1. Survival of ECP, n=557
Complete Censored
-5 0 5 10 15 20 25 30 35 40
Years after Esophagogastrectomies
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Cumulative
Proportion
Surviving
Cumulative Proportion Surviving (Kaplan-Meier)
P=0.000
Complete Censored
0 5 10 15 20 25 30 35 40
Years after Esophagogastrectomies
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Cumulative
Proportion
N 1 2
N 0
Cumulative Proportion Surviving (Kaplan-Meier)
P=0.000
Complete Censored
0 5 10 15 20 25 30 35 40
Years after Esophagogastrectomies
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Cumulative
Proportion
Surviving
Invasive Cancer
Early Cancer
Cumulative Proportion Surviving (Kaplan-Meier)
P=0.00084
Complete Censored
0 5 10 15 20 25 30 35 40
Years after Esophagogastrectomies
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Cumulative
Proportion
Surviving
A C H I R T
Surgery
Cumulative Proportion Surviving (Kaplan-Meier)
P=0.00294
Complete Censored
0 5 10 15 20 25 30 35 40
Years after Esophagogastrectmies
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Cumulative
Proportion
Surviving
Others
Upper/3
Cumulative Proportion Surviving (Kaplan-Meier)
P=0.000
Complete Censored
0 5 10 15 20 25 30 35 40
Years after Esophagogastrectomies
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Cumulative
Proportion
Esophageal Cancer
Cardioesophageal Cancer
Esophageal Cancer: Neural Networks, Complex System Analysis,
Statistics and Simulation Modeling for Best Management. #22247
Kshivets Oleg, MD, PhD Surgery Department, Roshal Hospital, Moscow, Russia
OBJECTIVE: 5-survival (5YS) and life span after radical surgery for esophageal cancer (EC) pa-tients (ECP) (T1-4N0-
2M0) was analyzed. The importance must be stressed of using complex system analysis, artificial intelligence (neural
networks computing), simulation modeling and statistical methods in combination, because the different approaches
yield complementary pieces of prognostic information.
METHODS: We analyzed data of 557 consecutive ECP (age=56.6±8.9 years; tumor size=6±3.5 cm) radically operated
(R0) and monitored in 1975-2023 (m=415, f=142; esophagogastrectomies (EG) Garlock=288, EG Lewis=269, combined EG
with resection of pancreas, liver, diaphragm, aorta, VCS, colon transversum, lung, trachea, pericardium,
splenectomy=168; adenocarcinoma=319, squamous=228, mix=10; T1=130, T2=115, T3=184, T4=128; N0=282, N1=70,
N2=205; G1=157, G2=142, G3=258; early EC=111, invasive=446; only surgery=425, adjuvant chemoimmunoradiotherapy-
AT=132: 5-FU+thymalin/taktivin+radiotherapy 45-50Gy). Multivariate Cox modeling, clustering, SEPATH, Monte Carlo,
bootstrap and neural networks computing were used to determine any significant dependence.
RESULTS: Overall life span (LS) was 1876.9±2219.8 days and cumulative 5-year survival (5YS) reached 52%, 10 years –
45.5%, 20 years – 33.4%, 30 years – 26.9%. 187 ECP lived more than 5 years (LS=4271±2411.9 days), 99 ECP – more than
10 years (LS=5883±2296.6 days). 228 ECP died because of EC (LS=629.8±324.1 days). AT significantly improved 5YS
(67.8% vs. 48.7%) (P=0.00084 by log-rank test). Cox modeling displayed that 5YS of ECP significantly depended on:
phase transition (PT) N0—N12 in terms of synergetics, cell ratio factors (ratio between cancer cells- CC and blood cells
subpopulations), T, G, histology, age, AT, localization, prothrombin index, hemorrhage time, residual nitrogen, protein
(P=0.000-0.019). Neural networks, genetic algorithm selection and bootstrap simulation revealed relationships between
5YS and
healthy cells/CC (rank=1), PT early-invasive EC (2); PT N0—N12 (3), erythrocytes/CC (4), thrombocytes/CC (5); stick
neutrophils/CC (6), lymphocytes/CC (7), segmented neutrophils/CC (8), eosinophils/CC (9), leucocytes/CC (10);
monocytes/CC (11). Correct prediction of 5YS was 100% by neural networks computing (area under ROC curve=1.0;
error=0.0).
CONCLUSIONS: 5-year survival of ECP after radical procedures significantly depended on: 1) PT “early-invasive
cancer”; 2) PT N0--N12; 3) Cell Ratio Factors; 4) blood cell circuit; 5) biochemical factors; 6) hemostasis system; 7) AT;
8) EC characteristics; 9) EC cell dynamics; 10) tumor localization; 11) anthropometric data; 12) surgery type. Optimal
diagnosis and treatment strategies for EC are: 1) screening and early detection of EC; 2) availability of experienced
thoracoabdominal surgeons because of complexity of radical procedures; 3) aggressive en block surgery and adequate
lymph node dissection for completeness; 4) precise prediction; 5) adjuvant chemoimmunoradiotherapy for ECP with
unfavorable prognosis.
Factors (n=413) Rank Sensitivity
Phase Transition Early---Invasive Cancer 1 24671
Healthy Cells/Cancer Cells
Erythrocytes/Cancer Cells
Phase Transition N0—N12
Thrombocytes/Cancer Cells
Segmented Neutrophils/Cancer Cells
Stick Neutrophils/Cancer Cells
Lymphocytes/Cancer Cells
Monocytes/Cancer Cells
Leucocytes/Cancer Cells
Eosinophils/Cancer Cells
2
3
4
5
6
7
8
9
10
11
23071
18425
18188
12766
7861
7164
6424
6229
5643
4982
Significant Factors (Number of Samples=3333) Rank Kendal Tau-A P
Tumor Size 1 -0.308 0.000
Healthy Cells/Cancer Cells 2 0.305 0.000
T1-4 3 -0.299 0.000
Erythrocytes/Cancer Cells 4 0.299 0.000
Leucocytes/Cancer Cells 5 0.290 0.000
Thrombocytes/Cancer Cells 6 0.285 0.000
Lymphocytes/Cancer Cells 7 0.281 0.000
Residual Nitrogen 8 -0.275 0.000
Segmented Neutrophils/Cancer Cells 9 0.273 0.000
Phase Transition N0---N12 10 -0.239 0.000
Hemorrhage Time 11 -0.228 0.000
Monocytes/Cancer Cells 12 0.227 0.000
Phase Transition Early---Invasive Cancer 13 -0.222 0.000
Esophageal/Cardioesophageal Cancer 14 -0.191 0.000
Operation Type 15 -0.187 0.000
Eosinophils/Cancer Cells 16 0.173 0.000
Stick Neutrophils/Cancer Cells 17 0.144 0.001
G1-3 18 -0.140 0.001
Tumor Growth 19 -0.113 0.01
Erythrocytes 20 0.100 0.01
Combined Procedure 21 0.095 0.01
Weight 22 0.092 0.01
Localization 23 0.069 0.05