Surgical resection for hepatocellular carcinoma: Moving from what can be done to what is worth doing

Potential conflict of interest: Dr. Mazzaferro is on the speakers’ bureau for Bayer and BTG. Dr. Bruix consults for Bayer, BTG, Syrtex, and Terumo. Supported by the Italian Ministry of Health (to R.R., V.M.), the Italian Association for Cancer Research (to V.M.), Istituto Nazionale Tumori 5per1000 funds, and a grant from the Instituto de Salud Carlos III (to J.B., PI 14/00962); CIBERehd is funded by Instituto de Salud Carlos III. See Article on Page 440. Radical resection is the mainstay of treatment for organ tumors. The same could apply to hepatocellular carcinoma (HCC) if frequent multifocality and coexisting cirrhosis did not limit its role. Current guidelines1 recommend resection only for single nodules of any size in patients without tumor‐related symptoms and clinically significant portal hypertension (CSPH) and with normal bilirubin (≤1 mg/dL). If this profile is not fulfilled, postoperative morbidity increases and long‐term survival is significantly reduced. An extension of the recommendation has been repeatedly suggested because in patients with CSPH3 multiple nodules or intrahepatic vascular invasion resection can be attempted with high rates of technical success in experienced centers, even though tumor elimination by surgery translates into improved survival only in properly selected candidates. Actually, while tumor removal would be technically feasible in patients with a large tumor burden or impaired liver function, resection may not be worth attempting as survival could even be decreased. In real life the decision to resect HCC is based on individual patient components and local conditions that are not captured by guidelines. Debate about resection is fueled by several publications in which the outcome in suboptimal candidates is still felt to be acceptable because it appears to be better than with other treatment options or no treatment. The controversy will grow as improvement in surgical techniques and new drugs for hepatitis C virus will determine a reduction in postoperative morbidity, as happened with hepatitis B virus. Finally, the growing epidemic of HCC in metabolic syndrome may also prime resections in large/multifocal tumors occurring in noncirrhotic liver, although comorbidities and advanced age may preclude safe surgery in many of these patients. In this issue of Hepatology, Roayaie et al.5 describe the surgical management of HCC within the Bridge database, which collects information about new incident cases worldwide. In a cohort of 8656 patients the authors evaluated how frequently guidelines for resection were followed and whether straying from them impacted survival. A total of 862 (10%) patients were classified as ideal resection candidates, and more than 80% of them underwent surgery; the remaining were mostly treated by ablation or embolization and experienced a two‐fold increased mortality risk. These results could be used to support the superiority of resection, but they could also reflect the fact that associated conditions (advanced age, comorbidities) excluded those less healthy subjects from surgery. The main study interest is focused in the 7794 (90%) patients who were classified as nonideal candidates, 20% of whom were resected despite a suboptimal profile. Unfortunately, the database did not register intrahepatic tumor location and type of hepatectomy performed. As known, the propensity of hepatologists to refer and that of surgeons to operate on nonideal patients with cirrhosis are influenced by the extent of planned resection and the perceived risk of the intervention. Therefore, the reader is left with the feeling that drivers of the decision to resect or not a nonideal candidate were not thoroughly recorded and, consequently, that nonideal resected patients formed a not fully reproducible cohort. The fact that just few resected cases had severe liver disease (Child‐Pugh C, bilirubin >2 mg/dL, severe portal hypertension reflected by platelets <50,000/mm3) or very compromised general conditions (performance status 3‐4) suggests that clinicians took into account liver function, tumor location, and general health status when favoring resection. Indeed, the <5% postoperative mortality rate indicates a quite appropriate evaluation of short‐term surgical risks. However, while remaining acceptable, the risk of 90‐day perioperative death was almost four times higher in nonideal patients than in ideal ones (4.5% versus 1.2%), and the fact that resection can be performed without excessive early mortality should not be understood as if long‐term survival would be as good as in optimal candidates. The analysis in all resected patients (ideal and nonideal) showed that CSPH and bilirubin >1 mg/dL were not associated with mortality if taken separately, while their association conditioned a worse prognosis. Multivariate analyses were performed using both a more sensitive (platelets <100,000/mm3 or evidence of varices/splenomegaly) and a more accurate (platelets <100,000/mm3 and evidence of varices/splenomegaly) definition of CSPH. Even though not significant in both models, CSPH was shown to increase the hazard ratio (1.17‐1.24) with a trend toward significance (P = 0.12‐0.08) moving from the first to the second analysis. Instead, multivariate models in nonideal candidates (resected and nonresected) exposed CSPH as a negative prognostic factor, stressing once more the significant impact of CSPH on prognosis in patients with cirrhosis. Certainly, different methods of defining CSPH (hepatic vein pressure gradient versus surrogate markers) and different study designs fuel the debate about the role of CSPH. A Japanese study6 confirmed the impact of portal hypertension on long‐term outcome after resection, while Italian studies reporting their data3 or using propensity‐score matching4 suggested a lack of impact. Single‐center data should be carefully assessed in order to ensure that postoperative mortality and survival rates reflect the figures observed in referral centers. In addition, propensity‐score studies need to be validated to avoid the flaw due to the exclusion of outliers that are indeed detected by the assessment of the key parameter, which in liver resection may be portal hypertension. This relevant role of CSPH is reinforced by a recent Barcelona7 and two Asian8 meta‐analyses. Interestingly, the study by Roayaie et al.5 raises the value of bilirubin >1 mg/dL for a better stratification of the patients classified as having CSPH using suboptimal definitions in comparison with hepatic vein catheterization. The bilirubin role was identified in the seminal study in Barcelona and is included in guidelines. Such a “breakpoint” is frequently dismissed, yet it appears to sense the surgical risk of nonideal candidates per CSPH, while serving to predict the survival of resected patients. Bilirubin is also a covariate of the Model for End‐Stage Liver Disease score, which is a reliable predictor of postoperative liver failure if higher than 10.4 Roayaie et al.5 report a 15% mortality risk reduction in ablated patients compared to resected ones, thus endorsing the recommendation to give priority to ablation in patients with small tumors that can be effectively and safely ablated. As expected, nonideal patients who were transplanted had the best outcome. This reinforces the role of transplantation in guidelines: first‐line treatment for nonideal resection candidates fulfilling criteria for transplant. Long‐term survival was better with resection than with transarterial therapies, but as mentioned before, it is impossible to ascertain whether the driver for an improved outcome was surgery by itself or the bias introduced by clinical decisions and/or available expertise in nonsurgical options. After all, Roayaie et al.'s article basically argues how guidelines should be used in an individual patient who could undergo different treatments and whether the principle should be endorsed that all “curative” options must be explored before relying on “palliative” ones. This suggestion tends to privilege initial activity (initial cure with risk of recurrence) rather than long‐term survival, which is the real goal. We believe that guidelines constitute a useful framework in which clinicians’ decisions can be tested, especially for patients with a well‐defined profile. They help to protect us from overtly incorrect choices, safeguarding health system quality. Instead, patients with characteristics that are not within the strict recommendation require assessment by experts, who may make nonideal decisions on nonideal patients but always consider individual and general factors favoring either of the available treatments (Table 1). In such a perspective, guidelines are like the fundamental rules for playing a sport, whose ignorance does not allow admission to practice. However, as in professional sports where the bare application of fundamentals does not forcibly imply winning the game, the too rigid interpretation of guidelines could stop experimentation on new potentially successful strategies and possibly preclude some patients from benefiting from more effective treatments chosen by expert judgment. Thereby, guidelines expose the current recommended practice, and data such as those provided by Roayaie et al. may open the door to prospective and robust investigations in the area of surgical resection. These may result in data leading to modification of the current guidelines or may just reinforce them. In the absence of such clarifying information, the debate will continue. Table 1 - Factors Concurring to Select Treatment for HCC Confined to the Liver Drivers of Treatment Selection In Favor of Resection In Favor of Transplantation In Favor of Ablation In Favor of Transarterial Therapies Patient Age ≤75 years ≤70 years No limit No limit Performance status 0 Any grade (high MELD) 0 0 Comorbidities Absent/minor Absent Major Major Tumor Size ≥3 cm Single ≤5 cm ≤3 cm Any size Number Single Up to 3 nodules ≤3 cm Up to 3 nodules Large/multinodular Location within liver Peripheral/exophytic growth Any site Central, far from vessels, Bile tract and viscera Central Vascular invasion (branch/segment) Not relevant by some Absent Absent Not relevant by some Satellites Not relevant only in anatomic resections Not counted when <1 cm Absent Not relevant Alpha‐fetoprotein The lower the better <1000 ng/mL Any level Any level Perceived antitumor efficacy High Very high High Moderate Liver disease Quality of parenchyma Noncirrhotic Cirrhotic Cirrhotic Cirrhotic Portal hypertension Absent/mild Any degree Any degree Any degree Bilirubin (NV ≤1 mg/dL) Normal Any level Normal/≤ 2 × NV Normal/≤ 2 × NV MELD score Very low Any value Low Low Local factors Specialized surgical expertise Available Available Not available Not available Interventional nonsurgical expertise Not available Not relevant Available Available Organ donation rate Low High Low Low Competition with non‐HCC patients on transplant waiting list High Low High High Abbreviations: MELD, Model for End‐stage Liver Disease; NV, normal value. Author names in bold denote shared co‐first authorship.