Abstract

Oral bisphosphonates are the primary medication for osteoporosis, but concerns exist regarding potential bone-quality changes or low-energy fractures. This cross-sectional study used artificial intelligence methods to analyze relationships among bisphosphonate treatment duration, a wide variety of bone-quality parameters, and low-energy fractures. Fourier transform infrared spectroscopy and histomorphometry quantified bone-quality parameters in 67 osteoporotic women treated with oral bisphosphonates for 1 to 14 years. Artificial intelligence methods established two models relating bisphosphonate treatment duration to bone-quality changes and to low-energy clinical fractures. The model relating bisphosphonate treatment duration to bone quality demonstrated optimal performance when treatment durations of 1 to 8 years were separated from treatment durations of 9 to 14 years. This may be due to a change in relationship of bone-quality parameters with treatment duration. This model also showed that the effects of bisphosphonate treatment duration were most highly correlated with changes in means and standard deviations of infrared spectroscopically derived mineral and matrix parameters and histomorphometric bone turnover parameters. A second model related treatment duration to bone fracture in all 22 patients who fractured while on treatment with bisphosphonates for more than 8 years. This second model showed that bisphosphonate treatment duration, not hip bone mineral density (BMD), was the most strongly correlated parameter to these low-energy bone fractures. Application of artificial intelligence enabled analysis of large quantities of structural, cellular, mineral, and matrix bone-quality parameters to determine relationships with long-term oral bisphosphonate treatment and fracture. Infrared spectroscopy provides clinically relevant bone-quality information of which bone mineral purity is among the most relevant. Nine or more years of bisphosphonate treatment was associated with abnormal bone mineral purity, matrix abnormalities, and low-energy fractures. These data justify limiting bisphosphonate treatment duration to 8 years.

Document Type

Article

Publication Date

9-6-2021

Notes/Citation Information

Published in JBMR Plus, v. 5, no. 11, e10549.

© 2021 The Authors

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Digital Object Identifier (DOI)

https://doi.org/10.1002/jbm4.10549

Funding Information

We acknowledge funding support from the National Institutes of Health-National Institute of Arthritis and Musculoskeletal and Skin Diseases (award R01AR061578) and the Kentucky Nephrology Research Trust.

Related Content

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

jbm410549-sup-0001-figures1.tiff (12659 kB)
Supplemental Fig. S1. Graphical depiction of estimated linear coefficients of various bone-quality relevant parameters as they relate to bisphosphonate treatment duration. This figure shows the full list of parameter features used in the machine-learning model relating bisphosphonate treatment duration and bone quality, ie, “the duration model.” The magnitude of the estimated linear coefficient of each model parameter is proportional to horizontal length. Signs of these linear coefficients are depicted by position versus the centerline. Red bars extending to the left denote parameters negatively correlated to bisphosphonate treatment duration; blue bars extending to the right denote parameters positively correlated to bisphosphonate treatment.

jbm410549-sup-0002-figures2.tiff (8440 kB)
Supplemental Fig. S2. Graphical representation showing linear coefficients of various bone-quality relevant parameters as they relate to bone fracture in patients treated with bisphosphonates. This figure shows the full list of parameters used in the machine-learning model relating bone fracture to bisphosphonate treatment duration and bone quality, ie, “the fracture model.” The magnitude of the linear coefficient of each model parameter is proportional to horizontal bar length. Signs of these linear coefficients are depicted by position versus the centerline. Red bars extending to the left denote parameters negatively correlated to bisphosphonate treatment duration; blue bars extending to the right denote parameters positively correlated to bisphosphonate treatment duration.

Share

COinS