Significance of Des-Gamma-Carboxy Prothrombin in Selection Criteria for Living Donor Liver Transplantation for Hepatocellular Carcinoma

Authors


* Corresponding author: Yasutsugu Takada, takaday@kuhp.kyoto-u.ac.jp

Abstract

Des-gamma-carboxy prothrombin (DCP) levels reportedly correlate with histological features of hepatocellular carcinoma (HCC). We examined serum DCP as a predictor of HCC recurrence in 144 patients who underwent living donor liver transplantation. Receiver operating characteristics (ROC) analysis revealed superiority of DCP and AFP over preoperative tumor size or number for predicting recurrence. Multivariate analysis revealed tumor size >5 cm, ≥11 nodules, and DCP >400 mAU/mL as significant independent risk factors for recurrence. Incidence of microvascular invasion (62% vs. 27%, p = 0.0003) and poor differentiation (38% vs. 16%, p = 0.0087) were significantly higher for patients with DCP >400 mAU/mL than for patients with DCP ≤400 mAU/mL. In ROC analysis for patients with ≤10 nodules all ≤5 cm to predict recurrence, area under the curve was much higher for DCP than for AFP (0.84 vs. 0.69). Kyoto criteria were thus defined as ≤10 nodules all ≤5 cm, and DCP ≤400 mAU/mL. The 5-year recurrence rate for 28 patients beyond-Milan but within-Kyoto criteria was as excellent as that for 78 patients within-Milan criteria (3% vs. 7%). The preoperative DCP level offers additional information regarding histological features, and thus can greatly improve patient selection criteria when used with tumor bulk information.

Abbreviations: 
DCP

des-gamma-carboxy prothrombin

AFP

α-fetoprotein

OLT

orthotopic liver transplantation

HCC

hepatocellular carcinoma

MC

Milan criteria

LDLT

living donor liver transplantation

DDLT

deceased donor liver transplantation

CT

computed tomography

MDCT

multidetector computed tomography

UCSF

University of California, San Francisco

MELD

model for end-stage liver disease

HCV

hepatitis C virus

HBV

hepatitis B virus

Introduction

Des-gamma-carboxy prothrombin (DCP), also known as protein induced by vitamin K absence or antagonist II (PIVKA-II), is an abnormal form of prothrombin. Since 1984, when Liebman et al. (1) first identified DCP in the plasma of patients with hepatocellular carcinoma (HCC), DCP has been used as an important tumor marker in the diagnosis of HCC, particularly in Japan (2–4).

Several groups have recently reported correlations between DCP level and histopathological findings such as portal venous invasion (4–6) or high grade of HCC differentiation (7). These findings may indicate that pretreatment serum DCP levels have a great impact on patient outcomes, because portal vein invasion increases the risk of intrahepatic metastasis or distant recurrence (8,9). Indeed, previous studies have demonstrated that a high DCP level at the time of HCC diagnosis is associated with poor prognosis (5,10–12). Moreover, the high DCP levels have been considered as a predictor of recurrence in patients treated with conventional treatments such as surgical resection (9,10,13,14), transcatheter arterial embolization (TAE) (15,16) and percutaneous ablation therapy (17,18).

In this decade, orthotopic liver transplantation (OLT) has been added to the current options for the treatment of HCC. Several groups have recently proposed new extended criteria beyond Milan criteria (MC) (19) for HCC based on tumor number and size (20–25). However, few attempts have been made to properly clarify the usefulness of DCP as a predictor of HCC recurrence after OLT. In the field of transplantation as well as in other treatments, pathological findings such as portal invasion (23,26–32) and poor differentiation of HCC (20,27,28,33) have been associated with poor outcome. We believe that prior to transplantation, DCP can provide further information related to the pathological features of HCC and patient outcomes. The present study analyzed relationships between serum DCP and histopathological findings of HCC in the explanted liver and assessed the utility of DCP as a predictor of recurrence after living donor liver transplantation (LDLT).

Patients and Methods

Patients

Between February 1999 and August 2007, a total of 785 patients underwent LDLT at the Kyoto University. Of these, 144 patients with HCC were analyzed in this study, excluding 16 patients with incidental HCC that had not been diagnosed by pretransplant imaging modalities. Of the 144 patients, 104 patients (72%) displayed a history of previous treatment for HCC using various nontransplant methods including TAE, percutaneous ablation therapy or hepatic resection. These treatments were performed not as a bridge to transplant or for down staging, but with intent for curative ablation. The patients suffering from uncontrolled recurrent HCC were subsequently referred to our department for LDLT as a second-line treatment (33).

Our institutional selection criteria for LDLT for HCC included (i) HCC unsuitable for resection or local ablation therapies; and (ii) exclusion of extrahepatic metastases or macroscopic venous invasion on preoperative imaging (34). The standard immunosuppression protocol comprised tacrolimus and low-dose steroid (34). As of the end of February 2009, the median duration of follow-up was 41 months (range, 1–120 months).

Staging classification

Staging of HCC was based on pretransplant imaging. Evaluation of the extent of tumor involvement using contrast-enhanced multidetector computed tomography (MDCT) was usually performed within 1 month before LDLT. Tumor staging was determined by counting only viable and enhancing nodules on MDCT. For patients who had undergone previous treatment, resected tumors or nodules that were judged as nonviable were not counted. The individual HCC was staged according to MC (19), as solitary tumor ≤5 cm, or more than or equal to three nodules all ≤3 cm, and University of California and San Francisco (UCSF) criteria (25) as solitary tumor ≤6.5 cm, or more than or equal to three nodules all ≤4.5 cm and total tumor diameter ≤8 cm. A total of 79 patients (54%) met MC and 83 patients (58%) met UCSF criteria. Of the 104 patients with a history of previous nontransplant treatment, records of tumor stage at first diagnosis of HCC were available for 99 patients. At the time of LDLT, tumor stage was up-staged in 32 patients, down-staged in 7 and unchanged in 60, according to MC.

Preoperative measurement of serum DCP and α-fetoprotein (AFP) levels

Serum DCP levels were determined by electrochemiluminescence immunoassay (Picolumi PIVKA-II kit; Sanko Junyaku, Tokyo, Japan). The cut-off value is set by the manufacturer at 40 mAU/mL for HCC diagnosis. Serum AFP levels were determined by EIA and the cut-off value is set at 20 ng/mL by the manufacturer. Serum DCP and AFP were measured within 1 month prior to LDLT.

Histological analysis

Histological examination was performed in the explanted liver. The length of the longest axis of HCC was defined as the tumor diameter. When multiple HCC were present, the largest tumor was measured and taken as a representative HCC diameter. HCC number was determined by counting viable nodules only.

Each HCC was histologically graded into one of three categories according to the modified Edmondson criteria: well differentiated; moderately differentiated; or poorly differentiated. When the tumor consisted of more than or equal to two grades of histological differentiation, the most progressive grade was used.

Receiver operating characteristic (ROC) analysis

ROC analysis was used to evaluate the ability of tumor variables to predict postoperative recurrence and to choose the optimal cut-off value for subsequent analysis.

For selection of optimal cut-off values, the concordance index (C-index) was calculated for each cut-off point on the ROC curve (35,36). The C-index for a cut-off point is defined as the area of the quadrilateral with vertices on the cut-off point on the ROC curve and points (0,0), (1,0) and (1,1) on the ROC graph. This value estimates the probability that the predictions and outcomes are concordant. The C-index was calculated using the following formula:

image

For indication of LDLT, high specificity is necessary to avoid excluding a large number of patients who would not develop recurrence. We therefore defined the optimal cut-off value as the point showing the highest C-index among values with specificity ≥0.85.

Statistical analysis

Cumulative overall survival and recurrence rates were calculated using Kaplan–Meier methods, and differences between curves were evaluated using log-rank testing. The χ2 test or Student's t-test was used to compare differences between subgroups. The Cox proportional hazards model was used to identify the independent risk factors for recurrence. For comparison of two areas under the curve (AUCs), the nonparametric method developed by Hanley and McNeil was employed, and the Z-statistic was calculated (37,38). The values of p < 0.05 were considered significant. All statistical analyses were performed using the StatView 5 statistical software (Abacus Concepts, Berkeley, CA, USA).

Results

Preoperative patient profile

Patient profiles and pretransplant clinical characteristics are shown in Table 1. The median age for the 144 patients (103 men, 41 women) was 54.5 years. Hepatitis C was the most common etiological factor, followed by hepatitis B. The median serum DCP was 47.5 mAU/mL, and the median serum AFP was 30.5 ng/mL. Forty-four percent of patients with HCC had a normal DCP level (<40 mAU/mL) and 38% had a normal AFP level (<20 ng/mL). The median size and the number of HCCs detected on pretransplant imaging were 3.0 cm (range, 0.8–22.0 cm) and three (range, 1–186), respectively.

Table 1.  Patient characteristics
VariableValue
  1. 1Data are given as median (range).

Age (years)54.5 (23–69)1
Gender (male/female)103/41
DCP (mAU/mL)47.5 (5.0–20 600.0)1
AFP (ng/mL)30.5 (0.9–212 220.0)1
Tumor size (cm)3.0 (0.8–22.0)1
No. of tumor nodules3 (1–186)1
Etiology 
 HCV79
 HBV46
 HCV and HBV 6
 Alcohol 4
 PBC 3
 Others 6

Only one patient showed a tumor ≥10 cm. This patient with a 22.0 cm tumor had undergone ligation of the tumor-feeding artery before referral to our hospital for LDLT. Since pretransplant computed tomography (CT) showed that a large area of the 22 cm tumor was necrotic and no sign of extrahepatic metastasis or macroscopic venous invasion was identified, LDLT was performed. The patient developed recurrent HCC at the diaphragm 8 months after transplant and died 29 months following LDLT.

Two patients showed ≥100 nodules (100 and 186 nodules) on pretransplant imaging. The largest tumors were <1 cm in diameter for the patient with 100 nodules, and 2.9 cm in diameter for the patient with 186 nodules. Both patients were referred to our hospital for LDLT after several treatments, including tumor resection or TAE. Both patients are currently alive without HCC recurrence after follow-up of 48 and 16 months, respectively.

Postoperative patient survival

As of the end of February 2009, a total of 98 patients remained alive, with 20 patients dying of recurrent HCC and 26 patients dying of tumor-unrelated causes, mainly infectious complications. Overall patient survival rate at 5 years was 68%. Five-year survival rates were 77% for patients within MC and 62% for those beyond MC (p = 0.1408).

Postoperative HCC recurrence and ROC curve analysis

Postoperative recurrence of HCC was identified in 22 patients. The ability of preoperative tumor variables to predict HCC recurrence was analyzed by ROC curves (Figure 1). For ROC analysis, only patients followed for ≥24 months without HCC recurrence were considered to be free of recurrence, since 19 of the 22 recurrences (86%) occurred within 24 months after LDLT. ROC analysis was performed using 22 patients with recurrence and 95 patients free of recurrence. AUCs for DCP, AFP, tumor size and tumor number were 0.77, 0.76, 0.68 and 0.69, respectively. AUC was largest for DCP followed by that of AFP, although no significant differences were seen among the four tumor variables (DCP vs. tumor size, p = 0.288; DCP vs. tumor number, p = 0.358; AFP vs. tumor size, p = 0.340; AFP vs. tumor number, p = 0.334). Among the cut-off values with sufficient specificity, the cut-off point with the highest C-index was chosen as the optimal cut-off value for subsequent analysis. The selected cut-off values were 400 mAU/mL for DCP ([C-index, sensitivity, specificity]=[0.71, 0.55, 0.88]), 800 ng/mL for AFP (0.68, 0.45, 0.93), 5 cm for tumor size (0.64, 0.32, 0.96) and 10 for tumor number (0.68, 0.41, 0.95).

Figure 1.

Receiver operating characteristic (ROC) curves of DCP (A), AFP (B), tumor size (C) and number of tumors (D) for predicting posttransplant HCC recurrence. The size and number of tumors were based on preoperative imaging. Among cut-off values with sufficient specificity, the value with the highest C-index was selected as the optimal cut-off value for risk analysis for HCC recurrence. The figure near the square indicates the selected cut-off value. Areas under the curve (AUCs) were 0.76 (A), 0.74 (B), 0.68 (C) and 0.69 (D), respectively. When cut-off values were set at 400 mAU/mL for DCP, 800 ng/mL for AFP, 5 cm for tumor size and 10 for tumor number, C-index was 0.71, 0.68, 0.64 and 0.68, respectively. Sensitivity and specificity were 0.55 and 0.88, 0.45 and 0.93, 0.32 and 0.96 and 0.41 and 0.95, respectively. Z-statistics were 1.06 (p = 0.288) for DCP vs. tumor size, 0.919 (p = 0.358) for DCP vs. tumor number, 0.956 (p = 0.34) for AFP vs. tumor size and 0.964 (p = 0.334) for AFP vs. tumor number.

Preoperative variables as risk factors for recurrence

For the total of 144 patients, overall cumulative recurrence rate was 18% at 5 years. Univariate analysis was performed to identify risk factors for recurrence among preoperative variables. First, for continuous variables, univariate Cox proportional hazard modeling revealed DCP (p < 0.0001) and AFP (p = 0.0030) as significant risk factors for recurrence. Subsequent analysis was then performed using the cut-off values selected by ROC analysis. Among preoperative variables, exceeding MC (p = 0.0002), exceeding UCSF criteria (p = 0.001), tumor size >5 cm (p < 0.0001), number of tumors ≥11 (p < 0.0001), DCP >400 mAU/mL (p < 0.0001) and AFP >800 ng/mL (p < 0.0001) were identified as significant risk factors for recurrence (Table 2).

Table 2.  Univariate analyses of recurrence of HCC on preoperative variables
Preoperative variable (n)pIncidence of recurrence (%)
1 year3 years5 years
Age0.1466   
 ≤60 years (117) 112021
 >60 years (27)  4 8 8
Gender0.6073   
 Male (103)  91919
 Female (41) 111116
Anti-HCV Antibody0.0699   
 Positive (85)  51212
 Negative (58) 152326
HBs antigen0.7007   
 Positive (52) 131720
 Negative (91)  81717
Milan criteria0.0002   
 Meeting (79)  1 5 7
 Exceeding (65) 193232
UCSF criteria0.0012   
 Meeting (84)  4 7 9
 Exceeding (60) 173131
Tumor size<0.0001    
 ≤3 cm (96)  71213
 >3 cm, ≤5 cm (35)  71414
 >5 cm (13) 407373
No. of tumors<0.0001    
 ≤3 (94) 61113
 ≥4, ≤10 (32) 71010
 ≥11 (18) 326363
DCP<0.0001    
 ≤100 mAU/mL (87)  5 8 9
 >100 mAU/mL, ≤400 mAU/mL (26)  51616
 >400 mAU/mL (29) 295151
AFP<0.0001    
 ≤200 ng/mL (107) 3 910
 >200 ng/mL, ≤800 ng/mL (16) 212121
 >800 ng/mL (21) 356060

Furthermore, multivariate Cox regression analyses revealed that, of the four variables identified as risk factors of recurrence in univariate analysis, DCP >400 mAU/mL (p = 0.0006), number of tumors ≥11 (p = 0.0147) and tumor size >5 cm (p = 0.0269) were independently associated with recurrence, whereas AFP >800 ng/mL was not. DCP >400 mAU/mL was the strongest predictor of recurrence (relative risk = 4.863), followed by tumor size >5 cm (Table 3).

Table 3.  Multivariate Cox proportional hazard analysis of recurrence of HCC with preoperative variables
Preoperative variableRelative risk95%CIp-Value
DCP >400 mAU/mL4.863 1.982–11.9340.0006
AFP >800 ng/mL2.0840.732–5.9280.1688
Tumor size >5 cm3.3241.147–9.6350.0269
No. of tumors ≥113.2191.258–8.2320.0147

Based on these results, to minimize the risk of recurrence, we defined new extended criteria (Kyoto criteria) as tumor size ≤ 5 cm, ≤10 nodules and DCP ≤400 mAU/mL. The 5-year recurrence rate for patients within Kyoto criteria (n = 97) was significantly lower than that for patients beyond Kyoto criteria (n = 45) (3% vs. 54%, p < 0.0001). In addition, to compare the additional benefit of DCP with that of AFP in the 5–10 rule (tumor size ≤5 cm, ≤10 nodules) for predicting HCC recurrence, ROCs for DCP and AFP among patients within the 5–10 rule were analyzed (Figure 2). AUC was markedly higher for DCP than for AFP (0.84 vs. 0.69).

Figure 2.

ROC curves of DCP and AFP for predicting posttransplant HCC recurrence for patients within the 5–10 rule. Among patients within the 5–10 rule, 8 patients with HCC recurrence and 88 patients free of recurrence with follow-up ≥24 months after LDLT were used for ROC analysis. AUC tended to be markedly higher for DCP than for AFP, although no significant difference was apparent (0.84 vs. 0.69, p = 0.26).

Postoperative histopathological variables and relationship to tumor markers

Histological grade was well differentiated in 24 patients (17%), moderately differentiated in 91 patients (63%) and poorly differentiated in 29 patients (20%). Microvascular invasion was present in 49 patients (34%).

Correspondence of preoperative MC with the results of pathological analysis was evaluated in postoperative examinations of explanted livers. Of the 79 patients within preoperative MC, 61 patients met so-called pathological MC and 18 did not. HCC recurred in four of these 18 underestimated patients, whereas no recurrences were seen in the 61 patients within ‘pathological MC’. Of the 65 patients beyond preoperative MC, 58 patients were beyond and 7 patients were within ‘pathological MC’. HCC recurred in 18 of 58 patients beyond both preoperative and pathological MC, and no recurrences were noted in the 7 patients within ‘pathological MC’.

Univariate analysis revealed that among the postoperative pathological variables, size >5 cm (p < 0.0001), number of tumors ≥11 (p < 0.0001), presence of microvascular invasion (p = 0.0001) and poorly differentiated HCC (p < 0.0001) represented significant risk factors for recurrence (Table 4).

Table 4.  Univariate analyses of recurrence of HCC on postoperative variables
Variable (n)p-ValueIncidence of recurrence (%)
1 year3 years5 years
Postoperative variable    
 Tumor size in explanted<0.0001   
   liver    
   ≤3 cm (89)  1 5 7
   >3 cm, ≤5 cm (36) 172222
   >5 cm (19) 356969
 Number of tumors in<0.0001   
   explanted liver    
   ≤3 (83)  3 4 4
   ≥4, ≤10 (39) 112428
   ≥11 (22) 315555
 vp Status<0.0001   
   negative (95)  2 810
   positive (49) 233636
 Pathological grading<0.0001   
 Well or moderately  2 911
   differentiated (115)    
 Poorly differentiated (29) 425050

To investigate relationships between pathological variables and tumor markers, microvascular invasion (vp)-positive status and poor differentiation were compared among the three groups stratified according to serum DCP or AFP levels (Figure 3). Tumors with higher serum DCP or AFP showed a higher incidence of vp-positive status and poorly differentiated grade. In particular, tumors with DCP > 400 mAU/mL showed a significantly higher incidence of vp-positive status (62% vs. 27%, p = 0.0003) and poorly differentiated grade (38% vs. 16%, p = 0.0087) compared with tumors with DCP ≤400 mAU/mL. Furthermore, tumors with DCP >400 mAU/mL showed a significantly higher incidence of vp-positive status (p = 0.0036) and a markedly higher incidence of poorly differentiated grade (p = 0.0609) compared with tumors showing DCP 100– 400 mAU/mL (Figure 3A). Conversely, incidence of vp-positive status and poor differentiation increased gradually as AFP increased (Figure 3B). When comparing tumors with AFP >800 ng/mL and tumors with AFP 200–800 ng/mL, differences in incidence of these histological findings were not significant (vp-positive status, p = 0.419; poorly differentiated grade, p = 0.4708).

Figure 3.

Incidence of microvascular invasion and poorly differentiated grade according to serum DCP (A) and AFP level (B). Figures near bars indicate the percentage of each pathological feature. (A) Patients with DCP >400 mAU/mL showed significantly higher incidence of positive microvascular invasion and markedly higher incidence of poorly differentiated grade compared to patients with DCP ≤100 mAU/mL (p < 0.01, p < 0.01, respectively) and patients with DCP 100–400 mAU/mL (p < 0.01, p = 0.06, respectively). (B) Patients with AFP >800 ng/mL showed a significantly higher incidence of positive microvascular invasion and poorly differentiated grades compared to patients with AFP ≤200 ng/mL (p < 0.01, p = 0.01, respectively), but differences in these histological findings were not significant compared to patients with AFP 200–800 ng/mL (p = 0.42, p = 0.47, respectively).

Multivariate analysis revealed that among three preoperative variables of tumor size and number (size >5 cm or number ≥11), DCP >400 mAU/mL and AFP >800 ng/mL, only the DCP level was independently associated with vp-positive status and poorly differentiated grade (Table 5).

Table 5.  Multivariate logistic regression analysis to investigate the predictability of preoperative tumor variables for (A) microvascular portal invasion and (B) poorly differentiated grade
VariableRelative risk95%CIp-Value
A) Microvascular invasion   
 Tumor size >5 cm or no.2.4290.887–6.6570.0843
   of tumor ≥11   
 DCP >400 mAU/mL3.4331.389–8.4830.0075
 AFP >800 ng/mL2.8020.922–8.5200.0693
B) Poor differentiation   
 Tumor size >5 cm or no.1.0760.355–3.2630.897 
   of tumors ≥11   
 DCP >400 mAU/mL2.5981.006–6.7130.0487
 AFP >800 ng/mL2.9400.955–9.0490.0601

Histological features and various selection criteria

Correlations between the two histological features and various criteria of OLT for HCC based on pretransplant evaluation including MC, UCSF, 5–10 rule and Kyoto criteria are summarized in Table 6. In comparing the prevalence of these histological features between 78 patients meeting MC and 64 patients exceeding MC, a significant difference was apparent for microvascular invasion (p < 0.0001), but not for poorly differentiated grade (p = 0.2203). When tumor criteria were extended to UCSF criteria, the difference in the prevalence of poorly differentiated tumor between meeting and exceeding criteria remained subtle (p = 0.39) and the difference for microvascular invasion remained significant (p < 0.0001). Correspondingly, the difference in the prevalence of poorly differentiated tumor for the 5–10 rule was also not significant (p = 0.1476). Contrasting with these results, when comparing 97 patients meeting Kyoto criteria and 45 patients exceeding Kyoto criteria, differences in prevalences of vp-positive status and poorly differentiated grade were both highly significant (p < 0.0001, p = 0.0023, respectively).

Table 6.  Pathological features of HCC according to various criteria
Tumor criteriavp-positive statusp-ValuePoorly differentiated gradep-Value
%Number%Number
  1. The difference in the prevalence of microvascular invasion was significant for each set of criteria, whereas the difference in the prevalence of poorly differentiated grade of tumor was significant only for Kyoto criteria (p = 0.0023).

Milan criteria      
 Meeting2116 of 78<0.00011613 of 780.2203
 Exceeding5032 of 64 2516 of 64 
UCSF criteria      
 Meeting2017 of 83<0.00011815 of 840.3900
 Exceeding5331 of 59 2414 of 59 
5–10 Rule (size ≤5 cm and no. ≤10)      
 Meeting2832 of 1160.00091821 of 1160.1476
 Exceeding6216 of 26 318 of 26 
Kyoto criteria (size ≤5 cm, no. ≤10 and DCP ≤400 mAU/mL)      
 Meeting2120 of 97<0.00011313 of 970.0023
 Exceeding6228 of 45 3616 of 45 

Analysis of recurrence and survival rates according to MC and Kyoto criteria

Patients were stratified into three groups according to MC and Kyoto criteria: patients within MC (n = 78); patients beyond MC but within Kyoto criteria (n = 28); and patients beyond Kyoto criteria (n = 36). Recurrence rates for these groups are shown in Figure 4A. The 5-year recurrence rates were similar for patients within MC (7%) and patients beyond MC but within Kyoto criteria (4%). However, the recurrence rate was significantly higher for patients beyond Kyoto criteria (55%) than for the other two groups (vs. within MC, p < 0.0001; vs. beyond MC but within Kyoto criteria, p < 0.0001).

Figure 4.

Cumulative recurrence rate (A) and patient survival rate (B) according to Milan and Kyoto criteria. Patients were stratified into three groups of ‘within Milan criteria’ (n = 79), ‘beyond Milan but within Kyoto criteria’ (n = 28) and ‘beyond Kyoto criteria’ (n = 36). (A) Five-year recurrence rates were similar for patients within Milan criteria (7%) and beyond Milan but within Kyoto criteria (4%). However, the recurrence rate was significantly higher for those beyond Kyoto criteria (55%) than for the other two groups (vs. within Milan, p < 0.0001; vs. beyond Milan but within Kyoto; p < 0.0001). (B) Five-year survival rates were 89%, 78% and 40% for patients beyond Milan but within Kyoto criteria, within Milan criteria and beyond Kyoto criteria, respectively. The survival rate was significantly lower for patients beyond Kyoto criteria than for the other two groups (vs. within Milan, p < 0.0001; vs. beyond Milan but within Kyoto criteria, p < 0.0001).

Patient survival rates for the three groups are shown in Figure 4B. The 5-year survival rates were 89%, 78% and 40% for patients beyond MC but within Kyoto criteria, patients within MC and patients beyond Kyoto criteria, respectively. The survival rate was significantly lower for patients beyond Kyoto criteria than for the other two groups (vs. within MC, p < 0.0001; vs. beyond MC but within Kyoto criteria, p < 0.0001).

Discussion

Since a substantial subset exists beyond MC with the potential for good patient outcomes after OLT for HCC, several institutions have recently proposed new criteria. These new criteria have extended the MC mainly based on tumor number and size (21,24,25), but new markers of biological behavior are needed in addition to morphological tumor size and number to explore optimal criteria that can reasonably predict the risk of recurrence. Several groups have identified microvascular invasion and poorly differentiated grade as independent predictors of recurrence after OLT (26–32). However, these histological findings of HCC have been left out of newly proposed criteria, as histological information is difficult to ascertain prior to OLT. Attempts to obtain histological information by the preoperative needle core biopsy have been discouraged due to the risk of tumor seeding along the biopsy tract (39). The ability to obtain histological information prior to OLT without the need for invasive maneuvers would thus be highly beneficial for patient selection. We have identified pretransplant serum DCP as strongly correlated with histological findings in the explanted liver, along with high recurrence rate of HCC after LDLT.

ROC analysis revealed that the abilities of pretransplant DCP and AFP to predict HCC recurrence after LDLT tended to be superior to tumor size or number, although the difference was not significant. According to C-index analysis based on ROC, optimal cut-off values to predict recurrence were set at 400 mAU/mL for DCP, 800 ng/mL for AFP, 5 cm for tumor size and 10 for tumor number. Patients with DCP >400 mAU/mL showed significantly higher recurrence rates of HCC after LDLT compared to patients with DCP ≤400 mAU/mL. In addition, HCC with DCP >400 mAU/mL tended to recur earlier than HCC with DCP ≤400 mAU/mL. The latest recurrence of patients with DCP >400 mAU/mL was observed 22 months after LDLT, whereas that with DCP ≤400 mAU/mL was seen 44 months after LDLT. In multivariate analysis, DCP >400 mAU/mL was found to be the most significant and independent predictor for HCC recurrence among four risk factors regarded as predictors of recurrence in univariate analysis (Table 3). Conversely, AFP was not considered as an independent predictor of recurrence in this study. In comparing DCP with AFP in ROC analysis for patients within the 5–10 rule, AUC of DCP was much higher at 0.84 than that of AFP at 0.69, which resembled the AUC of tumor size or tumor number alone (0.68 or 0.69, respectively). This indicates that AFP would offer little on addition to criteria based on tumor size and number. Given these results, DCP, but not AFP, was associated with recurrence, independent of tumor size or number.

These findings may indicate that tumors with DCP >400 mAU/mL have more malignant characteristics that allow HCCs to metastasize more frequently and grow faster. DCP is reportedly correlated with the development of portal invasion and early intrahepatic recurrence after treatment (6,40). Koike et al. (6) reported a serum DCP level at the time of initial HCC diagnosis as a significant indicator for subsequent development of portal invasion in 227 HCCs treated with percutaneous ethanol injection or microwave coagulation therapy. Likewise with portal invasion, the serum DCP level reportedly correlates with tumor size and high histological grade of HCC in the surgically resected liver (7,41). In the present study, portal invasion and poor differentiation of tumor were risk factors for recurrence in univariate analysis. These histopathological variables were not analyzed in multivariate analysis, because only preoperative variables should be used for preoperative patient selection.

When we looked at correlations between preoperative serum tumor markers and histological findings, univariate analysis revealed serum DCP >400 mAU/mL and AFP >800 ng/mL as significantly associated with incidence of microvascular invasion and poor differentiated grade. However, multivariate analysis using logistic regression analysis demonstrated serum DCP as the only factor significantly correlated with microvascular invasion and poorly differentiated grade. AFP or tumor size/number was not independently related to these pathological characteristics of HCC. This was compatible with the finding of DCP as the most significant predictor of recurrence after LDLT among other preoperative factors.

Criteria based only on tumor size and number, such as MC, UCSF criteria and 5–10 rules, failed to show a correlation with poorly differentiated grade (Table 6). Yao et al. (42) claimed that UCSF tumor stage is strongly correlated with histological features, showing higher prevalence of microvascular invasion in 10 patients exceeding UCSF criteria than in 60 patients meeting these criteria. However, differences in the prevalence of poorly differentiated grade between meeting and exceeding UCSF criteria were not significant (p = 0.08), while those between meeting and exceeding MC were significant (p = 0.02). We observed the same tendency in the current study for UCSF criteria, which included more poorly differentiated tumors than MC (18% vs. 16%). Thus, extending criteria according to tumor size or number alone may contain the risk of including tumors with high potential for recurrence. In contrast, Kyoto criteria containing a DCP cut-off value showed a strong correlation with both poorly differentiated grade (p = 0.0023, within Kyoto vs. beyond Kyoto criteria) and microvascular invasion (p < 0.0001). Of the 116 patients within the 5–10 rule, 19 patients were excluded from Kyoto criteria because of DCP >400 mAU/mL. Among the 19 patients who met the 5–10 rule but exceeded Kyoto criteria, 6 patients showed recurrence, resulting in 1- and 5-year recurrence rates of 25% and 38%, respectively. This clearly indicates that DCP offers additional benefit to the 5–10 rule in patient selection by excluding tumors with advanced histological grade. We believe that our analysis is reliable and has a great impact, as the sample was based on a larger volume of patients showing a larger prevalence of poorly differentiated HCC in total (29 of 144, 20%) than described by Yao et al. (10 of 70, 11%). The reason for the higher prevalence of poorly differentiated tumor may be due to our wider indications for LDLT and thus the inclusion of more advanced HCC than seen in cases undergoing deceased donor liver transplantation (DDLT).

When expansion of criteria is discussed, the first thing to be considered is how many patients beyond MC are included in the new criteria. Kyoto criteria included as many as 28 (44%) of 64 patients beyond the MC, maintaining excellent outcome. This is far more cases than included in UCSF criteria (8% of patients beyond MC) or other reports about this subject. Secondly, outcomes should be compared between patients beyond MC but within the new criteria and those within MC. A significant shortcoming of some studies is that they show acceptable outcomes for an entire group within the new criteria but do not assess that of patients beyond MC but within the new criteria (41). In such studies, the entire outcome is diluted by a large pool of patients within MC. We performed direct comparison of recurrence and survival rates between patients beyond MC but within Kyoto criteria and those within MC, showing an excellent 5-year recurrence rate (4% vs. 7%) and 5-year patient survival (89% vs. 78%) (Figure 3).

In conclusion, the present findings suggest that the preoperative DCP level offers additional information on tumor biological behaviors related to histological findings usually obtained by explant pathology. The proposed Kyoto criteria, incorporating DCP in addition to morphological tumor size and number, are useful and effective as expanded selection criteria for LDLT in patients with HCC.

Ancillary