Multi-state modeling associates proliferation genes with early estrogen receptor-positive breast cancer recurrence and survival

Background To identify baseline gene expression programs associated with recurrence timing in estrogen receptor-positive (ER+) breast cancer (BC) using a multi-state modeling framework. Methods We analyzed high-risk ER + BC patients with surgical tumor specimens profiled using a custom NanoString panel (145 genes). Recurrence was classified as none, early (\(\:<\)5 years after surgery), or late (\(\:\ge\:\)5 years). Differential gene expression was assessed using negative binomial regression. Transition-specific risks of recurrence and survival were estimated using semi-Markov multi-state Cox models, with sensitivity analyses conducted under alternative definitions of early and late recurrence. Results Among 79 patients analyzed, 28% developed recurrence (14 early, 9 late), and all BC-specific deaths occurred following early recurrence. In transition-specific models, higher baseline expression of proliferation-related genes, including UBE2C, EZH2, CCNB1, PTTG1 , and MKI67 , was consistently associated with increased risk of early recurrence (hazard ratio [HR] range: 1.91–2.14; all p < 0.05), whereas SNAI2 expression was protective (HR: 0.43; 95% CI: 0.18–0.99). In contrast, late recurrence was associated with signaling-related genes such as IGF1R, LEF1 , and WNT7B (HR range: 1.52–3.40; all p < 0.05). Early recurrence-associated genes showed concordance across differential expression analyses and sensitivity analyses, supporting biological coherence. Conclusions This study identifies a potential proliferative gene assay associated with early recurrence in ER + BC, while late recurrence appears linked to distinct signaling pathways. These findings prioritize candidate molecular biomarkers for future validation to improve risk stratification and surveillance.