Diagnostic accuracy of knee CT Hounsfield units for osteopenia and osteoporosis

Diagnostic accuracy of knee CT Hounsfield units (HUs) for bone mineral density remains unclear. This study assessed diagnostic accuracy of regional knee CT HU for osteopenia and osteoporosis in a consecutive institutional cohort, while controlling for CT energy.

A search was conducted at the study institution for all patients aged ≥ 50 years with non-contrast-enhanced phantomless knee CT acquired at 120 kilovoltage and dual-energy X-ray absorptiometry (DXA), including hip and lumbar spine, performed within a one-year interval, with imaging performed between 1 January 2010 and 15 October 2023. Data were retrospectively collected. Individuals were excluded with ipsilateral fractures, pathological lesions, or previous surgery of the femur, tibia, fibula, or patella. Hip and lumbar spine DXA T-scores were collected. Mean trabecular bone HU were assessed on three consecutive axial knee CT slices for nine knee regions within the patella, fibula, femur, and tibia, including metaphyseal and epiphyseal, and medial and lateral epiphyseal subregions. Diagnostic accuracy was assessed for each region.

A total of 126 patients met the eligibility criteria, with a mean age 68.7 years (51 to 89; SD 9.1), including 104 females (82.5%) and 22 males (17.5%). On central DXA (hips and lumbar spine), 15.1% of patients had normal bone mineral density (n = 19), 46.8% had osteopenia (n = 59), and 38.1% had osteoporosis (n = 48). Good diagnostic accuracy, indicated by an area under the curve (AUC) of 0.8 to 0.9, was found for seven of nine knee CT HU regions of interest versus central DXA T-scores for impaired bone density and osteoporosis (p < 0.001; e.g. proximal tibia epiphysis AUC = 0.860; p < 0.001, and AUC = 0.842; p < 0.001, respectively).

Knee CT HU demonstrated good diagnostic accuracy for bone mineral density, and therefore opportunistic screening with knee CT can extend the diagnostic value of CT obtained for other clinical purposes. This technique can be used to identify orthopaedic patients, including those undergoing robotic-assisted hip and knee arthroplasty, who could additionally benefit from further assessment and potential osteoporosis care, for optimization of orthopaedic outcomes.