AI-Driven Digital Twin and Drug Simulation for Esophageal Cancer

This is a prospective, single-center, observational cohort study integrating virtual clinical trial technology. The study is designed to validate the feasibility and accuracy of AI-driven digital twin models for predicting treatment responses in esophageal cancer patients.

Background and Rationale:

China accounts for approximately half of the global esophageal cancer cases, with the highest incidence rates concentrated in the Taihang Mountain region, particularly Henan Province. Despite advances in early diagnosis and treatment, tumor heterogeneity and individual patient variability continue to limit therapeutic efficacy, especially in advanced-stage patients. Virtual clinical trial technology, which uses supercomputing to simulate drug efficacy and safety across virtual patient populations, offers a promising approach to accelerate drug screening while reducing costs and ethical constraints associated with conventional trials.

Study Population:

The study will enroll 400 patients with locally advanced (T1N1-3M0 or T2-3N0-3M0) thoracic esophageal squamous cell carcinoma (8th UICC-TNM staging), aged 18 to 75 years, with ECOG performance status 0-1 and adequate organ function. All patients must have received no prior antitumor therapy for esophageal cancer and be candidates for R0 resection.

Virtual Patient Modeling:

For each enrolled patient, peripheral blood samples will be collected prior to treatment initiation for whole-exome sequencing (WES). Multi-omics data will be integrated with clinical information to generate a personalized digital twin for each participant. These virtual patients will undergo simulated treatment and efficacy prediction using the drug efficacy prediction model.

Treatment Arms:

• Experimental Group (n=200): Neoadjuvant immunochemotherapy consisting of toripalimab (JS001) 240 mg IV on day 3, paclitaxel 175 mg/m2 IV on day 1, and cisplatin 75 mg/m2 IV on day 1, administered in 21-day cycles for 2 cycles, followed by surgery and postoperative toripalimab maintenance therapy (240 mg Q3W for up to 8 cycles or 6 months).
• Control Group (n=200): Neoadjuvant chemotherapy consisting of paclitaxel 175 mg/m2 IV on day 1 and cisplatin 75 mg/m2 IV on day 1, administered in 21-day cycles for 2 cycles, followed by surgery.

Surgery and Follow-up:

Surgery will be performed 4-8 weeks after the last neoadjuvant cycle using McKeown minimally invasive or open esophagectomy with two-field lymphadenectomy. Postoperative surveillance includes contrast-enhanced CT of the chest and upper abdomen and contrast-enhanced CT of the neck, performed every 3 months for 2 years, then every 6 months until year 5. Adverse events will be monitored throughout and graded according to CTCAE v5.0.

Outcomes:

The primary endpoint is Event-Free Survival (EFS), defined as the time from randomization to the first occurrence of inoperability, distant metastasis, local recurrence, or death from any cause, whichever occurs first. Secondary endpoints include pathological complete response (pCR) rate, overall survival (OS), disease-free survival (DFS), objective response rate (ORR), R0 resection rate, major pathological response (MPR) rate, quality of life (EORTC QLQ-C30 and QLQ-OES18), and safety (incidence of adverse events per CTCAE v5.0).

Virtual Clinical Trial Component:

Parallel to the real-world IIT, a virtual clinical trial will be conducted using the AI-powered drug simulation platform. The virtual predictions will be compared with actual clinical outcomes to evaluate the accuracy and reliability of the digital twin model. The study will assess the concordance between predicted and observed responses, including sensitivity, specificity, positive predictive value, and negative predictive value of the model.

Statistical Analysis:

Based on the primary endpoint EFS, a sample size of 400 patients (1:1 allocation) is planned, providing 90% power at a two-sided alpha level of 0.05 to detect a hazard ratio of 0.68, assuming a median EFS of 20 months in the control group. Efficacy analyses will be performed on the intention-to-treat (ITT) population using stratified log-rank tests and Cox proportional hazards models. Kaplan-Meier methods will estimate survival distributions. Safety analyses will be conducted on the safety set.

Study Duration:

The study is anticipated to enroll participants from January 2026 to June 2028, with final data collection completed by December 2028.

Inclusion criteria

• Treatment-naïve patients with clinical stage locally advanced (T1N1-3M0 or T2-3N0-3M0) thoracic esophageal squamous cell carcinoma, according to the 8th UICC-TNM staging system.
• Cervical color Doppler ultrasound shows no suspicious metastatic lymph nodes, or patients with suspected lymph node metastasis on ultrasound who are eligible for three-field lymphadenectomy.
• No prior antitumor therapy for esophageal cancer (including chemotherapy, radiotherapy, or immunotherapy).
• Preoperative evaluation of organ function indicates no surgical contraindications.
• Adequate bone marrow, liver, and renal function, confirmed by the following laboratory tests performed within 7 days prior to enrollment:

• Hemoglobin ≥ 90 g/L;
• White blood cell count ≥ 4.0 × 10^9/L;
• Absolute neutrophil count (ANC) ≥ 1.5 × 10^9/L;
• Platelet count ≥ 100 × 10^9/L;
• Total bilirubin ≤ 1.5 × upper limit of normal (ULN);
• ALT and AST ≤ 2.5 × ULN;
• International normalized ratio (INR) of prothrombin time ≤ 1.5 × ULN, with activated partial thromboplastin time (APTT) within normal range;
• Serum creatinine ≤ 1.5 × ULN.
• No prior chemotherapy, radiotherapy, or hormone therapy for malignant tumors; no history of other malignancies, except for prostate cancer patients who received hormone therapy and have achieved disease-free survival (DFS) > 5 years.
• Expected to achieve R0 resection.
• Eastern Cooperative Oncology Group (ECOG) performance status score 0 to 1.
• Life expectancy ≥ 3 months.
• Women of childbearing potential must agree to use effective contraceptive measures (e.g., intrauterine device, oral contraceptives, or condoms) during the study and for 6 months after study completion. Serum or urine pregnancy test must be negative within 7 days prior to enrollment, and patients must not be breastfeeding. Male patients must agree to use contraceptive measures during the study and for 6 months after study completion.
• Patients must voluntarily participate in the study, provide written informed consent, demonstrate good compliance, and agree to follow-up.

Exclusion criteria

• Patients not meeting the inclusion criteria.
• Presence of multiple primary malignancies (synchronous or metachronous).
• Active infections requiring systemic treatment.
• Requirement for continuous systemic corticosteroid therapy (>10 mg/day prednisone or equivalent) for comorbid conditions.
• Unstable angina within 3 months or myocardial infarction within 6 months prior to enrollment.
• Psychiatric disorders that may affect study compliance.
• Known or concurrent hemorrhagic disorders.
• Female patients who are pregnant or breastfeeding.
• Patients with pre-existing or concurrent pulmonary fibrosis, interstitial pneumonia, pneumoconiosis, radiation pneumonitis, or other severe pulmonary impairment.
• Autoimmune diseases, immunodeficiency states, or history of organ transplantation.
• Known hypersensitivity to the study drugs (toripalimab, paclitaxel, cisplatin) or their excipients.
• Abnormal coagulation function (PT > 16s, APTT > 53s, TT > 21s, Fib < 1.5 g/L), bleeding tendency, or patients receiving thrombolytic or anticoagulant therapy.
• Bronchial asthma requiring intermittent use of bronchodilators or medical intervention.
• Active hepatitis B (HBV) or hepatitis C (HCV) infection. (Patients who are HBsAg positive or HBcAb positive may be eligible if HBV DNA is below the lower limit of detection/quantification; patients who are HCV antibody positive may be eligible if HCV RNA is below the lower limit of detection/quantification.)
• Human immunodeficiency virus (HIV) positivity.
• Any other condition that, in the investigator's judgment, may compromise patient safety, interfere with study compliance, or preclude successful completion of the study.