Deep learning enables to clarify the association of osteoporosis risk with bone metastasis in premenopausal women after surgery for early-stage breast cancer: multicenter retrospective cohort study

BackgroundAdjuvant use of bone-modifying agents (BMAs) to early-stage breast cancer (eBC) aims to maintain bone density, leading to prevention of bone metastasis (BM) in postmenopausal women; its mechanism remains unknown. Clinically, one-quarter of premenopausal women develop osteopenia. To this, deep learning (DL) enables to evaluate the risk of osteoporosis (RO) using perioperative chest radiography. The aim of the study was to clarify the association of RO with BM after eBC surgery and perform stratified analyses by menopausal status. MethodsIn this multicentre retrospective cohort study, we enrolled 785 women who underwent surgery for estrogen receptor-positive eBC between 2007 and 2019, excluding those with adjuvant BMAs. Perioperative chest radiography was interpretated into the DL-producing RO score (0-1), classifying them into 359 women with low risk of osteoporosis (LRO) and 426 women with high risk (HRO). The 5-year BM-free survival (BMFS) was compared. FindingsUnivariate stratification by menopausal status for the enrolled women (median [interquartile range] age, 56 [46-66] years) revealed: the number of premenopausal women, 325; HRO (23.1% [75/325]) in premenopausal women was associated with low BMFS (hazard ratio [HR] 2.37, 95% confidence interval [CI]; 1.23-4.57, p<0.01), whereas no association was observed in the others (p=0.78). Multivariate analyses identified RO as a poor prognostic factor in the enrolled women (HR 3.08, 95%CI; 1.23-7.47, p=0.01). Conduction rate of bone density tests was lower in premenopausal women (31/325 vs. 317/460, p<0.01). InterpretationRO is an independent prognostic factor for BM in women with eBC. Our study revealed association between RO and BM, specifically in premenopausal women, whose bone health is poorly studied. Expanded indications of adjuvant BMA are suggested for at-risk premenopausal women. FundingNone.