5 Key Takeaways

• 1

  Across UK Biobank, NHANES, and Tianjin cohorts, participants with AMD were older and had lower BMD than those without AMD.

• 2

  Higher BMD was associated with lower AMD incidence in adjusted analyses, with a UK Biobank hazard ratio of 0.89 and similar inverse associations in the other cohorts.

• 3

  Age and BMD were repeatedly important features in machine-learning models, but their added predictive value is uncertain and the models lack external validation.

• 4

  Mendelian randomization supported a modest contribution of higher BMD to lower AMD risk and identified GZMA, NELL1, and COL2A1 as potential intermediates.

• 5

  Rats with glucocorticoid-induced low BMD showed outer retinal thinning, narrower retinal vessels, and delayed visual-spatial performance, but no hallmark features of human AMD.