Abstract:Objective To evaluate the auxiliary diagnostic value of metabolic indicators and tumor markers combined with imaging in liver metastasis from pancreatic cancer，and to construct a nomogram model to improve diagnostic accuracy of radiologically indeterminate lesions. Methods A total of 255 patients with pancreatic cancer confirmed at Nanjing Drum Tower Hospital Affiliated to Nanjing University Medical School between January 2019 and August 2025 were retrospectively enrolled，including 46 patients with liver metastasis and 209 without. All cases were randomly divided into a training set（n=177，34 with liver metastasis）and a validation set（n=78，12 with liver metastasis）at a ratio of 7∶3. Clinical information，including CT/MRI imaging and routine laboratory data［lipid profile，glycated hemoglobin（HbA1C），tumor markers，liver enzymes］，was collected. Independent associated factors were identified in the training set through logistic regression analysis，and a combined nomogram model integrating these factors with imaging information was constructed，and its performance was assessed in both sets using ROC analysis，calibration，and decision curve analysis. Results Patients with liver metastasis had significantly higher levels of low - density lipoprotein cholesterol（LDL-C），HbA1C，lactate dehydrogenase（LDH），as well as higher proportion of patients with carcinoembryonic antigen（CEA）>10 ng/mL and CA125>30.2 u/mL compared with those without metastasis（P<0.05）. Multivariate analysis of the training set identified that elevated LDL-C（OR=2.449，95%CI：1.085-5.525，P=0.031），elevated HbA1C（OR=1.625，95%CI：1.107-2.384，P=0.013），and CA125>30.2 u/mL（OR=3.540，95%CI：1.006-12.450，P=0.049）were independent predictors of liver metastasis. The nomogram incorporating these factors and imaging information demonstrated excellent performance in the training set［area under the curve（AUC）=0.961，95%CI：0.907-1.000；sensitivity 89.9%，specificity 94.4%］，and outperformed imaging alone（AUC=0.806）. In the validation set，the model achieved an AUC of 0.907（95%CI：0.779-1.000），with sensitivity of 77.8% and specificity of 95.8%. Calibration curves indicated good agreement between predicted and observed outcomes，and decision curve analysis confirmed that the nomogram model provided a higher net clinical benefit across a wider range of threshold values. Conclusion Elevated LDL-C，HbA1C，and CA125>30.2 u/mL are significantly associated with liver metastasis in pancreatic cancer. A nomogram model integrating these indicators with imaging findings can markedly improve diagnostic accuracy，particularly in patients with atypical imaging results or limited pathological sampling.