Following fracture, BMD is often assessed by DXA, but cost, availability and time constraints restrict wider screening. DXR evaluation is rapid and can be performed on unmodified digital mammography equipment. This study assesses the correlation of BMD by DXR and DXA, and the potential for DXR screening of women undergoing follow up mammography. Participants were consenting post-menopausal women over age 50. After mammography, an X-ray of their non-dominant hand using an imaging preset was de-identified and transmitted for DXR analysis, and they underwent DXA evaluation at radial, lumbar and proximal femur sites. The mean age of 45 participants was 64±6 years and 77% had been diagnosed with breast cancer. Their BMI was 28.9±5.8, calcium supplements were taken by 42%, vitamin D by 63% and type 2 diabetes or osteoporosis were each reported by 17%. Prior fractures were reported by 33% with 10% falling in the previous year. BMD by DXR and DXA correlated significantly at the 1/3, total and ultradistal radius (r=0.63, r=0.67, r=0.63 respectively; all p<0.001), total hip and femoral neck (r=0.38 for both; P=0.007 and 0.008 respectively) and lumbar spine L1-4 (r=0.45; p=0.001). DXR BMD correlated to forearm BMD more closely than hip or spine BMD by DXA. Neither DXR nor DXA BMD correlated significantly to trabecular bone scores. Predictably, FRAX and Garvan fracture risk results were similar when DXR T-scores were substituted for DXA T-scores (Garvan 5 and 10 year hip fracture risk; r=0.74 for both, p<0.001; major osteoporotic fracture; r=0.76 and 0.75, p<0.001 for both). The area under the ROC curve for the 1/3 radius BMD by DXR and DXA was 0.84 (95% CI; 0.72-0.96), with DXR T-scores <-1.5 being 85% sensitive and 82% specific for predicting osteopenia/osteoporosis by DXA. Targeted DXR screening identifies women at risk of low BMD by DXA, particularly at forearm sites, and may assist in fracture reduction strategies by identifying patients in need of further evaluation.
Disclosure: The authors declared no competing interests. This work was supported by: 1. Sydney West Translational Cancer Research Centre Grant funded by the Cancer institute of NSW; 2. Sectra OneScreen Australia and Sweden for DXR reports.