Development and validation of a fluoride bone injury risk prediction model

Main Article Content

Jun Tu
Ke-jian Liu
Dan Kuang
Liang Wang

Keywords

Abstract

Background: Fluoride bone injury affects millions of people exposed to fluoride worldwide, and has no treatment - prevention is the only solution. Objectives: A risk prediction model was developed to identify workers at high risk for fluoride bone injury in aluminum production. Methods: We collected data from the Molecular Epidemiology Study of Fluoride Bone Injury. 120 fluoride bone injury cases and 120 controls were involved in the study. Logistic regression was used to determine variables in the risk prediction model. Predictive accuracy was validated with bootstrap method. Potential risk cut-offs was evaluated with receiver operating characteristic curve. Results: Working history, urinary fluoride, osteocalcin, bone alkaline phosphatase and calcitonin receptor gene polymorphism were included in the final prediction model. The model had very good calibration and discrimination (C index=0.986; Brier score 0.014). Conclusions: Our fluoride bone injury risk prediction model performed well in the present data, and the working history, urinary fluoride, osteocalcin, bone alkaline phosphatase, and calcitonin receptor gene polymorphism were identified as predictors. The model could be used to assess the fluoride bone injury risk, and identify the susceptible workers.

Abstract 310 | PDF Downloads 19