Nonmodifiable (e.g., genetics, sex, age) and modifiable (e.g., diet, physical activity, smoking) factors influence bone health. However, it is not known the relative impact of these factors on bone health. Chin and colleagues conducted a study aimed at defining the impact of these factors on bone mineral content (BMC) and bone mineral density (BMD) in healthy US men and women. They published their observations in the November 2021 issue of The Journal of Nutrition.
Several machine-learning models (ridge, lasso, elastic net, and random forest) were used to predict whole-body, femoral neck, and spine BMC and BMD. Subjects were between 18 and 66 years of age with a BMI of 18-44 kg/m2. Nonmodifiable variables included in the analyses were anthropometric, physiological, and demographical ones. Directly modifiable lifestyle variables (physical activity and smoking) and dietary variables (from FFQ), as well as variables that approximate directly modifiable factors (circulating 25-hydroxycholecalciferol and stool pH) were included in the models.
More of the variation in BMC and BMD was explained by models using nonmodifiable variables compared to those only using directly modifiable variables. The machine-learning models were also more effective than multivariate linear regression models reliant only on directly modifiable variables. Predictors of BMC and BMD included BMI, body fat percentage, height, and menstruation history, as were modifiable variables, such as cholesterol, eggs, cheese, cured meat, refined grains, and alcohol consumption. BMC and BMD were also predicted by low stool pH. The authors concluded that modifiable factors, like diet, explain less variation in bone health than nonmodifiable factors, except that low stool pH, a proxy for fermentable fiber intake, did predict higher BMC and BMD.
Elizabeth L Chin, Marta Van Loan, Sarah S Spearman, Ellen L Bonnel, Kevin D Laugero, Charles B Stephensen, Danielle G Lemay, Machine Learning Identifies Stool pH as a Predictor of Bone Mineral Density in Healthy Multiethnic US Adults, The Journal of Nutrition, Volume 151, Issue 11, November 2021, Pages 3379–3390, https://doi.org/10.1093/jn/nxab266.
Images via canva.com.