American Association for the Study of Liver Disease (AASLD) guidelines advocate surgery as the mainstay curative treatment for resectable HCC [3,5,6].While large HCC is not an absolute contraindication for surgery, larger tumor size is associated with unfavorable prognostic indicators like macro or microvascular invasion (MVI), and technical difficulties posing the risk of threatened (positive/involved) resection margins [7].
Recent studies have shown satisfactory outcomes for liver resection (LR) in patients with large HCC and Child A liver cirrhosis [8][9][10].However, lower 5year survival and higher recurrence have been reported for tumor size �10 cm [9,10].Transarterial chemoembolization (TACE) induces tumor necrosis through selective hepatic arterial injection of chemotherapeutic drugs [6].Preoperative neoadjuvant TACE (PN-TACE) may devascularise and shrink large and/or huge HCC, thus potentially enhancing survival by reducing tumor recurrence [11].
A meta-analysis by Si et al. [12] in 2016 including 5 randomized controlled trials (RCTs) and 430 patients (PN-TACE group: 212, LR alone group: 218) showed no disease-free survival (DFS) and overall survival (OS) difference between PN-TACE vs LR alone for resectable HCC for tumor sizes <5 cm versus �5 cm.However, more recent studies that explored PN-TACE use in large HCC show promising results for PN-TACE [13].Hence, this study aims to provide an updated meta-analysis on PN-TACE versus upfront LR in large (�5 cm) resectable HCC.Our primary outcomes are OS, DFS, and recurrence.Our secondary outcomes are postoperative morbidity, in-hospital mortality, and intra-operative outcomes.

Study selection and search strategy
This systematic review and meta-analysis were performed according to the PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) guidelines [14].This study was registered at PROSPERO (Ref no: CRD42021286751).A systemic search of the following databases (PubMed, Embase, The Cochrane Library and Scopus) was conducted for studies published from inception to 22 September 2022.A combination of the following search terms was used: 'hepatocellular carcinoma', 'large', 'huge', 'transarterial chemoembolization', 'liver resection' and 'hepatectomy'.The search was restricted to title, abstract and keywords.The complete search strategy is appended in Supplementary Table S1.Search strategies for other databases were modified accordingly from the initial search strategy based on the database requirements.
Included studies were randomized controlled trials (RCTs) and non-RCTs (cohort study and casecontrol study) on adult population (�18 years old) with large (� 5 cm) resectable HCC who underwent either PN-TACE and LR, versus upfront LR.LR was defined as anatomical or non-anatomical wedge resection, segmentectomy or major hepatectomy.Exclusion criteria were studies (1) irrelevant to our study question, (2) single-arm studies, (3) use of other preoperative interventions other than TACE (such as transarterial radioembolization, portal vein embolization), (4) comparing adjuvant/postoperative TACE with LR versus LR alone, (5) comparing TACE only versus LR alone, (6) studies where HCC was < 5 cm, or had HCC > 5 cm but did not perform subgroup analysis for HCC > 5 cm, (7) other types of liver LR (liver transplantation, Associating Liver Partition and Portal vein Ligation for Staged hepatectomy (ALPPS)) and (8) based on article type (non-English studies, conference abstracts, studies which have incomplete information, case report or series, editorials, expert opinions and review articles without original data).
All cross-references were screened for potentially relevant studies not identified by the initial literature search.After removal of duplicates, abstracts were screened for potential inclusion by two independent authors (WXT, FYC).Full texts of included studies were reviewed in entirety and selected based on the inclusion and exclusion criteria.If two or more studies reported on the same patient cohort, only the latest study was included in the meta-analysis.Discrepancies were solved by discussion with the senior author (VGS).

Data extraction
Data extraction was performed by three authors independently (KSC, WXT, FYC).The following variables were extracted from each study: publication details (name of first author, publication year and country), study characteristics (sample size, sex, age, Child-Pugh score, alpha-fetoprotein (AFP) levels, largest tumor size, number of lesions, mean number of TACE sessions, TACE regime and time from the last TACE to LR), presence of portal vein invasion (PVI) or MVI, long-term oncological outcomes (OS and DFS, intrahepatic recurrence and extrahepatic recurrence), short-term outcomes (in-hospital mortality, overall morbidity, post-hepatectomy liver failure (PHLF), bile leak, and wound infection) and intra-operative outcomes (operating time, estimated blood loss (EBL) and need for blood transfusion).For this study, upfront LR without PN-TACE (control group) will be referred to as Up-LR.AFP level was defined as the AFP recorded prior to the first TACE (for TACE þ LR group), and AFP recorded prior to LR (for LR group).Major LR was defined as resection of �3 segments.Non-anatomical resections were excluded from classification under major/minor LR, unless the number of segments were specified in the respective studies.

Assessment of study quality
Two authors (WXT, FYC) independently performed a quality assessment of the finalized studies.RCTs were assessed using the Revised Cochrane risk-ofbias tool for randomized trials (RoB 2.0) tool (Supplementary Table S2) [15].Observational studies were assessed using the modified Newcastle-Ottawa scale (Supplementary Table S3) [16].Disagreements were resolved by discussion with the senior author (VGS).

Statistical analysis
Study variables were extracted to Microsoft V R Excel 365 (Microsoft V R , Washington, USA).Mean and standard deviation of time from the last TACE to LR were estimated from median and range values using methods described by Wan et al. [17].Metaanalysis was performed using RevMan 5.4 (Review Manager 5.4,The Nordic Cochrane Centre, Copenhagen, Denmark).For cumulative OS and DFS, hazard ratio (HR) and standard error (SE) was estimated indirectly according to the methods described by Parmar et al. [18].Pooled HR was calculated through the inverse-variance method using the natural logarithm of HR (ln(HR)) and SE [19].Outcomes with dichotomous outcomes were expressed as odds ratios (ORs) with 95% confidence interval (CI), which were calculated using the Mantel-Haenszel method.Outcomes with continuous outcomes were expressed as mean difference (MD) with 95% CI using the inverse-variance method.The level of statistical significance was set at p < 0.05.Heterogeneity was assessed using Cochrane's Q and quantified by I 2 .Heterogeneity was defined by I 2 >50%.A random-effect model was used when I 2 >50%, and a fixed-effect model was used when I 2 �50%.Statistical significance was defined as p < 0.05.Publication bias was investigated using funnel plots (Supplementary Figure S1) [20].Subgroup analysis was not performed for HCC �10cm as there were only two studies that performed analysis for HCC �10cm [13,21].Sensitivity analysis was performed using Stata (version 17.0, StataCorp).

Results
The systematic search identified 4070 articles from the four databases.Existing meta-analyses by Zhou et al. [22] in 2013, Si et al. [12] in 2016, and Mi et al. [23] in 2022 on the use of PN-TACE in HCC were also screened for potential references for inclusion in our study, with additional 18 articles included for screening.Titles and abstracts of all the identified articles were screened and 42 full-text articles were reviewed for inclusion in the study.A total of 12 articles were included in the final analysis [13,21,[24][25][26][27][28][29][30][31][32][33], with a total of 15 data sets (3 data sets by Wang et al. [30] in 2022, and 2 data sets by Zhang et al. [32] in 2022).Figure 1 shows the PRISMA diagram for the study selection process.One study used propensity score matching (PSM) analysis [13]; only the PSM sample was analyzed in our study.Two studies analysed patients with HCC �10 cm [13,21].

Study characteristics
There were 15 data sets including 3960 patients (PN-TACE n ¼ 2447, Up-LR n ¼ 1513).There were 2 RCTs with 160 patients (PN-TACE n ¼ 76, Up-LR n ¼ 84) [21,33], 1 prospective study with 108 patients (PN-TACE n ¼ 54, Up-LR n ¼ 54) [26], and 12 retrospective data sets with 3692 patients (PN-TACE n ¼ 2317, Up-LR n ¼ 1375) [13,24,25,[27][28][29][30][31][32], of which, one study by Li et al. in 2019 used PSM analysis [13].Majority of patients included in the studies had Child-Pugh A liver cirrhosis (n ¼ 1250/1397, 89.5%).Mean tumor size ranged from 7.8 to 14.3 cm in the PN-TACE group and 7.3 to 14.5 cm in the Up-LR group.The mean time from the last TACE to LR ranged from 3.7 weeks to 14.9 weeks.There were only two studies which described the number of patients who failed to complete surgery in the PN-TACE group [21,33]; in the study by Wu et al. [21], all of the patients (n ¼ 24) completed surgery, while 91.0%(n ¼ 47/52) of the patients in the study by Zhou et al. completed surgery and were analyzed using intention-to-treat analysis [33].The remaining ten studies did not explicitly state whether patients who failed to complete surgery in the PN-TACE group were included or excluded in the final analysis [13,[24][25][26][27][28][29][30][31][32].Table 1 summarizes the baseline characteristics and patient demographics of the included data sets.Table 2 summarizes the clinical characteristics and details of treatment received in the included data sets.

Patient demographics and histopathological findings
Patient demographics of those who underwent PN-TACE were comparable with those with Up-LR in terms of age, sex, tumor size and pre-operative AFP value (�400 ng/mL) (Table 3).Number of patients who underwent major LR were also similar between the two groups (Table 3).Incidence of Child's B cirrhosis was higher in PN-TACE compared to Up-LR (OR 1.69, 95% CI: 1.18, 2.41, p ¼ 0.004).There were 4.6% (n ¼ 152/3316) with PVI and 54.6% (n ¼ 435/797) with MVI.The median incidence of complete tumor necrosis was 7.1% (n ¼ 4 studies, range 1.1-13.6%).One study reported 51.8% of patients with �70% tumor necrosis [28], while another study reported 45.2% with �75% tumor necrosis respectively [31].Incidence of PVI and MVI was comparable between PN-TACE and Up-LR.Comparison of patient demographics and histopathological findings between PN-TACE and Up-LR is shown in Table 3.

Intra-operative outcomes
Operating time (Figure 5

Discussion
Our study demonstrated that DFS for PN-TACE is superior compared to Up-LR in patients with large HCC.However, operating time and EBL are higher in PN-TACE compared to Up-LR.There was also a trend towards higher overall morbidity in PN-TACE, though this did not reach statistical significance.Patient demographics, OS, recurrence, and other post-operative outcomes were similar between PN-TACE and Up-LR.The BCLC staging system incorporates functional status, resectability of a lesion, and function of the liver remnant (Child-Pugh status) to guide treatment options [2].The updated BCLC 2022 classification system recommends TACE as first-line potentially curative treatment for BCLC-B HCC with preserved portal flow [5, 34,35].However, a metaanalysis by Hyun et al. [36] on 18 studies with 6508 patients with BCLC stage B and C HCC showed significantly better OS (HR: 0.59, 95% CI: 0.51-0.67,p < 0.01, I 2 : 84%) and 5-year survival (OR 2.20, 95% CI: 1.52-3.20,p < 0.001, I 2 : 53%) in LR compared to TACE.This raises the question on whether combination of PN-TACE and LR may improve oncological outcomes.
A recent meta-analysis by Mi et al. [23] in 2022 on 29 studies with 22,023 patients (2 RCTs and 27 cohort studies) showed superior DFS in PN-TACE compared to upfront LR (HR 0.80, 95% CI 0.73, 0.88, p < 0.001) in the overall cohort; superior OS was also demonstrated in patients with BCLC stage B HCC (HR 0.76, 95% CI: 0.60, 0.96, p ¼ 0.024).However, size of HCC was heterogenous (mean tumor size ranging 3.7-12.7 cm).Superior survival outcomes shown in their study may not be applicable to subgroups of patients with large HCC.Large HCC is associated with higher incidence of MVI and moderate/poorly differentiated HCC [7].HCC � 5 cm is also associated with inferior 5-year survival as compared to HCC < 5 cm (56.7% vs 86.8%) [37].Hence, our review should not be seen as a duplicate of previous meta-analyses [12,22,23].It is important to recognise the limitations of existing meta-analyses and raise the need for a metaanalysis on patients with large HCC only.
Compared to the study by Mi et al. [23], our study similarly showed superior DFS for PN-TACE (however, only in retrospective studies).Improved DFS following PN-TACE may be due to the almost exclusive blood supply of primary HCC through the hepatic arteries.Use of TACE may shrink the tumor, induce tumor necrosis, improve resection margins and reduce recurrence [6,11,38].Patients who had downstaging of HCC had clinically better 5-year DFS (29%) compared to those with no response to TACE (10%, p ¼ 0.08) or did not receive TACE (11%, p ¼ 0.10) [38].Patients who had previously unresectable tumors could also be resected after downstaging [38].The potential benefits conferred by PN-TACE is therefore an important area of exploration to improve OS and DFS, especially for large tumors (� 5 cm).However, our review showed median complete necrosis rate of only 7.1%.It is possible that only massive necrosis, but not complete necrosis (i.e.100%) may be necessary for improved survival outcomes.For instance, the study by Murakami et al. [28] had 51.8% of large HCC with �70% tumor necrosis, of which superior 5-year OS and DFS was noted compared to those with <70% tumor necrosis.
After subgroup analysis, our study showed that superior DFS was only shown in retrospective cohort studies, but not in RCTs (Figure 2  level, PVI and MVI were comparable between PN-TACE and Up-LR (Table 3) [7,41].The lack of statistical significance in subgroup analysis for RCTs and the prospective study may be due to small sample size instead.Another pertinent finding is the nonsuperiority of OS for PN-TACE compared to Up-LR.A plausible explanation for lack of improvement in OS may be due to higher incidence of Child's B cirrhosis in the PN-TACE group (OR 1.69, 95% CI: 1.18, 2.41, p ¼ 0.004).Unlike Child's A cirrhosis which represent well-compensated disease, Child's B cirrhosis signifies significant functional compromise with 1year survival of 80% (compared to 100% survival in Child's A cirrhosis) [42].
The idea of neoadjuvant treatment for downstaging and reducing tumour size to improve resectability of large HCC is an attractive idea.Apart from TACE, other treatment options which were previously used for locally advanced HCC have been explored on their use as neoadjuvant therapies.Transarterial radioembolization (TARE) have been shown to be feasible option prior to liver resection or orthotopic liver transplantation, with 5-year overall survival rate of 86% [43].Use of pre-operative systemic therapy with sorafenib has also been explored; initial results show good safety profile, but long-term survival outcomes have yet to be explored [44].A recent phase 3 trial demonstrated better OS and DFS in patients with unresectable HCC treated with atezolizumab plus bevacizumab [45].Further research should explore the use of the above discussed options as neoadjuvant therapy to downstage large HCC and improve survival outcomes.
Though the use of TACE has been demonstrated to be safe, side effects of TACE include pain, nausea, vomiting, lethargy, and fever.Rarely, TACE may be complicated by liver abscess, acute cholecystitis, post embolization syndrome, and liver failure [46].Ahmed et al. [47] conducted a systematic review and reported that HCC patients treated with palliative intent by TACE have reduced quality of life in the early post-treatment period.A concern with PN-TACE is the resultant liver inflammation causing increased intra-operative difficulties [29].We showed that patients who had PN-TACE required 32 min longer for LR (p ¼ 0.03) and had higher EBL (mean difference 190 mL, p ¼ 0.04), which may be due to adhesions from PN-TACE.It has been reported that a long interval between the last TACE and LR will have little influence on subsequent LR [48].A PSM study by Zhou et al. [49] which investigated on patients with intermediate-stage HCC who underwent PN-TACE with LR at least 4 weeks after the last TACE similarly showed no statistically significant increase in intraoperative blood loss (PN-TACE: 673.3 ± 677.1 mL, Up-LR: 571.2 ± 637.2 mL, p ¼ 0.13) or blood transfusion requirements.The mean time interval from last TACE to LR was 3.7-14.9weeks in our study.However, a plausible reason for increased operating time and EBL may be due to higher incidence of Child's B liver cirrhosis, where patients may have thrombocytopenia and/or coagulopathy.Additionally, patients who received PN-TACE also had higher overall morbidity, though this did not reach statistical significance (OR 1.33, 95% CI: 0.96, 1.86, p ¼ 0.09).This may similarly be due to resultant inflammation from PN-TACE as described above, and lack of statistical significance may be due to a small sample size [50].Further prospective studies will need to be conducted to investigate the intra-operative and post-operative outcomes of PN-TACE.
It has been postulated that postoperative mortality is higher for patients who received PN-TACE due to intraoperative difficulties [29].Our study did not show any significant difference in in-hospital mortality.Uchida et al. [51] similarly reported comparable in-hospital mortality (PN-TACE: n ¼ 5/60 (8.3%) versus upfront LR: n ¼ 4/68 (5.9%), p > 0.05).In addition, our study demonstrated similar incidence of PHLF following LR in both PN-TACE and without PN-TACE.TACE induces tumor cell necrosis through hypoxemia and ischemia of locally embolised tissue, releasing large amounts of free radicals and cell toxins [52,53].This may accelerate the apoptosis process of normal hepatocytes resulting in the injury of normal hepatocytes in non-neoplastic areas of the liver [54,55].This was demonstrated by Miksad et al. [56] whom analysed 572 patients; they showed deterioration of bilirubin in acute (0-29 days) and chronic (30-90 days) period in 30% and 23% of patients following a single TACE.It is possible that perioperative steroid use may modulate the inflammatory response and dampen the hyperbilirubinemia or increased morbidity associated with liver resection [57].However, the authors did not report PHLF separately.PHLF is variably defined and based on the criteria used, various authors have reported predictive factors.In a study including 244 HCC patients, Lei et al. [58] reported that 50-50 criteria on POD 5 and peak serum bilirubin >119 mmol/L (p ¼ 0.007) on POD 3 predicted 90-day mortality.In the studies included in our review, Li et al. [13] defined PHLF as bilirubin >100 mmol/L with prolonged clotting time >6s, or presence of hepatic encephalopathy, while Wu et al. [21] defined PHLF as transient Child's C status on POD 7. Standardized definitions are required in future trials to determine the true incidence of PHLF between patients who receive PN-TACE versus Up-LR.
The strength of our study is that this is the first meta-analysis on PN-TACE þ LR versus Up-LR in large HCC alone.This study had strict inclusion criteria and excluded small size HCC from PN-TACE unlike previous meta-analyses [12,22,23].While in most oncologic situations, DFS is a surrogate of OS, in HCC patients, it is not the case.As survival is not only determined by oncology, but also the underlying liver (dys)function, it is likely that a larger sample is necessary for an adequately powered study to evaluate the difference in OS.While our results were not statistically significant (p ¼ 0.37), there is a relative difference of 13% improvement in OS, which is clinically significant.Further studies are therefore necessary to confirm this finding.
Our study has its limitations.Majority of the included studies were retrospective observational studies with inherent selection bias.We assessed the study qualities by using the modified Newcastle-Ottawa scale and all studies were at least of moderate quality.There was also significant heterogeneity in included studies, and the experience of multidisciplinary oncology teams may affect outcomes.Most of the studies were conducted in Asia, which may limit the generalizability of our results.This is, however, likely due to the epidemiology of HCC, where 75% of liver cancer occurs in Asia [59].We also did not perform subgroup analysis for HCC � 10cm because of the low number of studies reporting on outcomes.Lastly, this study does not clearly define the subgroup of HCC patients that would benefit from PN-TACE as it is possible that other biological and clinical factors determine the outcomes.For example, multidisciplinary team discussions may determine a patient with large HCC and extensive PVTT to be unsuitable for PN-TACE.

Conclusion
In patients with large HCC, DFS may be superior with PN-TACE compared to Up-LR.This is confounded by the largely retrospective nature of included studies.RCTs alone did not show any superiority in DFS.OS was comparable between PN-TACE and Up-LR, while operating time and EBL were higher with PN-TACE.This may be due to resultant liver inflammation from PN-TACE, but may be confounded by higher number of patients with Child's B liver cirrhosis who received PN-TACE.Further well-designed randomized studies should be conducted to evaluate the survival advantage conferred by PN-TACE in selected large and huge HCC patients.

Figure 1 .
Figure 1.PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) flowchart for study selection.
(A)) and EBL (Figure 5(B)) were significantly longer in PN-TACE compared to Up-LR.Need for blood transfusion was however similar between PN-TACE and Up-LR (Figure 5(C)).
(B)).Outcomes of HCC are dependent on patient factors, disease factors, liver (dys)function, and treatment factors [2,39,40].A possible explanation for superior DFS only in retrospective studies may be due to inherent selection bias.Hence, we evaluated this by additionally comparing patient demographics and tumor characteristics.Confounding factors on survival outcomes such as age, sex, tumor size, AFP

Table 1 .
Baseline characteristics and patient demographics of the included studies (n ¼ 12) with 15 data sets.
All continuous variables are expressed in mean ± standard deviation, or median (interquartile range) unless otherwise stated.All categorical variables are expressed as n (%) unless otherwise stated.aValuesincluded is this study is obtained after propensity score matching.bValuesincluded is this study is only for the subgroup of patients who had HCC >8cm.cValuesincluded is this study is only for the subgroup of patients who had large HCC (data on patients with HCC < 5 cm were excluded).dValuesincludedare representative of the entire cohort of patients with HCC (i.e. for all tumor sizes, where there were 120 patients who underwent PN-TACE þ LR and 153 patients who underwent LR only; there was no specific demographics provided for patients with HCC > 5 cm only).AFP: Alpha-fetoprotein; CTC: Circulating tumor cells; HCC: Hepatocellular carcinoma; NR: Not reported; RCT: Randomized controlled trial; PN-TACE: Preoperative neoadjuvant transarterial chemoembolization; Up-LR: Upfront liver resection.

Table 2 .
Clinical characteristics and details of treatment received in the included studies (n ¼ 12) with 15 data sets.
a Values included is this study is obtained after propensity score matching.b Values included is this study is only for the subgroup of patients who had HCC >8cm.c Values included is this study is only for the subgroup of patients who had large HCC (data on patients with hepatocellular carcinoma < 5 cm were excluded).d Only data on patients with �70% necrosis was reported.e Only data on patients with �75% necrosis was reported.f Mean and standard deviation were calculated from median and range/interquartile range using methods described by Wan et al. [15].g Values included are representative of the entire cohort of patients with HCC (i.e. for all tumor sizes, where there were 120 patients who underwent PN-TACE þ LR and 153 patients who underwent LR only; there was no specific demographics provided for patients with HCC > 5 cm only).DFS: Disease-free survival; HCC: Hepatocellular carcinoma; LR: Liver resection; MVI: Microvascular invasion; NA: Not applicable; NR: Not reported; OS: Overall survival; PV: Portal vein; PN-TACE: Preoperative neoadjuvant transarterial chemoembolization; Up-LR: Upfront liver resection.

Table 3 .
Summary of effect size of different outcome variables between preoperative neoadjuvant transarterial chemoembolization (PN-TACE) and upfront liver resection in the included data sets.Odds ratio and 95% CI was used for dichotomous outcomes, mean difference and 95% CI was used for continuous outcomes, and hazards ratio and 95% CI was used for time-to-event outcomes.Values in bold indicate statistical significance where p is <0.05.