e15035 Background: Lung metastasis is common in patients with colorectal cancer (CRC). However, diagnosing lung metastasis in CRC is challenging due to slow-growing pulmonary lesions. Circulating tumor DNA (ctDNA) testing provides a direct indication of the presence of minimal residual disease (MRD) which contributing to cancer recurrence. Yet, there are no reports on the potential indications of ctDNA for the site of metastasis. The objective of this study is to develop a predictive model for lung metastasis in CRC patients who have received adjuvant therapy by integrating ctDNA with clinicopathological factors. Methods: Two independent cohorts' series were evaluated including 299 patients from China as training cohort and 134 patients from Spain as validation cohort from 2012-2020. We specifically focused on the detection of ctDNA in these patients. We employed LASSO regression for optimal feature selection and built a predictive model by multivariable logistic regression. Results: In training cohort, 59/299(19.7%) patients have detectable ctDNA. In validation cohort, 24/134 (18.6%) patients have detectable ctDNA. Patients who tested ctDNA-positive had a significantly higher incidence of lung metastasis (33.9% vs. 11.2%, P<0.001) compared to ctDNA-negative patients. A predictive model for lung metastasis was developed by multivariable logistic regression analysis. The model incorporated factors such as ctDNA status, preoperative carcinoembryonic antigen (CEA) level, presence of cancerous nodes, and the mutation status of Phosphatidylinositol-4,5-Bisphosphate 3-Kinase Catalytic Subunit Alpha (PIK3CA) in plasma. The lung metastasis prediction nomogram demonstrated a favorable ability to distinguish with a concordance index of 0.775 (95% CI: 0.706-0.775) during internal validation. The Hosmer-Lemeshow test, with p-value of 0.967, indicated that prediction nomogram model had a good fit. In addition, the prediction model of validation cohort demonstrated a discriminative performance with a concordance index of 0.724(95% CI: 0.698-0.724). Furthermore, the DCA (Decision Curve Analysis) and CIC (Clinical Impact Curve) in both cohorts indicated that the lung metastasis nomogram exhibited a favorable clinical net benefit. Finally, through multi-model comparison, it was found that the predictive performance of the model, which includes ctDNA and blood PIK3CA mutation status, was significantly superior to all clinical prediction models. Conclusions: A lung metastasis prediction nomogram was developed by integrating postoperative ctDNA and clinicopathological risk factors. The purpose of this nomogram is to assist clinicians in predicting the probability of lung metastasis in patients who have undergone surgery for CRC, with the aim to facilitate prompt intervention.