On-treatment change in bone turnover markers predicts 2-year bone mineral density after sequential therapy following romosozumab: a real-world cohort study.

Background: Romosozumab significantly increases bone mineral density (BMD); however, identifying patients who will achieve the greatest long-term BMD gains after sequential therapy remains challenging. BTMs decrease markedly during romosozumab treatment; however, whether the magnitude of this on-treatment change predicts 2-year cumulative BMD has not been established. Methods: In this retrospective single-center cohort study (n=315; April 2019-April 2025; ethics approval RO7-5), we analyzed 129 patients with complete on-treatment BTM pairs and 24-month BMD data. On-treatment changes in procollagen type I N-terminal propeptide (ΔP1NP) and tartrate-resistant acid phosphatase 5b (ΔTRACP-5b) were calculated as the difference between post-treatment and pre-treatment values. The primary outcome was the cumulative percentage change in lumbar spine BMD (LS-BMD) from baseline to 24 months. Spearman's rank correlations and multiple linear regression adjusting for age, baseline LS-BMD, prior treatment, sequential therapy type, and 12-month BMD change were performed. ROC analysis defined an optimal ΔP1NP cutoff for predicting a good response, and subgroup analyses examined the predictive value by sequential therapy type and prior treatment status. Results: ΔP1NP was significantly correlated with the 24-month cumulative LS-BMD change (ρ=-0.375, p<0.0001, n=129), and a greater decline in P1NP during romosozumab treatment predicted greater cumulative BMD gain. ΔTRACP-5b showed a similar association (ρ=-0.348, p=0.0001, n=114). In multiple linear regression adjusted for all covariates, ΔP1NP remained an independent predictor (β=-0.061, p=0.010; model R²=0.796, n=113). The optimal ΔP1NP cutoff was -40.3 μg/L (AUC = 0.761; sensitivity=0.545, specificity=0.854). Patients in the highest ΔP1NP decline tertile (T1) achieved 23.5% cumulative LS-BMD gain versus 12.3% in the lowest tertile (T3; p=0.0003). This association was replicated in the denosumab sequential therapy subgroup (ρ=-0.354, p=0.0010, n=84). The predictive value was stronger in treatment-naïve patients (ρ=-0.359, p=0.0021) than in previously treated patients (ρ=-0.207, p=0.120; interaction p=0.024). Conclusion: On-treatment ΔP1NP is an independent predictor of 2-year cumulative BMD after sequential therapy following romosozumab, with the greatest predictive value in treatment-naïve patients. These findings support the clinical utility of BTM monitoring during romosozumab treatment; however, given the single-center retrospective design, predominantly female Japanese population (96.9%), and unvalidated ROC cutoff, prospective multicenter validation is required before clinical implementation.