Fracture prediction by bone trait dis-integration using DXA among a clinical cohort of adults with cerebral palsy

Multi-trait vs. single trait approaches, such as dis-integration patterns of key size-mass traits, may better capture the heterogeneity of bone strength profiles for skeletally complex populations like adults with cerebral palsy (CP). The objective was to assess if dis-integration of dual-energy x-ray absorptiometry (DXA)-derived bone traits predict fracture incidence among adults with CP.

This was a retrospective cohort study including n=75 adults with CP with a hip DXA from 01/01/2012-03/05/2021 from a single Medical Center; individuals were followed through 9/12/2023 for fracture incidence. Logistic regression estimated the odds ratio (OR) of fracture by the exposure, an interaction between (1) the residual of the BMC-area linear regression and (2) bone area, after adjusting for confounders. Discrimination (c-statistic) was assessed to compare whether the primary exposure or BMD better predicted incident fracture.

Femoral neck BMC-area residual was associated fracture incidence (n=19 fractures) but was conditional on bone area (P-for-interaction, 0.026-0.067). A lower residual was associated with increased OR for smaller areas (e.g., at 10^th percentile of area, OR = 1.18; 95 %CI = 0.96-1.45), but a lower OR for larger areas (e.g., at 90^th percentile of area, OR = 0.88; 95 %CI = 0.77-1.02). The primary exposure had higher discrimination of incident fracture compared to BMD across all unadjusted and adjusted models (c-statistic range 0.69-0.84 vs. 0.49-0.79, respectively).

Dis-integration of key size-mass bone traits was associated with incident fracture and was a stronger predictor of fracture compared to BMD in this clinical cohort of adults with CP.