Hounsfield unit values in patients with diffuse idiopathic skeletal hyperostosis and thoracolumbar injuries - a single-center retrospective study

Abstract

Purpose

This study compared Hounsfield unit (HU) values on computed tomography (CT) scans at fractured sites and each vertebral level of the whole spine among three groups: individuals with diffuse idiopathic skeletal hyperostosis (DISH), young individuals without DISH, and older individuals without DISH.

Methods

A total of 71 patients with thoracolumbar fracture treated at our hospital from 2011 to 2022 were enrolled. Patients were divided into three groups: those with DISH (22 cases), young without DISH (Y non-DISH; age < 50 years, 24 cases), and older without DISH (O-non-DISH; age ≥ 50 years, 25 cases). The total spine HU values from C2 to S1 were measured from CT axial images, as follows: C2-C6, cervical spine; C7-T4, upper thoracic spine; T5-T9, lower thoracic spine; T10-L2, thoracolumbar spine; and L3-S1, lumbar spine. We compared (1) the HU values between fractured and non-fractured sites in each group (2) the HU values by site within each group, and (3) the HU values at each vertebral level of the whole spine between the three groups.

Results

In all groups, the mean HU values were significantly higher in the cervical spine than in the other regions (p < .05) and HU values increase at the fracture site (p < .05). The mean HU values for the lower thoracic, thoracolumbar, and lumbar spine were not significantly different in each group. The mean HU values below T6 were significantly lower in the DISH and O-non-DISH groups than in the Y-non-DISH group (p < .05). The mean HU values, except at T3, did not differ significantly between the DISH and O-non-DISH groups.

Conclusion

Among patients with thoracolumbar injuries, spinal HU values were lower in older patients with and without DISH than in younger patients.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12891-025-09382-5.

Background

Diffuse idiopathic skeletal hyperostosis (DISH) is a condition characterized by the development of bone bridges linking connecting adjacent vertebrae in the interbody region. Reflecting the super-aging society, the number of patients with DISH is increasing worldwide. DISH occurs in 2.9% to 25% of selected populations [1–3]. Vertebral fractures tend to occur with even minor trauma in patients with DISH and often require surgical treatment because they are destabilized due to the concentration of stress at the fracture site [4–7]. Vertebral fractures associated with DISH require long-range fixation, typically spanning three vertebrae above and three vertebrae below the fracture site [8], and are prone to pedicle screw loosening because of bone fragility [3, 6].

Osteoporosis is defined by reduced bone mass, bone structure deterioration, and heightened bone fragility, which contribute to an increased risk of fracture [9]. The gold standard for measuring bone mineral density (BMD) is dual-energy X-ray absorptiometry (DXA), which is useful as a screening test [4]. DXA testing of the lumbar spine, however, is strongly affected by osteophytes and vascular calcification; the BMD of the lumbar spine is reported to be falsely elevated in patients with DISH and in older patients with comorbidities such as vertebral compression fractures and osteoarthritis [4, 10].

Hounsfield unit (HU) values in computed tomography (CT) are the difference in contrast between black and white regions on a CT image and represent a normalized index of X-ray attenuation based on a scale defined as −1000 for air and 0 for water at standard pressure and temperature [11]. Because HUs are a normalized index of X-ray attenuation that eliminates the need for phantoms or calibration to compensate for X-ray tube drift, unlike M-scores from quantitative CT or magnetic resonance imaging [12]. HU values are reported to strongly correlate with BMD based on DXA and are proposed as an alternative method of assessing bone quality [13]. The HU measurements are often obtained at L1 or L1-L4 [13]. Pickhardt et al. [14] reported a sensitivity and specificity for osteoporosis of 75% for HU values of ≤ 135, respectively. The advantage of using HU values is that they are not subject to measurement errors like DXA, but patient exposure to radiation is a disadvantage. Although radiation exposure in CT examination is a crucial problem [4], CT examination is routinely performed in patients with spinal disease and spinal trauma. Few studies, however, have reported HU values of the cervical and thoracic spine and fracture sites. Furthermore, the standard HU values in patients with DISH compared to those without also remain unclear. This study compared the HU values at all vertebral levels among three groups of patients with thoracolumbar injuries: those with DISH, and young and old patients without DISH.

Methods

Participants

This retrospective cohort study was conducted at a single tertiary emergency and critical care center in Japan and utilized medical records of patients with spinal trauma diagnosed by orthopedic spine surgeons. The study protocol was reviewed and approved by the Institutional Review Board of Hirosaki University Hospital (Approval No. 2018–1002) and conducted in accordance with the Declaration of Helsinki and the Ethical Guidelines for Medical and Health Research Involving Human Subjects in Japan. Because the study involved only retrospective analysis of anonymized clinical data and posed minimal risk to participants, the requirement for individual informed consent was waived by the ethics committee. Instead, an opt-out notice was posted on the official website of Hirosaki University Hospital to inform patients of the study objectives and to provide the opportunity to decline participation. Patients were free to refuse participation or request the removal of their data at any time without disadvantage. A total of 261 patients with spinal injury who visited our hospital between 2011 and 2022 were screened, and 71 patients with thoracolumbar fracture (T11–L2) were finally included. Exclusion criteria were: single cervical fracture, thoracic fracture (T1–T10), lumbar fracture (L3–L5), ankylosing spondylitis, age under 18 years, or missing data. Patients were divided into three groups: the DISH group (n = 22), the young non-DISH group (Y-non-DISH; age < 50 years, n = 24), and the older non-DISH group (O-non-DISH; age ≥ 50 years, n = 25) (Fig. 1).

Fig. 1.

Flow diagram in this study. Participants were 71 patients with thoracolumbar fracture (T11 -L2) who visited our hospital from 2011 to 2022; 22 were classified into the DISH group and 49 into the non-DISH group. Among those the non-DISH group, 24 were further classified into the Y-non-DISH group (< 50 years of age) and 25 into the O-Non-DISH group (≥ 50 years of age)

Data collection

Demographic data included sex, age, body mass index (BMI), spinal level of fracture, fracture type based on the AO Spine Thoracolumbar Injury Classification System (AO Classification) [15], and the HU value. The AO classification is based on three main injury patterns: type A (compression), type B (tension band disruption), and type C (displacement/translation) injuries. Type A injuries are further divided into five subtypes and type B injuries are further divided into three subtypes [15].

HU value assessments

All CT scans were taken within two days of injury. The CT acquisition parameters were as follows: tube voltage 120 kVp, tube current 100–300 mA, slice thickness 1 mm, and a bone reconstruction kernel. The HU values of all vertebrae from C2-S1 were measured in axial CT images (GE Health Care, Waukesha, WI, USA) using a picture archiving and communication system (PACS) with an oval region of interest (ROI) that included only the cancellous bone of the vertebral body on the axial image in which the pedicles demonstrated the greatest diameter [11] (Fig. 2). The ROI was carefully placed to include cancellous bone only, avoiding the cortical rim and osteophytes. Instrumented vertebrae and severe imaging artifacts were excluded from analysis. C1 was not measured because it was difficult to measure only cancellous bone. To clarify the difference in the HU values of fractured vertebrae and non-fractured vertebrae, we compared the HU values of fractured and non-fractured areas in the thoracolumbar spine. Furthermore, to clarify differences between regions of the spine, the spine was divided into C2-C6 (cervical spine), C7-T4 (upper thoracic spine), T5-T9 (lower thoracic spine), T10-L2 (thoracolumbar spine), and L3-S1 (lumbar spine). The HU values of fractured vertebrae were excluded from the comparison of total spine HU values because the HU values of fractured vertebrae were higher than those of non-fractured vertebrae.

Fig. 2.

Representative images of axial slices of computed tomography scan. The HU values were measured within an elliptical region of interest on the axial image in which the pedicles demonstrated the greatest diameter. Arrowhead indicates vertebral fracture. CT parameters: 120 kVp, 100–300 mA, 1 mm slice thickness and a bone reconstruction kernel

Two blinded examiners, a spine surgeon (GK, 21 years of experience) and an orthopedic surgeon (AS, 4 years of experience), measured the HU values.

Statistical analysis

Statistical examinations were performed using SPSS version 29 (IBM Corp., Armonk, NY, USA). Values are presented as the mean ± standard deviation (SD) unless otherwise noted, and statistical significance was set at p <.05. Patient background variables and HU values at each vertebral level were compared among the three groups (DISH, Y-non-DISH, and O-non-DISH) using one-way analysis of variance (ANOVA) followed by Tukey’s honestly significant difference (HSD) post hoc test; no further multiple comparison correction was applied. Effect sizes were calculated as η² (eta squared) for ANOVA and Cohen’s d for pairwise comparisons, and mean differences with 95% confidence intervals (95% CI) were reported for major comparisons. The chi-square test was used for categorical variables. For regional comparisons of HU values (cervical, upper thoracic, lower thoracic, thoracolumbar, and lumbar), a two-sample t-test was applied after confirming normality by the Kolmogorov–Smirnov test. The correlation between age and mean HU values from L1 to L4 was analyzed using Spearman’s rank correlation coefficient. HU values of each vertebra were measured independently by two examiners, and interobserver reliability was evaluated using the intraclass correlation coefficient (ICC, two-way random-effects model, absolute agreement) with 95% confidence intervals, as summarized in Supplementary Table S1.

Results

Characteristics of patients with thoracolumbar injury

The mean (SD) age of the Y-non-DISH group was significantly lower than that of the DISH and O-non-DISH groups (76.0 ± 9.3 years in the DISH group, 34.5 ± 9.5 years in the Y-non-DISH group, and 68.6 ± 9.9 years in the O-non-DISH group, [95%CI, 0.688 to 0.838], P <.001). Although the age difference between the DISH group and the O-Non-DISH group was 7.4 years, it was not statistically significant. No significant differences in sex, BMI, or spinal level of fracture were detected among the three groups (Table 1).

Table 1.

Characteristics of patients with thoracolumbar injury

	DISH (n=22)	Y-non-DISH (n=24)	O-non-DISH (n＝25)	P value	Effect size (η²)	95% CI for group differences
Male, n (％)	14 (64%)	13 (52%)	15 (60%)	0.578	-	-
Mean age (years) (SD)	76.0(9.3)	34.5(9.5)	68.6(9.9)	<0.001	0.788	[0.688–0.838]
Min age (years)	56	18	50	-	-	-
Max age (years)	93	49	89	-	-	-
Mean BMI (kg/㎡) (SD)	24.1(5.0)	24.0(5.0)	23.5(3.8)	0.876	0.004	[0.000–0.047]
Fracture level T11, n (%)	10 (45%)	2 (8%)	1 (4%)	-	-	-
T12, n (％)	7 (32%)	7 (29%)	13 (52%)	-	-	-
L1, n (％)	5 (23%)	11 (46%)	11 (44%)	-	-	-
L2, n (％)	5 (23%)	7 (29%)	6 (24%)	-	-	-

One-way analysis of variance (ANOVA) followed by tukey’s honestly significant difference (HSD) post hoc test was performed; no further multiple comparison correction was applied. DISH Diffuse idiopathic skeletal hyperostosis; SD Standard deviation, BMI Body mass index. Effect size (η²) and 95% confidence intervals were calculated using one-way ANOVA

Type B2 was the most common fracture type in all groups based on AO type classification. The DISH group had the highest AO type, with B2 and B3 accounting for > 70% of the fractures, whereas in the non-DISH group, B1 and B2 accounted for more > 70% of the fractures (Table 2).

Table 2.

AO Spine thoracolumbar injury classification system of patients

	DISH (n=22)	Y-non-DISH (n=24)	O-non-DISH (n=25)
Type A2 n (%)	0 (0%)	1 (4%)	0 (0%)
A3, n (%)	2 (9%)	5 (21%)	1 (4%)
A4, n (%)	1 (5%)	0 (0%)	5 (20%)
B1, n (%)	1 (5%)	6 (25%)	6 (24%)
B2, n (%)	13 (59%)	12 (50%)	11 (44%)
B3, n (%)	3 (14%)	0 (0%)	1 (4%)
C, n (%)	2 (9%)	0 (0%)	1 (4%)

DISH Diffuse idiopathic skeletal hyperostosis

Comparison of HU values between fractured and non-fractured areas in the thoracolumbar spine in each group

In the DISH group, HU values were significantly higher in fractured vertebrae (117.9 ± 51.7) than in non-fractured vertebrae (86.1 ± 72.7), with a mean difference of 31.8 HU (95% CI: 6.8–56.8, p =.014, Cohen’s d = 0.47). In the Y-non-DISH group, HU values were significantly higher in fractured vertebrae (198.3 ± 55.9) than in non-fractured vertebrae (163.1 ± 43.3), with a mean difference of 35.2 HU (95% CI: 13.0–57.3, p =.003, Cohen’s d = 0.75). In the O-non-DISH group, HU values were significantly higher in fractured vertebrae (127.0 ± 62.7) than in non-fractured vertebrae (96.0 ± 51.4), with a mean difference of 31.0 HU (95% CI: 5.9–56.1, p =.017, Cohen’s d = 0.57). HU values in fractured vertebrae were significantly higher than HU values in non-fractured vertebrae in all groups (P <.05, Fig. 3).

Fig. 3.

Comparison of HU values between fractured and non-fractured areas in each group. All groups had significantly higher HU values at fractured levels than at non-fractured levels, and significantly higher values when compared overall. Error bars represent 95% confidence intervals. * P <.05, t-test

Comparison of HU values by spinal site for each group

The intraclass correlation coefficients of the HU values were greater than 0.800 and 95% confidence intervals confirming high precision at all vertebral levels from C2 to S1, indicating high interobserver reliability (Supplemental table). Mean (SD) HU values of the cervical, upper thoracic, lower thoracic, thoracolumbar and lumbar spine, respectively, were 276.4 ± 127.6, 194.0 ± 79.4, 129.7 ± 66.4, 87.9 ± 67.1, and 93.1 ± 66.7 in the DISH group; 242.6 ± 95.1, 165.7 ± 60.4, 122.4 ± 49.2, 94.2 ± 51.4, and 84.1 ± 49.5 in the O-non-DISH group; and 295.1 ± 82.1, 206.9 ± 43.4, 175.3 ± 45.0, 163.5 ± 43.3, and 163.0 ± 56.9 in the Y-non-DISH group (Fig. 4 with 95% confidence intervals). In all groups, the HU values were significantly higher for the cervical spine than for the other sites (P <.05, Fig. 4). Furthermore, HU values in all groups were significantly higher for the upper thoracic spine than for the thoracolumbar transition region, and lumbar spine (P <.05, Fig. 4). No significant differences in the HU values were detected among the lower thoracic, thoracolumbar, and lumbar spine in each group (Fig. 4).

Fig. 4.

Comparison of site-specific HU values by site in each group. HU values were significantly higher in the cervical spine than in the other regions in all groups. No significant differences were detected between the lower thoracic, thoracolumbar, and lumbar spine in any of the groups. Error bars represent 95% confidence intervals. * P <.05, ANOVA Tukey test

Comparison of mean HU values for each group

Mean HU values at each vertebral level with 95% confidence intervals are presented in Fig. 5. The HU values in all groups decreased in order from the cervical to the lumbar spine (Fig. 5). The HU values below T6 were significantly higher in the Y-non-DISH group than in the other two groups at all levels (P <.05, Fig. 5). The HU values were significantly higher in the Y-non-DISH group than in the O-non-DISH group at C6, T1, and T5 (P <.05, Fig. 5), whereas they were significantly higher in the DISH group than in the O-non-DISH group only at T3 (P <.05, Fig. 5). Furthermore, to investigate the effect of age on the HU values, we examined the correlation between age and the mean HU values from L1 to L4. The results revealed significant negative correlations between age and the lumbar HU values in all groups (DISH, R = -.462, P =.03; O-Non-DISH, R = -.657, P <.001; Y-Non-DISH, R = -.580, P =.003, Fig. 6).

Fig. 5.

Comparison of mean HU values at each spinal level for each group. The fractured vertebra was excluded. The rhomboid indicates the DISH group, the square indicates the Y-non-DISH group, and the triangle indicates the O-non-DISH group. Dotted line indicates 135 HU, a threshold reported to have a sensitivity and specificity of 75% for diagnosing osteoporosis respectively [14]. Error bars represent 95% confidence intervals. ༊∗Y-non-DISH vs. DISH, O-non-DISH, # Y-non-DISH vs. O-non-DISH, ♭ DISH vs. O-non-DISH, ANOVA Tukey test P <.05

Fig. 6.

Spearman’s correlation between age and mean HU value from L1 to L4 for each group. All groups showed significant negative correlations in all groups

Discussion

The present study elucidated spinal HU values in DISH patients. We demonstrated that the HU values were increased at fracture sites. In all groups, the HU values were higher in the cervical spine compared with all other spinal levels. In the lower thoracic spine, the HU values were significantly lower in the DISH and O-non-DISH groups compared with the Y-non-DISH group.

HU values at fracture sites are higher than HU values at non-fractured sites. This may be because the fracture sites contain cortical bone and other non-cancellous bone. This finding indicates that the HU values at fracture sites do not accurately reflect HU values at other regions of the spine and should not be used to assess bone quality in trauma patients.

HU values were higher in the cervical spine than at all other spinal levels in all groups. Razzouk et al. [7] reported higher HU values in the cervical spine, with a mean (SD) HU of 266.26 (88.69) for C3-C7, 165.57 (55.06) for T8-T12, and 166.45 (51.38) for L1-L4, similar to the results in the present study. This may be due to the fact that the vertebral volume gradually increases from C2, resulting in different pressures per unit volume [16]. One reason for this may be the different anatomic shape of the cervical spine compared with the thoracic and lumbar spine regions [17]. Cervical spine HU values were significantly higher in the Y-non-DISH group than in the O-non-DISH group only at C6, and no significant differences were detected among the other groups. This suggests that HU values in the cervical spine are not suitable for bone quality assessment because no age-related differences are detected.

No significant differences in the HU values were detected between the O-non-DISH and DISH groups, except at T3. These results indicate that HU values, unlike DXA-derived BMD, may serve as a potential surrogate indicator of bone quality in patients with DISH, as HU measurements are obtained from cancellous bone and are less affected by osteophytes, ossification, or vascular calcification. However, because this study did not include direct comparisons with DXA or biomechanical testing, the present findings should be interpreted with caution. HU cannot replace DXA for the diagnosis of osteoporosis, and prospective validation studies are required to confirm its clinical utility. Our findings also imply that bone quality may be relatively low in patients with DISH, in contrast to previous studies reporting higher or comparable BMD values in age-matched controls.

The HU values below T6 were significantly higher in the Y-non-DISH group than in the DISH and O-non-DISH groups. This may be because HU values assess bone quality, which decreases with age, as reported by Schreiber et al., who found that lumbar spine HU values did not differ by sex but correlated with age, decreasing with older age [11]. On the other hand, Razzouk et al. [7] reported that lumbar spine HU values did not differ significantly, even by age. The findings of the present study suggest that lumbar spine HU values correlate with age-related loss of BMD. Pickhardt et al. [14] reported that osteoporosis has a specificity of 91% with an HU value of ≤ 110 in L1. Buckens et al. [18] also reported that 94% of patients with compression fractures had an HU of ≤ 160. In the present study, the DISH and O-non-DISH groups, which may include many osteoporotic patients, had low HU means of 73.5 and 90.7 for L1, respectively.

The mechanism by which DISH patients are prone to fracture is thought to be osteoporosis plus stress concentration due to a long lever arm, which leads to fracture and instability [5]. Our study supports this mechanism. Our findings revealed that bone quality may be low in patients with DISH, in contrast to previous studies, which found that BMD in patients with DISH is either higher than or similar to that in age-matched healthy controls [19–22]. In this study, DXA data were not available for the analyzed cohort, and therefore direct comparison between HU values and DXA-derived BMD could not be performed. However, previous studies have demonstrated significant correlations between vertebral HU values and BMD measured by DXA, supporting the potential of HU as a surrogate indicator of bone quality [13, 14]. Future studies should integrate opportunistic DXA or quantitative CT data to validate the relationship between HU values and bone mineral density in patients with DISH.

This study has the following limitations. First, only one slice in the axial position per vertebra was used to measure the HU. Schreiber et al. [11], however, stated that they found no difference in HU values when using the average HU of one slice and triaxial slices. In a comparison of ROI sizes, Romme et al. [23] found excellent agreement in HU values between ROIs of 50 mm², 100 mm², and 200 mm², with interclass coefficients >0.99 for each comparison. Lee et al. [24] retrospectively compared HU values in the sagittal versus transverse planes and reported that HU measurements in the sagittal and transverse planes were in agreement. Therefore, although only one slice in the axial position was measured in this study as well, we do not think this is problematic. Second, artifacts caused by cortical disruption, hemorrhage, or callus formation at fracture sites may have influenced local HU measurements. To minimize this effect, fractured vertebrae were excluded from the analysis of overall spinal HU values. Third, although this study suggests potential clinical utility of HU assessment, further studies are needed to confirm its relevance for evaluating bone fragility, predicting vertebral fracture risk, and optimizing screw fixation strategies in spinal instrumentation surgery. Fourth, there was no control group of non-fractured cases. Fifth, we did not compare HU values with DXA-derived bone mineral density. Sixth, the precise mechanisms of fracture in patients with DISH remain unclear. Seventh, this was a retrospective, single-center study with a relatively small sample size, and all participants were Asian (Japanese). Therefore, the generalizability of our findings to other populations may be limited. Razzouk et al. [7], however, reported no significant differences in HU values between races. Despite these limitations, the present study provides clinically meaningful insights into vertebral HU distribution patterns in DISH and non-DISH patients and supports the potential role of HU measurement as a convenient surrogate marker of bone quality in routine clinical practice.

Conclusion

HU values decrease in the order of the cervical spine, upper thoracic spine, lower thoracic spine, and thoracolumbar spine, while they increase at the fracture site. Our findings suggest that patients with DISH and older patients without DISH have lower HU values below the lower thoracic spine than younger patients without DISH. HU values other than those of fractured vertebrae may serve as a convenient surrogate indicator of bone quality in patients with DISH, pending further validation using DXA or mechanical testing.

Abbreviations

HU

Hounsfield unit

CT

Computed tomography

DISH

Diffuse idiopathic skeletal hyperostosis

BMD

Bone mineral density

DXA

Dual-energy X-ray absorptiometry

AO Classification

AO Spine Thoracolumbar Injury Classification System

PACS

Picture archiving and communication system

ROI

Region of interest

SD

Standard deviation

Authors’ contributions

Conceived and designed the experiments: AS, GK, KW and YI. Performed the experiments: AS, GK and WK. Analyzed the data: AS. Wrote the paper: AS.

Funding

This research was supported by a Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (23K08603).

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to institutional data protection policies, but are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Review Board of Hirosaki University Hospital (Approval No. 2018-1002). Written informed consent to participate was obtained from all individual participants included in this study, in accordance with the Declaration of Helsinki and relevant national guidelines.

Consent for publication

Not applicable. (No individual identifying personal or clinical information is included in this article.).

Competing interests

The authors declare no competing interests.