Risk Factors Analysis of Bone Mineral Density Based on Lasso and Quantile Regression in America during 2015–2018
Chao Sun, Boya Zhu, Sirong Zhu, Long Jiang Zhang, Xiaoan Du, Xiaodong Tan
Abstract
This study aimed to explore the risk factors of bone mineral density (BMD) in American residents and further analyse the extent of effects, to provide preventive guidance for maintenance of bone health. A cross-sectional study analysis was carried out in this study, of which data validity was identified and ethics approval was exempted based on the National Health and Nutrition Examination Survey (NHANES) database. Candidates’ demographics, physical examination, laboratory indicators and part of questionnaire information were collected and merged from NHANES in 2015–2016 and 2017–2018. The least absolute shrinkage selection operator (lasso) was used to select initial variables with “glmnet” package of R, quantile regression model to analyze influence factors of BMD and their effects in different sites with “qreg” code in Stata. Among 2937 candidates, 17 covariates were selected by lasso regression (λ = 0.00032) in left arm BMD, with 16 covariates in left leg BMD (λ = 0.00052) and 14 covariates in total BMD (λ = 0.00065). Quantile regression results displayed several factors with different coefficients in separate sites and quantiles: gender, age, educational status, race, high-density lipoprotein (HDL), total cholesterol (TC), lead, manganese, ethyl mercury, smoking, alcohol use and body mass index (BMI) (p < 0.05). We constructed robust regression models to conclude that some demographic characteristics, nutritional factors (especially lipid levels, heavy metals) and unhealthy behaviors affected BMD in varying degrees. Gender and race differences, Low-fat food intake and low exposure to heavy metals (mostly lead, manganese and mercury) should be considered by both clinical doctors and people. There is still no consensus on the impact of smoking and alcohol use on bone mineral density in our study.