Conditional Survival after Hepatic Resection for Hepatocellular Carcinoma in Cirrhotic Patients

Hepatocellular carcinoma (HCC) in cirrhosis is one of the leading causes of death from cancer worldwide, especially in Western countries where an increased incidence has been observed over the last decade (1, 2). Treatment allocation and assessment of prognosis is very challenging for clinicians, as these patients suffer from 2 potentially lethal diseases: cancer and cirrhosis. It is well known that liver function and degree of liver dysfunction condition both the therapeutic strategy and assessment of prognosis (3, 4). However, the impact of these variables is well recognized and established when patients are observed for the first time, whereas it is unclear what their impact is over time when early events have not occurred. For instance, AASLD guidelines, following the Barcelona Clinic Liver Cancer (BCLC) staging system, do not recommend trans-arterial chemoembolization (TACE) in the decompensated patients with HCCs in Child–Pugh B class, nor any treatment at all in Child–Pugh C, apart from transplantation, even when tumors are small, highlighting the dominating prognostic role of compromised liver function due to cirrhosis. For this reason, in trials for HCC, it is recommended to include overall survival as an endpoint and to avoid recurrence-free survival (5), and practically no trial on HCC ever includes decompensated Child–Pugh B or C patients. However, in patients who are cured for HCCs by resection, corresponding to those with preserved liver function, the forecast of future events leading to death is more difficult. In fact, life expectancy at 2 years, according to the natural history, is very good in Child–Pugh A patients, being more than 90% (6). Thus, one would expect that these patients will almost invariably die of tumor recurrence. However, the risk of tumor recurrence and associated prognosis changes over time. In the first 2 years, tumor recurrences most likely represent intrahepatic metastatic dissemination of the initial resected tumor (7–9) and consequently imply a worse prognosis. After 2 years, recurrences more often indicate de novo tumors, implying a better prognosis and more risk of dying due to worsening of the underlying liver function, which might have progressed in the meanwhile. The latter case becomes even more likely in patients who do not present recurrence at all. Thus, the prognostic estimations made at the time of the initial surgical treatment are usually valid for describing groups but not likewise useful to define individual prognosis, especially when this is to be reassessed in the future. Moreover, this prognostic assessment, based mainly on clinical experience, does not appear to have ever been the object of a formal statistical analysis in the literature. A more accurate individual prognosis would be important in clinical practice not only to better manage patients (for instance, deciding on antiviral strategies or tailoring surveillance strategies) but also in the research setting to better understand the comparability of groups in treatment trials, especially in the setting of adjuvant strategies, such as the ongoing “Sorafenib as adjuvant treatment in the prevention of recurrence of hepatocellular carcinoma” (STORM) trial (www.clinicaltrials.gov/ct2/show/NCT00692770). For these patients, the prognosis can be more accurately assessed using the conditional survival analysis (10). Conditional survival is derived from the concept of conditional probability, and this concept is applied to determine the probability that a patient, who has survived for a specific period, will still be alive at another fixed interval. As in many other malignancies, hazard rates for death from HCCs have been found to be relatively higher in the first few years after diagnosis and treatment but thereafter start to decrease.

Representing the changing likelihood of demise over time, the conditional survival analysis offers meaningful prognostic information for those patients who are fortunate enough to have survived the initial management, by providing stronger estimates of their survival probability (10, 11). Several previously published cancer data revealed distinct patterns of conditional survival that can vary substantially among diagnoses (11, 12); however, to the best of our knowledge, no investigations have explored the conditional survival pattern of HCCs.

The aim of the present study was to describe how conditional survival probability and prognostic factors for HCCs in cirrhotic patients (including all those known to influence prognosis) can change over time, taking into consideration those patients submitted to a radical curative technique of the neoplastic disease, such as surgical resection.

Conditional survival is derived from the concept of conditional probability in biostatistics (10). It can be calculated from the traditional Kaplan–Meier or actuarial life table survival data. The mathematical definition of conditional survival (CS) can be expressed as follows: CS(y | x) is the probability of surviving for an additional y years, given that the person has already survived x years. Let S(t) be the traditional actuarial life table survival at time t. Conditional survival can be expressed as: CS (y | x) = S(x + y)/S(x). For example, in computing the 5-year conditional survival for a patient who has already survived 2 years, that is, the survival at 5 + 2 years, S(7), is divided by the survival at 2 years, S(2). When a survival curve shows a changing hazard rate over time, this will be reflected as a change in conditional survival as more time elapses from the time of diagnosis.

After local institutional review board approval, a review was made of the clinical data of all those patients who had undergone hepatectomy, in our center, between January 1, 1997, and December 31, 2008. A total of 300 patients who underwent hepatic resection due to the presence of HCCs and histologically proven cirrhosis (13), were identified with adequate clinical data for review. Exclusion criteria were incomplete clinical data (25 patients), presence of severe comorbidities that could affect life expectancy (3 patients with a history of severe cardiac disorders and 1 patient with a hematologic disorder who died of acute myocardial infarction and pulmonary embolism, 1 patient with a severe cardiac disorder who is at present still alive but in New York Heart Association class III), and preoperative portal vein embolization (1 patient). None of the patients in the study group were submitted to surgical portosystemic shunts before or at the same time as hepatic resection, were treated as an emergency, had a palliative resection or had presented, at pathologic examination, tumor invasion into a major branch of the portal or hepatic veins, or direct invasion of adjacent organs, or spread to the lymph nodes of the hepatic hilum. All resections considered in the present analysis were curative resections (R0) at histology. The policy adopted in our center about indications for hepatic resection, has been published elsewhere (14, 15). Briefly, patients were selected for surgery on the basis of the technical feasibility of success that was established when the residual liver volume was expected to be sufficient in view of curative resection. In our center, esophageal varices, a platelet count of <100,000/mm³, and the presence of multiple nodules were not considered absolute contraindications and the judgment on residual liver function is not based only on Child–Turcotte–Pugh (CTP) Class. For many years, our approach has been based on Model for End-stage Liver Disease (MELD) score and a MELD score more than 11 represents a contraindication for surgery except for the possibility of salvage transplantation (14, 15) and a MELD score of 10 is only occasionally accepted depending upon the type of resection. However, also following these rules, only 13 patients in the present study had a CTP Class B (11 with a score of 7 and 2 with a score of 8, both the latter having been operated upon in the 90s) and the vast majority (>95%) were CPT class A.

The day before surgery, the following clinical and biochemical data were collected for each patient: age, gender, cause of cirrhosis, serum levels of albumin (g/dL), creatinine (mg/dL and μmol/L), total bilirubin (mg/dL and μmol/L), sodium (mmol/L), and international normalized ratio. The CTP score was obtained according to the classification proposed by Pugh and colleagues (16) and the MELD score was calculated using the appropriate formula (17). Clinical signs of portal hypertension were defined on the basis of the BCLC description (4) as (i) esophageal varices detectable at endoscopy or (ii) splenomegaly (major diameter, >12 cm) with a platelet count of <100,000/mm³.

Intraoperative management has already been described elsewhere (14); intraoperative ultrasound was conducted systematically to detect the presence of any additional nodules not revealed preoperatively and to obtain a tumor-free margin of at least 1 cm; major hepatic resection was defined as the removal of more than 2 segments: the extent of the hepatectomy was based on the International Hepato-Pancreato-Biliary Association Classification (18). Tumors were staged on the basis of preoperative imaging, according to the United Network for Organ Sharing (UNOS)—TNM classification (19). Tumor grade was defined using the modified nuclear grading scheme outlined by Edmondson and Steiner (20, 21): specifically, modified Edmondson–Steiner grades I and II were defined as well-differentiated and grades III and IV as moderately/poorly differentiated. In all cases, tumor grade was defined by the poorest degree of differentiation, identified within the tumor, upon pathologic analysis of the entire specimen; microscopic vascular invasion (MVI) was defined according to the presence of tumor emboli within the central hepatic vein, the portal, or the large capsular vessels (22).

Following discharge, all patients were observed periodically at follow-up to exclude possible recurrence of HCCs: biochemical liver function tests, serum α-fetoprotein level measurement, and ultrasound were conducted 3 and 6 months after discharge and then according to an annual surveillance program; more recently, the surveillance program has become semiannual for almost all patients (23, 24). Recurrence was diagnosed on the basis of 2 coinciding imaging techniques or the combination of increased alpha-fetoprotein and consistent ultrasound, computed tomographic or MRI findings according to the latest guidelines released. None of the patients in this study group received adjuvant chemotherapy. Patients presenting recurrence were managed with various therapeutic modalities, including re-resection, when possible, for resectable recurrence, and salvage liver transplantation, for selected patients with transplantable recurrence. The patient selection criteria for second hepatic resection, however, were the same as for primary resection. Patients with non-resectable recurrence were submitted to TACE in the case of multiple or large nodules and/or clinical signs of light or moderate liver dysfunction; radiofrequency ablation was used in the presence of deep and/or multiple nodules <3 cm; percutaneous ethanol injection was preferred when nodules were adjacent to a major vessel. Systemic chemotherapy was conducted in selected patients with extrahepatic metastatic and, from the end of 2008, sorafenib (Nexavar; Bayer) therapy was also adopted, either alone or in combination with percutaneous approaches (25, 26).

Summary statistics were obtained using well-known statistical methods and presented as percentages or median and range. Patient survival was computed from the day of surgery until the most recent follow-up visit or until death, follow-up of patients submitted to salvage transplantation was censored the day before the procedure. Follow-up data were collected until December 31, 2010. Survival rates were calculated with the Kaplan–Meier method, compared with the log-rank test and used in the calculation of the 5-year conditional survival. Only variables that were significantly related to patient survival at Kaplan–Meier analysis were used for 5-year conditional survival calculation. The conditional survival differences observed between subgroups were compared with the calculation of standardized differences (d). Standardized difference is a measure of the effect size: when outcomes are expressed as proportions, such as 5-year conditional survival, the standardized difference in proportions is calculated as (P₂ − P₁)/√[P(1 − P)] where P is the weighted mean of P₁ and P₂ (27): d values lower than |0.1| indicate very small differences between means; d values between |0.1| and |0.3| indicate small differences, d values between |0.3| and |0.5| indicate moderate differences, and d values greater than |0.5| indicate considerable differences. A significance level of 0.05 was used in all analyses. The statistical analysis was conducted using SPSS Version 10.0 software for PC computer (SPSS).

Baseline characteristics of the study population are presented in Table 1. The median follow-up of the entire study population, comprising 300 patients, was 3.8 years (range, 1 month–12 years); during this time period, 116 patients died (38.6%). The 30-day mortality was 1.7% (5 cases) and the 90-day mortality was 4.0% (12 cases). During follow-up, 165 patients experienced tumor recurrence (55.0%): the 3-, 5-, and 10-year recurrence rates were 52.0%, 67.1%, and 80.9%, respectively. The therapeutic approach most frequently used in cases of tumor recurrence was TACE, accounting for 37.6% of treatments (62 cases); 23.6% of patients underwent re-resection (39 cases); radiofrequency ablation or percutaneous ethanol injection were attempted in 12.7% of patients (21 cases); liver transplantation was conducted in 6.0% (10 cases); systemic chemotherapy with unproven benefit in the literature or best supportive therapies were adopted in 16.9% (28 cases), and sorafenib in 3.0% of patients (5 cases). Tumor recurrence, with or without liver failure as a consequence of end-stage cirrhosis, represented the most frequent cause of death (55 cases; 18.3%), followed by liver failure without tumor recurrence (42 cases; 14.3%) and other causes that accounted for the remaining patients (18 cases; 8.0%). The 3-, 5-, and 10-year survival rates were 69.0%, 57.7%, and 25.3%, respectively (Fig. 1). Of the 165 patients who experienced tumor recurrence, early recurrence (<2 years) was observed in 113 patients (68.5%) and late recurrence (>2 years) in 52 patients (31.5%). Of these, 74 patients died during follow-up (57 with early recurrence and 17 with late recurrence), but tumor progression was not significantly more often the cause of death (P = 0.349) in those with early (44 of 57 patients, 77.2%) than in those with late recurrence (11 of 17 patients, 64.7%). Nontumor-related liver failure accounted, respectively, for 15.8% and 23.5% of the cases.

Actuarial survival rates, in relationship to the clinical and tumoral characteristics taken into consideration, are reported in Table 2. As expected, the following variables were found to be significantly related to a decrease in actuarial patient survival in agreement with reports in the literature, namely, presence of clinical signs of portal hypertension (P = 0.005); MELD score ≥9 (P = 0.001); UNOS-T3 tumor (P = 0.001); presence of GIII–GIV tumors (P = 0.001), and presence of MVI (P = 0.001). Thus, the expected actuarial patient survival was determined by factors related both to the underlying liver disease and the presence of the tumor. However, current knowledge is unable to predict the role of each of these variables at different time points after resection, depending on whether one or the others had already impacted or not. Instead, conditional survival analysis shows that these variables had a different role on patient probability of survival over time.

The 5-year conditional survival of the entire study population and in relation to the degree of the underlying liver disease and tumor characteristics is outlined in Table 3. In particular, it should be noted that the conditional survival of the entire study population showed an increase in the first 2 years and subsequently a decrease from the third year onward: this characteristic was explored by analyzing the relationship between conditional survival and the degree of liver disease and tumor features. The 5-year conditional survival of patients without clinical signs of portal hypertension always remained better by at least 10% at all time points in comparison to that of patients with this clinical feature. Likewise, patients with an MELD score <9 had a 5-year conditional survival that remained always better by at least 5.5% in comparison to MELD ≥9. Unlike findings observed for variables related to the degree of the underlying liver disease, tumor factors had an impact on shortening survival only for the first 2 years. In particular, from 2 years after hepatic resection, the probability of survival of recurrence-free patients with a more advanced tumor stage or with less favorable tumor characteristics was very similar to that of those with less advanced stages or with either well-differentiated tumors or without MVI. Differences observed in the 5-year conditional survival of patients with T3 tumor start to decrease from the third year onward: the probability of surviving an additional 5 years, given that such patients have already survived 3 years, was only 2.7% less than that of patients with T1–2 tumor and 0.2% if these patients have already survived 5 years. Likewise, 3 years after surgery, the probability of surviving an additional 5 years of patients with GIII–GIV tumor was only 0.9% less than that of patients with a GI–GII tumor and remains stable in the following years. Finally, 3 years after surgery, the probability of patients with microvascular invasion surviving for an additional 5 years was only 1.9% less than that of patients with a GI–GII tumor and 0.2% if these patients had already survived 5 years. Thus, from the third year after hepatic resection onward, the probability of survival no longer seems to be affected by tumor characteristics.

Another conditional survival analysis was conducted comparing patients undergoing repeated hepatectomy versus first hepatectomy. Interestingly, patients undergoing second hepatectomy for tumor recurrence showed a very high 5-year conditional survival in comparison to first hepatectomy over time. For instance, 3 years after surgery, the probability of those patients undergoing second hepatectomy, surviving for an additional 5 years, was 24.4% higher than that of patients submitted to the first hepatectomy, suggesting that patients in whom recurrence takes place with a tumor stage and liver function suitable for re-resection should be offered this second opportunity. To better understand this issue, those patients submitted to a second resection were assessed according to whether recurrence took place before (21 patients) or after 2 years (18 patients). Conditional survival could not be calculated, as the groups were small with not enough patients, in each subgroup that had been followed-up for 10 years. However, the calculated overall survival at 1, 3, and 5 years after the second resection, taken as time zero, was, respectively, 90.5%, 85.7%, 63.5% and 95.2%, 90.9%, 75.8% (P = 0.061).

Conditional survival information is potentially of great interest for patients, clinicians, and researchers, as it quantifies a patient's changing profile of risk over time (10–12). This makes it different from standard survival predictors, which are less able to take into consideration the changes of the weight of individual variables at subsequent times after initial assessment. In this study, we showed that conventional tumor features, known to impact on patient prognosis, such as tumor stage, differentiation degree, and microvascular invasion, are indeed, as expected, predictors of survival but only in the first 2 years. Patients remaining tumor recurrence-free for the first 2 years return to the same survival estimates as patients with more favorable tumor features. This difference was not captured by conventional assessment modalities of survival. This information might be of use for further reasoning in the field of the best timing for transplantation in patients with HCCs within the conventional tumor transplant criteria (i.e., the Milan criteria) who are also eligible to curative tumor treatments, such as hepatic resection. One possibility would be to “de principe” transplant all of them, given the risk of recurrence outside transplantability criteria and overall poor survival of such patients with their naive livers over the long term (28–30). Another possibility, gaining popularity, is to resect those eligible to surgery and leave the opportunity of transplantation as a salvage procedure, with immediate listing only for those who experience early recurrence within the first 2 years and possible re-resection for those who recur later (31). Such strategy considers the transplant benefit of the whole patients population rather than the single individual benefit only. Our findings provide scientific support to the latter possibility of salvage transplantation, previously based mainly on expert opinions and pathologic data on the type of recurrences (31), showing that those patients who do not recur within the first 2 years maintain a satisfactory survival with their own naive liver. In our series, in greater detail, patients whose recurrence was amenable for surgery and who where re-resected had a 5-year life expectancy of more than 80% when observed recurrence free at 2 and 3 years after initial resection in our series. Even more interestingly, their survival remained >75% also at 5 years after the second operation, when conducted for late recurrence, with a very high survival after the initial operation. These figures are absolutely comparable with those of transplantation, at least in a 5- to 7-year perspective. Thus, even a patient whom someone might theoretically consider a transplant candidate due to adverse tumor features (28,29) is no longer a good transplant recipient if he/she had remained recurrence free with his/her own native liver for more than 2 years from resection. This is particularly true, when some liver or general characteristics might suggest a shorter survival specifically with transplantation, such as in the case of hepatitis C virus–infected patients, who nearly invariable experience hepatitis C virus recurrence and shorter survival than patients with other etiologies. Cirrhotic patients with HCCs represent a particular case in oncology as their prognosis relies not only on the tumor features but also on the severity underlying liver disease. Despite this situation is well known to clinicians managing patients with HCCs, very little scientific evidence is available to assess the relative role of the 2 conditions during follow-up and not only at the time of initial observation, when the treatment strategy is set up.

It is accepted in general that survival after hepatectomy, in cirrhotic patients with HCCs, is mainly affected by tumor recurrence, as only those with preserved liver function are operated, so that survival in the short- and mid-term shout not be greatly influenced by liver function. Accordingly, patients with early recurrence mostly died of cancer and were usually predicted by adverse tumor features (7, 8, 15, 32). However, in the present investigation, long-term survival was affected by tumoral and clinical characteristics in different ways. The conditional survival calculation showed that from the third year after surgery, the possibility of survival is no longer affected by tumor stage or tumor differentiation or presence of MVI. In the present study, the clinical oncologic course after HCC recurrence in patients with bad or good resected tumor features was not specifically analyzed to further confirm that the long survival in those remaining recurrence free from 2 years onward was exactly due to the fact of remaining recurrence free and not to a different clinical course after recurrence. That is because the only logical interpretation for the longer survival appears to be the absence of recurrence; moreover, the needed clinical information would have been difficult to be categorized in connection to the different possible treatment modalities, and the consequent statistical analysis was considered redundant and ultimately confusing. However, we acknowledge that the specific impact of recurrence on survival in our study remains not definitively proven by a dedicated statistical analysis. In addition, a question on the role of different variables at baseline on survival might theoretically emerge. It should be noted that the calculation of conditional survival limited the application of multivariate analyses because of the observed nonlinear contribution on survival of the different variables over time and the fact that conditional survival was calculated on groups of patients and not for each single patient, thus it cannot be suitable for both linear and nonlinear regression models.

Having more precise figures on the role played by different variables on survival at different time points after cure of HCCs could be extremely important in judging on the opportunity of antiviral treatment, particularly in hepatitis C infection, and in modulating the role of adjuvant treatments with new drugs, including length of therapy. Until recently, this was not an issue, but in very recent years, adjuvant trials were started (www.clinicaltrials.gov/ct2/show/NCT00692770) and the problem might become relevant, especially as very little or no information exists on the impact of prolonged systemic treatments on liver function and cirrhotic complications. To better judge such effects, the individual roles of different variables, including clinical parameters and their relative weight, have to be known more precisely at different time points from initial cure, in a way that only conditional survival can do.

F. Piscaglia has honoraria from speakers bureau and is a consultant/advisory board member of Bayer. L. Bolondi has honoraria from speakers bureau is a consultant/advisory board member of Bayer and BMS. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Cucchetti, F. Piscaglia, G. Ercolani, A.D. Pinna

Development of methodology: A. Cucchetti, A.D. Pinna

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Cescon, G. Ercolani, E. Terzi, M. Zanello, A.D. Pinna

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Cucchetti, A.D. Pinna

Writing, review, and/or revision of the manuscript: F. Piscaglia

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Terzi, M. Zanello

Study supervision: M. Cescon, G. Ercolani, L. Bolondi, A.D. Pinna

Writing assistance was provided by Mrs. Marian Shields.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.