Immunochemotherapy has been recommended as a first-line treatment for patients with de novo metastatic nasopharyngeal carcinoma (dmNPC), but there is a lack of effective ways to predict survival benefits. The objective of this study was to utilize multimodal data to construct a prognostic model for patients undergoing immunochemotherapy, to facilitate subsequent treatment decisions.A total of 268 patients with dmNPC who received first-line programmed death-1(PD-1) inhibitor in combination with a chemotherapy regimen were enrolled. Radiomics and pathomics features were extracted from magnetic resonance (MR) images and whole slide images (WSI) of nasopharyngeal hematoxylin-eosin staining (HE) slides to construct radiomics and pathomics scores. Clinical and morphological features were collected for analysis. A clinical prognostic model, a clinical-radiomics (CR) prognostic model, and a clinical-radiomics-pathomics (CRP) prognostic model were constructed using independent prognostic features, radiomics scores, and/or pathomics scores. The performance of these models was described and compared. The survival benefits of subsequent radiotherapy were investigated in patients with different risk levels based on the most effective model.The C-indexes of the CRP model in the training cohort, internal and external validation cohort were 0.823 (0.751-0.895), 0.787 (0.675-0.899), and 0.772 (0.569-0.974), respectively. After the integrated discrimination improvement test, the CRP model demonstrated superior efficacy compared to the CR model in the training cohort, internal validation cohort, and external validation cohort (all p < 0.001). In the low-risk group, patients receiving locoregional radiotherapy achieved better survival outcomes compared to those who did not (p = 0.009).The CRP model is capable of accurately predicting the survival of patients with dmNPC after first-line immunochemotherapy. In patients with low risk, the addition of subsequent locoregional radiotherapy may further improve the survival outcomes.Copyright © 2025. Published by Elsevier B.V.