Asia–Pacific clinical practice guidelines on the management of hepatocellular carcinoma: a 2017 update

The most common risk factor for HCC is chronic HBV infection, which accounts for more than 50% of all cases globally and 60–80% in some Asian countries such as China, Korea, and Vietnam [120]. Strategies to prevent HBV-related HCC include universal hepatitis B vaccination to reduce new infection as primary prevention, antiviral treatment to prevent disease progression by effectively suppressing HBV replication and regular surveillance to detect HCC in earlier stage as secondary prevention, and combination of curative therapies and adjuvant antiviral treatment to increase survival and prevent recurrence for HCC patients as tertiary prevention [2, 121, 122].

In chronic HCV infection, a meta-analysis of retrospective studies implies that the risk of HCC is reduced among patients with HCV who achieve SVR with antiviral therapy with interferon or interferon plus ribavirin [139‐142]. Findings on the effect of SVR on liver-related clinical outcomes are similar to those of retrospective, and often smaller, studies from Japan [143‐147], the results of which supported an approximately 70–90% reduction in the risk of liver-related clinical outcomes over a follow-up period of 2–6 years in patients achieving SVR. This was reaffirmed by a recent study on 33,005 HCV-infected individuals who received treatment, whose authors concluded that the risk of HCC after HCV cure, though considerably reduced, remains relatively high at 0.33% per year [148].

According to the HALT-C and EPIC studies, a continued elevated risk of HCC in patients with advanced chronic HCV, even in those who achieved SVR, was evident [139]. The 5-year risk of HCC developing in noncirrhotic patients was 4.8%. It has been suggested that the incidence of HCC in patients with cirrhosis from HCV only increases substantially once the platelet count is less than 100 × 10⁹/L. Furthermore, older age and presence of cirrhosis at the point of SVR are associated with a higher risk of HCC, warranting surveillance [148].

Although these studies have validated that there is a reduced incidence of HCC in treated patients, there are no data that demonstrate that treating or eradicating HCV completely eliminates the risk for HCC. Thus, it seems prudent to continue surveillance of patients with HCV and cirrhosis who have achieved viral clearance on therapy. Surveillance is recommended in SVR patients with any histologic stage of HCV with comorbidities, such as alcohol abuse and DM, all of which are established independent risk factors for HCC disease progression [148‐151].

While the most commonly observed clinical benefits in SVR patients were the consequence of the arrest of fibrosis progression, regression of preexisting cirrhosis could be documented [152‐157]. However, regressed cirrhosis is not a reason to withhold surveillance. The recent new arrival of direct-acting antiviral (DAA) therapy allowed for the achievement of SVR in over 90% of treated patients, irrespective of liver fibrosis stage [158‐160]. However, there is no evidence that successful DAA therapy reduces the incidence of HCC in patients with HCV cirrhosis. Maintaining surveillance for SVR patients with advanced liver fibrosis, independently of the histologic response to therapy, is highly recommended.

The estimated prevalence of NASH in the general population ranges from 2 to 3% [161]. Contrary to earlier studies, recent data revealed that up to 44% of cases of NAFLD can progress to NASH even in the absence of inflammation at baseline [162] and approximately 23% of cases of NASH progress to cirrhosis over the next 10–15 years [163]. In general, up to 30–40% and 10–15% of cases of NASH do have advanced fibrosis and cirrhosis, respectively, at initial diagnosis [163]. Although the overall incidence of HCC depends on the stage of underlying NAFLD and associated comorbid conditions, HCC can develop in the absence of cirrhosis in such cases [164]. The incidence of HCC in patients with NASH cirrhosis has been reported to be 2.3–4.0% per year [165, 166].

Metabolic syndrome and components of metabolic syndrome have been associated with the development of HCC [167]. In a meta-analysis of 38,940 cases of cancers (43 articles), Esposito et al. [168] found an association of metabolic syndrome with HCC (RR = 1.43, p < 0.0001), and the association was stronger in Asians (p = 0.002). Moreover, the estimated risk of HCC was high in overweight (RR = 1.48; 95% CI 1.31–1.67) and obese (RR = 1.83; 95% CI 1.59–2.11) individuals. Tanaka et al. [169] reported that overweight or obesity increased the risk of liver cancer among Japanese population in the meta-analysis. Consistent with these findings, a significant association of DM and HCC was estimated (RR = 2.31; 95% CI 1.87–2.84) by Wang et al. in a meta-analysis of 17 case–control and 32 cohort studies [170]. Longer duration of DM and treatment with insulin or sulfonylureas was also associated with a higher risk of HCC, while a lower risk of HCC was found with metformin treatment. In another meta-analysis of 25 cohort studies, a higher incidence of HCC was found in 17 studies [summary relative risks (SRRs) = 2.01; 95% CI 1.61–2.51] among diabetics compared with nondiabetic patients. However, due to the presence of significant heterogeneity among studies (Q = 136.68, p < 0.001, I ² = 87.6%), a subgroup analysis was performed to control for confounders and DM was associated with a higher HCC-related mortality (SRR = 1.56; 95% CI 1.30–1.87) [51].

Because HBV and HCV are predominant risk factors responsible for a high burden of liver diseases in the Asia–Pacific region, Chen et al. [171] evaluated the influence of obesity, DM and HBV/HCV infections on the risk of HCC. This meta-analysis found that the positive association with obesity was independent of DM or infections with HBV/HCV.

In general, the relationship between DM, obesity, metabolic syndrome, and HCC is linked with development of NAFLD, and theoretically all are interlinked with each other. Considering the significant burden of NASH (i.e., 2–3% of the global population) and a global rise in the burden of obesity and DM, it is expected that there will be a further increase in the burden of NASH and HCC in the foreseeable future unless considerable preventive measures are taken [172]. Most of the studies have used the Western criteria for obesity, and one can expect an even higher risk if similar estimates are calculated using the Asian criteria for obesity. Higher HCC-related mortality rates have been reported in DM compared with non-DM patients (RR = 2.43; 95% CI 1.66–3.55) [170]. Furthermore, due to limited available treatment options, patients with NASH-related cirrhosis carry a significant risk for HCC development. Moreover, many other aspects must be explored; for instance, the role of impaired glucose metabolism, dyslipidemia, and the effect of concomitant use of alcohol must be evaluated [167]. Hence, it is imperative to implement measures to reduce the burden of factors associated with NAFLD/NASH.

So far, the major key preventive measures here include “healthier diet” and lifestyle modification, which should be explained and promoted to every individual who is at risk of or has suffered from such metabolic derangements. Dietary modifications according to underlying risk factors, such as DM, obesity, dyslipidemia, and hypertension, should be promoted. Regular walking and exercise also have a major role in the control of metabolic syndrome and NAFLD/NASH. Treatment of concomitant metabolic conditions with statins and metformin may also have beneficial effects on portal hypertension, complications of liver cirrhosis, and HCC prevention [173]. The efficacy of metformin as a preventive agent in a clinically relevant rat model of HCC was evaluated and was associated with a reduction in fibrotic and inflammatory markers and a 44% decrease in HCC incidence when administered in an early phase by suppressing the receptors for advanced glycation end products and inhibiting activation of hepatic progenitor cells [174]. However, these preclinical findings must be confirmed in clinical studies.

Bariatric surgery could be recommended for patients with morbid obesity, which may reduce liver fibrosis but carries a risk of decompensation in patients with advanced liver cirrhosis [173]. Furthermore, periodic screening for HCC in patients with NASH will help to identify HCC cases at early stage. Patients with NASH cirrhosis should be considered for HCC screening according to the American Association for the Study of Liver Diseases (AASLD)/American College of Gastroenterology (ACG) practice guidelines [175]. The other systemic review of NAFLD recommended biannual imaging screening in cirrhosis patients [176]. However, screening recommendations have not yet accounted for the increasing number of patients suffering from HCC even though up to 50% of cases may occur in the absence of cirrhosis [59]. The latest European guidelines suggested that the PNPLA3 rs738409 C > G gene polymorphism has been associated with an increased HCC risk and might provide patient risk stratification for tailored HCC surveillance in NAFLD. However, no recommendation can currently be made on the timing of surveillance or its cost-effectiveness [177].

A small RCT was performed to evaluate the safety and efficacy of 16 weeks of interferon α-2b therapy after hepatic resection in a group of patients with predominantly HBV-related HCC [181]. The RR of death after interferon treatment was 0.42 (95% CI 0.17–1.05, p = 0.06). Subset analysis showed that adjuvant interferon had no survival benefit for pTNM stage I/II tumors (5-year survival 90% in both groups, p = 0.91), but prevented early recurrence and improved the 5-year survival of patients with stage III/IVA tumors from 24 to 68% (p = 0.04). After this study, another similar RCT was conducted by Chen et al. [182]. A total of 268 patients were allocated randomly to receive either 53 weeks of adjuvant interferon α-2b treatment or observation alone. The primary endpoint of this study was RFS. The median RFS in the interferon α-2b and control arms was 42.2 and 48.6 months, respectively (p = 0.83). In this study, adjuvant interferon α-2b did not reduce the postoperative recurrence of HBV-related HCC. HCC recurrence after ablative treatment modalities is also common. Although patients who received medical ablation usually exhibit compensated hepatic functional status, the frequent recurrence of HCC after successful ablation contributes to short-term survival. A small RCT was conducted to evaluate the effectiveness of interferon therapy in preventing HCC recurrence after successful medical ablation therapy for primary tumors [183]. The cumulative HCC recurrence rate of the patients treated with interferon-α and the control group was 25 and 40% at the end of 1 year, and 47 and 90% at the end of 4 years, respectively (p = 0.01). Furthermore, this study also showed that the prevention of HCC recurrence using interferon-α was effective in HBV-related HCC [183].

A retrospective study was conducted to evaluate the efficacy with or without using nucleos(t)ide analogs in patients following curative treatments for HBV-related HCC [184]. Cumulative OS rates of HCC were significantly different between the two groups (p < 0.01), and cumulative RFS rates of HCC were also significantly different (p < 0.01). Yin et al. [185] also reported that nucleos(t)ide analogs improved not only liver function but recurrence and OS rates. In another study, improvements of liver function and OS rates were reported even if the recurrence rates were not significantly different between groups treated with and without nucleos(t)ide analogs [186, 187]. In addition, a large-scale nationwide cohort study was reported from the Taiwan National Health Insurance Research Database [188]. Among 100,938 newly diagnosed HCC patients, they identified 4569 HBV-related HCC patients who received curative liver resection for HCC. The risk of first tumor recurrence was compared between patients who did not (untreated cohort, n = 4051) and did (treated cohort, n = 518) take nucleos(t)ide analogs. The treated cohort had a higher prevalence of liver cirrhosis when compared with the untreated cohort (48.6 versus 38.7%, p < 0.001), but had a lower risk of HCC recurrence [n = 106 (20.5%) versus n = 1765 (43.6%), p < 0.001] and lower overall risk of death [n = 55 (10.6%) versus n = 1145 (28.3%), p < 0.001]. After adjusting for competing mortality, the treated cohort had a significantly lower 6-year HCC recurrence rate (45.6%; 95% CI 36.5–54.6% versus untreated, 54.6%; 95% CI 52.5–56.6%, p < 0.001). Six-year overall mortality was 29.0% (95% CI 20.0–38.0%) for the treated and 42.4% (95% CI 40.0–44.7%, p < 0.001) for the untreated cohort. On modified Cox regression analysis, administration of nucleos(t)ide analogs (HR 0.67; 95% CI 0.55–0.81, p < 0.001) was independently associated with a reduced risk of HCC recurrence. One study elucidated the superior choice of nucleos(t)ide analogs [189]. A total of 865 HBV-related HCC patients received antiviral treatment at diagnosis or immediately following surgery (adefovir 10 mg per day in 300 patients, entecavir 0.5 mg per day in 325 patients, and lamivudine 100 mg per day in 240 patients). The 1-, 2-, and 3-year resistance rates were 0.9, 1.8, and 2.5%, respectively, for the entecavir group; 3.0, 8.3, and 12.0%, respectively, for the adefovir group; and 21.7, 31.7, and 39.6%, respectively, for the lamivudine group. The 3-year RFS for the entecavir group also differed significantly compared with the adefovir and lamivudine groups (HR 0.810; 95% CI 0.656–0.999, p = 0.049 and HR 0.737; 95% CI 0.591–0.919, p < 0.01). A randomized, placebo-controlled trial by Jang et al. [190] also showed that preemptive lamivudine therapy in patients receiving transarterial chemoembolization (TACE) significantly reduced the incidence of HBV reactivation (p < 0.01), overall hepatitis (p = 0.02), and severe hepatitis (p = 0.035) due to HBV reactivation after repeated TACE. However, prevention of HCC by preemptive lamivudine therapy was not shown because of advanced stage of HCC in patients receiving TACE in that trial [190]. Further prospective RCTs using a larger number of patients are required to assess its role in tertiary prevention of HCC.

Antiviral therapy, such as interferon, might reduce the overall incidence of recurrence by preventing de novo carcinogenesis. Indeed, several small-sized RCTs, performed in Japan or Taiwan, showed that the incidence of recurrence was reduced in HCV-related HCC by interferon therapy subsequent to initial HCC treatment [194, 195]. Other RCTs, also performed in Japan or Taiwan, failed to find a significant delay in the first recurrence with interferon therapy, but the second or third recurrence was significantly reduced especially in sustained responders, and the OS was improved [196‐198]. Another RCT in Italy did not detect effects of interferon therapy on early recurrence but did find an effect for late recurrence: after an interval of more than 2 years, the rate seemed to be reduced among interferon responders [199]. These data are compatible with the hypothesis that de novo carcinogenesis was prevented by successful antiviral therapy. On the other hand, three reports on long-term observation of recurrence after interferon therapy following HCC treatment showed that the recurrence rate in interferon-treated patients decreased over time, suggesting that the growth of residual microscopic tumors had been delayed by interferon (in fact, the two presumed mechanisms are not necessarily mutually exclusive) [200‐202]. Most of these studies used interferon monotherapy and suffered from low sustained response rates because most patients had advanced fibrosis or cirrhosis. Preventive effects of interferon on HCC recurrence have yet to be reevaluated using current, more efficient protocols.

DAA therapies are promising pan-genotypic agents used to eradicate HCV. However, there is no evidence that DAA therapy will prevent HCC recurrence. One study by Reig et al. [203] reported that eradication of HCV with DAAs led to an unexpected HCC recurrence in some cases, and others reported DAAs did not lead to an unexpected HCC recurrence [204‐206], although there is no evidence that SVR after DAA therapy reduces the incidence of recurrence in HCC patients receiving curative treatments. Furthermore, prospective studies are needed.

Hasegawa et al. [207] reported RCT using oral administration of uracil-tegafur after curative hepatic resection but found no beneficial effects on recurrence and a possible adverse effect on OS. Bruix et al. [208] assessed the efficacy and safety of sorafenib versus placebo as adjuvant therapy in patients with HCC after surgical resection or local ablation. It was a double-blind, placebo-controlled study of patients with HCC with complete radiologic response after surgical resection (n = 900) or local ablation (n = 214) at 202 sites (hospitals and research centers) in 28 countries. At final analysis, 464 RFS events had occurred (270 in the placebo group and 194 in the sorafenib group). Median follow-up for RFS was 8.5 months in the sorafenib group and 8.4 months in the placebo group. There was no difference in median RFS between the two groups (33.3 versus 33.7 months, respectively; HR 0.940; 95% CI 0.780–1.134; one-sided p = 0.26) [208]. In 1996, Muto et al. [209] reported that administration of polyprenoic acid, an acyclic retinoid, reduced the recurrence of HCC in RCT. Updated, long-term data were published subsequently [210], postulating that the eradication of premalignant or latent malignant clones was the mechanism of action. However, in a large-scale RCT, the superiority of acyclic retinoid over placebo could not be validated, 600 mg per day was shown to be the optimal dose, and treatment may possibly reduce the recurrence of HCV-related HCC, particularly after 2 years [211]. The investigators concluded that administration of 600 mg per day of acyclic retinoid to patients with HCV-related HCC who have completed curative therapy might improve survival for those classified as having Child–Pugh class A disease, for whom liver function was relatively stable in subanalysis [212]. Other adjuvant treatments have not been shown to prolong RFS (Table 2).

As the prognosis of HCC depends largely on the stage at which the tumor is detected, detection of HCC early in its development is critical to improve the survival of affected patients [218‐220]. Although ultrasonography (US) is the most widely used modality for HCC screening and surveillance, the reported sensitivity of surveillance US is in the range of 40–81% with specificity of 80–100% [221‐225]. According to a recent meta-analysis study, among B-mode US, contrast-enhanced US (CEUS), contrast-enhanced (CT), and gadolinium-enhanced (MRI), B-mode US has the lowest sensitivity and positive predictive value (59.3, 77.4%) while the other three imaging modalities show similar pooled per-lesion sensitivity and positive predictive value (73.6–84.4%, 83.6–89.3%) [226]. Therefore, US is not advocated as a diagnostic test for confirmation due to overlapped imaging features of benign and malignant cirrhotic nodules on US.

Key alterations during hepatocarcinogenesis include angiogenesis, changes in cellularity, the transporters of hepatocytes, and decrease in the number and function of Kupffer cells [227]. Among them, hemodynamic alteration of the nodules, composed of increased arterial flow and decreased portal flow, is the most important change for the diagnosis of HCC [228‐230]. However, B-mode US cannot demonstrate tumor vascularity, and color Doppler imaging and power Doppler imaging have low sensitivity for detecting the microflow in the nodules [231‐235]. CEUS using microbubble contrast agents and low mechanical index (MI) contrast-specific imaging techniques has been proved to be useful for characterizing liver tumors [235, 236]. Moreover, as Sonazoid microbubbles are phagocytosed by Kupffer cells, Kupffer imaging can be achieved [237]. CEUS can provide superior sensitivity to detect arterial hypervascularity and better demonstration of rapid wash-out for non-HCC malignancy and very late wash-out of HCC compared with dynamic CT or dynamic MRI [238‐240]. In addition, CEUS has several other advantages including relative inexpensiveness, no nephrotoxicity of the contrast agents, and no ionizing radiation. In general, CEUS shares many features with dynamic CT and dynamic MRI, but as they are purely intravascular, in cholangiocarcinoma a discordant enhancement pattern is observed on CEUS [236, 238, 241‐243]. The AASLD removed CEUS from their guidelines in part because of the perceived possibility of false-positive HCC diagnosis in patients with intrahepatic cholangiocarcinoma [63, 242], but according to recent studies, wash-out time >55 s identified patients with HCC with the highest level of accuracy (92.7%) while wash-out time ≤55 s correctly identified the vast majority of non-HCC malignancies (diagnostic accuracy 98.3%) [244, 245]. In terms of diagnostic accuracy of CEUS for small HCC, recent meta-analysis studies demonstrated that pooled per-lesion sensitivity and positive predictive value of CEUS are similarly high (84.4 and 89.3%) compared with CT (73.6 and 85.8%) and MRI (77.5 and 83.6%), with better cost-effectiveness than CT or MRI [226, 246, 247]. A comparison of the diagnostic ability for hepatic nodules between CEUS using Sonazoid and contrast-enhanced CT showed that the sensitivity and accuracy were significantly higher for the former (95.4 and 94.7%) than the latter (85.2 and 82.3%) [248]. In real clinical practice, however, given that cirrhotic liver has a limited sonic window for whole-liver evaluation and that there is a strong need for CT or MRI for tumor staging, use of CEUS as a first-line diagnostic approach, albeit possible, may not be more cost-effective than CT or MRI. As of now, it is generally accepted that CEUS is a cost-effective second-line imaging modality for rapid diagnosis of HCC once the liver focal lesion is detected on US, although dynamic CT or dynamic MRI is the gold standard for characterization of small nodules at high risk for HCC in cirrhotic liver in Western guidelines [2, 249].

Once a screening test result is abnormal or there is clinical suspicion of HCC, imaging plays a very important role for diagnosis and staging of this tumor [218, 250‐252]. The radiological stage is used to inform clinical decision-making, optimize treatment strategies, and determine eligibility and priority for LT [252, 253]. As mentioned above, in addition to typical hemodynamic changes of HCC such as increased arterial flow and decreased portal flow, several pathologic changes can occur during development of HCC, including changes in cellularity, the transporters of hepatocytes such as organic anionic transporting polypeptides (OATP), and a decrease in the number and function of Kupffer cells [218, 227, 229, 254‐257]. Accumulating data demonstrate that OATP8 expression level decreases during hepatocarcinogenesis prior to reduction in portal venous flow and prior to complete neoarterialization and to elevation of arterial flow, which may allow higher sensitivity for detection of malignant changes [257, 258]. The most reliable diagnostic tests for HCC diagnosis are quadruple-phase, multidetector CT (MDCT) and dynamic MRI including late hepatic arterial, portal venous, and delayed phase imaging at about 3–5 min after contrast administration [259‐261]. Dynamic CT and dynamic MRI with extracellular gadolinium agents permit diagnosis and staging of HCC based mainly on assessment of vascularity [218, 251]. Presence of arterial enhancement followed by wash-out has sensitivity and specificity of 90 and >95%, respectively, and positive predictive value approximating 100% among the group having high risk for developing HCC, e.g., those with liver cirrhosis [262‐265]. When extracellular agents are used, dynamic CT and dynamic MRI permit diagnosis and staging of HCC based mainly on assessment of vascularity, and the hallmark of HCC on CT or MRI is presence of arterial hyperenhancement (wash-in) followed by wash-out of the tumor in the portal venous and/or delayed phases [224, 266‐268]. Sangiovanni et al. [269] also reported that the sensitivity of contrast-enhanced MDCT, and MRI using extracellular contrast medium for 1–2-cm HCCs was 44 and 44%, with 100% specificity. This may be explained by the fact that its diagnostic hallmark is often unseen in small HCCs (≤2 cm), resulting from incomplete neoangiogenesis [224]. According to several recent meta-analysis studies on the diagnostic accuracy of US, CT, and MRI [225, 226, 270, 271], the per-lesion sensitivity of MRI for nodular HCC of all sizes is 77–100%, while that of CT is 68–91%, and MRI showed at least equivalent or higher per-lesion sensitivity compared with MDCT and therefore could be the preferred imaging modality for diagnosis of HCCs [269, 272‐275]. The per-lesion sensitivity, stratified by size, was 100% for both modalities for nodular HCCs > 2 cm, 44–47% (MRI) and 40–44% (CT) for 1–2 cm HCCs [269, 272, 276], and 29–43% (MRI) and 10–33% (CT) for HCCs < 1 cm [269, 273, 276]. To date, there are insufficient data regarding the specificity of the combined criteria of wash-in and wash-out appearance in subcentimeter cirrhotic nodules for HCC diagnosis on dynamic CT or dynamic MRI.

More recently, cell-specific contrast agents other than nonspecific extracellular gadolinium-based contrast media such as superparamagnetic iron oxide (SPIO) particles or in conjunction with gadolinium-based contrast agents (double contrast) or gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid (Gd-EOB-DTPA) have been shown to be highly sensitive for detection of HCC, particularly for small tumors [218, 270, 277‐285]. Several studies demonstrated that hepatobiliary contrast media, gadoxetate disodium (Gd-EOB-DTPA, Primovist, Bayer Healthcare, Berlin, Germany) and gadobenate dimeglumine (Gd-BOPTA, Multihance, Bracco, Milan, Italy), have higher overall sensitivity than dynamic CT or dynamic MRI using nonspecific gadolinium chelates [218, 270, 281‐285]. A recent meta-analysis study demonstrated that Gd-EOB-DTPA-enhanced MRI showed significantly higher per-lesion sensitivity than MRI performed with other contrast agents (87 versus 74%) [270]. However, it should be noted that approximately 10–20% of HCCs may appear as iso- to hyperintense nodules on hepatobiliary phase (HBP) images [268, 286]. Despite the great advantage of Gd-EOB-DTPA-enhanced MRI for detection of liver malignancies, one possible pitfall of Gd-EOB-DTPA-enhanced MRI arises from absence of its equilibrium phase, which can show better wash-out of HCC than portal phase of dynamic CT or dynamic MRI. Indeed, hypointensity relative to the liver in the transitional phase (1–5 min) of Gd-EOB-DTPA-enhanced MRI may reflect hyperenhancement of liver parenchyma rather than deenhancement of a mass (“pseudo-wash-out”), thereby lowering the specificity for HCC diagnosis [252, 287]. Therefore, to maintain specificity, only portal venous phase “wash-out” should be used for a noninvasive HCC diagnosis, because malignant lesions other than HCCs, such as intrahepatic cholangiocarcinoma as well as hemangioma, can show hypointensity on the transitional phase and/or HBP [224, 288, 289].

Furthermore, diffusion-weighted imaging (DWI) may improve the diagnostic performance of MRI for small HCCs by demonstrating higher cellularity of HCC [290‐292]. However, although hypointensity on the HBP and diffusion restriction could improve the sensitivity for the diagnosis of HCC, these findings are not specific to HCC and can be found in other hepatic tumors [224]. Other ancillary imaging features favor HCC diagnosis, including presence of intralesional fat, mild to modest hyperintensity on T2-weighted images [259, 293, 294], and morphologic findings such as intratumoral hemorrhage, fatty metamorphosis, and nodule-in-nodule architecture [252, 259]. However, great caution is still required when applying these ancillary imaging features for atypically enhancing cirrhotic nodules in order to retain high specificity, facing clinicians with the dilemma of balancing sensitivity and specificity [224].

Although CT hepatic arteriography (CTHA) and CT during arterial portography (CTAP) images have been used as the gold-standard diagnostic method for estimating the malignancy grade based on hemodynamic alteration, this has fallen out of favor in most practice settings except in some countries due to its invasiveness and high false-positive diagnosis rates [229, 295‐297].

Hypovascular nodules associated with liver cirrhosis include low-grade dysplastic nodule (LGDN) or high-grade dysplastic nodules (HGDN), early HCCs, and well-differentiated HCCs [218, 255, 266, 298‐302]. As there are significant overlaps in enhancement patterns on dynamic CT or dynamic MRI [295‐297, 303, 304], the sensitivity of dynamic CT or dynamic MRI in detection of borderline nodules is quite low [301]. When detectable, most borderline lesions have a low–low–low, iso–low–low, or iso–iso–low enhancement pattern compared with adjacent background liver parenchyma on CT or MRI during the hepatic arterial, portal venous, and delayed phases [218, 301, 305, 306]. However, because expression of OATP8 decreases during hepatocarcinogenesis before complete neoarterialization, early HCCs may be more frequently visible on the HBP images of Gd-EOB-DTPA-enhanced MRI as hypointense nodules [289, 296, 297, 300, 307‐313]. HBP hypointensity is a strong predictor of premalignancy or malignancy, and its presence favors HGDN or early HCC over LGDN or cirrhotic nodule [310, 314‐316]. Although several imaging features are reported to be associated with interval progression to hypervascular HCCs, including large (>9–10 mm diameter) nodule size on initial imaging, nodule growth speed, hyperintensity on T2-weighted images or DWI, hyperintensity on pre-T1-weighted imaging, and intratumoral fat components [289, 315, 317, 318], it is quite challenging to differentiate early HCC from HGDN based on MRI findings [315]. More recently, when hypovascular nodules are detected by MDCT and MRI, the guidelines published by the (JSH) recommend use of CEUS using Sonazoid and Gd-EOB-DTPA-enhanced MRI [319]. The guidelines published by the JSH stated that hypovascular nodules that are hypointense in the hepatobiliary phase of Gd-EOB-DTPA-enhanced MRI and hypoechoic in the Kupffer phase of CEUS using Sonazoid can almost always be diagnosed as early HCC even without biopsy. However, it is highly likely that there may be some overlaps between early HCCs and HGDNs on both Gd-EOB-DTPA-enhanced MRI and CEUS using Sonazoid, so these noninvasive diagnostic criteria for hypovascular nodules need to be confirmed with further studies. As of now, those hypovascular nodules showing hypointensity on HBP of Gd-EOB-DTPA-enhanced MRI and decreased uptake in the Kupffer phase of CEUS using Sonazoid require biopsy and pathologic confirmation, as they possess high malignant or premalignant potential.

Tumor markers for HCC are used in diagnosis and treatment evaluation and during follow-up after treatment. The diagnostic performance of tumor markers is evaluated in terms of sensitivity, specificity, and likelihood ratios for positive and negative results (LR+/LR−) [320]. There is an inverse relationship between sensitivity and specificity according to cut-off values. Setting a lower cut-off value increases sensitivity and decreases specificity, and vice versa. A tumor marker with high LR+ is useful in confirming diagnosis, whereas a tumor marker with high LR− is useful in exclusive diagnosis. Those likelihood ratios also change according to cut-off values. The serum level of a tumor marker usually increases as the total tumor volume increases. This fact indicates that the sensitivity of a tumor marker essentially decreases as the target tumor size gets smaller when the cut-off value is fixed.

Since surveillance with alpha-fetoprotein (AFP) alone is only acceptable in a population-based setting and not recommended for high-risk population for HCC [321], the optimal cut-off value of AFP for surveillance should be determined on the premise that it is examined simultaneously with US. In such a situation, lower cut-off value increases the frequency of recall procedures and subsequent negative results and decreases the efficiency of the program.

Combination of two or more tumor markers may contribute to increased sensitivity without decreasing specificity when the correlation among them is small enough. However, to date the efficacy of adding another tumor maker to a surveillance program with US and AFP has not been fully assessed, especially in terms of cost-effectiveness.

This section of the guidelines is markedly revised from the APASL 2010 guidelines [4]. Various studies have verified the usefulness of Gd-EOB-DTPA-enhanced MRI for diagnosis of HCC [268, 286, 296, 297, 300, 311‐313, 317, 318, 344‐363], although this method is not yet included in the AALSD or European Association for Study of the Liver (EASL) guidelines [2, 63, 364, 365]. Only the updated APASL diagnostic algorithm includes Gd-EOB-DTPA-enhanced MRI as a first-line diagnostic tool for HCC, similar to the JSH-LCSGJ guideline [366].

Many institutions use US to screen for HCC, followed by dynamic CT or dynamic MRI for subsequent examinations. When a lesion is intensely enhanced in the arterial phase and shows hypoenhancement in the equilibrium phase by dynamic CT or transitional phase by Gd-EOB-DTPA-enhanced MRI, a diagnosis of HCC is unproblematic; however, benign hypervascular lesions (such as high-flow-type hemangioma), cholangiocarcinoma or combined HCC must be ruled out. When the hepatobiliary phase of Gd-EOB-DTPA-enhanced MRI or the Kupffer phase of CEUS using Sonazoid confirms a defect in these hypervascular nodules, the lesion is diagnosed as HCC.

When a lesion shows low attenuation in the equilibrium phase of dynamic CT, even though it is not intensely enhanced during the early arterial phase, it is possible that a more sensitive tool may diagnose it as hypervascular HCC; thus, either Gd-EOB-DTPA-enhanced MRI or CEUS is necessary. Gd-EOB-DTPA-enhanced MRI is useful for differentiating HCC (even early HCC) from a dysplastic nodule (DN) [299, 309, 367, 368].

Among the nodular lesions associated with liver cirrhosis, LGDN and HGDN (both of which are considered to be precancerous lesions), early HCC, and nodule-in-nodule liver cancer are regarded as nonhypervascular [299, 309, 367, 368]. The most sensitive modalities that can objectively depict the early carcinogenic process are (1) Gd-EOB-DTPA-enhanced MRI, followed by (2) CTAP/CTHA [369, 370], and (3) CEUS [237, 362, 371]. Portal blood flow may be maintained in some cases of DN and early HCC, but is reduced in other nodules, although arterial blood flow in cases of DN and early HCC will not have increased yet.

The hepatobiliary phase of Gd-EOB-DTPA-enhanced MRI can detect the earliest initial changes suggestive of HCC. The second earliest initial carcinogenic changes are detected by CTAP and the third earliest by CTHA or CEUS (an increase in intranodular arterial blood flow). However, because CTHA and CTAP are invasive tests, they are only performed in a few countries. Indeed, they are not common in the majority of countries in the Asia–Pacific region. Since Gd-EOB-DTPA-enhanced MRI can identify initial carcinogenic changes earlier than CTHA and CTAP [258, 357, 372], the latter have been almost completely replaced by Gd-EOB-DTPA-enhanced MRI. Hypervascular lesions depicted as nodule-in-nodule or as entire hypervascular nodules can be interpreted as advanced cancer, even though they are very small (<2 cm).

Dynamic CT and Gd-EOB-DTPA-enhanced MRI show high sensitivity for arterial blood flow, but cannot detect arterial vascularity in some nodules (detection depends on acquisition timing, tumor location, and liver function), even though lesions appear hypervascular on CEUS. Nodules showing intense enhancement on dynamic CT and Gd-EOB-DTPA-enhanced MRI are assumed to exhibit high intensity on T2-weighted images and DWIs of MRI.

It is recommended that institutions specializing in liver cancer use Gd-EOB-DTPA-enhanced MRI rather than dynamic CT, even when no tumor is detected on US. Institutions that cannot perform Gd-EOB-DTPA-enhanced MRI as the first-line modality may use dynamic CT as a first screening/diagnostic step, even when no nodule is evident on US; however, it is absolutely essential that Gd-EOB-DTPA-enhanced MRI or CEUS be performed when dynamic CT does not identify hallmarks of HCC (i.e., arterial enhancement with venous wash-out) in the detected nodule.

If Gd-EOB-DTPA-enhanced MRI (or dynamic CT) identifies a hypervascular nodule with venous wash-out, a definitive diagnosis of HCC can be made. If Gd-EOB-DTPA-enhanced MRI (or dynamic CT) shows a hypervascular nodule without venous wash-out, a diagnosis of HCC can be made if the nodule shows hypointensity in the hepatobiliary phase of Gd-EOB-DTPA-enhanced MRI. Also, in this case, another modality or MRI sequence should be used to rule out high-flow-type hemangioma, because the latter can exhibit characteristics similar to HCC. If the hepatobiliary phase of Gd-EOB-DTPA-enhanced MRI identifies the nodule as isointense or hyperintense, biopsy is necessary to confirm the diagnosis (Fig. 1a).

Isointense or hyperintense nonhypervascular nodules in the hepatobiliary phase of Gd-EOB-DTPA-enhanced MRI can enter the routine surveillance protocol. However, nonhypervascular hypointense nodules have high potential for malignant transformation [296, 297, 311‐313, 317, 318, 345, 346, 351, 355, 356, 373‐379], therefore CEUS study using Sonazoid is highly recommended. HCC can be correctly diagnosed by CEUS if hypervascularity and/or a defect in the Kupffer phase is observed. Even when a nodule is hypovascular on CEUS with an evident defect in the Kupffer phase, a finding of hypointensity in the hepatobiliary phase of Gd-EOB-DTPA-enhanced MRI is highly suggestive of malignancy [300]. Accordingly, biopsy is recommended for nodules 1.0 cm or larger to make a differential diagnosis between early HCC and a DN. If a nodule is diagnosed as a DN or a borderline lesion, intensive follow-up (every 3–6 months) with Gd-EOB-DTPA-enhanced MRI (or dynamic CT) is recommended. Intensive follow-up is also recommended for nodules smaller than 1.0 cm (Fig. 1a). In field practice, multiple imaging modalities are not available at all institutions. Thus, a diagnostic algorithm based on using only dynamic CT (or MRI) is shown in Fig. 1b.

Surveillance is continuous monitoring for disease occurrence and includes application of a diagnostic test in subjects who are predisposed to develop a given disease. The primary motive of a surveillance strategy is to achieve a reduction in disease-related mortality through prompt diagnosis (stage migration), which could, in turn, increase the cost-effectiveness and applicability of certain curative therapies. To consider an intervention effective, it must result in an increase in longevity of approximately 90 days, and if this goal can be attained at a cost of less than approximately US $50,000 per year of life gained, it can be deemed cost-effective [380].

US is widely used for surveillance of HCC; its widespread popularity is due to its potential advantages of being noninvasive, an absence of risks associated with the procedure, and good acceptance by patients at a relatively moderate cost. A meta-analysis that included 19 studies showed US to be less effective in detecting early-stage HCC (demonstrating sensitivity of only 63%). However, it could detect the vast majority of HCCs before the disease would present clinically (depicting pooled sensitivity of approximately 94%) [223]. In a study by Sato et al. [381] including 1431 patients with chronic HCV, US-based surveillance performed by trained operators resulted in early detection of HCC with average tumor size of 1.6 ± 0.6 cm and only 1.4% of cases exceeded tumor size of 30 mm. Thus, it was suggested that US-based surveillance performed biannually was adequate for early detection of HCC at size smaller than 3 cm [381]. The performance of US in an HCC surveillance strategy strongly depends on the quality of the equipment and the expertise of the performing operator. Thus, special training is warranted for ultrasonographers.

Existing evidence does not support routine use of CT scan as part of the surveillance strategy for HCC. Patients with a ≥1 cm nodule in the liver are recommended to undergo a contrast-enhanced CT scan of the abdomen as a confirmatory test for the diagnosis of HCC, including unenhanced, arterial, venous, and delayed phase imaging. A randomized trial in 2013 examined 163 patients who had compensated cirrhosis. These patients were tested with either annual CT plus biannual AFP measurements or biannual US plus serum AFP measurements. The combination of biannual US and AFP was marginally more sensitive at detecting HCC compared with annual CT (sensitivity and specificity of 71.4 and 97.5%, respectively, versus 66.7 and 94.4%, respectively). This approach was deemed more cost-effective as well [222].

Similar to the recommendations made for CT scanning of the abdomen, no existing evidence is available to recommend use of abdominal MRI as part of a routine surveillance strategy for detecting HCC. For patients with a ≥1 cm nodule in the liver, dynamic imaging, such as contrast-enhanced MRI of the abdomen, is often needed as a confirmatory test.

Measurement of serum AFP levels is a commonly used strategy for surveillance of HCC because it is widely available, inexpensive, and easy to perform. However, AFP has suboptimal performance as a serological test for surveillance of HCC because it depicts fluctuating levels in patients with cirrhosis with a flare of HCV or HBV infection, in exacerbations of the underlying liver disease, or with the occurrence of HCC [382]. Abnormal serum AFP levels can be detected in only a meager proportion of early-stage HCC tumors (10–20%), which has been correlated now with a particular subtype of HCC depicting an aggressive behavior (S2 class, EpCAM positive) [383]. AFP level of 10.7 ng/ml showed the best combination of specificity (78.1%) and sensitivity (77.2%), a cutoff that approaches the routine limits of normalcy [384].

Other serum markers, such as (DCP), α-fucosidase, AFP-L3%, and GPC3, are used predominantly in the diagnostic rather than surveillance setting. The presence of elevated AFP-L3% is correlated with an HCC tumor with shorter doubling time, and raised serum DCP levels might be indicative of microinvasion [327, 385]. The HALT-C trial studied the AFP and DCP levels of 39 patients with HCC at diagnosis and 1 year before diagnosis. Neither test alone, nor the combination of the two, was adequate for HCC surveillance because the sensitivity of these two markers was very low when they were used either alone or in combination for the strategy to be considered efficacious and cost-effective in detecting HCC at an early stage [386]. Thus, at present, other than AFP, none of these markers can be recommended routinely as part of a surveillance strategy in patients at risk for HCC [327, 387].

Conflicting results have been obtained in studies regarding combination of imaging modalities with serum biomarkers for surveillance of HCC.

The pooled data of a meta-analysis that included 19 studies revealed the combination of US and serum AFP measurement versus US alone to be less specific, no better at detecting subclinical and early-stage HCC, and also not cost-effective. Although the combination of US and serum AFP resulted in marginally increased sensitivity of 69% compared with 63% for US, this result was not statistically significant [209].

In contrast, a recent study demonstrated that serum AFP at cut-off of 20 ng/ml had specificity and sensitivity of 93.3 and 52.9%, respectively, whereas US had specificity and sensitivity of 74.2 and 92.0%, respectively. A combination of US and AFP demonstrated specificity and sensitivity of 68.3 and 99.2%, respectively. It was shown that, when using a cutoff level at 20 ng/ml and AFP level increase of ≥2 times from its nadir in the past 12 months, the combination of AFP and US depicted improved specificity of 71.5% and sensitivity of 99.2% [388].

The benefit of surveillance was demonstrated in a subset of patients with chronic HBV by Zhang et al. [389], in which biannual US and serum AFP measurement decreased mortality from HCC by 37%. Compliance with scheduled tests was depicted to be approximately 58.2%.

Several reports indicated the cost-effectiveness of HCC surveillance, and that US combined with AFP has been shown to increase quality-adjusted life years in patients who suffered from HCC, especially those who underwent resection or transplantation [390, 391]. The cost-effectiveness of HCC surveillance depends on the potential of receiving curative therapy in high-risk patients. Thus, if patients are ineligible for treatments due to severe liver disease or other comorbidities, HCC surveillance is not necessary.

Studies depicting cost-effectiveness suggest that an incidence of HCC of ≥1.5% per year would require implementing a surveillance strategy in patients with cirrhosis [395], which would be irrespective of the etiology involved [396, 397]. The presence of late-stage decompensated cirrhosis (Child–Pugh class C) prohibits use of potentially curative therapies, thus implementing surveillance strategies may not be a cost-effective approach in this subset of patients [398]. Cost-effectiveness of HCC surveillance depends on the potential of receiving curative therapy in high-risk patients. Thus, if patients are ineligible for treatments due to severe liver disease or other comorbidities, HCC surveillance is not necessary. An exception to this is patients who are on a wait list for liver transplantation, who should undergo screening for HCC regardless of their liver functional status, because detecting tumors that exceed the conventional criteria may help formulate priority policies for liver transplantation. A recent Danish nationwide cohort study of patients suffering from alcohol-related cirrhosis of the liver demonstrated that the 5-year cumulative risk of HCC was only 1.0% [399]. Thus, it was suggested that a surveillance strategy in this subset of patients might not prove to be cost-effective. However, further studies are needed to verify these findings.

Patients with chronic HBV infection are also prone to HCC development in the noncirrhotic stage. The cut-off for the annual incidence of HCC is still ill defined in this subset of patients, although opinions from expert groups suggest that surveillance strategies are needed if the incidence of HCC is at least 0.2% per year [63]. The incidence of HCC developing in adult African or Asian active chronic HBV carriers or those having a history of HCC in the family exceeds this value, and Asian patients having high HBV-DNA levels (>10,000 copies/mL) in serum are linked to a yearly risk of more than 0.2%/year [127].

A recent study by Lok et al. [400] showed that HCC can occur in noncirrhotic patients with chronic HCV who suffer advanced fibrosis (METAVIR F3). Because the transition to cirrhosis from advanced fibrosis cannot be determined accurately, patients with chronic HCV with bridging fibrosis can be considered for surveillance; however, further data are needed before making this recommendation. Noninvasive methods to ascertain liver fibrosis, such as transient elastography (TE), appear to be novel tools to stratify patients at different HCC risks [401]. On the other hand, HCV-infected patients without cirrhosis remain at risk for HCC even after achieving SVR. Fibrotic stage (F2 or 3), old age, gamma-glutamyl transferase (γGT) levels, and DM carry high risk of HCC occurrence in noncirrhotic patients, and these patients should be followed carefully for HCC after SVR [402, 403].

Patients with NAFLD who do not have underlying cirrhosis might also benefit from surveillance strategies, because emerging evidence suggests an increased risk of HCC development in this subset of patients [404]; however, more data on this aspect are needed before this strategy is recommended routinely [63]. Groups for whom HCC surveillance is uncertain are shown in Table 4.

Patients who achieve sustained HBV-DNA suppression or HBeAg seroconversion in chronic HBV and SVR in chronic HCV have increased; however, those treatments do not eliminate the risk of HCC completely [405, 406]. Thus, surveillance can be offered to treated patients with chronic HCV who have advanced fibrosis or cirrhosis even after achieving SVR and also to patients with chronic HBV who remain at risk of HCC due to various baseline factors.

The optimal surgical strategy for HCC has been controversial so far, as indicated by the great difference in the indication for liver resection (LR) and LT for HCC among major algorithms worldwide [407]. When considering LR for HCC, the extent of radical resection to remove the tumor, as well as the functional reserve of the diseased liver and the volume of the future liver remnant, must be taken into account. LT is now an established surgical treatment for HCC patients. In contrast to LR, there is no restriction for the indication of LT, at least in terms of liver function, and LT, which could potentially cure both the diseased liver and HCC, is superior to any other conventional therapeutic options from an oncologic point of view. It is now a matter of debate how best to select those to be offered LR or LT among HCC patients [408]. This section summarizes the current opinions regarding LR and LT for HCC.

Image-guided percutaneous ablation therapies include ethanol injection [449‐451], microwave ablation (MWA) [452], radiofrequency ablation (RFA) [453‐455], and others. They are potentially curative, minimally invasive, and easily repeatable for recurrence. They are mainly performed on patients with small HCC, generally in Child–Pugh class A or B patients with three or fewer tumors each 3 cm or less in diameter.

Percutaneous ethanol injection was first reported in the early 1980s [449‐451], and was long the standard in ablation. Survival of patients treated with ethanol injection has been reported to be 38–60% at 5 years [456‐459]. Local tumor progression after percutaneous ethanol injection has been reported to occur in 6–31% depending on the tumor size [456, 458, 460, 461]. Percutaneous ethanol injection has been considered a safe procedure, with mortality and morbidity of 0–3.2% and 0–0.4%, respectively [458‐460, 462]. Nowadays, ethanol injection is a treatment of choice only in cases in which RFA cannot be performed safely because of either enterobiliary reflux, adhesion between the tumor and the gastrointestinal tract, or other reasons.

Percutaneous MWA, in which cancer tissue is ablated by dielectric heat produced by microwave energy emitted from the inserted bipolar-type electrode, was introduced into clinical practice in the 1990s [452]. The first-generation MWA has been replaced by RFA in Japan [463], because of small volume of ablation. New-generation MWA systems incorporating antenna cooling and high-power generation have received considerable attention [464]. New-generation MWA may create a more predictable ablation zone, and a larger ablation volume in a shorter procedure time. However, its cumulative reported experience is limited. Further studies are needed, especially from the viewpoint of long-term survival.

In RFA, radiofrequency energy emitted from the exposed portion of the electrode is converted into heat, which causes necrosis of the tumor. RFA has recently been the most widely used ablation technique for HCC. Its survival has been reported to be 39.9–68.5% at 5 years and local tumor progression to be 2.4–27.0% [465‐470]. Mortality and morbidity of RFA have been reported to be 0.9–7.9% and 0–1.5%, respectively [465‐469]. Compared with RFA alone, combination of RFA with TACE may increase the volume of necrosis [471, 472], and might improve overall survival [473, 474]. Likewise, hepatic arterial balloon occlusion during RFA might extend the area of ablation and decrease tumor recurrence from the same subsegment as the ablated tumor [475].

Irreversible electroporation (IRE) is a nonthermal tumor ablation technique that uses electric pulses to induce cell death, while preserving the structural integrity of bile ducts and vessels. IRE seems to be useful for tumors near a major Glisson’s sheath [476].

There have been five RCTs comparing RFA with ethanol injection. Four of them demonstrated superiority of RFA over ethanol injection, in terms of treatment response, recurrence, and OS [455, 477‐479], while the other trial showed that OS was not significantly different between RFA and ethanol injection [480]. Ethanol injection, however, does not require special instruments and is cheaper. Ethanol injection might be a treatment of choice in very small HCC.

An RCT comparing RFA with first-generation MWA demonstrated that the number of treatment sessions was fewer in RFA, although complete therapeutic effect, major complications, and local tumor progression were not statistically different between the two therapies [481].

It is not easy to compare outcomes between RFA and surgical resection; the indications are different between the two treatments. Furthermore, indications for each treatment are different from institution to institution. Thus, a case adjudged to be treatable by RFA or surgical resection at an institution may not be given the same treatment at another. There have been four RCTs comparing RFA with surgical resection. Three of them showed that OS was similar between RFA and surgical resection. A trial on patients with a solitary HCC 5 cm or smaller showed that OS and disease-free survival (DFS) were not statistically different between RFA and resection, but complications were more frequent and severe after surgery [482]. Another trial on patients with nodular diameters of less than 4 cm and up to 2 nodules showed that there were no statistically significant differences between RFA and surgical resection in terms of OS and RFS [483]. In another trial on patients with HCC 3 cm or smaller in diameter, there was no significant difference in DFS or OS between RFA and hepatectomy, although the incidence of postoperative complications and hospital stay were significantly greater in hepatectomy [484]. The remaining study on patients within the Milan criteria showed that OS and RFS were significantly lower in RFA than in surgical resection [485].

Concerning OS, some nonrandomized comparative studies reported that RFA had similar survival to resection [486‐497], while others found that resection was associated with higher survival [426, 498‐502]. Even in studies which reported that surgical resection was superior to RFA, there was no significant difference in OS between RFA and surgical resection in patients with HCC 2 cm or smaller in diameter [426] or 3 cm or smaller in diameter [499‐501]. In one study, RFA had better long-term survival than surgical resection after propensity score analysis [503]. RFA was associated with fewer major complications [494, 500] and shorter hospital stay [494]. RFA may be more cost-effective than surgical resection [504]. Most studies reported that RFS was higher in surgical resection than in RFA, although OS was not significantly different between RFA and surgical resection in them. This is probably because surgical resection removes a much larger volume of liver parenchyma, which may result in removal of some occult metastases and reduction of new carcinogenesis but may be prone to liver decompensation. In addition, most recurrence can be treated curatively by iterative RFA [469] but not by repeated surgical resection. Although further RCTs are warranted to compare ablation with surgical resection [505], data available at present suggest that OS is not significantly different between RFA and surgical resection. Various innovations, such as CEUS [506] and multimodality fusion imaging [507], would improve outcomes in ablation.

Ablation is less invasive and less expensive. Because patients with HCC have been markedly aging, minimally invasive therapies such as ablation would play a more important role. Because many Asian countries are still developing, from the viewpoint of medical economics, highly cost-effective therapies such as ablation should have priority. Ablation techniques, especially RFA, may be an alternative to surgery in selected cases.

Although the normal liver receives a dual blood supply from the hepatic artery and the portal vein, HCC is supplied almost exclusively by the hepatic artery [508]. TACE exploits the preferential hepatic arterial supply of HCC for targeted delivery of chemotherapeutic agents, usually mixed with lipiodol, followed by embolization or reduction in arterial flow using various types of particles (e.g., gelfoam particles), while sparing the surrounding liver parenchyma [509]. TACE is currently considered as the mainstay of therapy for unresectable, large/multifocal HCCs without vascular invasion or extrahepatic spread [510]. TACE provided a significant survival benefit in selected HCC patients with preserved liver function and adequate performance status [511‐514]. Therefore, the guidelines published by the EASL and AASLD recommend TACE as a first-line, noncurative therapy for nonsurgical patients with large/multifocal HCC who do not have vascular invasion or extrahepatic spread [2, 63]. In addition, according to the guidelines published by the JSH [515], hepatectomy or TACE is recommended if there are 2 or 3 tumors of less than 3 cm, and TACE or hepatic arterial infusion chemotherapy is recommended if there are 4 or more tumors. In addition, TACE can be performed in patients at early stage in whom RFA is difficult to perform because of tumor location or medical comorbidities [516]. TACE is also the first-line therapy for downstaging tumors that exceed the criteria for LT.

As TACE usually does not induce significant liver dysfunction even in cirrhotic patients and treatment-related mortality is less than 5% [516, 517], the benefits of TACE procedure should not be offset by treatment-induced liver failure. TACE is associated with transient postembolization syndrome, but incidence of severe events has been reported to be less than 5%, including hepatic insufficiency, liver abscess, acute cholecystitis or gastrointestinal bleeding [517, 518]. Important predisposing factors are major portal vein obstruction, compromised hepatic functional reserve, biliary obstruction, previous biliary surgery, excessive amount of iodized oil, and nonselective embolization [519]. Therefore, selective or superselective TACE should be attempted to maximize tumor necrosis and to minimize procedure-related complications by preserving nontumorous liver parenchyma [520, 521].

However, there is no standardized protocol for TACE in terms of treatment schedule or type and dosage of anticancer agent. In addition, predictions of its therapeutic efficacy are limited by the use of nonstandardized embolic material. TACE performed with drug-eluting beads (DEB-TACE) loaded with doxorubicin has been shown to modify the pharmacokinetics of the injected chemotherapy, allowing longer intratumoral exposure and less systemic exposure of the drug, reducing toxicity [522, 523]. In prospective clinical trials, liver toxicity and systemic adverse effects occur less frequently after DEB-TACE than conventional TACE. Although there is no significant benefit of DEB-TACE over conventional TACE with respect to objective response, selected patient groups such as those with Child–Pugh class B, ECOG performance status 1, bilobar disease, and recurrent disease showed a significant increase in objective response in DEB-TACE group [524]. Furthermore, DEB-TACE was associated with improved tolerability, with a significant reduction in serious liver toxicity and side-effects [524]. Despite these promising results, use of DEB-TACE in Asia has been relatively low compared with Western countries [2, 415, 525]. So far, in Asia, there has been no robust evidence favoring use of DEB-TACE in terms of efficacy and cost-effectiveness. Therefore, further research is required to address this issue.

Although TACE is considered the standard of care for nonsurgical HCCs that are also ineligible for percutaneous ablation, those with so-called bulky tumor burden (tumor size >5 cm) and Child–Pugh class B showed the worst survival outcomes (median OS of about 9 months) [526]. Furthermore, several scoring systems [i.e., SNACOR [527], hepatoma-embolisation prognostic (HAP) score [528], modified HAP score [529], Selection for TrAnsarterial chemoembolisation TrEatment (STATE) score [530], and the Chiba HCC in intermediate-stage prognostic (CHIP) score [531]] have been developed, identifying a subgroup with unfavorable outcomes. Among them, STATE score based upon tumor burden (up-to-7 criteria), albumin level, and C-reactive protein level was suggested as an objective point score to guide the decision regarding the first treatment, showing that lower STATE score was associated with worse outcome [530]. In a similar context, the Assessment for Retreatment with TACE (ART) score, an objective point score to guide the decision regarding retreatment with TACE, was developed based upon an increase of aspartate aminotransferase by >25%, Child–Pugh score increase, and absence of radiological response [532]. Higher ART score was associated with major adverse events after the second TACE (P = 0.011) [532]. Based upon these findings, sequential use of the STATE and ART scores was suggested to identify the most suitable and unsuitable patients for multiple TACE sessions [530]. So, for such a population with relatively unfavorable outcomes primarily owing to tumor burden and/or liver function, other treatment options based upon multidisciplinary approaches, including a switch of treatment modality from TACE to sorafenib or hepatic arterial infusion chemotherapy, might also be considered. Vice versa, even for large/multinodular HCC, active curative treatments including LT (e.g., within up-to-7 criteria), or so-called downstaging strategies might be tried in selected cases [533‐535].

According to conventional size-based response evaluation criteria, i.e., World Health Organization (WHO) criteria [536], the reported rate of objective response ranges between 16 and 60% [513, 517]. The Response Evaluation Criteria in Solid Tumors (RECIST) [537] generally ignore tumor necrosis, and thus may underestimate treatment response [538]. In contrast, two enhancement criteria, the EASL criteria [539] and the modified RECIST (mRECIST) [540], have demonstrated superior efficacy for assessing treatment response and predicting survival outcome compared with the WHO criteria or RECIST [541‐543]. The objective response rate using enhancement criteria ranges between 58 and 86%, and 20–41% achieve complete response [538, 541‐543]. In addition, the biological response based upon changes in tumor markers after treatment might be used as an ancillary method for assessment of overall response [544, 545].

Another issue related to TACE is the concept of “failure” or “refractoriness” to TACE. So far, several studies have tried to address this [366, 525, 546‐548]. The JSH has provided a definition of TACE failure/refractoriness as two or more consecutive ineffective responses seen within the treated tumors, two or more consecutive progressions in the liver (including an increase in the tumor number), continuous elevation of tumor markers right after TACE, appearance of vascular invasion, and appearance of extrahepatic spread [366]. Similarly, according to Raoul et al. [546], a switch of treatment modality from TACE to others including sorafenib might be considered for those who have progression after two sessions of TACE. However, there is still no consensus regarding the definition of TACE failure or refractoriness. Moreover, there is no proven therapy for the purpose of rescue, although sorafenib rescue might improve survival in patients who experience TACE failure, compared with those who continue TACE [549, 550]. Other treatment modalities including internal or external radiotherapy and new molecular targeted agents have been studied as potential rescue therapies for patients with TACE failure.

Many attempts have been made to improve the treatment outcomes of TACE. Combination of sorafenib and TACE might be an eligible option [551]. However, a RCT comparing the efficacy in HCC treated with sorafenib or placebo plus DEB-TACE showed that combination therapy did not improve outcome [552].

Transarterial radioembolization (TARE) involves injection of implantable radioactive microspheres into tumor-feeding arteries in order to expose the tumor to highly concentrated radiation while protecting the normal parenchyma. TARE using yttrium-90 is an evolving and promising regional therapy, which can complement or replace TACE [2, 415, 553]. In a European phase II study of patients with intermediate or advanced HCC, TARE resulted in 40.4% objective tumor response rate with median survival of 15 months [554]. In another large retrospective cohort study conducted in the USA, the median survival of TARE-treated patients with portal vein invasion was significantly shorter than those without invasion (10 versus 15.3 months) [555, 556]. In a recent prospective multicenter Korean study, the 3-month tumor response rate was 57.5% and the 3-year OS rate was 75% [557]. Although there is not enough evidence confirming clinical benefit of TARE compared with conventional TACE, TARE might be recommended to patients who are not good candidates for TACE due to bulky tumor and/or portal vein invasion, based on published data.

Although HCC is considered to be a radiosensitive tumor, it also is located in a radiosensitive organ. Due to the development of three-dimensional conformal radiation therapy (3D-CRT), radiotherapy (RT) can be performed more safely for patients with HCC without severe toxicity. Technological developments for targeting HCC precisely with RT [intensity-modulated RT (IMRT) and image-guided approaches, including stereotactic body radiotherapy (SBRT)] can improve the benefit and reduce the risk. However, they do not alter the high recurrence rates in other nontreated areas of the liver. There are no large-scale RCTs demonstrating an effect of any form of RT on survival and no consensus regarding the optimal use of this therapy. Thus, RT is not recommended in the AASLD and EASL guidelines for treating HCC [2, 63]. Even though strong evidence is lacking, RT may be one of the promising treatment options for HCC.

With the development of 3D-CRT techniques, RT can be performed more safely to the HCC with less liver toxicity. Most available data are from retrospective or single-center studies [562, 563]. A phase II trial in France reported sustained local tumor control in 78% of patients with early-stage HCC (one nodule and ≤5 cm, or two nodules and ≤3 cm) who were treated with 3D-CRT [564]. One of the problems with RT is the high intrahepatic recurrence rate outside of the high-dose irradiation area, which may be caused in part by difficulties with accurately targeting HCC during conventional RT treatment planning [565].

SBRT (sometimes called stereotactic radiosurgery) is a technique in which a limited number of high-dose RT with hypofractionation (typically 3–6) are delivered to a small, definite target using multiple, nonparallel radiation beams. The beams converge on the target lesion, minimizing radiation exposure to other normal tissue or organs. This targeting makes it possible to treat a lesion in either a single or limited number of dose fractions. Experience with SBRT for HCC is increasing [566‐569]. In the largest series, 93 patients (Child–Pugh A: 69 patients; Child–Pugh B: 24 patients) with small HCCs (median 2 cm; range 1–6 cm) who were not eligible for surgical resection or RFA were treated with SBRT [566]. The in-field complete response rate was 16%, but the in-field progression-free survival at 3 years was 92%.

There is a growing body of evidence, primarily from Japan, supporting use of proton beam and carbon ion beam irradiation, particularly for patients with large tumors or portal vein thrombus [570‐573]. In one study, 162 patients with 192 HCCs were treated with proton beam irradiation [570]. Most tumors had diameter of 3–5 cm. The majority of the patients had past history of receiving other nonsurgical treatments. The 5-year local control and 5-year survival rates were 87 and 24%, respectively.