The paper introduces TWIN-GPT, a large language model-based approach for creating personalized digital twins to enhance the accuracy and efficiency of clinical trial outcomes. TWIN-GPT leverages ChatGPT to generate digital twins that simulate patient physiological states, disease progression, and treatment effects, addressing the challenges of limited data and data inconsistencies in electronic health records (EHRs). The model is fine-tuned on clinical trial datasets to generate unique digital twins for different patients, improving the prediction accuracy of clinical trial outcomes. Experiments demonstrate that TWIN-GPT can produce high-fidelity trial data, closely aligning with real-world scenarios, and enhance the privacy protection of patient information. The method is evaluated through various metrics, including dimension-wise probability, counterfactual digital twin evaluation, severe outcome prediction, and adverse event prediction, showing superior performance compared to existing methods. The study highlights the potential of TWIN-GPT in advancing clinical research and patient care.The paper introduces TWIN-GPT, a large language model-based approach for creating personalized digital twins to enhance the accuracy and efficiency of clinical trial outcomes. TWIN-GPT leverages ChatGPT to generate digital twins that simulate patient physiological states, disease progression, and treatment effects, addressing the challenges of limited data and data inconsistencies in electronic health records (EHRs). The model is fine-tuned on clinical trial datasets to generate unique digital twins for different patients, improving the prediction accuracy of clinical trial outcomes. Experiments demonstrate that TWIN-GPT can produce high-fidelity trial data, closely aligning with real-world scenarios, and enhance the privacy protection of patient information. The method is evaluated through various metrics, including dimension-wise probability, counterfactual digital twin evaluation, severe outcome prediction, and adverse event prediction, showing superior performance compared to existing methods. The study highlights the potential of TWIN-GPT in advancing clinical research and patient care.