Opportunistic Osteoporosis Screening with Cardiac CT: Can We Predict Future Fractures?

See also the article by Therkildsen et al in this issue.

Dr Miriam A. Bredella is professor of radiology at Harvard Medical School and vice chair of the Department of Radiology at the Massachusetts General Hospital. She directs the National Institutes of Health (NIH)-funded Harvard KL2 Program and serves on the NIH Clinical and Translational Science Award Program Steering Committee. She is a NIH-funded investigator on bone and body composition assessment across the weight spectrum, with special focus on the bone-fat connection. She is the deputy editor of the musculoskeletal imaging section for Radiology.

Osteoporosis is a prevalent disease that is characterized by low bone mineral density (BMD) and skeletal fragility, resulting in an increased risk of fractures. Osteoporotic fractures are a major public health threat to millions of Americans and are associated with significant morbidity, mortality, and decreased quality of life. Moreover, the financial burden of osteoporosis is substantial, with annual costs ranging from $17 billion to $20 billion. Because of our aging population, the prevalence of osteoporosis and the number of osteoporotic fractures are expected to increase. However, despite the clinical significance of osteoporotic fractures, the growing size of our elderly population, and the availability of effective treatments, osteoporosis is underdiagnosed and undertreated (1,2).

Screening for osteoporosis is typically performed by using dual-energy x-ray absorptiometry (DXA). DXA is a planar imaging technique that assesses areal BMD and can be subject to artifact from degenerative changes, bowel contents, and arterial calcification. Almost half of patients with osteoporotic fractures have normal BMD at DXA. The majority of women and men who are eligible do not undergo osteoporosis screening, and many osteoporotic fractures occur in individuals who have never been screened (2,3). Therefore, it is of paramount importance to accurately identify individuals with osteoporosis to initiate therapy, with the goal to prevent future fractures. Given the low rates of osteoporosis screening by using DXA, the identification and validation of other techniques might increase the number of individuals undergoing screening.

Quantitative CT, which can be performed with or without a calibration phantom, helps to assess true volumetric BMD and can overcome some of the limitations of DXA. In the United States, more than 80 million CT examinations are performed each year, many of which could be used to screen for osteoporosis without additional costs or radiation exposure. Several studies have validated abdominal or chest CT to perform so-called opportunistic osteoporosis screening (4,5). These studies focused on assessment of the lumbar spine and were able to help predict future osteoporotic fractures (6–8). CT examinations that are suitable for opportunistic osteoporosis screening are preferably performed in individuals at risk for osteoporosis without the administration of intravenous contrast material. One of those studies is CT colonography, for which there is overlap between individuals undergoing colorectal cancer screening and those at risk for low BMD. CT colonography is performed without contrast material. Therefore, several studies have successfully performed opportunistic osteoporosis screening by using CT colonography (5,7).

Cardiac CT is a technique used to screen for coronary artery calcifications and to predict cardiovascular events. Noncontrast-enhanced scans are routinely obtained to estimate coronary artery calcium content, and cardiac CT is typically performed in individuals at risk for osteoporosis (9). Cardiac CT includes parts of the thoracic spine, which could be exploited for osteoporosis screening.

In this issue of Radiology, Therkildsen and colleagues (10) conducted a prospective cohort study in patients undergoing cardiac CT to predict fractures. They studied 1487 participants of the Danish study of the Non-Invasive Testing in Coronary Artery Disease (Dan-NICAD) trial 1 who met indications for coronary CT angiography. Study participants were subsequently followed to determine any incident fracture and incident osteoporosis-related fractures. Cardiac CT was performed by using asynchronous calibration, in which a calibration phantom is imaged between participants and analysis is performed by using validated commercially available software. The origin of the main coronary artery was identified, and the thoracic vertebra at that level and the adjacent two caudal levels were used for analysis. Volumetric BMD (in milligrams per cubic centimeter) was calculated and categorized as normal (>120 mg/cm³), low (80–120 mg/cm³), and very low (<80 mg/cm³) by using recommendations for lumbar spine QCT published by the American College of Radiology.

The application of lumbar spine quantitative CT thresholds to the thoracic spine is innovative but has not been validated. To determine the predictive value of thoracic quantitative CT, the authors followed participants for a median of 3 years to identify incident fractures. National Danish patient registries were accessed to identify any incident fractures and incident osteoporosis-related fractures by using the International Classification of Diseases (10th edition) codes. Moreover, demographics, comorbidities, medication use, exercise status, and tobacco use were recorded and controlled for in the statistical models.

The study cohort was predominately White and included mostly postmenopausal women, with age ranging from 40 to 80 years (mean age, 57 years ± 9 [standard deviation]), a group at high risk for osteoporosis. During a median 3.1-year follow-up, 5.4% of participants were diagnosed with an incident fracture. The estimated annual rate of any fracture per BMD category was 3.4% (95% confidence interval [CI]: 2.1%, 5.4%) for very low BMD, 2.0% (95% CI: 1.4%, 2.7%) for low BMD, and 1.3% (95% CI: 0.9%, 1.9%) for normal BMD. In unadjusted and sex- and age-adjusted analyses, very low BMD was associated with a significantly increased risk of any fracture and very low and low BMD was associated with an increased risk of osteoporosis-related fractures (10).

These results indicate that the thoracic spine CT protocol, with three thoracic vertebral bodies, can be used to assess volumetric BMD and lumbar BMD thresholds can be applied to the thoracic spine and these results can predict incident fractures. However, the application of lumbar BMD thresholds to the thoracic spine might not be optimal for fracture prediction. Therefore, the authors performed receiver operating characteristic curve analyses to determine thoracic spine BMD thresholds to predict incident fractures. A BMD cutoff of 102.6 mg/cm³ to predict any fracture resulted in a sensitivity of 54% (95% CI: 42%, 65%), a specificity of 66% (95% CI: 63%, 68%), and an area under the receiver operating characteristic curve (AUC) of 0.60 (95% CI: 0.54, 0.65). For osteoporosis-related fractures, a BMD cutoff of 96.5 mg/cm³ resulted in a sensitivity of 65% (95% CI: 45%, 81%), a specificity of 73% (95% CI: 71%, 76%), and an AUC of 0.69 (95% CI: 0.60, 0.78). Although the sensitivity, specificity, and AUC were moderate to predict incident fractures, they were higher than when the BMD thresholds of the lumbar spine proposed by the American College of Radiology were used, which considers less than 80 mg/cm³ to be very low BMD. By using this threshold for any fracture, sensitivity was 21% (95% CI: 13%, 32%), specificity was 88% (95% CI: 87%, 90%), and AUC was 0.55 (95% CI: 0.50, 0.59); for osteoporosis-related fractures, sensitivity was 26% (95% CI: 12%, 45%), specificity was 88% (95% CI: 87%, 90%), and AUC was 0.57 (95% CI: 0.49, 0.65).

Therkildsen and colleagues (10) expanded opportunistic osteoporosis screening by including the thoracic spine. Many patients undergo noncontrast CT of the chest for lung cancer screening or coronary artery calcifications and these CT examinations can be used to screen for osteoporosis and assess fracture risk. In addition to the use of thoracic spine CT for fracture prediction, strengths of the study include the large sample size and prospective study design. Use of established and new BMD thresholds to predict fracture risk will have clinical implications and can extend the application beyond research and into clinical practice.

Limitations of the study include the predominately White, postmenopausal female population, which limits the applicability of the proposed thresholds to other populations. Moreover, the relatively small number of fractures makes the study more susceptible to type 2 errors. In addition, the sensitivity, specificity, and AUC for predicting fractures by using established or new thresholds were moderate, suggesting that factors other than BMD contribute to fracture risk. A shortcoming is the use of patient registries to diagnose incident fractures because osteoporotic spine fractures often do not come to clinical attention and may not have been identified. The use of asynchronous calibration requires the use of a calibration phantom between examinations and special software for analysis, which limits wide applicability.

In conclusion, Therkildsen and colleagues (10) expanded opportunistic osteoporosis screening to involve the thoracic spine. Established and new thresholds were able to predict incident fractures. I hope this study will ignite interest in using chest CT examinations performed for other purposes, such as lung cancer screening, for opportunistic osteoporosis screening and prediction of fractures in vulnerable populations.