Predictive Value of Preoperative Alpha-Fetoprotein

and Des-gamma-Carboxy Prothrombin

on Hepatocellular Carcinoma Recurrence after

Liver Transplantation

Juhan Lee1, Hyung Soon Lee1, Gi Hong Choi1,2, Dae Ryong Kang3,

Dong Jin Joo1,2, Myoung Soo Kim1,2, Jin Sub Choi1,2,

Soon Il Kim1,2

The 38th Congress of the Korean Association of HPB surgery

Department of Surgery1,

The Research Institute for Transplantation2 and

Graduate School of Public Health3

Yonsei University Health System, Seoul, Korea

Backgound

• Milan criteria have been accepted gold standard in LT for patients with HCC

• Radiological aspects alone are not fully able to select candidates

• Too restrictive and could be improved.

- more advanced HCCs either by expanding selection criteria or by offering

down-staging protocols.

Merani et al. J Hepatol. 2011;55(4):814-9

Backgound

• AFP and DCP seem to be a better surrogate markers for unfavorable tumor biology than

radiologic criteria.

• AFP and DCP are independently produced from HCC show no correlation with each other.

• These two markers might complement each other.

• Combined measurement of these two markers appears to be useful in predicting the

prognosis of HCC.

Tumor marker

Chon et al. Int J Cancer. 2012 15;131(10):2332-41

Purpose

• To assess whether the combined use of preoperative serum AFP and DCP levels

improve predictive performance of the Milan criteria for HCC recurrence after LT

- Design the Milan criteria based predictive model for HCC recurrence

- This model incorporated pre-transplant AFP and DCP respectively or simultaneously

Methods

• Retrospectively review with medical records

• Duration : 2001.10 - 2012.10

Inclusion criteria

Patients with HCC that received LT

Pathologically proven HCC

Exclusion criteria

Incidental HCC

Combined HCC-CCC

Without elevation of AFP and DCP

(AFP <9 ng/mL, DCP <35 mAU/mL)

Immediate post-op. mortality

Methods

: Base model (Quasi-Milan criteria) with adjusted age, sex, MELD score and pre-transplant treatment

Model 1. Base model + AFP

Model 2. Base model + DCP

Model 3. Base model + AFP + DCP

We compared a global concordance probability of the each models.

(integrated area under the curve, or iAUC)

New predictive model for HCC recurrence incorporating AFP and DCP

Methods

• To evaluate the predictive performance of each predictive model

: we employed the time-dependent receiver operating characteristic (ROC) curve

• As the recurrence is time dependent, time-dependent ROC curves are more

appropriate than conventional ones.

: disease onset usually varies by length of observation period

Results

Scheme of Patients Selection

LT for HCC

(n=210)

Post-op. mortality

(n=20)

Without

AFP and DCP elevation

(n=6)

HCC-CCC

(n=9)

Incidental HCC

(n=13)

Study population

(n=162)

Within the Milan criteria

(n=129)

Beyond the Milan criteria

(n=33)

Duration : 2001.10 - 2012.10

Median follow-up time: 23.5 months (range, 1-132)

Patient Characteristics

Characteristic Categories Patients (n=162)

Age (yrs) * 53 (37-67)

Gender Male/Female 137/25

Etiology of cirrhosis HBV 137 (84.5%)

HCV 15 (9.3%)

Non-B, Non-C 10 (6.2%)

Living donor LT 123 (75.9%)

Pre-transplant treatment No treatment 45 (27.8%)

TACE 75 (46.3%)

RFA 18 (11.1%)

Resection 17 (10.5%)

CCRT 7 (4.3%)

Within Milan criteria 129 (79.6%)

MELD score* 10 (5-34)

Tumor size (cm)* Radiologic 2.0 (0.8-8.3)

Tumor number* Radiologic 1.0 (1-7)

AFP (ng/mL)* 12.2 (0.8-3900)

DCP (mAU/mL)* 33.0 (5-2017)

E-S grade (major) I 46 (28.4%)

II 80 (49.4%)

III 11 (6.8%)

Total necrosis 100% necrosis 25 (15.4%)

Satellite nodule 20 (12.3%)

Microscopic vascular invasion 39 (24.1%)

Base model (Quasi-Milan criteria with adjusted variables)

: Age + Sex + MELD score + Pre-transplant treatment + Milan criteria

Model 1. Base model + AFP

Model 2. Base model + DCP

Model 3. Base model + AFP + DCP

Estimated ROC curves at 1 year after LT

1- specificity

Se

ns

itiv

ity

New predictive model of HCC recurrence incorporating AFP and DCP

Comparison of iAUC up to 40 months after LT in Each Models at Time-dependent ROC curve

• Model 0. Base

: iAUC = 0.794

• Model 1. Base + AFP

: iAUC = 0.826

• Model 2. Base + DCP

: iAUC = 0.821

• Model 3. Base + AFP + DCP

: iAUC = 0.855

Time after LT (Months)

Comparison of iAUC up to 40 months after LT in Each Model at Time-dependent ROC curve

Base model vs Model 1

: Not significant

(Estimated difference = -0.044, 95% CI = -0.131 to 0.004)

Base model vs Model 2

: Not significant

(Estimated difference = -0.032, 95% CI = -0.098 to 0.001)

Base model vs Model 3

: Significant

(Estimated difference = -0.069, 95% CI = -0.159 to -0.007)

Time after LT (Months)

Cut-off point of AFP and DCP

AFP = 140 ng/mL

DCP = 65 mAU/mL

Univariate Cox Regression Analysis of Recurrence (n = 162)

Variable Categories

Disease-free Survival

Hazard Ratio 95% Confidence Interval P

Age (continuous) 1-year increase 0.936 0.869-1.007 0.076

Gender Female vs Male 1.587 0.517-4.868 0.415

Type of transplantation Living donor vs Deceased donor 0.591 0.218-1.599 0.295

MELD score (continuous) 1 unit increase 0.987 0.902-1.079 0.766

Etiology of cirrhosis Other vs HBV 3.201 0.424-24.145 0.233

Milan criteria Within Milan vs Beyond Milan 8.276 3.118-21.966 <0.001

Pre-transplant treatment No vs Yes 2.759 0.631-12.067 0.160

AFP (ng/mL) and DCP (mAU/mL) AFP ≤140, DCP ≤65 vs AFP >140 and/or DCP >65 10.984 3.152-38.282 <0.001

Satellite Nodule No vs Yes 11.495 4.353-30.358 <0.001

Microscopic vascular invasion No vs Yes 11.651 3.795-35.770 <0.001

Total necrosis No vs Yes 0.292 0.039-2.206 0.204

Multivariate Cox Regression Analysis of Recurrence (n = 162)

Variable Categories

Disease-free Survival

Hazard Ratio 95% Confidence

Interval P

Milan criteria Within Milan vs Beyond Milan 2.732 0.906-8.233 0.074

AFP (ng/mL) and DCP (mAU/mL) AFP ≤140, DCP ≤65 vs AFP >140 and/or DCP >65 4.535 1.183-17.382 0.027

Satellite Nodule No vs Yes 3.241 1.071-9.806 0.037

Microscopic vascular invasion No vs Yes 2.917 0.740-2.917 0.126

Impact of Pre-transplant AFP and DCP Level on Disease-free Survival

Within the Milan criteria (n=129) Beyond the Milan criteria (n=33)

Group Number 1 yr (%) 3 yrs (%) 5 yrs (%) P

Within Milan criteria 129 95.4 94.1 92.3 -

AFP ≤140, DCP ≤65 98 98.6 98.6 98.6 0.083

AFP >140 and/or DCP >65 31 86.2 81.4 74.7 0.017

Group Number 1 yr (%) 3 yrs (%) 5 yrs (%) P

Beyond Milan criteria 33 76.5 53.0 53.0 -

AFP ≤140, DCP ≤65 13 100 83.3 83.3 0.506

AFP >140 and/or DCP >65 20 65.0 37.9 37.9 <0.001

Combined use of preoperative serum AFP and DCP levels better predict the HCC

recurrence after LT when compared with the other predictive model.

Our results show that combined use of preoperative serum AFP and DCP is an

important prognostic factor for HCC recurrence after LT.

Summary

Thus, LT candidates with HCC can be assessed more accurately by adding pre-

transplant AFP and DCP levels to the Milan criteria.

Conclusion

Thank you for your attention