Diagnostic efficiency of chest CT for opportunistic screening for osteoporosis and low BMD in postmenopausal women

Abstract

Early diagnosis and treatment of osteoporosis and low bone mineral density (BMD) may help avoiding vertebral or other fractures and improving the outcomes. While computed tomography (CT) scans are beneficial for diagnosis, the optimal and most convenient method for opportunistic screening is not yet determined. This retrospective study evaluated how well chest CT can diagnose osteoporosis and low BMD at the T11 and T12 levels, using dual-energy X-ray absorptiometry (DXA) as the standard reference in participants who received CT and DXA scans. We included 147 postmenopausal women. Based on the lowest T-score from the DXA test, all patients were classified into normal, osteopenia, and osteoporosis groups. Both T11 HU and T12 HU showed a strong correlation with T-scores (P < 0.01). They are similar in their ability to diagnose osteoporosis (AUC, 0.806 vs. 0.841; P = 0.210) and low BMD (AUC, 0.824 vs. 0.825; P = 0.977), which may serve as a readily accessible screening tool for early diagnosis.

Introduction

Postmenopausal osteoporosis is caused by estrogen deficiency, which leads to increased osteoclast differentiation and accelerated bone resorption beyond bone formation, resulting in bone loss. This results in low bone mineral density, deterioration of bone microstructure, decreased bone strength, and increased risk of fragility fractures. Roughly 20% of women in the United States who are over 50 and 30% who are over 65 meet the diagnostic standards for osteoporosis¹. In China, the prevalence of osteoporosis and vertebral fracture in postmenopausal women is 32.1% and 4%. And it increases with age². Worldwide, 40.4% of women aged 50 years and older suffer from osteopenia³.

Identifying and treating osteoporosis early can significantly decelerate bone resorption, lower the risk of fragility fractures, and decrease the occurrence of complications related to osteoporosis, and reduce the social burden⁴. Despite much work done by clinicians, it is worrying that a significant number of high-risk female fracture patients are not diagnosed before the fracture and do not receive effective treatment⁵. Therefore, it is crucial to conduct early screening and monitoring for timely prevention and treatment of osteoporosis.

Measurement of vertebral bone mineral density is an important indicator recommended by the World Health Organization for the diagnosis and screening of osteoporosis. At present, the DXA method is the gold standard for assessing bone mineral density, and it is also the most commonly used bone densitometry measurement method in clinical and scientific research, which can be used for the diagnosis of osteoporosis, the prediction of fracture risk, and the evaluation of drug efficacy. In addition, there is a quantitative computed tomography (QCT) that provides three-dimensional volumetric bone density⁶. Nonetheless, early detection of osteoporosis is challenging and insufficient. In America, only about one-third of eligible patients are tested for DXA⁷. Solomon’s survey found that in the United States, even after a hip fracture, the percentage of patients treated for osteoporosis fell from 41% in 2005 to 21% in 2011⁸. This is due to the need for additional bone densitometry equipment, insufficient understanding of fragility fractures, and the additional cost of manual operation required to measure bone densitometry⁹.

Opportunistic osteoporosis screening refers to the use of routine CT scan data from chest, abdomen, etc., for other indications to assess volumetric bone mineral density (vBMD) and osteoporosis screening at no additional cost and without additional equipment, time, or radiation exposure^10–12. In addition, early intervention can significantly reduce osteoporotic fractures and mortality^13,14.

Chest CT is the most widely used examination in clinical practice and physical examination, and is mainly used for screening for lung diseases such as tumors and inflammation. CT images offer vertebral HU values that are useful for diagnosing osteoporosis^10,11,15. The majority of earlier research focused on the relationship between HU values and osteoporosis have involved the whole population, rather than the postmenopausal women with the highest risk, the largest number and the highest incidence^10,11. And most of them are concentrated in L1. However, in our experience, not all chest CT scans L1. This study aimed to explore the link between thoracic spine attenuation and osteoporosis and to determine the HU threshold.

Methods

Study design and participants

This retrospective study received approval from the Ethics Committee of Henan Provincial People’s Hospital, and informed consent was obtained from all the participants. Additionally, it was conducted following the ethical principles of the World Medical Association Declaration of Helsinki.

We recruited 325 patients hospitalized in the Department of Spine and Spinal Cord Surgery at Henan Provincial People’s Hospital between January 2024 and December 2024: (1) postmenopausal women; (2) DXA and chest CT are carried out in our hospital at intervals of no more than 2 weeks. Exclusion criteria: (1) Patients with severe heart failure, liver and kidney damage; (2) Patients with conditions impacting bone metabolism including diabetes, multiple myeloma, hyperparathyroidism, etc. (3) Patients who consume alcohol excessively or use medications like glucocorticoids that influence bone metabolism; (4) Previous lumbar internal fixation surgery or vertebroplasty surgery; (5) Previous diagnosis of osteoporosis; (6) T11/12 fracture or surgery. A total of 147 individuals were included in the study (Fig. 1).

Fig. 1.

Flow-chart of patient inclusion and grouping.

Dual-energy X-ray absorptiometry

For measuring bone mineral density, a dual-energy X-ray absorptiometry (Lunar iDXA, GE Healthcare, USA) was used. All patients underwent DXA scans of the lumbar spine(L1-4) and proximal femur. Areas with severe degeneration or significant osteophyte formation that cause markedly increased local bone mineral density are excluded. Typically, fractured lumbar vertebrae exhibit significantly higher bone density and are also excluded from the analysis. BMD results presented as T-scores and g/cm². In accordance with the World Health Organization’s guidelines, participants were categorized into three groups: osteoporosis (T-score ≤ − 2.5), osteopenia (− 2.5 < T-score < − 1.0), or normal (T-score ≥ − 1.0)¹⁶. This categorization was determined based on the lowest T-score.

CT scan acquisition

Patients were hospitalized due to spinal fractures, lumbar disc herniation, lumbar spinal stenosis, or cervical spondylosis and required surgical intervention. To assess potential pulmonary conditions—such as infections or inflammatory diseases—that could influence surgical decisions, chest CT scans were performed. Chest CT examinations were done on a 64-row spiral CT (Somatom Definition Flash, Siemens). The patient was in the supine position, and a CT scan was performed. The CT scanning parameters are as follows: tube voltage: 120 kV, tube current: automatic adjustment, layer thickness: 1 mm.

CT values of the 11th and 12th thoracic vertebrae were measured in the axial position. Initially, the median sagittal position of the CT scan was chosen as the reference point for measurement. Next, create a region of interest (ROI) that is as large as possible in the axial position of the corresponding vertebra, while leaving out the vertebral cortical bone and nutrient foramen. Finally, separate measurements of the HU values for the mesial upper, middle, and mesial inferior endplates were conducted and then averaged.

Focal abnormalities can be identified on CT scans by their characteristically elevated or reduced Hounsfield units, and such cases were excluded from analysis. Cases involving T11/T12 fixation or vertebroplasty were also excluded due to their significant effects on HU measurements.

The identification of vertebral fractures

The subjects of this study include acute vertebral fragility fractures or chronic vertebral fractures. Acute vertebral fragility fractures are defined as: (1) caused by low-energy trauma, such as a fall from standing height or lower; (2) thoraco-low back pain associated with postural changes; (3) obvious tenderness or percussion pain on physical examination. (4) characteristics of acute vertebral fracture found on MRI or CT. A chronic vertebral fracture is calculated when vertebral height is compressed (> 25%) according to the Genant semi-quantitative assessment but there are no typical symptoms, signs, or MRI features of an acute vertebral fracture¹⁷. fractured vertebrae exhibit significantly higher bone density and are excluded from the analysis.

Statistical analysis

All patients were divided into normal, osteopenia, and osteoporosis groups according to the DXA T-score. Continuous variables express as mean ± standard deviation (SD). The receiver operating characteristic (ROC) curve was used to evaluate the efficacy of CT attenuation values in predicting osteopenia or osteoporosis. The area under the ROC curve (AUC), sensitivity, and specificity were calculated. All statistical results were obtained using IBM SPSS Statistics for Windows ver. 26.0 (Armonk, NY, USA).

Results

Patient characteristics

A total of 147 patients were included in this study. Patient characteristics are shown in Table 1. The patients were classified according to the lowest T-score, and were divided into three groups: normal, osteopenia, and osteoporosis. There were 45 patients (30.6%), 56 (38.1%) and 46 (31.2%) in the normal group, osteopenia group and osteoporosis group, respectively, and there were differences in age, BMI, T-score and HU value among the three groups (P < 0.05). The T-scores represent the average BMD values for each lumbar vertebra or proximal femur subregion. While the T-scores of the osteopenia group may appear normal due to confounding factors such as degenerative changes or fractures, etc.

Table 1.

Patient characteristics. Data are presented as the mean ± standard deviation.

Characteristic	Normal	Osteopenia	Osteoporosis
Age(year)	56.5 ± 8.3	63.8 ± 8.6	69.3 ± 9.1
BMI (kg/cm2)	26.6 ± 4.6	27.7 ± 4.5	22.0 ± 4.2
Lumbar vertebra T-scores	0.2 ± 0.4	−0.2 ± 1.2	−2.8 ± 1.2
Femur T-scores	0.1 ± 0.5	−0.9 ± 0.8	−2.6 ± 0.7
T11 Hu	175.9 ± 24.3	117.5 ± 45.1	74.8 ± 35.5
T12 Hu	158.2 ± 25.4	111.2 ± 42.6	69.7 ± 30.3

Correlation between CT Hounsfield units (HU) and T-scores

Both T11 HU and T12 HU showed a strong correlation with T-scores (Table 2). The femur showed a stronger correlation compared to the lumbar spine. The correlation coefficient for T11 HU and lumbar spine T-scores (r = 0.616, P < 0.01) was lower than that for T12 (r = 0.624, P < 0.01), while the correlation between T11 HU and femur T-scores was greater than T12 HU (0.715 vs. 0.708, respectively).

Table 2.

T-Scores and HU-Value person correlation analysis. ^** is P < 0.01.

	Lumbar vertebra T-scores	Femur T-scores
T11HU	0.616**	0.715**
T12HU	0.624**	0.708**

Diagnostic performance for osteoporosis and low BMD

Figure 2 demonstrates the diagnostic efficacy of T11 and T12 for osteoporosis and low BMD (osteoporosis and osteopenia) based on the area under the curve (AUC) of the ROC line. T11 showed similar performance in diagnosing osteoporosis (T11 0.806 vs. T12 0.841; P = 0.210) and low BMD (T11 0.824 vs. T12 0.825; P = 0.977), compared with T12. Based on the lowest T-scores being located at either the femoral neck or total hip, we reclassified the patients and performed ROC analysis. Consistent with prior results, T11 and T12 showed similar diagnostic efficacy for osteoporosis (T11 0.869 vs. T12 0.888; P = 0.544)/low BMD (T11 0.851 vs. T12 0.896; P = 0.367).

Fig. 2.

Receiver operating characteristic curve comparisons for (a, c) osteoporosis and (b, d) low BMD (osteoporosis and osteopenia). Patients were grouped according to the lowest T-scores in the lumbar spine (a, b) or femur (c, d).

We also found the cut-off HU values for T11 and T12 HU in diagnosing osteoporosis and low BMD. For osteoporosis detection, T11 showed a threshold of 106.1 HU (sensitivity 0.924, specificity 0.644), while T12 had a threshold of 92.8 HU (sensitivity 0.797, specificity 0.794). For low BMD detection, T11 demonstrated a threshold of 140.5 HU (sensitivity 0.657, specificity 0.848), compared to T12 with a threshold of 133.9 HU (sensitivity 0.686, specificity 0.843). Following patient reclassification as previously described, we identified distinct HU thresholds: for osteoporosis, 101.4 HU (sensitivity 0.675, specificity 0.878, T11) vs. 82.6 HU (sensitivity 0.822, specificity 0.823, T12); for low BMD, 153.1 HU (sensitivity 0.650, specificity 0.875, T11) vs. 140.2 HU (sensitivity 0.751, specificity 0.874, T12). In summary, T11 and T12 exhibited comparable diagnostic performance in detecting both osteoporosis and low BMD.

Correlation between CT HUs and the number of fractures

94 individuals had no fractures, 13 had fractures in 1 vertebral body, 6 had fractures in 2 vertebral bodies, 14 had fractures in 3 vertebral bodies,12 had fractures in 6 vertebral bodies, 5 had fractures in 7 vertebral bodies, and 3 had fractures in 10 vertebral bodies. In total, there were 204 fractured vertebral bodies, comprising 73 lumbar vertebrae and 131 thoracic vertebrae. The increasing fracture numbers show a significant decrease in Hu values (Fig. 3a). A statistically significant difference (P < 0.01) was found between the no-fracture and single-fracture groups, while an even more pronounced distinction (P < 0.001) emerged when comparing the no-fracture group to the group with > 1 fractures. For patients with more than one fracture, T11 HU values were consistently under 100 (Fig. 3b). For fracture detection, T11 demonstrated significant discriminative capacity (AUC = 0.869) and showed a threshold of 101.7 HU (sensitivity 0.770, specificity 0.800) (Fig. 3c).

Fig. 3.

Correlation between the CT values and the number of fractures. (a) The higher the number of fractures, the lower the CT value which is shown as the mean ± standard deviation. (b) The mean values of the density measurement in HU show a significant difference between patients without and with fracture. **means P < 0.01. ***d means P < 0.001. (c) Receiver operating characteristic analysis of the CT values’ ability to diagnose fracture.

Discussion

The article examines the feasibility of using chest CT data to diagnose osteoporosis and low BMD. Budoff et al. suggested that the T10-T12 vertebral region serves as an effective surrogate for lumbar spine in BMD evaluation due to their high measurement correlation¹⁸. Although a small number of scholars have studied the opportunistic screening effect of chest CT for osteoporosis, most of the population is the whole population. In fact, there is still a big difference between different groups of people. The number of women is large, the incidence of osteoporosis after menopause is increasing, and the proportion of secondary fractures is also increasing¹. Most people only get the first diagnosis of osteoporosis after a fracture and then start treatment. Therefore, early diagnosis of osteoporosis is becoming increasingly important⁵.

We found that bone mass decreases with age, and this is due to the withdrawal of hormones after menopause, which affects women’s bone metabolism¹. At the same time, we saw that patients with osteoporosis had significantly lower mean BMI, which was consistent with previous studies¹⁹. We found that the HU value of T12 was lower overall than that of T11, which was similar to previous studies^10,15. We can also find that the T-scores and HU values of the three groups of patients are correlated, and the smaller the T-score, the lower the HU value. Interestingly, we then studied the relationship between T11/12 attenuation and the T-score of the lumbar spine/femur, and found that the correlation coefficient between the thoracic HU value and the T-score of the femur was larger, but we still chose to use the T-score of the lumbar spine as the criterion for the evaluation of normal bone mass, low bone mass and osteoporosis, because we mainly focused on spine-related diseases such as spinal fractures, in order to enhance the precision of the study’s measurements. Moreover, previous study has shown that the lumbar spine is more sensitive to metabolic changes because it contains more cancellous bone and is therefore important in early osteoporosis detection²⁰.

We focused on the diagnostic performance of thoracic HU values of osteoporosis. Through ROC analysis, we saw no significant difference in AUC between T11/12 for the diagnosis of osteoporosis, suggesting that the diagnostic value of T11/12 was similar. T11 showed a threshold of 106.1 HU (sensitivity0.924, specificity0.644), while T12 had a threshold of 92.87 HU (sensitivity 0.797, specificity 0.794). The results show that the advantages of the HU thresholds of T11 in diagnosing osteoporosis are more due to its very high sensitivity, while T12 is more balanced in terms of sensitivity and specificity. Other study reported that the HU threshold of T12 was 114.5, and 122.5 for T11, which was quite different from the HU value of T12 or T11 in this study¹⁰. This may be due to the specific postmenopausal osteoporosis population in this study.

For low BMD detection, T11 demonstrated a threshold of 140.5 HU (sensitivity 0.657, specificity 0.848), compared to T12 with a threshold of 133.9 HU (sensitivity 0.686, specificity 0.843), which shows that the HU thresholds of T11 and T12 have their respective advantages in the identification of normal and low BMD. After patient reclassification, T12 appeared to marginally outperform T11 in both sensitivity and specificity for detecting osteoporosis and low BMD, which may require further investigation in future studies.

Since the previous results showed that T11 attenuation has a higher predictive sensitivity for osteoporosis, we investigated the relationship between the HU value of T11 and fractures. Obviously, the increasing fracture numbers show a significant decrease in Hu values (Fig. 3). Interestingly, we further evaluated the discriminative value of T11 HU measurements for vertebral fracture identification and found that fractures may occur below a threshold of 101.7 HU. Akane et al. measured The HU values of the anterior 1/3 of the vertebral bodies based on the sagittal plane at T11–L2 in patients with rheumatoid arthritis (RA), pointing out that T11 HU value < 125 was a risk factor for vertebral fracture within 2 years²¹. Graffy et al. found results comparable with ours, which showed significant difference of L1 attenuation value among patients having different numbers of vertebral fractures, suggested that L1 attenuation value ≤ 90 HU may be the optimal threshold for determining osteoporotic vertebral fracture²². Park et al. found fracture-free survival was significantly lower in breast cancer patients with L1 attenuation value ≤ 90 HU on CT and T-scores ≤ −2.5 on DXA, and pointed out CT can be used for predicting osteoporosis and discriminating incidental fracture risk²³. Most of the subjects studied were lumbar spine. However, the relationship between CT values and osteoporosis and fracture risk is similar. Therefore, the HU value of T11 may be of great significance in predicting the occurrence of fractures.

Some limitations should be mentioned. First of all, the sample size is small. Secondly, in order to measure the HU value of T11/12 to reduce the experimental error, we excluded patients with T11/12 fractures, which may have an impact on the results. Thirdly, due to factor of manual measurement, it may be necessary to combine artificial intelligence to better solve the problem.

Conclusion

We show the feasibility of identifying patients with osteoporosis and low BMD, using chest CT obtained for other conditions. Vertebral CT attenuation analysis could offer a practical screening modality that can be seamlessly integrated into clinical workflows to proactively detect patients susceptible to future osseous abnormalities, especially to prevent fractures.

Author contributions

Zhiqiang Hou, Yanzheng Gao and Guangquan Zhang performed study conception and design, Zhiqiang Hou, and Fuqiang Shao performed data analysis of the CT-scans, Zhiqiang Hou, Shuai Ding, and Guangquan Zhang contributed to data interpretation. All authors contributed to the manuscript and all authors approved the final manuscript.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.