Opportunistic screening for osteoporosis using abdominal computed tomography scans obtained for other indications

Abstract

Background: Osteoporosis is a prevalent but underdiagnosed condition.

Objective: To evaluate computed tomography (CT)-derived bone mineral density (BMD) assessment compared with dual-energy x-ray absorptiometry (DXA) measures for identifying osteoporosis by using CT scans performed for other clinical indications.

Design: Cross-sectional study.

Setting: Single academic health center.

Patients: 1867 adults undergoing CT and DXA (n = 2067 pairs) within a 6-month period over 10 years.

Measurements: CT-attenuation values (in Hounsfield units [HU]) of trabecular bone between the T12 and L5 vertebral levels, with an emphasis on L1 measures (study test); DXA BMD measures (reference standard). Sagittal CT images assessed for moderate-to-severe vertebral fractures.

Results: CT-attenuation values were significantly lower at all vertebral levels for patients with DXA-defined osteoporosis (P < 0.001). An L1 CT-attenuation threshold of 160 HU or less was 90% sensitive and a threshold of 110 HU was more than 90% specific for distinguishing osteoporosis from osteopenia and normal BMD. Positive predictive values for osteoporosis were 68% or greater at L1 CT-attenuation thresholds less than 100 HU; negative predictive values were 99% at thresholds greater than 200 HU. Among 119 patients with at least 1 moderate-to-severe vertebral fracture, 62 (52.1%) had nonosteoporotic T-scores (DXA false-negative results), and most (97%) had L1 or mean T12 to L5 vertebral attenuation of 145 HU or less. Similar performance was seen at all vertebral levels. Intravenous contrast did not affect CT performance.

Limitation: The potential benefits and costs of using the various CT-attenuation thresholds identified were not formally assessed.

Conclusion: Abdominal CT images obtained for other reasons that include the lumbar spine can be used to identify patients with osteoporosis or normal BMD without additional radiation exposure or cost.

Primary funding source: National Institutes of Health.

Appendix Figure 1. CT-attenuation data relative to DXA T-scores

Mean trabecular CT-attenuation values (SDs) at each vertebral level, stratified by osteoporosis, osteopenia, and normal bone mineral density according to the DXA reference standard. The differences between mean attenuation for each bone mineral density group at each level are significant (P < 0.001). On average, CT attenuation tends to be lowest at the L3 level and increases slightly at higher and lower levels. CT = computed tomography; DXA = dual-energy x-ray absorptiometry.

Appendix Figure 2. PPV and NPV, according to L1 CT-attenuation threshold

The graph shows the PPV for osteoporosis for L1 CT-attenuation values at or below each threshold; the NPV for excluding osteoporosis refers to L1 CT-attenuation values at or above each threshold. The relatively high NPV throughout is driven partly by the lower overall prevalence of osteoporosis (22.9%). CT = computed tomography; NPV = negative predictive value; PPV = positive predictive value.

Appendix Figure 3. Receiver-operating characteristic curves for predicting osteoporosis by using CT values (region of interest method)

The AUCs are similar at each individual vertebral level, using a multilevel T12 to L5 average, and in a multivariable model incorporating measures from all levels simultaneously, with broad overlap of 95% CIs. Osteopenia was considered a false-positive result for these calculations. The total number of assessable CT measurements per level was 2016 for T12, 2040 for L1, 2048 for L2, 2046 for L3, 2021 for L4, and 1943 for L5. AUC = area under the receiver-operating characteristic curve; CT = computed tomography.

Figure 1. Trabecular L1 CT-attenuation values for BMD assessment on body CT scans

Example of axial CT images at the L1 vertebral level in 4 patients (A through D) viewed in standard soft tissue (row 1) and bone (row 2) window settings. Trabecular bone CT-attenuation values are shown in red for each oval region of interest; note that the attenuation measure (in HU) does not change according to the CT window for viewing. The 4 patients represent sample BMDs ranging from low (osteoporosis) (A) to high (normal) (D), which is more visually apparent on the soft tissue window setting (row 1). Assuming a study-derived CT-attenuation threshold for osteoporosis of ≤145 HU (see Results section for details), patient A has osteoporosis by both CT (attenuation value, 20 HU; L2 vertebral fracture [not shown]) and DXA (T-scores for both lumbar spine and hip, −4.0). Patient B has osteoporosis by CT (attenuation value, 93 HU; severe L4 vertebral fracture [not shown]) and osteopenia by DXA (lumbar spine T-score, −2.2; hip T-score, −1.6). Patient C has osteopenia by CT (attenuation value, 148 HU) and DXA (lowest T-score, −1.6). Patient D has normal BMD by CT (attenuation value, 210 HU) and DXA (lowest T-score, 0.1). BMD = bone mineral density; CT = computed tomography; DXA = dual-energy x-ray absorptiometry.

Figure 2. Opportunistic osteoporosis screening at abdominal CT in a 59-year-old woman undergoing colorectal cancer screening (with CT colonography)

CT = computed tomography; DXA = dual-energy x-ray absorptiometry. A. Axial CT image at the L1 vertebral level viewed in a bone window setting shows appropriate placement of the oval region of interest in the trabecular bone. The CT-attenuation value of 109 HU places this patient in the lowest quintile, raising concern for osteoporosis. B. Sagittal CT view shows a moderate T12 compression fracture (arrow). Note that higher thoracic vertebral bodies are also sometimes included on abdominal CT scans. C and D. DXA evaluation of the hips (C) and lumbar spine (D) performed 3 mo later demonstrated osteopenic T-scores ranging from −1.1 to −1.9 (lowest T-score of −1.9 from L1 to L4 evaluation). Therefore, this represents a DXA false-negative result.

Figure 3. Distributions of L1 CT-attenuation values in normal, osteopenic, and osteoporotic cohorts

Based on lowest central DXA T-score. BMD = bone mineral density; CT = computed tomography; DXA = dual-energy x-ray absorptiometry.

Figure 4. L1 CT-attenuation values and corresponding DXA T-scores

The black circles represent patients with CT-detected T12 to L5 compression fractures (not shown here are 15 patients with L1 compression fractures where reliable CT-attenuation measurement at this level was not possible). Note the broad range of DXA T-scores among patients with vertebral fracture, including normal scores. Overall, more than half of all patients with fractures had a nonosteoporotic DXA T-score, but 97% had an L1 or mean vertebral attenuation ≤145 HU. CT = computed tomography; DXA = dual-energy x-ray absorptiometry.

Figure 5. Receiver-operating characteristic curves for predicting osteoporosis by using CT attenuation at L1

Receiver-operating characteristic curves for the L1 vertebral level show no statistically significant difference in AUC for CT scans performed with and without intravenous contrast (P = 0.91). However, further investigation is needed to determine whether any adjustment in specific CT-attenuation thresholds is necessary. AUC = area under the receiveroperating characteristic curve; CT = computed tomography.