A meta-analysis of the diagnostic accuracy of Hounsfield units on computed topography relative to dual-energy X-ray absorptiometry for the diagnosis of osteoporosis in the spine surgery population

Abstract

Background: The preoperative identification of osteoporosis in the spine surgery population is of crucial importance. Limitations associated with dual-energy x-ray absorptiometry, such as access and reliability, have prompted the search for alternative methods to diagnose osteoporosis. The Hounsfield Unit(HU), a readily available measure on computed tomography, has garnered considerable attention in recent years as a potential diagnostic tool for reduced bone mineral density. However, the optimal threshold settings for diagnosing osteoporosis have yet to be determined.

Methods: We selected studies that included comparison of the HU(index test) with dual-energy x-ray absorptiometry evaluation(reference test). Data quality was assessed using the standardised QUADAS-2 criteria. Studies were characterised into 3 categories, based on the threshold of the index test used with the goal of obtaining a high sensitivity, high specificity or balanced sensitivity-specificity test.

Results: 9 studies were eligible for meta-analysis. In the high specificity group, the pooled sensitivity was 0.652 (95% CI 0.526 - 0.760), specificity 0.795 (95% CI 0.711 - 0.859) and diagnostic odds ratio was 6.652 (95% CI 4.367 - 10.133). In the high sensitivity group, the overall pooled sensitivity was 0.912 (95% CI 0.718 - 0.977), specificity was 0.67 (0.57 - 0.75) and diagnostic odds ratio was 19.424 (5.446 - 69.275). In the balanced sensitivity-specificity group, the overall pooled sensitivity was 0.625 (95% CI 0.504 - 0.732), specificity was 0.914 (0.823 - 0.960) and diagnostic odds ratio was 14.880 (7.521 - 29.440). Considerable heterogeneity existed throughout the analysis.

Conclusion: In conclusion, the HU is a clinically useful tool to aide in the diagnosis of osteoporosis. However, the heterogeneity seen in this study warrants caution in the interpretation of results. We have demonstrated the impact of differing HU threshold values on the diagnostic ability of this test. We would propose a threshold of 135 HU to diagnose OP. Future work would investigate the optimal HU cut-off to differentiate normal from low bone mineral density.