PURPOSE: To estimate the associations of prevalent vertebral fracture (PVFx) and abdominal aortic calcification (AAC) with incident ASCVD (myocardial infarction, fatal or non-fatal cerebrovascular accident, or coronary heart disease death).
METHODS: 2799 older men (mean [SD] age 76.3 [5.5] years) enrolled in the MrOS sleep ancillary study had PVFx (SQ grade 2 or 3) assessed by human reader and AAC by automated convolutional neural networks on baseline lateral spine radiographs. Auto-AAC was categorized as low, moderate, or high (24-point scale score < 2, 2 to <6, or  ≥  6). Men were contacted every 4 months for ascertainment of possible ASCVD events over a mean (SD) follow-up of 8.4 (3.1) years. Associations of PVFx and auto-AAC category with incident ASCVD were estimated with modified proportional hazards models accounting for non-CVD mortality.
RESULTS: PVFx was present in 7.3% of the cohort; 34.5% had moderate auto-AAC, 28.8% had high auto-AAC, and 396 (14.1%) had an incident ASCVD event. Compared to men with low auto-AAC, those with moderate (HR 1.34, 95% CI 1.01, 1.77) and high (HR 1.55, 95% CI 1.16, 2.06) auto-AAC had increased risk of ASCVD events adjusted for PVFx and other risk factors. Compared to men with no PVFx, those with PVFx had a higher risk of ASCVD events (HR 1.51, 95% CI 1.07, 2.14) adjusted for auto-AAC level and other risk factors.
CONCLUSION: AAC assessed by automated methods and PVFx are independently associated with incident ASCVD events and may aid in ASCVD risk stratification in older men, but confirmatory studies are needed.