This cross sectional study compared the validity of BMD/OSTA/FRAX (without BMD) and BFH-OSTM for identifying POVCF in Chinese men aged 50 and over. According to diagnostic criteria of POVCF issued by our criteria, these tools are suitable for comparison between healthy men and patients with PNOVCF.
Our study showed that for BMD, the AUC values for assessing the risk of PNOVCF of the femoral neck, hip and lumbar vertebral were 0.779, 0.776 and 0.650, respectively, and the corresponding optimal critical values were −1.4, −1.4 and −0.7. This indicates that BMD measurements of the femoral neck, hip and lumbar vertebral are only moderately predictive. Among the three parts, the identifiability of the lumbar spine was higher, and its sensitivity was 79.53%. As a screening method, the specificity of total hip was 77.33%. However, our results show that the sensitivity of BMD in assessing the risk of fracture is moderate, with a sensitivity of 71.93% in the femoral neck, 64.91% in the total hip and 79.53% in the lumbar spine, respectively, and the specificity in the lumbar spine is only 44.3%, so the specificity is less acceptable in the lumbar spine. Our previous research showed that the AUC for estimating the risk of fracture at the femoral neck, hip, and lumbar spine were 0.706, 0.711, and 0.706, respectively, with optimal cutoffs of −2.5, −1.4, and −1.6, with had a sensitivity of 42.34%, 67.57%, 52.25% and specificity of 89.87%, 65.45%, 77.14%. Due to the high cost of central dual-energy X-ray absorptiometry, BMD is not suitable as a preliminary screening tool in primary hospitals. In the fracture group, 17.5%, 23.4% and 38.0% of the patients had normal BMD at the three parts of femoral neck, total hip and lumbar spine respectively, so the value of bone mineral density as a predictor of PNOVCF was limited. Therefore, we urgently need a screening tool with higher accuracy and simplicity than bone mineral density measurement to identify PNOVCF. Because there is no such equipment in primary hospitals, this tool should not be used as a screening tool.
As showed in Table 1, the average weight, height and BMI of the Fracture group were lower than those of the non-Fracture group. Patients in the Fracture group experienced more fragility fractures than the non-Fracture group, so we think that low weight and previous fragility fracture history of osteoporosis are also risk factors for PNOVCF. This is consistent with the traditional clinical view [27]. If the height is lower than that of the general healthy people, it may be due to the physiological characteristics of the spine, the vertebral body of the osteoporosis patient is more likely to be compressed, and the morphological changes of the vertebral body and intervertebral space lead to the shortening of the length of the spine [28]. The calculation of OSTA is very simple, based only on the two influencing factors of age and body weight, which is simpler than BMD measurement, and is suitable for the risk assessment of osteoporosis in postmenopausal Asian women. In recently published reports, data show that the OSTA index can also be used to predict the risk of osteoporosis in elderly Chinese men, but the prediction of new osteoporotic vertebral compression fractures in this population has not been confirmed [29]. Our previous and current study shows that there is a significant difference in the distribution of OSTA score between the Fracture group and the non-Fracture group. Its ability to recognize OVCF (AUC = 0.752) is slightly lower than that of the femoral neck and hip, but better than that of the lumbar spine. However, the disadvantage that cannot be ignored is low sensitivity (50.88%). For screening tools, we focus more on high sensitivity than on high specificity, as fewer patients will undergo unnecessary treatment or invasive tests [30]. so the OSTA index may not be applicable to the prediction of PNOVCF in Chinese elderly men, which run counter to the purpose of our screening.
The FRAX algorithm was developed to assess the risk of osteoporotic fractures of the hip, spine, distal forearm and shoulder in 10 years and has been recommended by the World Health Organization. Because it does not need BMD measurement, the patient data collection is more comprehensive, so its ability to distinguish PNOVCF is indeed stronger than BMD measurement and OSTA score [23]. The AUC value of FRAX in the diagnosis of OVCF risk was 0.796, and its sensitivity and specificity were 74.85% and 78.52% respectively at the optimal critical value. Among the tools are tested in the present study, FRAX had a higher discriminating ability for identifying PNOVCF, followed by OSTA and BMD. In clinical practice, FRAX needs to include many risk factors and be equipped with corresponding hardware and software, so it has certain limitations [26, 31].
Our BFH-OSTM is a calculation model based on multiple regression analysis of data from multiple centers, and two risk factors, body weight and previous fragility fracture history are selected. Compared with FRAX, BFH-OSTM model may be an effective tool to determine the risk of PNOVCF in this elderly Chinese men population. The performance value of BFH-OSTA is better than FRAX (p < 0.05). The formula is (body weight [kg] − history of previous fracture [no = 0, yes = 1] × 7). BFH-OSTM can not only predict osteoporosis, but also can be used for early detection of PNOVCF, and the value of cutoff may be different. The optimal cutoff value for identifying PNOVCF is 69, and the sensitivity and specificity are 81.29% and 70.67%, respectively, and the area under the curve AUC is 0.841. The ability to identify PNOVCF is significantly improved, and the sensitivity is also higher (Fig. 5). Compared with FRAX, the predictive value of our BFH-OSTM is obviously better than that of FRAX. We think that the cutoff value may be slightly different in different people and regions, which need to be further confirmed.
Among the tools tested in this study, BFH-OSTM had the highest discriminative validity in identifying PNOVCF in the elderly men population, with approximate high identification, followed by FRAX, with better BMD in the femoral neck and hip than OSTA. In this study, compared with OSTA and BMD T scores, FRAX score (without BMD) includes more related risk or protective factors and adapts to local conditions, so it has higher identification value, but its universality also limits its promotion and application. Despite the availability of computer software to simplify the calculation. However, the complexity of data collection also makes clinical evaluation difficult, which is not suitable for large-scale screening and community application. Our BFH-OSTM model is a multi-factor analysis model, which accurately captures the two influencing factors of the most critical body weight and previous brittle fracture history, which is simple enough, but at the same time obtains the best predictive value of PNOVCF and ensures the sensitivity and specificity of screening, so it is easy to popularize and apply [32].
Our research presents several noteworthy advantages. First of all, our research is a cross-sectional study, so the information obtained is not retrospective. Second, due to the rigor of data collection, our study shows that the age and weight of subjects are recorded while measuring bone mineral density, and all diagnoses and results are made by experienced physicians. Third, we imposed strict inclusion and exclusion criteria to exclude the effects of other factors. Finally, our research has important clinical significance; it can help inexperienced doctors in primary hospitals or community health service centers to detect PNOVCF as early as possible. More importantly, there is no need for the learning curve, compared with the traditionally recognized ability of FRAX to assess the risk of fracture, BFH-OSTM is a more simple, direct and effective model for clinicians.
However, current research still has some limitations. First of all, we only collected the subjects recruited from a hospital, which is a single-center study, so we cannot fully represent the entire demographic data of the people's Republic of China. Secondly, we suggest that more centers should participate to improve big data and verify the accuracy of the model at the same time.