Hidefumi Kasai (1), Yoko Mori (2), Atsushi Ose (2), Masataka Shiraki (3) and Yusuke Tanigawara (1)
(1) Keio University School of Medicine, Japan, (2) Asahi Kasei Pharma Corporation, Japan, (3) Research Institute and Practice for Involutional Diseases, Japan
Objectives: Prevention of fractures is the final goal of osteoporosis treatment. The efficacy of drug treatment turns out long after the beginning of the treatment because improvement of bone strength takes long time. On the other hand, changes in bone turnover markers (BTMs) are observed within a few weeks after the start of treatment with bisphosphonates. Bone mineral density (BMD) is considered as a surrogate endpoint in osteoporosis drug development. But the changes in BMD are observed later than that of BTMs. Zoledronate acid is a bisphosphonate which, with once-yearly intravenous infusion, has been shown to cause an increase in BMD and reduces bone fracture risk [1]. We have already reported that BMD profiles up to 2 years can be predicted using patients’ background characteristics and the early response of TRACP-5b [2]. This study aimed to develop a mathematical model predicting long-term fracture risk after two annual administration of 5 mg of zoledronic acid using the early response of a bone resorption marker or BMD in osteoporosis patients.
Methods: All data used in this analysis were obtained from 656 patients with primary osteoporosis from a randomized, placebo-controlled, double-blind, 2-year study of zoledronic acid [1]. Zoledronic acid or placebo was once-a-yearly administered. Individual bone resorption marker (tartrate-resistant acid phosphatase 5b, TRACP-5b) profiles for 2-year were predicted using the baseline and the first observation (3 months) data with the previously developed pharmacodynamic model [2] and each patient’s parameter values as Bayesian maximum a posteriori (MA) estimates. Individual BMD profiles were also predicted using the baseline and the first (6 months) observations by the same approach with the model [2]. A parametric time-to-event model for clinical fracture was developed. Parametric hazards were assumed as exponential, Weibull, or log-normal distribution. The hazard for the fracture was predicted with TRACP-5b or BMD profiles and number of baseline vertebral fractures. All the analyses were performed using the Phoenix NLME software Version 8.1 (Certara, LP, Princeton, NJ).
Results: The parametric fracture risk model up to 2 years were constructed with significant covariates of patients’ TRACP-5b profile as a time-varying covariate and number of prevalent vertebral fractures at baseline. The model using BMD profiles also succeeded to predict fracture risk. Both the models gave similar predictive ability, but, although the model using BMD gave a slightly better prediction (lower Akaike Information Criterion), the model using TRACP-5b was considered to be better because the change in TRACP-5b concentration can be detected earlier than BMD and model prediction can be performed at an earlier stage. No other covariates, such as gender and age, found to be significant. The simulated 90% prediction interval almost covered the observed Kaplan-Meier fracture profiles, and the predictions were comparable to the observed fracture rate both for zoledronic acid and placebo groups. The final model showed that fracture risk was predicted to increase when (1) higher levels of TRACP-5b or (2) larger number of baseline vertebrate fractures.
Conclusions: Fracture risk could be predicted by TRACP-5b or BMD profile. Early (3 months) TRACP-5b measurement is highly valuable to predict a patient’s 2 years fracture risk.
References:
[1] Nakamura T et al. Osteoporos Int (2017) 28:389-398.
[2] Mori Y et al. Osteoporos Int (2018) 29:1155-1163.
Reference: PAGE () Abstr 9417 [www.page-meeting.org/?abstract=9417]
Poster: Drug/Disease Modelling - Other Topics