Liver transplantation versus surgical resection for HCC meeting the Milan criteria: A propensity score analysis.

The optimal treatment (liver transplantation [LT] vs surgical resection [SR]) for early-stage hepatocellular carcinoma (HCC) remains controversial.A total of 209 SR patients and 129 LT patients were identified at our institution. After eliminating 27 patients with Child-Pugh C, the data from 209 SR patients and 102 LT patients were analyzed using a propensity score matching (PSM) model. Forty-six pairs were generated. A subgroup analysis was conducted based on the alpha-fetoprotein (AFP) level or platelet count (PLT). A survival analysis was performed using the Kaplan-Meier method.Gender, satellite lesions, and the treatment method were predictors of HCC recurrence. The Ishak score and treatment methods were associated with long-term survival after surgery. Before PSM, LT patients had a better prognosis than those treated by SR. Among HCC patients with childhood A/B cirrhosis, after PSM, SR achieved similar overall survival outcomes compared with LT. LT and SR resulted in comparable long-term survival for patients with or without thrombocytopenia. Patients with an AFP ≥ 400 ng/mL might achieve more survival benefits from LT.Our propensity score model provided evidence that, compared with transplantation, surgical resection could result in comparable long-term survival for resectable early-stage HCC patients, except for the AFP ≥ 400 ng/mL HCC subgroup. Surgical resection might not be a contraindication for early-stage HCC patients with thrombocytopenia due to their similar prognosis after transplantation.

Introduction

Hepatocellular carcinoma (HCC) is one of the most common malignancies worldwide. In China, it is the second-most common cancer.[ Surgical resection (SR) has been validated as a highly effective approach for HCC. In 1996, Mazzaferro et al[ proposed the Milan criteria (a tumor 5 cm or less in diameter or more than 3 tumor nodules, each 3 cm or less in diameter, with no evidence of macrovascular invasion or extrahepatic disease) for the selection HCC patients as candidates for liver transplantation (LT) due to their satisfactory long-term survival. Theoretically, LT is the best therapy for HCC patients with underlying liver cirrhosis because it eliminates the tumor and the underlying liver disease simultaneously. However, in an era of donor shortage, for those with resectable early stage HCC, there has been some controversy about which is the optimal treatment (transplantation vs SR). Based on an analysis of 408 HCC patients with the Milan criteria, Fan et al[ suggested that the 5-year survival rate after SR could reach up to 72.8%, which is comparable to the 5-year survival rate after transplantation. This finding is consistent with those of previous studies.[ In contrast, some authors contended that transplantation was superior to resection for HCC with the Milan criteria.[ According to an intention-to-treat analysis, Baccarani et al[ concluded that LT had advantages over SR for early-stage HCC (5-year estimated OS rates: 72% vs 27%). Tumor characteristics, such as number and tumor size, and liver function were the main factors influencing the choice of treatment method and the prognosis of liver disease. For example, for single HCC with a tumor size of 3 to 5 cm, some prior studies suggested that the prognosis after LT or SR was comparable,[ whereas multiple HCCs might achieve superior survival benefits from LT.[ Another study from Lu et al[ suggested that tumor size was an independent prognostic factor of HCC patients and that an HCC ≤ 3 cm presented with relatively benign pathological features. Unfortunately, most investigators tended to ignore these differences in the baseline characteristics. The propensity score matching (PSM) analysis can be used to balance the covariates and thus reduce this bias in the control and treatment group.[ More investigators set out to adopt this method and to reach more convincing conclusions in liver disease research.[ In this study, we performed a survival analysis between the LT group and the SR group using a propensity model.

Alpha-fetoprotein (AFP) is an oncogenic protein is often elevated in HCC. Elevated AFP levels are associated with HCC recurrence or long-term survival after resection or transplantation, according to some studies.[ An AFP level of approximately 400 ng/mL was validated as the optimal cutoff value for predicting the prognosis.[ A high AFP level contributes to tumor cell proliferation through the activation of the cAMP-PKA pathway.[ AFP has been demonstrated to be related to intra-hepatic or extra-hepatic metastasis through the up-regulation of keratin 19 (K19), epithelial cell adhesion molecule (EpCAM), matrix metalloproteinase 2/9 (MMP2/9), and CXC chemokine receptor 4 (CXCR4) expression.[ There is little information available from survival analyses about differences between transplantation and hepatectomy when stratified by the preoperative AFP level. The platelet count is known to be an indirect indicator of portal hypertension, especially in HBV-associated HCCs. According to the Barcelona Clinic Liver Cancer stage (BCLC) classification system, early-stage HCC with portal hypertension is recommended for LT but not SR. Some research has suggested that a low preoperative platelet count (PLT) is independently associated with an increased risk of major complications and mortality after resection,[ whereas other studies showed that thrombocytopenia has no impact on survival.[ There are few studies comparing the survival outcomes of transplantation and hepatectomy based on the preoperative platelet level.

Patients and methods

Patients who underwent SR or LT at the West China Hospital, Sichuan Province, China, from January 1, 2001, to December 31, 2014, were identified from a prospectively maintained database. The diagnosis of HCC was based on either coincident findings using at least 2 techniques (an abdominal ultrasound or computed tomography [CT] or magnetic resonance imaging [MRI]) showing characteristic features of HCC or 1 positive image study with an AFP > 400 ng/mL).[ HCCs were histopathologically confirmed by experienced liver pathologists at the West China Hospital. Satellite lesions are defined as lesions smaller than 2 cm and located within 2 cm of the main tumor.[ A total of 209 patients with Child–Pugh A/B who underwent curative hepatectomy were classified into the SR group. A total of 129 patients (27 patients with Child–Pugh C) who underwent transplantation were categorized in the LT group. Demographic data, blood tests, liver function tests, hepatitis B virus (HBV) infection status, AFP measurements, and tumor characteristics including the liver cirrhosis status, the number of tumors, the maximum tumor size, microvascular invasion (MVI), satellite lesions, and the degree of tumor differentiation were collected. Our inclusion criteria were as follows: (1) primary early-stage HCC within the Milan criteria (a solitary tumor 5 cm or less in diameter or no more than 3 nodules 3 cm or less in diameter without macrovascular invasion or extrahepatic metastasis); and (2) a serum creatinine level less than 124 mmol/L. Exclusion criteria: (1) re-resected HCC in the SR group; (2) salvage LT in the LT group; (3) second primary malignant tumors and (4) poor data integrity. Written informed consent was obtained from all patients and was stored in the hospital database and used for research purposes. This study was approved by the ethics committee of the West China Hospital, and it was conducted in accordance with the Declaration of Helsinki.

Follow-up

Patients with a positive hepatitis B virus (HBV)-DNA load took antiviral drugs orally (Entecavir or adefovir dipivoxil or lamivudine) before and after surgery. After surgery, all patients were regularly followed-up every 1–3 months during the first 2 years and every 3–6 months in the subsequent years. Blood tests, liver function tests, and abdominal ultrasound or CT or MRI were performed, and AFP levels and HBV-DNA levels (if the patient had an HBV infection) were measured in the follow-up examinations. HCC recurrence was confirmed by CT and MRI and/or a rising AFP level or by biopsy when necessary. Re-resection, radiofrequency ablation (RFA), transcatheter arterial chemoembolization (TACE), systematic therapy, and Best Care Support (BCS) were recommended for HCC recurrence by the Multidisciplinary Team (MDT) based on the tumor status and the general condition of the patients. Overall survival (OS) was defined as the time from surgery to death or the latest date of follow-up. Disease-free survival (DFS) was defined as the time from surgery to the day of HCC recurrence including intra-hepatic recurrence and/or distant metastases. The last follow-up date occurred at the end of May 2016. The median follow-up length was 51.3 months, and ranged from 0 to 170 months.

Statistical analysis

Continuous data are expressed as the median ± standard deviation (SD) and were analyzed by the Student's t test. Categorical variables are presented as a number (percent) and were analyzed by the χ2 test or Fisher's exact test. The Kaplan–Meier method was utilized to calculate OS and DFS, and the log-rank test was used to assess differences between survival curves. Cox proportional hazards regression was performed to evaluate the risk factors associated with prognosis. Variables with values of P < 0.1 in the univariate analysis were further included in the multivariate Cox proportional hazards regression analysis.[ AFP cut-off values of 200 and 400 ng/mL were adopted in the assessment of the relationship between AFP levels and survival in both groups.[ Similarly, PLT cut-off values of 80 and 100 ×109/L were used.[ Two-tailed P < 0.05 values were considered statistically significant.

Based on the propensity score, one-to-one nearest-neighbor matching without replacement was adopted to overcome selection bias in both groups using a 0.1 caliper. The propensity score calculated by a logistic regression model represents the probability of each patient being assigned to each surgical approach. Variables possibly affecting postoperative outcomes were entered into the PSM, including age, gender, HBV infection status, Ishak score, total bilirubin (TBIL), alanine aminotransferase (ALT), albumin (ALB), PLT, AFP levels, the number of tumors, the maximum tumor size, the degree of tumor differentiation, MVI and satellite lesions.[ The degree of covariate imbalance in the unmatched and matched samples was measured using the standardized difference.

Results

Clinicopathological characteristics before PSM

As shown in Table 1, the patients in the transplant group were younger and had a higher rate of HBV infection and a lower rate of positive HBV-DNA loads. In the LT group, patients had poorer liver function with higher TIBL and ALT levels and lower ALB levels. In addition, 27 patients had Child–Pugh C liver function in the LT group. As expected, most of the patients in the SR group had Child A cirrhosis. The LT group had a higher rate of patients with thrombocytopenia. Regarding the tumor characteristics, the LT group had a higher rate of MVI, and more patients suffered from multiple HCC and having an Ishak score greater than 5. In contrast, the tumor sizes were larger in the SR group. No other difference was observed between the groups (Table 1).

Table 1

Comparison of preoperative clinicopathologic data of patients receiving liver transplantation (LT) or surgical resection (SR) before propensity score matching.

Survival analysis in both groups

Until the last follow-up, recurrence occurred in 107 patients in the SR group and 8 patients in the LT group (P < 0.001). Most patients developed intrahepatic recurrence (86 [41.1%] in the SR group and 5 [3.9%] in the LT group). After the detection of HCC recurrence, salvage LT (SR:6 ([2.9%] vs LT:0), re-resection (SR: 22 [10.5%] vs LT: 0 [0%]), RFA (SR: 14 [6.7%] vs LT: 1 [0.8%]), TACE (SR:37 [17.7%] vs LT: 1 [0.8%]) and BCS (SR: 23 [11.0%] vs LT: 6 [4.7%]) were preformed (Table 1). Prognostic factors that had a significant association (P < 0.05) with DFS in the univariate and multivariate analyses are shown in Table 2. Age, gender, a positive HBV-DNA load, Child–Pugh, TBIL, ALT, ALB, satellite lesions, and treatment method were statistically significant in the univariate model. The Cox model included gender (Hazard ratio [HR] 2.158, 95% confidence interval [CI] 1.089–4.276, P = 0.028), satellite lesions (HR 2.053, 95%CI 1.067–3.950, P = 0.031) and treatment method (HR 8.108, 95%CI 3.929–16.731, P < 0.001) (Table 2). Similarly, the univariate analysis suggested that the treatment method was significantly associated with OS. The multivariate analysis showed that the treatment method (SR vs LT: HR 1.888, 95% CI 1.133–3.147, P = 0.015) and Ishak score (HR 1.196 95%CI 1.034–3.551, P = 0.039) were independently associated with OS in early-stage HCC patients (Table 3). In the LT group, the 1-, 3- and 5-year OS rates were 88.9, 84.5 and 83.1%, respectively, and the DFS rates were 98.3, 92.6 and 92.6%, respectively. In the SR group, the 1-, 3-, and 5-year OS rates were 98.6, 84.3, and 67.1%, respectively, and the DFS rates were 85.4, 61.1, and 48%, respectively. Both the OS and DFS rates were significantly higher in the LT group (P = 0.032 and P < 0.001). In particular, the OS and DFS rates were no different in the LT group when stratified by liver function (Fig. 1).

Table 2

Univariate and multivariate analyses to identify factors related to disease-free survival in all patients.

Table 3

Univariate and multivariate analyses to identify factors related to overall survival in all patients.

Figure 1

Survival analysis before propensity score matching. The disease-free survival (DFS) (A) and overall survival (OS) rates (B) were higher in the liver transplantation (LT) group than in the surgical resection (SR) group. Child–Pugh classification had no impact on the DFS (C) and OS (D) in the LT group. DFS = disease-free survival, LT = liver transplantation, OS = overall survival, SR = surgical resection.

Propensity score matching analysis in both groups

Since LT is the only choice to cure Child–Pugh C liver disease, we excluded patients with Child–Pugh C cirrhosis in the LT group in subsequent PSM analyses. Finally, 102 patients with Child–Pugh A/B in the LT group and 209 patients with Child–Pugh A/B in the SR group were included in the PSM model. Patients treated with LT or SR were matched one-to-one using PSM to eliminate confounding factors. Clinicopathological variables entered into the PSM analysis were age, gender, status of HBV infection, Ishak score, TBIL, ALT, ALB, PLT, AFP level, the number of tumors, the maximum tumor size, the degree of tumor differentiation, MVI and satellite lesions. In total, 46 pair patients were matched in both groups. There were no significant differences in the clinicopathological variables between the 2 matched groups (Table 4). The matching outcome of the propensity model was good (Fig. 2).

Table 4

Comparison of preoperative clinicopathologic data patients receiving liver transplantation (LT) or surgical resection (SR) after propensity score matching.

Figure 2

Covariate balance was improved in the matched samples. (A) Parallel line plot of the standardized difference in means before and after PSM in HCC patients within the Milan criteria. (B) Even distribution of propensity scores in the matched groups. HCC = hepatocellular carcinoma, PSM = propensity score matching.

Comparisons of the OS between both groups after PSM are shown in Fig. 3: after PSM, in the LT group, the 1-, 3-, and 5-year OS rates were 89.0, 83.2, and 83.2%, respectively, and the DFS rates were 95.3, 91.8, and 87.8%, respectively. In the SR group, the 1-, 3-, and 5-year OS rates were 97.8, 79.6, and 63.6%, respectively, and the DFS rates were 81.3, 61.1, and 36.4%, respectively. The DFS rate remained higher in the LT group (P < 0.001). However, there was no statistically significant difference in terms of the OS between patients receiving LT and SR (P = 0.124).

Figure 3

Survival analysis after propensity score matching. (A) The liver transplantation (LT) group had a higher rate of disease free survival (DFS). (B) The surgical resection (SR) group achieved a comparable overall survival (OS) rate. DFS = disease-free survival, LT = liver transplantation, OS = overall survival, SR = surgical resection.

Subgroup analysis by AFP and PLT levels

When the cutoff value of the AFP was 200 ng/mL, there were no statistically significant differences in terms of the OS between the LT and SR groups. When the cutoff value of the AFP was 400 ng/mL, in the AFP < 400 ng/mL subgroup, patients in the LT group achieved similar long-term survival compared to patients in the SR group. However, in the AFP ≥ 400 ng/mL subgroup, the 5-year OS rate for the 12 patients in the LT group was 93.3%, which was significantly greater than that of the 15 patients in the SR group (48.6%, P = 0.030) (Fig. 4).

Figure 4

Survival analysis stratified by the AFP level. (A–C) There was no difference in terms of overall survival (OS) when the AFP cutoff value was 200 ng/mL or 400 ng/mL. (D) Patients with an AFP ≥ 400 ng/mL had a better prognosis after LT than SR. AFP = alpha-fetoprotein, LT = liver transplantation, OS = overall survival, SR = surgical resection.

In the PLT < 80×109/L subgroup, the 3-year OS rates for the LT and SR groups were 87.5 and 82.8%, respectively (P = 0.402). In the PLT ≥ 80×109/L subgroup, the 5-year OS rates for the LT and SR groups were 80.3 and 76.0%, respectively (P = 0.157). In the PLT < 100×109/L subgroup, the 3-year OS rates for the LT and SR groups were 86.7 and 83.3%, respectively (P = 0.528). In the PLT ≥ 100×109/L subgroup, the 3-year OS rates for the LT and SR groups were 80.4 and 66.0%, respectively (P = 0.080) (Fig. 5).

Figure 5

Survival analysis stratified by the PLT count. There was no difference in terms of overall survival (OS) when the PLT count cutoff value was 80 or 100 × 109/L. OS = overall survival, PLT = platelet count.

Discussion

This study was designed to investigate the survival benefits of early-stage HCC patients after LT and SR treatments, with particular attention to similar baseline characteristics. Liver transplantation and hepatectomy are the main curative methods for early stage HCC. There is no consensus with regard to the optimal treatment for early stage HCC.[ In an era of extreme organ shortage, the status of tumor characteristics varies partly due to different waiting times in different geographic locations.[ The majority of the literature did not take similar baseline clinicopathological variables into account. We hold the view that the conclusions are more convincing if the outcomes are compared between the 2 treatments based on similar baseline patients’ characteristics. Therefore, we used a PSM analysis in the current study to balance the baseline clinical and pathological characteristics. Before the PSM analysis, there was a difference in some of the variables, such as the number of tumors and tumor size, among others, which might be associated with the prognosis. According to the Kaplan–Meier analysis, patients with HCC meeting the Milan criteria who received LT had a significantly better prognosis than patients receiving SR. However, after the PSM analysis, although the DFS rate in the LT patients remained higher than that in the SR group, the OS rate was comparable between both groups. Consistently, researchers have believed that SR for patients with HCC within the Milan criteria can achieve satisfactory long-term survival.[ The implications of the above findings are that, for patients with HCC within the Milan criteria, surgical resection may be an equivalent alternative treatment to transplantation. This is of great value in areas with an organ shortage. In this study, we also showed that the prognosis of Child–Pugh C status HCC patients within the Milan criteria was similar to that of Child–Pugh A/B status HCC patients after transplantation. Thus, we support the view that early-stage HCC patients with decompensating liver function should be given priority since transplantation is the only therapy for them.

In the present study, satellite lesions and treatment methods were predictors of recurrence. The presence of satellite lesions indicated that intrahepatic metastasis might spread by invading portal vein branches.[ Satellite lesions have been demonstrated to be associated with tumor recurrence.[ Wide surgical margins and the elimination of the underlying liver cirrhosis are beneficial factors. The transplantation group had a low rate of HCC recurrence. Recently, data from 2046 patients showed that liver cirrhosis might increase the risk of death after hepatectomy for HCC.[ Similarly, our study demonstrated this point with an increase in mortality of 1.331-fold.

In the clinic, thrombocytopenia predominantly caused by hypersplenism in the liver disease is reflective of portal hypertension. The current BCLC staging system recommends transplantation and resection for HCC with portal hypertension; however, surgical resection for HCC patients with portal hypertension remains controversial.[ Some authors reported that thrombocytopenia negatively impacts liver function recovery after hepatectomy, as well as after transplants.[ Two meta-analyses have also showed that thrombocytopenia acts as an unfavorable factor associated with the prognosis.[ However, in our study, the platelet count was not validated as a predictor of the prognosis. The survival rates between the transplantation and surgical resection groups were comparable. In our hospital, all HCC patients who underwent liver resection receive a liver function test (an ICG test) and residual liver volume measurements preoperatively. No perioperative mortality was observed in the SR group. Our previous study showed that for patients with HCC and hypersplenism, survival benefits are obtained by surgical resection and a splenectomy.[ Thus, for HCC patients with preserved liver function, thrombocytopenia might not be a contraindication for surgical resection.

Our study also proves that patients with an AFP ≥ 400 ng/mL achieved greater survival benefits from transplantation, whereas there was no statistically significant difference in terms of OS in patients with an AFP < 400 ng/mL after both treatments. This was consistent with the findings of a prior study that suggested that a serum AFP level above 400 ng/mL predicted a poor prognosis.[ Many studies showed that elevated AFP levels are a risk factor related to the prognosis via the promotion of HCC proliferation and metastasis through alterations in the cancer-related signal transduction pathway.[

There are some limitations to our study. First, this is a retrospective study. SR patients were mainly collected from 2007 to 2014, whereas LT patients were mainly collected from 2001 to 2014. The selected patients do not represent the entire early-stage HCC patient population, but our results still serve as valuable as references. Second, we used PSM analysis to reduce to selection bias. However, the small numbers in the propensity model might cause bias in the survival analysis. Third, the follow-up time was relatively short. This also might lead to unconvincing conclusions. Therefore, we need to conduct further studies.

Conclusions

Our propensity score model provided evidence that, compared with transplantation, surgical resection could result in comparable long-term survival for resectable early-stage HCC patients, except for HCC patients with an AFP ≥ 400 ng/mL. Surgical resection might not be a contraindication for early-stage HCC patients with thrombocytopenia due to similar prognoses after transplantation.