Efficacy and Safety of Neoadjuvant HAIC Combined With Tislelizumab and Lenvatinib in Patients With Resectable HCC Screened by a Multimodal Deep Learning Model: a Multicenter Randomized Controlled Trial.

The protocol was revised in April 2025 (V1.1). Primary liver cancer is one of the most common malignant tumors in the world, and about 80%~90% of primary liver cancers are pathologically characterized as hepatocellular carcinoma (HCC). Radical surgery is the main method for patients with HCC to obtain long-term survival. However, the early recurrence rate of high-risk HCC is very high, which seriously affects the overall therapeutic effect. Although the tumor is in BCLC-A stage, when the tumor diameter is more than 5cm, the effect of surgery is worse than that of single small HCC due to the large resection range, the high risk of surgery and residual disease. In addition, BCLC-stage B and C tumors have a high recurrence rate due to multiple lesions or macrovascular invasion. To address this issue, new tools are urgently needed to guide the selection of appropriate treatment regimens to reduce the risk of postoperative recurrence and improve overall survival.

The investigators multidisciplinary team used deep learning technology to construct an artificial intelligence prediction model of neoadjuvant therapy (Neoadj-Net) benefit based on pre-treatment genetic testing data, digital pathology slides and imaging data (enhanced MRI) of 536 intermediate-stage HCC patients treated with HAIC in combination with lenvatinib and PD-1 monoclonal antibody in six centers, and external center data validated the model's good ability to identify the beneficiary population of the combination regimen ( AUC 0.89, Accuracy 0.86). This study is to explore the effectiveness and safety of Neoadj-Net in reducing postoperative recurrence by observing the benefit of the combined neoadjuvant regimen in patients who are potentially benefited from neoadjuvant therapy and direct surgery from the perspective of precision therapy.