New research and recommendations were presented at ASBMR 2021, the American Society for Bone and Mineral Research annual meeting, from October 1-4 in San Diego and virtually. The features below highlight key presentations from the conference.

AI Outperforms DXA in Predicting Hip Fracture

Data indicate that automated analysis of plain radiographs using artificial intelligence (AI) could improve availability of diagnostic tests and potentially enhance the overall identification of patients at risk for hip fracture (HF). Researchers used AI methods based on deep convolutional neural networks (DCNNs) to examine digitized pelvic radiographs. They first conducted preprocessing steps, including placement of DXA like region of interest (ROIs) to guide DCNNs, and ROIs were placed automatically based on anatomical landmarks using a key-point detector CNN. AI models included one with additional image preprocessing steps and a second that needed limited supervision in training; the third AI model was based on DenseNet architecture. Dual X-ray absorptionmetry data were used as the reference standard, based on the femoral neck region (aBMD_FN). Among 7,964 women, preprocessing led to a dataset of 6,338 women for training and validation of the DCNNs, with 924 incident HFs over 14.0 years of follow-up. Among 1,252 women in the test dataset, 184 incident HFs occurred during 15.0 years of follow-up. aBMD_FN and all AI predictors were significantly associated with HF incidence in univariate models, as well as models adjusted for age and BMI, while DenseNet based predictors adjusted for age and BMI demonstrated significantly better predictive power than age-adjusted aBMD_FN in the same participants.

FREM Impacts Fracture Risk Regardless of FRAX Probability

Researchers who aimed to determine whether Fracture Risk Evaluation Model (FREM) scores affect 10-year fracture risk regardless of 10-year Fracture Risk Assessment Tool (FRAX) probability computed with BMD identified women and men who received a baseline BMD evaluation between 1996-2018; they determined FREM scores with 10 years of healthcare data and FRAX scores, including BMD, and identified incident major osteoporotic fracture (MOF) and hip fracture (HF) up to 10 years. Moderate correlations were observed between FREM and FRAX, with no significant collinearity in the models. During the follow-up period, incident MOF occurred in 7,957 women and 646 men; incident HF occurred in 2,554 women and 294 men. Increased FREM scores correlated with significantly greater FRAX-adjusted risk for MOF (highest vs middle quintile: HR for women, 1.49; HR for men, 2.06) and HF (highest vs middle quintile: HR for women, 2.15; HR for men, 2.20). The researchers also observed an inconsistent trend toward decreased risk for lower FREM quintiles and found that increased mortality from higher FREM scores affected the examination of FRAX calibration. For the highest FREM quintile, observed results slightly surpassed predicted probability for MOF (ratio, 1.05 in women and 1.49 in men) and HF (ratio, 1.29 in women and 1.34 in men).

Surrogate Threshold Effect Approach Supported for Future Osteoporosis Treatment Trials

Studies have shown that the surrogate threshold effect (STE)—the minimum treatment effect on the surrogate that is reliably predictive of a treatment effect on the clinical outcome—serves as a framework for applying a surrogate endpoint in a clinical trial. Investigators who sought to validate the STE for change in total hip (TH) BMD as a surrogate for fracture risk assessed data from 61,415 participants in 16 randomized controlled trials of bisphosphonates, selective estrogen receptor modulators, denosumab, odanacatib, and teriparatide and performed a random effects meta-regression of the log relative fracture risk reduction in each trial against 24-month change in TH BMD. They also calculated the STE as the intersection of the 95% prediction limits of this regression with the line of no fracture reduction for four categories—all, vertebral, hip, and non-vertebral fractures— and validated the STE method. The STE was 1.83% for all fractures, 1.42% for vertebral fractures, 3.18% for hip fractures, and 2.13% for non-vertebral fractures. In 10 of 11 sufficiently powered fracture outcomes, the anti-fracture efficacy was correctly determined by the STE.

Epidural Steroid Injections Decrease Bone Formation in Postmenopausal Women

Despite the efficacy of epidural steroid injections (ESI) for back pain, associated adverse effects may include bone loss. Patients who often receive ESI are older adults, many of whom have osteoporosis. With a lack of controlled studies examining the acute influence of ESI on bone turnover, investigators examined the acute and long-term consequences of ESI on bone turnover among menopausal women scheduled for an ESI and untreated controls. They collected data on fasting morning blood before ESI (baseline) and at weeks 1, 4, 12, 26, and 52, as well as bone turn over markers and serum cortisol at each timepoint to measure suppression of endogenous glucocorticoids by ESI. Among women who received an ESI, cortisol was reduced by 72% at week 1 and trended back to baseline at week 4. Bone formation was reduced by approximately 27% at week 1 and stayed suppressed at weeks 4 and 12. However, by 6 months, levels had returned to baseline, and bone resorption did not significantly change. No significant change in cortisol, bone formation, or resorption was seen in controls.

Vitamin D Deficiency Impacts COVID-19 Outcomes

Available evidence suggests that vitamin D has a protective role against COVID-19—and that low vitamin D may increase risk for severe COVID-19 infection. To examine whether patients requiring hospitalization for COVID-19 experience increased rates of vitamin D deficiency, and to analyze the relationship between vitamin D and COVID-19 severity, researchers prospectively obtained blood samples (within 48 hours of admission) from patients with COVID-19 severe enough to require hospitalization and measured 25D levels. A control group, matched by age and sex, included outpatients who had blood work done for routine clinical management from 2018-2020. Patients hospitalized for COVID-19 had a greater than 8-fold higher rate of vitamin D deficiency (<25 nmol/L) compared with the general population, a finding that remained for both sexes and both younger and older patients. Several cytokines correlated with 25D levels, with the most notable association for IL-8; in particular, 25D-deficient patients had 5-fold higher levels of IL-8 compared with 25D-sufficient patients. The study team also noted an inverse relationship between clinical COVID-19 severity and vitamin D status among hospitalized patients.