A New Model to Estimate Survival for Hepatocellular Carcinoma Patients Po-Hong Liu, Chia-Yang Hsu, Cheng-Yuan Hsia, Yun-Hsuan Lee Yi-Hsiang Huang, Chien-Wei Su, Fa-Yauh Lee, Han-Chieh Lin, Teh-Ia Huo Taipei Veterans General Hospital National Yang-Ming University TAIWAN stuartliu@gmail.com; tihuo@vghtpe.gov.tw
The Authors Have Nothing to Disclose Disclosure The Authors Have Nothing to Disclose
Staging System for HCC Staging: critical step in cancer management Prognosis of HCC is complex Tumor extent Liver dysfunction General medical condition HCC: hepatocellular carcinoma J Hepatol 2016;64:535-536
Current Scoring Systems Model Tumor Status Liver Function Performance Status Serum AFP Serum ALK-P TIS Yes CTP No CLIP Tokyo JIS MESIAH MELD NIACE French Bilirubin Karnofsky CUPI Symptoms AFP: alpha-fetoprotein; ALK-P: alkaline phosphatase; CLIP: Cancer of the Liver Italian Program; CUPI: Chinese University Prognostic Index; JIS: Japan Integrated Scoring; MESIAH: Model to Estimate Survival In Ambulatory HCC; TIS: Taipei Integrated Scoring J Hepatol 2016;64:601-608
Do We Need Another Score ? Model Tumor Status Liver Function Performance Status Serum AFP Serum ALK-P TIS Yes CTP No CLIP Tokyo JIS MESIAH MELD NIACE French Bilirubin Karnofsky CUPI Symptoms New Score ? AFP: alpha-fetoprotein; ALK-P: alkaline phosphatase; CLIP: Cancer of the Liver Italian Program; CUPI: Chinese University Prognostic Index; JIS: Japan Integrated Scoring; MESIAH: Model to Estimate Survival In Ambulatory HCC; TIS: Taipei Integrated Scoring J Hepatol 2016;64:601-608
Aim of the Study To establish a new prognostic Model to Estimate Survival for HCC (MESH Score) Patients Our approach Assess pre-treatment status Use common variables Simplistic approach & user-friendly Statistically robust
Methods & Study Flowchart All Patients Derivation Cohort (n = 1,591) Selecting Predictors Forward Cox Regression MESH Score Validation Cohort (n = 1,591) Kaplan-Meier Curve Discriminatory Ability Homogeneity Subgroup Analysis 1:1 Randomization Single-center 3,182 Patients (2002-2013)
Choosing Survival Predictors All candidate baseline survival predictors were included in initial analysis All predictors are dichotomized Clinical knowledge (Age, Milan, Single/Multiple) Conventional definitions (ALT, ALK-P, AFP) Youden index in ROC curve (CTP scores, AFP, PS) AFP, alpha-fetoprotein; ALK-P: alkaline phosphatase; ALT: alanine transaminase; CTP: Child-Turcotte-Pugh; ROC: Receiver-Operating-Characteristics, PS: performance status
Predictors of Survival Univariate analysis Multivariate analysis HR p value β CI Age (<65/≥65 years) 1.133 0.133 Sex (male/female) 0.926 0.127 HBsAg (negative/positive) 0.883 0.136 Anti-HCV (negative/positive) 0.774 0.005 Non-significant Alcoholism (no/yes) 1.330 0.007 ALT (<40/≥40 IU/L) 1.120 0.186 Platelet (≥150K/<150K/μL) 1.349 <0.001 ALK-P (<200/≥200 IU/L) 4.150 1.953 0.669 1.580-2.414 CTP score (5/6-15) 3.083 2.055 0.720 1.706-2.476 Performance status (0-1/2-4) 4.563 2.415 0.882 1.979-2.948 Serum AFP (<20/≥20 ng/mL) 2.035 1.540 0.432 1.282-1.849 Single/Multiple tumor 1.486 Early/Non-early tumor (Milan) 2.919 1.823 0.601 1.507-2.206 Vascular invasion + Metastasis 5.530 2.752 1.012 2.245-3.374
Predictors of Survival Univariate analysis Multivariate analysis HR p value β CI Age (<65/≥65 years) 1.133 0.133 Sex (male/female) 0.926 0.127 HBsAg (negative/positive) 0.883 0.136 Anti-HCV (negative/positive) 0.774 0.005 Non-significant Alcoholism (no/yes) 1.330 0.007 ALT (<40/≥40 IU/L) 1.120 0.186 Platelet (≥150K/<150K/μL) 1.349 <0.001 ALK-P (<200/≥200 IU/L) 4.150 1.953 0.669 1.580-2.414 CTP score (5/6-15) 3.083 2.055 0.720 1.706-2.476 Performance status (0-1/2-4) 4.563 2.415 0.882 1.979-2.948 Serum AFP (<20/≥20 ng/mL) 2.035 1.540 0.432 1.282-1.849 Single/Multiple tumor 1.486 Early/Non-early tumor (Milan) 2.919 1.823 0.601 1.507-2.206 Vascular invasion + Metastasis 5.530 2.752 1.012 2.245-3.374
Predictors of Survival Univariate analysis Multivariate analysis HR p value β CI Age (<65/≥65 years) 1.133 0.133 Sex (male/female) 0.926 0.127 HBsAg (negative/positive) 0.883 0.136 Anti-HCV (negative/positive) 0.774 0.005 Non-significant Alcoholism (no/yes) 1.330 0.007 ALT (<40/≥40 IU/L) 1.120 0.186 Platelet (≥150K/<150K/μL) 1.349 <0.001 ALK-P (<200/≥200 IU/L) 4.150 1.953 0.669 1.580-2.414 CTP score (5/6-15) 3.083 2.055 0.720 1.706-2.476 Performance status (0-1/2-4) 4.563 2.415 0.882 1.979-2.948 Serum AFP (<20/≥20 ng/mL) 2.035 1.540 0.432 1.282-1.849 Single/Multiple tumor 1.486 Early/Non-early tumor (Milan) 2.919 1.823 0.601 1.507-2.206 Vascular invasion + Metastasis 5.530 2.752 1.012 2.245-3.374
The MESH Score Score range from 0 to 6 Prognostic Factors 1 1 Tumor Burden (Milan) Small Large Vascular invasion or metastasis Absent Present Child-Turcotte-Pugh score 5 ≥ 6 Performance status 0-1 ≥ 2 Serum AFP level < 20 ng/mL ≥ 20 ng/mL Serum ALK-P level < 200 IU/L ≥ 200 IU/L Score range from 0 to 6
Validation of MESH Score
Kaplan-Meier Curve in Validation Cohort Significant Survival Differences across all MESH Scores
Comparing Prognostic Performances Model Homogeneity (Wald χ2) Corrected Akaike Information Criteria BCLC 386.478 3017.886 HKLC 475.030 2969.055 TIS 551.217 2926.460 CLIP 664.101 2852.240 MESIAH 804.692 2769.194 MESH 893.352 2697.351 Homogeneity: small difference in survival for patients among the same classification within each system Akaike information criterion: amount of information loss during model creation BCLC: Barcelona Clinic Liver Cancer; HKLC: Hong Kong Liver Cancer
Discriminatory Ability Model Death at 1-year Death at 3-year Death at 5-year BCLC 0.794 0.741 0.713 HKLC 0.821 0.766 0.735 TIS 0.832 0.768 0.724 CLIP 0.838 0.772 0.732 MESIAH 0.867 0.806 0.773 MESH 0.860 0.805 0.769 * * * * * * MESH Score: High Prognostic Accuracy in Validation Cohort Discriminatory ability: The ability to identify survivor and non-survivor * p<0.05
MESH Score in Different Clinical Settings HBV- & HCV-related HCC Curative & Non-curative Treatment BCLC & HKLC
MESH Score for Different Etiologies Model Homogeneity (Wald χ2) Corrected Akaike Information Criteria HBV-related HCC (41%) CLIP 276.524 1153.218 MESIAH 335.326 1120.017 MESH 388.941 1070.173 HCV-related HCC (23%) 92.821 638.981 117.745 622.728 117.521 617.654 HBV: hepatitis B virus; HCV: hepatitis C virus
MESH Score for Different Treatments Model Homogeneity (Wald χ2) Corrected Akaike Information Criteria Curative treatment (SR, RFA, transplantation, 44%) CLIP 42.873 1054.411 MESIAH 63.767 1033.436 MESH 60.457 1035.369 Non-curative treatment (All other treatment, 56%) 430.317 1706.867 472.195 1686.460 627.475 1594.824 RFA: radiofrequency ablation, SR, surgical resection
BCLC 0/A and HKLC I/II HCC MESH Score Discriminate Survival for Earlier HCC
BCLC B/C/D and HKLC III/IV/V HCC MESH Score Discriminate Survival for Later HCC
Limitations Choice of predictors and their cut-points Developed from “treated cohort” Lack of external validation
MESH Score - Summary Simple, common, and accurate Can be used in different clinical settings Supplementary to current staging systems
Thank You for Your Attention MESH Score 1 Tumor Burden (Milan) Small Large Vascular invasion or metastasis Absent Present Child-Turcotte-Pugh score 5 ≥ 6 Performance status 0-1 ≥ 2 Serum AFP level < 20 ng/mL ≥ 20 ng/mL Serum ALK-P level < 200 IU/L ≥ 200 IU/L stuartliu@gmail.com; tihuo@vghtpe.gov.tw
Supplementary Materials Cohort characteristics Detailed MESH score Choosing cut-off values Can scores guide treatment decisions ?
Benchmark - Detailed MESH Score Prognostic Factors Absent Original MESH Beta Coefficient Detailed MESH Tumor Burden (Milan Criteria) 1 0.601 1.5 Vascular invasion or metastasis 1.012 2.5 CTP score 0.720 Performance status 0.882 2 Serum AFP level 0.432 Serum ALK-P level 0.669 Based on relative ratios of β-coefficients Detailed MESH score range from 0 to 10
Performance of Detail MESH Score Model Homogeneity (Wald χ2) Corrected Akaike information criteria BCLC 386.478 3017.886 HKLC 475.030 2969.055 TIS 551.217 2926.460 CLIP 664.101 2852.240 MESIAH 804.692 2769.194 MESH 893.352 2697.351 Detailed MESH 940.837 2685.502
Choosing Survival Predictors All candidate baseline survival predictors were included in initial analysis All predictors are dichotomized Clinical knowledge (Age, Milan, Single/Multiple) Conventional definitions (ALT, ALK-P, AFP) Youden index in ROC curve (CTP scores, AFP, PS) ALT: alanine transaminase; CTP: Child-Turcotte-Pugh; PS: performance status
Alkaline Phosphatase Alk-P may be related to HCC growth Had been included in CUPI and French score Cut-points CUPI score: Alk-P ≥ 200 IU/L French score: Alk-P ≥ 2x upper limit Youden Index: 127 IU/L
Child-Turcotte-Pugh Score Child-Pugh class A: 73% patients Further sub-division Youden index for CTP score Cut-point: 5 vs 6-15 New marker for liver dysfunction ? Albumin-bilirubin (ALBI) grade (1 vs 2-3) Platelet-albumin-bilirubin (PALBI) grade (1 vs 2-3)
Performance Status ECOG PS 1 BCLC stage C PS is highly associated with survival PS 1 HCC benefits from aggressive therapy Youden index for performance status Cut-point: 0-1 vs 2-4 Hepatology 2013;57:112-119 ECOG: Eastern Oncology Cooperative Group
Can Score Guide Treatment Decisions ? Prognostic scores can stratify BCLC stages MESIAH score BCLC 0/A, B, C, D NIACE score BCLC A, B, C MESH score BCLC 0, A, B, C (Data not shown) Can prognostic scores guide treatment algorithm ? A proof-of-concept study Eur J Gastroenterol Hepatol 2016;28(4):433-40 Hepatology 2012;56(2)614-21
Nomogram for Recurrence after RFA (BCLC 0/A) Point Number of Tumor Largest Tumor Serum Albumin MELD Score Risk of recurrence after RFA (BCLC 0/A) Low risk: nomogram score < 9.8 High risk: nomogram score ≥ 9.8 Platelet Count Sum up Total Points Recurrence-free survival Medicine 94(43):e1808
Scores in Treatment Algorithm Very Early & Early HCC Nomogram Low-Risk RFA High-Risk SR
Cohort Characteristics Prospective cohort of 3,182 HCC in Taiwan Timespan: 2002-2013 HBV-related HCC: 41% HCV-related HCC: 23% Curative treatment: 44%
Percentages of Patients Score (%) 1 2 3 4 5 6 MESH 13.2 24.7 21.6 15.4 11.6 9.4 4.2 CLIP 30.5 27.4 15.1 11.4 9.5 5.0 1.0 BCLC 8.3 23.1 15.8 40.3 12.4 HKLC 31.5 27.1 10.1 9.3 22.1