Download - Kshivets O. Cardioesophageal Cancer Surgery
ARTFICIAL INTELLIGENCE, SYSTEM ANALYSIS AND SIMULATION MODELING IN PREDICTION OF
5-YEAR SURVIVAL OF CARDIOESOPHAGEAL CANCER PATIENTS AFTER COMPLETE LEFT
THORACOABDOMINAL ESOPHAGOGASTRECTOMIES
Oleg Kshivets, MD, PhDOleg Kshivets, MD, PhD Department of Surgery, Siauliai Public Hospital & Cancer Center, LithuaniaDepartment of Surgery, Siauliai Public Hospital & Cancer Center, Lithuania
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
AbstractAbstract• ARTFICIAL INTELLIGENCE, SYSTEM ANALYSIS AND SIMULATION MODELING IN PREDICTION OF 5-
YEAR SURVIVAL OF CARDIOESOPHAGEAL CANCER PATIENTS AFTER COMPLETE LEFT THORACOABDOMINAL ESOPHAGOGASTRECTOMIES
• Oleg Kshivets Department of Surgery, Siauliai Public Hospital & Cancer Center, Siauliai, Lithuania• OBJECTIVE: We examined the clinicomorphologic factors associated with the low- and high-risk of generalization of
cardioesophageal cancer (CEC) (T1-4N0-3M0) after complete esophagogastrectomies (EG) through left thoracoabdominal incision. METHODS: We analyzed data of 150 consecutive CEC patients (CECP) (age=54.9±0.7 years; tumor size=6.9±0.2 cm) radically operated and monitored in 1975-2006 (males=116, females=34; combined EG with resection of pancreas, liver, diaphragm, colon transversum, splenectomies=49; lymphadenectomy D2=59, D3=91; esophagogastroanastomosis=89, esophagoenteroanastomosis=61; adenocarcinoma=125, squamos=19, mix=6; T1=16, T2=32, T3=58, T4=44; N0=59, N1=18, N2=71; N3=2; G1=42, G2=30, G3=78). Variables selected for 5-year survival (5YS) study were input levels of 45 blood parameters, sex, age, TNMPG, cell type, tumor size. Survival curves were estimated by the Kaplan-Meier method. Differences in curves between groups of CECP were evaluated using a log-rank test. Multivariate Cox modeling, multi-factor clustering, discriminant analysis, structural equation modeling, Monte Carlo, bootstrap simulation and neural networks computing were used to determine any significant dependence. RESULTS: 44 CECP (life span: LS=3545.8±238.7 days) lived more than 5 years without any features of CEC progressing (5YS=29.3%). 106 CECP died because of generalization of CEC during the first 5 years after radical procedures (LS=593.5±32.6 days). Cox modeling displayed that 5YS of CECP (n=150) after complete EG significantly depended on: T1-4, combined procedures, histology, G1-3, blood lymphocytes, monocytes, neutrophils, lymphoid infiltration of CEC, age (P=0.000-0.038). Neural networks computing, genetic algorithm selection and bootstrap simulation revealed relationships between 5YS of CECP and combined procedures (rank=1), N0-3 (2), histology (3), gender (4), CEC growth (5), type of operations (6), P1-4 (7), T1-4 (8), adjuvant chemoimmunotherapy (9), G1-3 (10), blood coagulation time (11), blood lymphocytes (12), thrombocytes (13), blood rest nitrogen (14), hemorrhage time (15), ESS (16), age (17), weight (18), blood chlorides (19), tumor size (20).CONCLUSIONS: Correct prediction of CECP survival after radical procedures was 90.7% by logistic regression (odds ratio=86.7), 96% by discriminant analysis and 100% by neural networks computing (area under ROC curve=1.0; error=0.0012).
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Data• Males……………………………………………..116• Females………..…………………………………..34• Age=54.9±0.7 years• Tumor Size=6.9±0.2 cm• Only Surgery.…………………………………...132• Adjuvant Chemoimmunotherapy
(5FU+thymalin/taktivin, 5-6 cycles)…………….18
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Radical ProceduresRadical Procedures• Proximal Esophagogastrectomies with Single-
Stage Esophagogastroplasty.………………….81• Total Esophagogastrectomies with Single-Stage
Esophagoenteroplasty..………..………………69• Combined Esophagogastrectomies with Resection
of Diaphragm, Liver, Mesocolon, Colon Transversum, Splenectomy, Left Hemipancreatectomy, etc…………………….49
• Lymphadenectomy D2………………………...59• Lymphadenectomy D3………………………...91
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Schemas of ProceduresSchemas of Procedures
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
ProceduresProcedures
One-Stage Esophagogastroplasty or EsophagoenteroplastyOne-Stage Esophagogastroplasty or Esophagoenteroplasty
Schemas of Combined Procedures
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Schemas of D3 Lymphadenectomy
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Staging• T1……16 N0..…..59 G1…………42• T2……32 N1……18 G2…………30• T3……58 N2……71 G3…………78• T4……44 N3……..2• exophytic growth…………..53• endophytic growth………....85• mix growth..………………..12• adenocarcinoma……………93• squamos cell carcinoma……52• mix carcinoma……………….5
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Survival Rate• 5-Year Survivors…………..…….44 (29.3%) • 10-Year Survivors……………….18 (12.0%)• Losses……………………………106 (70.7%)• General Life Span= 1459.5±132 days• For 5-Year Survivors= 3545.8±238.7 days• For Losses= 593.5±32.6 days• Cumulative 5-Year Survival…….29.3%• Cumulative 10-Year Survival…...22.4%• Cumulative 15-Year Survival…...20.2%
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
General Cardioesophageal Cancer Patients Survival after Complete Left Thoracoabdominal Esophagogastrectomies (Kaplan-Meier) (n=150)
Survival FunctionComplete Censored
Survival Cardioesofageal Cancer Patients after Esophagogastrectomies, n=150
Years after Complete Left Thoracoabdominal Esophagogastrectomies
Cum
ulat
ive
Prop
ortio
n Su
rviv
ing
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 5 10 15 20 25
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Results of Univariate Analysis in Prediction of Cardioesophageal Cancer Patients Survival (n=150)
Cumulative Proportion Surviving (Kaplan-Meier)Complete Censored
Survival Cardioesophageal Cancer Patients with N0 & N1-3P=0.000 by Log-Rank Test
Years After Complete Esophagogastrectomies
Cum
ulat
ive
Prop
ortio
n Su
rviv
ing
-0.10.00.10.20.30.40.50.60.70.80.91.0
0 5 10 15 20 25
N0 N1-N3
Cumulative Proportion Surviving (Kaplan-Meier)Complete Censored
Survival Cardioesophageal Cancer PatientsP=0.001 by Log-Rank Test
Years After Esophagogastrectomies
Cum
ulat
ive
Prop
ortio
n Su
rviv
ing
-0.10.00.10.20.30.40.50.60.70.80.91.0
0 5 10 15 20 25
T2 T3
Results of Univariate Analysis in Prediction of Cardioesophageal Cancer Patients Survival (n=150)
Cumulative Proportion Surviving (Kaplan-Meier)Complete Censored
Survival Cardioesophageal Cancer PatientsP=0.0016 by Log-Rank Test
Years After Esophagogastrectomies
Cum
ulat
ive
Prop
ortio
n Su
rviv
ing
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 5 10 15 20 25
G1 G3
Cumulative Proportion Surviving (Kaplan-Meier)Complete Censored
Survival Cardioesophageal Cancer PatientsP=0.00001 by Log-Rank Test
Years After Esophagogastrectomies
Cum
ulat
ive
Prop
ortio
n Su
rviv
ing
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 5 10 15 20 25
exo endo
Results of Univariate Analysis in Prediction of Cardioesophageal Cancer Patients Survival (n=150)
Cumulative Proportion Surviving (Kaplan-Meier)Complete Censored
Survival of Cardioesophageal Cancer Patients, n=150P=0.021 by Log-Rank Test
Years After Complete Esophagogastrectomies
Cum
ulat
ive
Prop
ortio
n Su
rviv
ing
0.00.10.20.30.40.50.60.70.80.91.0
0 5 10 15 20 25
Only Surgery, n=132 Adjuvant Chemoimmunotherapy, n=18
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Results of Cox Regression Modeling in Prediction of Cardioesophageal Cancer
Patients Survival after Complete Esophagogastrectomies (n=150) • Factors Wald df P Exp(B) 95%CI for Exp(B)
Lower Upper• Eosinophils% 6.539 1 0.011 1.802 1.147 2.829• Stick.Neutrophils% 12.447 1 0.000 2.298 1.447 3.648• Seg.Neutrophils% 14.680 1 0.000 2.379 1.527 3.706• Lymphocytes% 11.949 1 0.001 2.212 1.410 3.469• Monocytes% 12.832 1 0.000 2.303 1.459 3.634• T1-4 12.579 3 0.006• T(1) 11.124 1 0.001 0.196 0.075 0.510• T(2) 6.870 1 0.009 0.349 0.159 0.767• T(3) 1.409 1 0.235 0.672 0.348 1.296• Ad.CHIT 0.648 1 0.421 0.744 0.362 1.529• Age 10.419 1 0.001 1.042 1.016 1.068• Comb.operation 15.523 6 0.017
The 60th Annual Meeting of Society of Surgical Oncology
March 15-18, 2007, Washington , DC, the USA
Results of Cox Regression Modeling in Prediction of Cardioesophageal Cancer
Patients Survival after Complete Esophagogastrectomies (n=150) • Factors Wald df P Exp(B) 95%CI for Exp(B) Lower
Upper• Histology 15.864 2 0.000• Histology(1) 13.556 1 0.000 8.361 2.700 25.894• Histology(2) 15.753 1 0.000 9.631 3.147 29.473• Lymphocytes abs 5.423 1 0.020 6.126 1.332 28.166• Seg.Neutrophils abs 4.899 1 0.027 0.495 0.265 0.923• G1-3 6.539 2 0.038• G(1) 5.907 1 0.015 0.519 0.305 0.881• G(2) 3.578 1 0.059 0.565 0.313 1.021• LIT 32.081 3 0.000• LIT(1) 25.546 1 0.000 6.648 3.189 13.858• LIT(2) 23.429 1 0.000 5.804 2.848 11.829• LIT(3) 1.084 1 0.298 1.424 0.732 2.768
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Results of Discriminant Analysis in Prediction of Cardioesophageal Cancer Patients Survival after Complete Esophagogastrectomies (n=150)
• Discriminant Function Analysis Summary • Wilks' Lambda: 0.305 approx. F (61,88)=3.281 p< 0.0000
• Wilks' Partial F-remove P-level • Lambda Lambda (1,88) • LIT 0.403 0.757 28.238 0.000• Comb.Operation 0.319 0.956 4.032 0.048• Erythrocytes 0.319 0.956 4.012 0.048• Weight 0.319 0.958 3.893 0.052• Glucose 0.318 0.962 3.503 0.065• Growth 0.312 0.979 1.904 0.171• G1-3 0.308 0.990 0.854 0.358• Ad.CHIT 0.307 0.995 0.402 0.528• T1-4 0.306 0.999 0.128 0.722• N0-3 0.306 0.999 0.064 0.801
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Results of Logistic Regression Analysis in Prediction of Cardioesophageal Cancer Patients Survival after Complete
Esophagogastrectomies (n=150)
• Est. S.E. Wald P Odds 95.0% C.I.for Odds Ratio Ratio Lower Upper
• Const.B 18.24 9.25 3.89 0.049 8.3e+7 0.95 5.8e+12• Seg.Neut.% -0.21 0.10 4.44 0035 0.81 0.67 0.99• Monocytes abs-5.46 2.51 4.74 0.030 0.00 0.00 0.608• LIT 1.99 0.42 22.38 0.000 7.28 3.18 16.70• G1-3 -0.59 0.42 2.02 0.155 0.55 0.24 1.26• Growth 1.54 0.71 4.75 0.029 4.69 1.15 19.04• Comb.Oper. -0.47 0.16 8.75 0.003 0.63 0.46 0.86• T1-4 -0.26 0.48 0.29 0.589 0.77 0.30 1.99• N0-3 -0.20 0.40 0.26 0.613 0.817 0.370 1.80
Chi2=105.11; df=15; P=0.0000; Odds ratio=86.70
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Results of Correspondence Analysis in Prediction of Cardioesophageal Cancer Patients Survival (n=150)
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Results of Multi-Factor Clustering of Clinicopathological Data in Prediction of Cardioesophageal Cancer Patients Survival (n=150)
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Neural Networks in Prediction of Cardioesophageal Cancer Patients Survival after Complete Esophagogastrectomies (n=150)
• Baseline Errors=0.0012• Area under ROC curve=1.00• Correct Classification Rate=100%
• Losses 5-year survivors • Total106 44
Correct106 44
Wrong 0 0
• Genetic Algorithm Selection• Useful for L ESS Haemor.Time Protein PI D Histology G1-3 Growth Sabs • Survival Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes• Useful for Lymabs Comb.Oper. P1-4 L/CC S/CC T1-4 N0-3 Chlorides Stot• Survival Yes Yes Yes Yes Yes Yes Yes Yes Yes
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Results of Kohonen Self-Organizing Neural Networks Computing in Prediction of Cardioesophageal Cancer Patients Survival after Complete Esophagogastrectomies (n=150)
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Results of Neural Networks Computing in Prediction of Cardioesophageal Cancer Patients Survival after Complete Esophagogastrectomies (n=150)
Error=0.0011; Area under ROC Curve=1.00; Correct Classification Rate=100%
– Factor Rank Error Ratio• Comb.Operat. 1 0.199 159.5• N0-3 2 0.185 148.0• Histology 3 0.172 138.0• Sex 4 0.166 133.1• Growth 5 0.164 130.9• Oper. Type 6 0.153 122.6• P1-4 7 0.117 94.0• T1-4 8 0.109 87.2• Ad.CHIT 9 0.106 84.7• G1-3 10 0.071 57.1• Coag.Time 11 0.029 23.5• Lymph.% 12 0.029 23.2
– Factor Rank Error Ratio• Thr.tot 13 0.023 18.2• Rest Nitrogens 14 0.019 15.2• Haem.Time 15 0.007 5.6• ESS 16 0.005 3.8• Age 17 0.003 2.6• Weight 18 0.003 2.5• Chlorides 19 0.002 1.9• D 20 0.002 1.8• St.Neutr.tot 21 0.002 1.6• Protein 22 0.002 1.5• L/CC 23 0.002 1.4• Erythrocytes 24 0.002 1.4
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Results of Bootstrap Simulation in Prediction of Cardioesophageal Cancer Patients Survival after Complete Esophagogastrectomies (n=150)
• Number of Samples=3333• Significant Factors Rank Kendall’s Tau-A P<• LIT 1 0.316 0.000• D 2 -0.196 0.000• P1-4 3 -0.189 0.000• Leucocytes/CC 4 0.188 0.000• HC/CC 5 0.187 0.000• Erythrocytes/CC 6 0.186 0.000• Lymphocytes/CC 7 0.185 0.000• T1-4 8 -0.184 0.000• N0-3 9 -0.176 0.000• Seg.Neutrophils/CC 10 0.167 0.000• Thrombocytes/CC 11 0.160 0.000• Monocytes/CC 12 0.146 0.000
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Results of Bootstrap Simulation in Prediction of Cardioesophageal Cancer Patients Survival after Complete Esophagogastrectomies (n=150)
• Number of Samples=3333• Significant Factors Rank Kendall’s Tau-A P<• Eosinophils/CC 13 0.124 0.000• Operation Type 14 -0.119 0.000• G1-3 15 -0.119 0.000• Histology 16 -0.118 0.000• Chlorides 17 0.106 0.000• Coagulation Time 18 -0.096 0.000• Growth 19 -0.092 0.000• Stick Neutrophils/CC 20 0.068 0.000• Protein 21 0.063 0.000• Combined Operation 22 0.048 0.000• Rest Nitrogens 23 0.042 0.001• Colour Index 24 0.041 0.000
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Results of Bootstrap Simulation in Prediction of Cardioesophageal Cancer Patients Survival after Complete Esophagogastrectomies (n=150)
• Number of Samples=3333• Significant Factors Rank Kendall’s Tau-A P<• ESS 25 0.038 0.002• Erythrocytes tot 26 0.038 0.002• Thrombocytes 27 -0.033 0.005• Adjuvant CHIT 28 0.031 0.005• Eosinophils tot 29 0.028 0.018• Leucocytes tot 30 0.027 0.023• Sex 31 0.027 0.023• Weight 32 0.026 0.024• Haemorrhage Time 33 0.026 0.024• Monocytes tot 34 0.024 0.05
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Prediction of Cardioesophageal Cancer Patients Survival after Complete Esophagogastrectomies (n=150)
• Classification of Cases by Logistic Regression, n=150, Odds Ratio=86.7• Observed Pred.Losses Pred.Survivors Correct• Losses 102 4 96.2%• 5-Year Survivors 10 34 77.3%• Total 112 38 90.7%• Classification of Cases by Discriminant Analysis, n=150• Observed Pred.Losses Pred.Survivors Correct• Losses 105 1 99.1%• 5-Year Survivors 5 39 88.6%• Total 110 44 96.0%• Classification of Cases by Neural Networks, n=150, Errors=0.0012• Observed Pred.Losses Pred.Survivors Correct• Losses 106 0 100%• 5-Year Survivors 0 44 100%• Total 106 44 100%
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Ratio of Lymphocytes to Cancer Cells & Glucose Level Ratio of Lymphocytes to Cancer Cells & Glucose Level in Prediction of Cardioesophageal Cancer Patients Survival after Complete
Esophagogastrectomies (n=150)
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Prognostic SEPATH-Model of Cardioesophageal Cancer Patients Survival after Complete Esophagogastrectomies, n=150
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Holling-Tenner Models of Cardioesophageal Holling-Tenner Models of Cardioesophageal Cancer Cell Population and Cytotoxic Cell Cancer Cell Population and Cytotoxic Cell
Population DynamicsPopulation Dynamics
0 2 4 6 8 10
0.1
10
Early CancerInvasive Cancer, Stage IIInvasive Cancer, Stage IIIGeneralization
Model "Early Cancer---Lymphocytes"
Gastroesophageal Cancer Cell Population
Lym
phoc
yte
Popu
latio
n
5
0.381
X1 3
X2 3
X3 3
X4 3
100.09 X1 2 X2 2
X3 2 X4 2
0 50 100 150 200 250
0.01
0.1
10
LymphocytesCancer Cells
Model "Early Cancer---Lymphocytes"
Time
Gas
tr.es
oph.
Cel
l Pop
ulat
ion
Dyn
amic
s
5
0.09
X1 2
X1 3
2000 X1 1
Cardioesophageal Cancer Dynamics
Conclusions:• 5-year survival and life span of cardioesophageal cancer
patients after complete esophagogastrectomies significantly depended on:
• 1) cell ratio factors: ratio of cancer cell population to blood cell subpopulations in integral patient organism;
• 2) cancer characteristics (cancer cell population number, TNMG-system);
• 3) the data of blood cell circuit, biochemical homeostasis and hemostasis system;
• 4) character of surgical procedure;• 5) anthropometric data.
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Conclusions:• Optimal treatment strategies for cardioesophageal
cancer patients are:• 1) screening and early detection of cardioesophageal
cancer; • 2) availability of very experienced surgeons because of
complexity of radical procedures;• 3) aggressive en block surgery for completeness; • 4) precise prediction; • 5) adjuvant chemioimmunotherapy for patients with • unfavorable prognosis.
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA
Address:Oleg Kshivets M.D., Ph.D.,
Consultant Thoracic, Abdominal, General Surgeon & Surgical Oncologist
• Surgery Department, Siauliai Public Hospital & Cancer Center• Tilzes:42-16, LT78206 Siauliai, Lithuania• Tel. 37041-416614• e-mail: [email protected] • http//:myprofile.cos.com/Kshivets
The 60th Annual Meeting of Society of Surgical Oncology March 15-18, 2007, Washington , DC, the USA