Measurement of Distal Ulnar Hounsfield Units Accurately Predicts Bone Mineral Density of the Forearm

Wagner, Scott C. MD1,2; Dworak, Theodora C. MD1; Grimm, Patrick D. MD1; Balazs, George C. MD1; Tintle, Scott M. MD1,2,3,a

Journal of Bone & Joint Surgery - American Volume: 19 April 2017 - Volume 99 - Issue 8 - p e38
doi: 10.2106/JBJS.15.01244
Scientific Articles
Disclosures

Background: Hounsfield unit (HU) measurement obtained from computed tomography (CT) scans of the wrist is a potential new screening method for low bone mineral density (BMD). We hypothesized that HU measurements of the ulnar head obtained from CT scans would correlate with BMD assessed with dual x-ray absorptiometry (DXA) scans of the forearm.

Methods: Patients with both upper-extremity CT and DXA scans performed at a single institution were included in the study. Hounsfield units were manually measured in the distal part of the ulna by 1 author blinded to the DXA results. Average values were then compared with forearm BMD values as determined with a DXA scan.

Results: Seventy-seven CT scans of 74 patients were included. Average HU values were significantly lower in the osteoporotic and osteopenic groups in comparison with the normal BMD group. The upper limit of the 95% confidence interval for osteopenic patients was 145.9 HU. The average forearm T-score for patients with an HU value at or below the cutoff of 146 was significantly lower than the average T-score for those with an HU value of >146 HU (p < 0.0001). Sensitivity and negative predictive value for low BMD using this cutoff value were calculated to be 91% and 89%, respectively.

Conclusions: Distal ulnar HU measurements accurately reflect the BMD of the forearm as diagnosed with a DXA scan. Our results suggest that distal ulnar HU measurements of ≤146 HU are strongly associated with low BMD and that values above this cutoff accurately rule out low forearm BMD with a high degree of sensitivity and negative predictive value. Utilizing this technique may improve the capture of at-risk patients and streamline the screening process for osteoporosis.

Level of Evidence: Diagnostic Level III. See Instructions for Authors for a complete description of levels of evidence.

1Department of Orthopaedics, Walter Reed National Military Medical Center, Bethesda, Maryland

2Department of Surgery, Uniformed Services University of the Health Sciences, Bethesda, Maryland

3Department of Orthopaedic Surgery, University of Pennsylvania Health System, Philadelphia, Pennsylvania

E-mail address for S.M. Tintle: scott.tintle@gmail.com

Article Outline

It is estimated that, in 2010, 53.6 million elderly adults in the United States met the criteria for having osteoporosis or low bone mass1. The U.S. Census Bureau estimates that, by 2050, the population of individuals 65 years and older will rise to 83.7 million2, and the current health-care costs associated with osteoporosis will only continue to increase3. Fragility fractures are often initially treated by orthopaedic surgeons, but the process of screening for and managing overall bone health is typically deferred to primary care physicians and is often inadequately performed4. Some researchers have estimated that 79% of older patients sustaining a distal radial fracture are at moderate risk for sustaining a subsequent fragility fracture within 10 years5; however, only 8.7% of these patients are even screened for osteoporosis at the time of injury, suggesting that neither complete workup nor appropriate medical treatment is performed for a vast majority of high-risk patients6. Thus, there has been a recent push in the orthopaedic literature promoting active screening on the part of the managing surgeon to assess bone health and, at a minimum, initiate appropriate referrals3.

A recent study by Schreiber et al.7 described “opportunistic osteoporosis screening” utilizing diagnostic computed tomography (CT) of the wrist to quantify bone mineral density (BMD) of the distal part of the radius after a fracture. As a technique, Hounsfield unit (HU) measurement on wrist CT scans has the potential advantage of providing a simple tool to evaluate BMD without the need for additional imaging8, as CT scans of the wrist may be obtained for a variety of diagnostic purposes outside the realm of distal radial fractures. Despite the widespread use and dependence on HU measurements in pivotal medical decisions on whether to include an evaluation of coronary arteries and whether a renal mass necessitates biopsy, orthopaedic use of HU measurement remains limited. No correlation between scores from wrist CT and dual x-ray absorptiometry (DXA) scans has been reported in the literature, to our knowledge. We hypothesized that the HU measurements obtained from wrist CT scans would accurately correlate with DXA scan T-scores and BMD of the forearm. Therefore, we set out to measure HU values of the distal part of the ulna, utilizing CT scans of the wrist obtained for a variety of indications, and to elucidate any potential association with forearm BMD as measured on DXA scans.

Back to Top | Article Outline

Materials and Methods

Following institutional review board approval, the Military Health System Management Analysis and Reporting Tool (M2) database was utilized for this study. The database includes medical data on active-duty service members, military retirees, and their dependents. The database was queried for Current Procedural Terminology (CPT) codes for patients with both DXA and CT scans of the wrist performed at a single institution between 2009 and 2015. An additional list of patients was generated using the PACS (picture archiving and communication system) electronic radiograph database (IMPAX software, version 6.4; Agfa Healthcare) at our institution, with a general search query for all wrist CT scans obtained between 2002 and 2009. Patients without DXA scans were excluded. Electronic medical records and final DXA scan reports were evaluated and recorded.

All CT scans of the wrist were performed with either a Philips Brilliance 64-slice machine (Royal Philips Healthcare) or a GE Discovery CT750 HD 64-slice machine (GE Healthcare) at 1 institution. The typical setting for peak kilovoltage (kVp) for wrist CT on our scanners is 120. The exposure software determines appropriate milliampere-seconds (mAs) on the basis of the size of the imaged body part and the density of the surrounding tissue; given the relative lack of soft tissue at the level of the distal part of the wrist and the small size of the region of interest, the typical value for CT of the wrist is 99 mAs at our institution. Patients whose CT scans did not include the distal ulnar head were excluded from the study. For our study, patients with distal radial fractures often had fracture fragments obscuring large portions of the distal radial metaphysis, which in turn made measurement of the HU values inconsistent. Therefore, we standardized our HU measurement to the distal ulnar head only, which was reproducible and simple to perform, as the cancellous bone of the distal part of the ulna was rarely fractured or deformed. HU values were measured by 1 author on 3 sequential coronal CT slices of the distal part of the ulna at the level of the distal radioulnar joint (DRUJ) to minimize variations in measurement (Figs. 1-A, 1-B, and 1-C). We performed the measurements using a technique similar to that described by Schreiber et al.7,8, in which cortical bone is avoided where seen by the unaided eye and an autogenerated circle is used to capture the region of interest (ROI). All measurements excluded cortical bone and were limited to the cancellous region, as per previously reported methodologies optimized for assessment of the lumbar spine8 and wrist7. The author performing the measurements was blinded to any DXA information during HU measurement, and averages were recorded for each patient.

Figure. No caption a...
Figure. No caption a...
Figure. No caption a...

Statistical analysis involved a 2-tailed t-test to compare averages between groups and determine any significant differences, and the chi-square test was used for categorical value comparisons. A 95% confidence interval (CI) for HU values for each group (osteoporotic, osteopenic, and normal BMD) as well as for each individual patient was also calculated. Sensitivity and specificity were calculated manually with use of our experimentally determined HU cutoff value, which was obtained by comparing average HU measurements with DXA forearm T-scores.

Back to Top | Article Outline

Results

A total of 77 CT scans of 74 patients with forearm DXA information were included. Fifty-six (75.7%) of the 74 patients were female, with only 18 males included. The average patient age was 57.4 years. Five patients (6.8%) had forearm DXA T-scores consistent with osteoporosis of the forearm (less than −2.5). Twenty-seven patients (36.5%) had forearm T-scores between −2.4 and −1.1, which is consistent with osteopenia. Forty-two patients (56.8%) had normal forearm BMD, which is defined as a forearm T-score greater than −1.0. The average HU measurement (and standard deviation) was significantly lower in the osteoporotic group compared with the group within normal limits: 98.1 ± 22.7 versus 198.6 ± 18.9 (p < 0.0001). The average HU value for the osteopenic group was also significantly lower than that for the within-normal-limits group: 126.9 ± 19.0 versus 198.6 ± 18.9 (p < 0.0001). When considering the low-BMD group as a whole, the average HU value was 122.5 ± 16.7 versus 198.6 ± 18.9 (p < 0.0001) (Table I). The upper limit of the 95% CI for the average HU measurement in osteopenic patients was 145.9.

For each individual patient, the lower limit of the 95% CI based on the 3 HU measurements was generated to yield the lowest likely true HU value for that patient. Based on these measurements, 49 patients (66.2%) had a minimum HU value of ≤146, which is consistent with low BMD. The average forearm T-score for patients with values at or below the 146-HU cutoff was significantly lower than for those with values above the cutoff (−1.33 versus +0.2; p < 0.0001), suggesting that average forearm BMD in patients with HU values at or below our cutoff is consistent with, at a minimum, osteopenia. Sensitivity and negative predictive value for low BMD were calculated to be 91% and 89%, respectively, when utilizing the cutoff threshold of 146 HU. These results are summarized in Table II.

Patients were then analyzed by sex. The average HU value was significantly lower in female patients than in male patients (153.2 versus 207.8; p = 0.02). Four female patients and 1 male patient had a forearm T-score of less than −2.5, while 22 females and 6 males were osteopenic. The remaining patients were within normal limits for BMD on DXA scans. For females, the average HU value for osteoporotic patients was significantly lower in comparison with that for females with normal BMD (103.6 ± 25.7 versus 180.3 ± 19.4; p = 0.002). The same was true for osteopenic patients (121.6 ± 19.0 versus 180.3 ± 19.4 for patients with normal BMD; p < 0.0001) and the low-BMD group as a whole (118.8 ± 16.6; p < 0.0001). The 18 male patients were analyzed separately. Only 1 male patient had a forearm T-score below −2.5, so all males with T-scores below −1.0 were considered in an aggregated low-BMD group. The average HU value in the male low-BMD group was 136.3 ± 51.5, which was significantly lower than the average HU value in the male within-normal-limits group (253.4 ± 31.5; p = 0.003). The upper limit of the 95% CI for male patients with low BMD was 188.0 HU, and the upper threshold for osteopenia in females was 141 HU.

Fourteen (18.9%) of the 74 patients had CT scans for distal radial fractures. We were unable to show a significant difference between the fracture and no-fracture groups: the average distal ulnar HU value was 149.9 in the fracture group and 169.6 in the no-fracture group (p = 0.33). The lower limit of the 95% CI for the fracture group (109.6 HU) fell below our cutoff, while the lower limit for the no-fracture group (153.0 HU) did not.

Back to Top | Article Outline

Discussion

A majority of at-risk patients are not appropriately evaluated for low BMD7-9. Even after sustaining a fragility fracture of the distal part of the radius, most patients are not screened for osteoporosis and <25% are managed appropriately4,9. A simple mechanism by which to identify at-risk patients and initiate workup would streamline the process, improve capture of these undiagnosed patients, and potentially prevent future fragility fractures7. Schreiber et al.7 compared the HU measurements of the distal part of the ulna, capitate, and radial fracture fragments in patients with distal radial fractures with the HU measurements of matched controls, and found decreased average HU values in the fracture group. We limited our study to only 2 CT scanners, both used at our institution and with the same kVp settings, to minimize any minute variations that may exist.

Currently, DXA is the gold standard for the diagnosis of osteoporosis, and it is the standard against which any novel BMD assessment tool should be compared7,10. To our knowledge, correlation of wrist CT scan measurements of HU values and forearm DXA measurements has not been reported in the literature. CT scans offer volumetric estimation of BMD, in contradistinction to simple planar values that are reported with DXA scans, which may be a potential advantage of CT measurement of BMD in comparison with DXA measurement7. In fact, there has been some suggestion that CT may be more accurate for estimation of BMD than DXA is9, but this hypothesis has not yet been definitively proven. While clinical CT scanning does not provide the same 3-dimensional information that is provided with high-resolution quantitative CT, volume-averaging effects can be minimized with thin-section scanning; further, because the surrounding soft tissue of the wrist is relatively minimal, obfuscation of the bone imaging is also relatively inconsequential8.

We found that measurement of the HU value of the distal ulnar head on CT scans of the wrist accurately corresponds to forearm BMD and T-scores (Figs. 2 and 3), similar to previously reported findings in the lumbar spine8. Schreiber et al.7 previously reported an interrater reliability of 0.947, indicating high reproducibility, for this technique. The utility of our experimentally determined threshold value of 146 HU for the forearm was demonstrably high, with sensitivity and negative predictive values of 91% and 89%, respectively. The importance of these high values cannot be overstated when considering CT scans of the wrist as an opportunistic screening mechanism for BMD, as high sensitivity and negative predictive values minimize the number of false-negative test results and capture a majority of truly affected patients. Most importantly, measuring HU values on a CT scan of the wrist does not involve a substantial time investment by the treating physician, it can be done with use of any currently available electronic radiographic software, and it can be performed at no increased cost or radiation exposure regardless of the indication for the scan7. Biswas et al.11 found that the average radiation dose from a wrist CT scan was 0.03 mSv, while the average exposure from a chest radiograph has been shown to be 0.1 mSv12, suggesting that the radiation exposure from a wrist CT scan may be negligible when considered as a screening tool. Screening for osteoporosis could thus be performed immediately, at the time of initial evaluation by the surgeon, bypassing the current screening algorithm completely. Indeed, obtaining CT scans of all distal radial fractures for at-risk patients may merit some consideration, at least in some settings, given the relative ease and minimal radiation exposure inherent to this technique. If otherwise medically appropriate, further interventions may include performing a DXA scan and obtaining laboratory markers for other causes of metabolic bone disease7.

We also found significant differences between the male and female patients included in our study. The females had lower average HU measurements than the males did. However, regardless of sex, patients with low BMD on DXA scans had lower average HU measurements. These findings confirm data published by Schreiber et al.7. The upper threshold of the 95% CI for osteopenia in females was shown to be 141 HU, which closely approximates the 146-HU threshold that we found when evaluating the population as a whole. The upper limit of the 95% CI for male patients with low BMD was 188.0 HU. However, because so few males were included in the study, any meaningful conclusions about cutoff values for screening males separately from females cannot be made. Fourteen of the CT scans (in 14 patients [18.9%]) were obtained after the patients sustained distal radial fractures. Comparison of the distal radial fracture group with the no-fracture group yielded no significant difference in average HU values between groups. However, the lower limit of the 95% CI for the HU value in the fracture group fell below our 146-HU cutoff, while the lower limit of the no-fracture group did not. These findings suggest that had distal ulnar HU measurement been used to screen the patients in the fracture group earlier, our mechanism may have been able to identify them as at-risk prior to fracture.

There are several important limitations to our study. It is a retrospective review of cases completed at a single institution, and the indications for a CT scan of the wrist were not standardized. Ideally, DXA and CT scans would be obtained at the same time, across a wide range of patient demographic characteristics, with an even male-to-female ratio, in a prospective manner. The small number of male patients included in our review does limit the applicability of our cutoff value, as the BMD of males will likely always be higher than that of females while the fracture risk may be similar7. It is also highly likely that the low proportion of males with DXA scans available for our study is a reflection of overall low BMD screening rates in the male population. Schreiber et al.7 found that male patients with distal radial fractures had higher wrist HU values than females did, and it is likely that male patients at risk for fractures will have higher average ulnar HU values than females do. Overall, while our population consisted of active-duty military personnel and their dependents, the average age was relatively young, and there was a preponderance of female patients, we do believe that our study is representative of an at-risk populace that is poorly screened for low BMD1. However, we were able to demonstrate very high sensitivity and negative predictive value for low BMD using the cutoff threshold of 146 HU, which suggests that despite the high false-positive rate, false-negative findings for low BMD occur very infrequently.

A possible concern with the generalization of these HU values obtained by our CT scanners lies in the reproducibility of these values using scanners from other manufacturers. Despite these concerns, HU measurements are frequently and effectively used in the assessment of coronary arteries and tumors as well as in the identification of renal calculi composition. As this science of opportunistic screening of patients with CT scans progresses, it will be imperative to factor in the differences between scanner brands, as the HU values are dependent on kVp and the scanner brand. For this reason, it is critical to report the manufacturer and the scanner settings in any article involving HU values. However, it is evident from the literature that the scanners are relatively consistent within a brand, thus supporting the fact that HU values determined in a well-controlled prospective study with well-defined parameters could be universally applied to all scanner brands with use of simple mathematical equations13.

It is also important to note that while BMD assessment is an important aspect of the evaluation of bone health, clinical diagnosis of osteoporosis is multidisciplinary and involves other clinical and laboratory evaluations. The clinical importance of isolated peripheral BMD, its association with central BMD (hip, femoral neck, and spine), and the relationship between these values and ulnar HU values remain to be elucidated. Utility of this technique as more than an opportunistic screening tool for low BMD can only be claimed when it has met specific milestones for development. After establishing that forearm HU values correlate with forearm T-scores, which was the goal of this current study, a correlation between ulnar HU values and central BMD must be determined. Ideally, BMD data, including DXA and wrist HU values, should be prospectively collected across a normally distributed population for all ranges of BMD. The interobserver and intraobserver reliability for the measurement technique must be validated. Experimentally determined cutoff values should then be assessed for predictive efficacy with regard to other fragility fractures. Finally, a cost-effectiveness analysis of wrist CT versus DXA scans must be performed if this technique is to supplant our current BMD-screening algorithm rather than be utilized as an opportunistic technique.

Investigation performed at the Walter Reed National Military Medical Center, Bethesda, Maryland

Disclosure: The authors indicated that no external funding was received for any aspect of this work. The Disclosure of Potential Conflicts of Interest forms are provided with the online version of the article (http://links.lww.com/JBJS/C249).

Disclaimer: The views expressed in this article are those of the authors and do not necessarily reflect the official policy or position of the Department of the Navy, Department of Defense, or the U.S. Government.

Back to Top | Article Outline

References

1. Wright NC, Looker AC, Saag KG, Curtis JR, Delzell ES, Randall S, Dawson-Hughes B. The recent prevalence of osteoporosis and low bone mass in the United States based on bone mineral density at the femoral neck or lumbar spine. J Bone Miner Res. 2014 ;29(11):2520–6.
2. US Census Bureau. An aging nation: the older population in the United States, population estimates and projections; current population reports. 2014 . http://www.census.gov/prod/2014pubs/p25-1140.pdf. Accessed 2016 Oct 21.
3. Bunta AD. It is time for everyone to own the bone. Osteoporos Int. 2011 ;22(Suppl 3):477–82. Epub 2011 Aug 17.
4. Freedman KB, Kaplan FS, Bilker WB, Strom BL, Lowe RA. Treatment of osteoporosis: are physicians missing an opportunity? J Bone Joint Surg Am. 2000 ;82(8):1063–70.
5. Beattie K, Adachi J, Ioannidis G, Papaioannou A, Leslie WD, Grewal R, MacDermid J, Hodsman AB. Estimating osteoporotic fracture risk following a wrist fracture: a tale of two systems. Arch Osteoporos. 2015;10(1):13. Epub 2015 May 9.
6. Baba T, Hagino H, Nonomiya H, Ikuta T, Shoda E, Mogami A, Sawaguchi T, Kaneko K. Inadequate management for secondary fracture prevention in patients with distal radius fracture by trauma surgeons. Osteoporos Int. 2015 ;26(7):1959–63. Epub 2015 Mar 20.
7. Schreiber JJ, Gausden EB, Anderson PA, Carlson MG, Weiland AJ. Opportunistic osteoporosis screening - gleaning additional information from diagnostic wrist CT scans. J Bone Joint Surg Am. 2015 ;97(13):1095–100.
8. Schreiber JJ, Anderson PA, Rosas HG, Buchholz AL, Au AG. Hounsfield units for assessing bone mineral density and strength: a tool for osteoporosis management. J Bone Joint Surg Am. 2011 ;93(11):1057–63.
9. Rozental TD, Makhni EC, Day CS, Bouxsein ML. Improving evaluation and treatment for osteoporosis following distal radial fractures. A prospective randomized intervention. J Bone Joint Surg Am. 2008 ;90(5):953–61.
10. Rozental TD, Deschamps LN, Taylor A, Earp B, Zurakowski D, Day CS, Bouxsein ML. Premenopausal women with a distal radial fracture have deteriorated trabecular bone density and morphology compared with controls without a fracture. J Bone Joint Surg Am. 2013 ;95(7):633–42.
11. Biswas D, Bible JE, Bohan M, Simpson AK, Whang PG, Grauer JN. Radiation exposure from musculoskeletal computerized tomographic scans. J Bone Joint Surg Am. 2009 ;91(8):1882–9.
12. Lin EC. Radiation risk from medical imaging. Mayo Clin Proc. 2010 ;85(12):1142–6; quiz 1146. Epub 2010 Dec 3.
13. Cropp RJ, Seslija P, Tso D, Thakur Y. Scanner and kVp dependence of measured CT numbers in the ACR CT phantom. J Appl Clin Med Phys. 2013 ;14(6):4417.

Supplemental Digital Content

Back to Top | Article Outline
Copyright 2017 by The Journal of Bone and Joint Surgery, Incorporated