Development and validation of a prognostic gene expression model for second primary malignancy in head and neck squamous cell carcinoma

Abstract Second primary malignancy (SPM) and local recurrence (LR) are linked to poor prognosis in head and neck squamous cell carcinoma (HNSCC), highlighting the need for early detection in the successful management of HNSCC patients. This study aimed to develop prognostic models for predicting time to SPM and LR. A retrospective cohort of 1,113 HNSCC patients (2008–2019) was analyzed. Gene expression data from 419 patients with primary HNSCC (76 with SPM, 67 with LR, and 289 with neither event) were evaluated. RNA was extracted from FFPE tissues of the index tumor and quantified by NanoString nCounter ® PlexSet™ assay. Predictive scoring algorithms were developed using the least absolute shrinkage and selection operator (LASSO) approach to derive sparse models for each outcome. The SPM risk score predicted earlier SPM in training (HR 4.47; C-index 0.79) and test (HR 7.19; C-index 0.81) sets. The LR risk score showed modest discrimination (C-index 0.62 in both sets). The score was significantly associated with LR in the training set (HR 39.62) but not in the test set (HR 25.96, p = 0.20). Kaplan–Meier curves separated high- versus low-risk groups for SPM and LR (all p < 0.05). The validated SPM risk score may support personalized surveillance frequency and tailored adjuvant treatment for high-risk patients, potentially improving intermediate- to long-term HNSCC management, whereas the LR findings remain exploratory and require confirmation in a larger independent cohort.