Comparing the Diagnostic Accuracy of CBCT Grayscale Values with DXA Values for the Detection of Osteoporosis

Cone beam computed tomography (CBCT) allows for the dimensional evaluation of jaw anatomic structures and the measurement of jawbone mineral density (BMD), to evaluate the relationship between the measured CBCT grayscale (GS) values and (dual X-ray absorptiometry) DXA scores among osteoporotic and normal dental implant participants. In addition, it can help to establish the CBCT GS values in different jaw regions that might indicate or associate osteoporosis. Eighty-one (81) participants (59 females and 22 males) were recruited into this clinical inquiry. Participants’ ages ranged between 36 and 75 years (mean age ± SD: 57.9 ± 9.3 years). The inclusion criteria included patients diagnosed with osteoporosis and were indicated for the dental implant-supported prosthesis.GS values were calculated for each participant using the CBCT radiograph and associated computer software. The average T values obtained from L1, L2, L3 and L4 were recorded and evaluated for one time only, and correlated with CBCT GS values at different jaw positions. Associations between age, gender, GS values, and T values were evaluated by Pearson correlation tests. The GS values at posterior maxilla were able to predict the presence of osteoporosis (P < 0.001, R square = 0.849). On the other hand, attempts to predict T value using GS at different jaw sites showed that GS values at posterior maxilla and anterior mandible were able to predict T value (P < 0.001, R square = 0.924). The CBCT GS values of BMD at different jaw bone regions are significantly related to the DXA T values in the presence of osteoporosis.


Introduction
Osteoporosis is a common disease that affects over two hundred million people worldwide [1]. It is associated with bone fractures among men and women, especially hip and vertebral column bone fractures [2]. It affects nearly 30% of postmenopausal women in the United States and Europe [2]. Dual X-ray absorptiometry(DXA) (previously known as DEXA) is a simple, safe and precise tool that uses two X-ray beams of different energy levels for the identification of peripheral and total body bone mineral density (BMD) [3]. This technique attained its popularity and is considered the gold standard paradigm for bone quality and quantity assessment, because it allows high resolution images, precise short time scanning and, stability of calibration [3]. The outcomes of a DXA scan would associate with osteoporosis, especially in the maxilla. Therefore, this study was carried out to provide more acquaintance in this regard.
This investigation aimed to measure the bone mineral density (BMD) of anterior maxilla, posterior maxilla, anterior mandible, and posterior mandible using CBCT GS values, and to find the relationship between the measured CBCT GS values and DXA scores among osteoporotic and normal dental implant participants. In addition, it helps to establish the CBCT GS values in different jaw regions that might indicate or associate with osteoporosis. The null hypothesis for this inquiry was set as that no correlation exists between CBCT GS values and DXA scores among normal or osteoporotic participants, and that CBCT GS values are not related to the existence of osteoporosis.

Study Design
In total, 81 participants (59 females and 22 males) were recruited into this clinical inquiry. Participants' ages ranged between 36 and 75 years (mean age ± SD: 57.9 ± 9.3 years). They were patients who were diagnosed with osteoporosis and were attending the dental implant clinic (College of Dentistry, Jouf University, KSA) for treatment with dental implant-supported prosthesis. The study was conducted throughout the period from March 2018 to March 2019.This clinical study was conducted according to the principles of the Helsinki Declaration (9th version, 2013). This research and the associated clinical procedures were ethically endorsed and approved by the Bioethics Committee, Jouf University, KSA (381-41). Before recruitment into this clinical study, each participant was requested to sign a written consent, after being provided with an explanation of this investigation and associated procedures. The study was conducted as per STARD guidelines [29].

Participants
The recruited participants in this research were distributed into two groups; the osteoporosis group (42 participants) and control group (39 participants). Participants were included in the osteoporosis group if they were recently diagnosed with osteoporosis and did not receive treatment for this condition yet. They should also have the results of their DXA assessment available for the investigators. The average T value information obtained from L1, L2, L3 and L4wasused for DXA assessment. On the other hand, participants were included in the control group if they were previously screened for osteoporosis by their medical doctors, and were confirmed to have no osteoporosis. In order to be included in this research; all participants should have no medical condition (other than osteoporosis for those recruited into the osteoporosis group) that might complicate implant treatment, including cardiovascular disease, bleeding tendencies, renal disease, immune problems, endocrine disease, gastrointestinal disease or hepatic problems. They should suffer no local disease or condition that might interfere with implant treatment, including bony lesions or problems affecting adjacent teeth, such as periodontal disease, endodontic problems, caries, occlusal problems, tooth fracture, or tooth wear.

Eligibility Criteria
Participants who were pregnant, alcoholics, or smokers were excluded from this study. In addition, participants who had any of the above medical problems (other than osteoporosis), had local disease or condition, or suffered problems with the adjacent teeth were also excluded from this research. In addition, participants who had a barium X-ray examination, received contrast material for CT or radioisotope scan, or had taken calcium supplements were also excluded from this study.

Test Methods
After being included in this inquiry; each participant was thoroughly evaluated to gather updated information regarding their medical and dental histories, complaints, and demographic data, including, age, gender, education, marital status, occupation, and, address. Moreover, the participants were clinically examined by a well-trained oral surgeon and prosthodontist following previous investigations [30][31][32][33][34]. A dental mirror (15/16 inch; Hahnenkratt, Königsbach-Stein, Germany) and a dental explorer probe (0700-9, anatomical handle single ended; ASA Dental, Bozzano, Italy) were used during the clinical examination of each participant on a dental unit. A CBCT (Index test) radiograph (SORDEX, Nahkelantie 160 Tuusula, Finland) was taken for the upper and the lower jaws of each participant, in order to check for local problems, to help during implant treatment planning [35], and to measure Hounsfield values of jaw bone density (machine settings: Medium field of view, 0.2mm voxel size,90KV). Then, the GS values were calculated for each participant using the CBCT radiograph and associated computer software, using OnDemand 3D, Yuseong-gu, Daejeon, Korea ( Figure 1). For each participant, the GS values were calculated at four regions; namely, anterior maxilla, posterior maxilla, anterior mandible, and posterior mandible. After that, each participant was requested to provide the investigator with documentation of the results of the reference test (DXA values) they had for screening, or the diagnosis of osteoporosis by medical professionals, as it allows high resolution images, precise short time scanning, and the stability of calibration ( Figure 2) [3]. The T value information from the reference test was not available to the performers of the index test, but the same was provided to a statistician directly, as submitted by each participant who underwent the reference test. The assessors of the reference test were also not aware of the clinical presentation and GS values of the participants. For all participants, the assessment of GS was performed by one investigator (MGS). Intra-examiner reliability was confirmed by the recalculation of 10 duplicate GS evaluations by the same investigator (MGS) (Kappa was 0.89-0.95). Moreover, inter-examiner reliability was confirmed by recalculation of the same 10 GS evaluations by another investigator (MKA) (Kappa was 0.85-0.92), which is considered satisfactory [36][37][38].
Appl. Sci.2020, 10, 4584 4 of 13 treatment planning [35], and to measure Hounsfield values of jaw bone density (machine settings: Medium field of view, 0.2mm voxel size,90KV). Then, the GS values were calculated for each participant using the CBCT radiograph and associated computer software, using OnDemand 3D, Yuseong-gu, Daejeon, Korea ( Figure 1). For each participant, the GS values were calculated at four regions; namely, anterior maxilla, posterior maxilla, anterior mandible, and posterior mandible. After that, each participant was requested to provide the investigator with documentation of the results of the reference test (DXA values) they had for screening, or the diagnosis of osteoporosis by medical professionals, as it allows high resolution images, precise short time scanning, and the stability of calibration ( Figure 2) [3]. The T value information from the reference test was not available to the performers of the index test, but the same was provided to a statistician directly, as submitted by each participant who underwent the reference test. The assessors of the reference test were also not aware of the clinical presentation and GS values of the participants. For all participants, the assessment of GS was performed by one investigator (MGS). Intra-examiner reliability was confirmed by the recalculation of 10 duplicate GS evaluations by the same investigator (MGS) (Kappa was 0.89-0.95). Moreover, inter-examiner reliability was confirmed by recalculation of the same 10 GS evaluations by another investigator (MKA) (Kappa was 0.85-0.92), which is considered satisfactory [36][37][38].

Analysis
The statistical analysis for this research was performed using SPSS computer software (IBM SPSS Statistics v19.0; IBM Corp., USA). The calculation of the sample size for this research was performed using computer software (G*Power 3.1.9.2; Heinrich-Heine-University Dusseldorf) after a pilot inquiry. For an effect size of 20%, statistic power of 80% and significant level of 5%; the sample size was found to be 35 participants per group. Additional numbers of participants were recruited into this research to balance any dropouts through this inquiry. Associations between age, gender, GS values and T values were evaluated by Pearson correlation tests. Assessments of the relationships between T and GS values within the osteoporosis group and within controls were evaluated by utilizing a paired samples t-test. A comparison between groups was carried out using Analysis of Variance (ANOVA) test. Significant findings were pointed at α = 0.05. The series mean was calculated and used to handle missing values of T and GS values.

Participants
In total, the data for 81 participants (59 females and 22 males, age range: 36-75 years old, mean age ± SD: 57.9 ± 9.3 years) were assessed and analyzed ( Figure 3). Forty-two participants suffered osteoporosis (6 males and 36 females with a mean age of: 60.0 ± 8.4 years) and 39 participants were controls and osteoporosis free (16 males and 23 females with mean age of: 55.6 ± 9.7 years).

Analysis
The statistical analysis for this research was performed using SPSS computer software (IBM SPSS Statistics v19.0; IBM Corp., USA). The calculation of the sample size for this research was performed using computer software (G*Power 3.1.9.2; Heinrich-Heine-University Dusseldorf) after a pilot inquiry. For an effect size of 20%, statistic power of 80% and significant level of 5%; the sample size was found to be 35 participants per group. Additional numbers of participants were recruited into this research to balance any dropouts through this inquiry. Associations between age, gender, GS values and T values were evaluated by Pearson correlation tests. Assessments of the relationships between T and GS values within the osteoporosis group and within controls were evaluated by utilizing a paired samples t-test. A comparison between groups was carried out using Analysis of Variance (ANOVA) test. Significant findings were pointed at α = 0.05. The series mean was calculated and used to handle missing values of T and GS values.

Participants
In total, the data for 81 participants (59 females and 22 males, age range: 36-75 years old, mean age ± SD: 57.9 ± 9.3 years) were assessed and analyzed ( Figure 3). Forty-two participants suffered osteoporosis (6 males and 36 females with a mean age of: 60.0 ± 8.4 years) and 39 participants were controls and osteoporosis free (16 males and 23 females with mean age of: 55.6 ± 9.7 years).

Test Results
GS values (Index test) were recorded using CBCT at different regions of jaws without any adverse effects. The results indicated that patients with osteoporosis were significantly older than in the normal group (P = 0.029). Moreover, the results indicated a significant association between gender and osteoporosis, with more females being affected with the condition (P = 0.007) ( Table 1). Table 2 demonstrates means, standard deviations, as well as the maximum and minimum values of T value of DXA analysis and GS values at different jaw regions for osteoporosis group and controls. In both groups; GS values were highest at anterior mandible and lowest at posterior maxilla. However, the controls scored higher T value and GS values at different jaw regions than the osteoporosis group (Table 2). Osteoporosis was associated with GS values that ranged between 400 and 630 at the anterior mandible, 290 and 420 at the posterior mandible, 310 and 422 at the anterior maxilla, and 50 and 233 at the posterior maxilla. Age and gender were correlated to T value and GS values at different jaw regions. Among participants within each group, no significant relationships were found between age and each of T value and GS values at different jaw regions, except that younger age was associated with higher GS values at anterior mandible (R = −0.363, P = 0.018) within the osteoporosis group. Among controls, gender had no significant relationships with T value or GS values at different jaw regions. In contrast, females from the osteoporosis group demonstrated lower GS values at anterior mandible, posterior mandible, and anterior maxilla (R = −.0435, −0.415, and −0.442; and P =0.004, 0.006, and, 0.005 respectively) than males. However, males and females with osteoporosis were not significantly different regarding T value and GS at posterior maxilla (P > 0.05). Table 3 shows the correlations between T value and GS values at different jaw regions for the osteoporosis group and controls. Higher T values were associated with higher GS values at all jaw regions (P < 0.001). Table  4 presents the associations between T value and GS values at different jaw regions within each group. Within each group, higher T values were associated with higher GS values at all jaw regions (P <

Test Results
GS values (Index test) were recorded using CBCT at different regions of jaws without any adverse effects. The results indicated that patients with osteoporosis were significantly older than in the normal group (P = 0.029). Moreover, the results indicated a significant association between gender and osteoporosis, with more females being affected with the condition (P = 0.007) ( Table 1). Table 2 demonstrates means, standard deviations, as well as the maximum and minimum values of T value of DXA analysis and GS values at different jaw regions for osteoporosis group and controls. In both groups; GS values were highest at anterior mandible and lowest at posterior maxilla. However, the controls scored higher T value and GS values at different jaw regions than the osteoporosis group ( Table 2). Osteoporosis was associated with GS values that ranged between 400 and 630 at the anterior mandible, 290 and 420 at the posterior mandible, 310 and 422 at the anterior maxilla, and 50 and 233 at the posterior maxilla. Age and gender were correlated to T value and GS values at different jaw regions. Among participants within each group, no significant relationships were found between age and each of T value and GS values at different jaw regions, except that younger age was associated with higher GS values at anterior mandible (R = −0.363, P = 0.018) within the osteoporosis group. Among controls, gender had no significant relationships with T value or GS values at different jaw regions. In contrast, females from the osteoporosis group demonstrated lower GS values at anterior mandible, posterior mandible, and anterior maxilla (R = −0.435, −0.415, and −0.442; and P = 0.004, 0.006, and, 0.005 respectively) than males. However, males and females with osteoporosis were not significantly different regarding T value and GS at posterior maxilla (P > 0.05). Table 3 shows the correlations between T value and GS values at different jaw regions for the osteoporosis group and controls. Higher T values were associated with higher GS values at all jaw regions (P < 0.001). Table 4 presents the associations between T value and GS values at different jaw regions within each group.
Within each group, higher T values were associated with higher GS values at all jaw regions (P < 0.001). An analysis of variance of T value and GS values at different jaw regions revealed significant differences between groups (Table 5). Osteoporosis group participants were associated with significantly lower T value and GS values at different jaw regions (P < 0.001). A stepwise regression analysis was carried out in an attempt to predict the presence of osteoporosis, using GS values at different jaw regions; this revealed that GS values at the posterior maxilla were able to predict the presence of osteoporosis (P < 0.001, R square = 0.849) ( Table 6). On the other hand, attempts to predict T value using GS values at different jaw sites showed that GS values at the posterior maxilla and anterior mandible were able to predict T value (P < 0.001, R square = 0.924) ( Table 7).

Discussion
This inquiry demonstrated that positive relationships did exist between the measured CBCT HU values and DXA scores among osteoporotic and normal dental implant participants. In addition, CBCT HU values were related to the existence of osteoporosis and were able to predict osteoporosis. Therefore, the null hypothesis was rejected. In this study, a digital 3D CBCT machine was used to calculate bone mineral density. It has nine fields of views (FOV) and various low-doses settings that offer great flexibility when choosing the exposure mode. Moreover, it has a wide range of sizes that allow choosing based on true clinical needs. Helena reported that the radiation dose decreased after decreasing the FOV, but the relation was not linear [39]. During this investigation, the acquisitions were made with medium FOV (10 × 14 mm), in order to reduce the radiation dose. The diagnostic accuracy of this CBCT machine has been confirmed previously [40]. The DXA machine used to assess osteoporosis for the purpose of this study was a central DXA device, as it is more sensitive and better standardized than peripheral ones [41]. In addition, it has controlled dose methods to minimize scatter radiation. This ensures that only the exposed parts are imaged, while other parts of the body receive minimal radiation.
Few studies have been conducted to evaluate the efficacy of CBCT in measuring bone density and to correlate it with DXA values [11,13,17,[42][43][44]. To the authors' knowledge, this is the first study that correlated the CBCT Hounsfield number to DXA T scores at different regions, in both the mandible and the maxilla. In the present study, GS values at the posterior maxilla were able to predict the presence of osteoporosis. This might be explained by the thin cortical bone and greater volume of cancellous bone in comparison to the mandible. Therefore, it is affected more by hormonal changes in osteoporotic conditions, as it has a richer blood supply [11]. This is in accordance with the findings of Naitoh et al., who concluded that the voxel values of mandibular trabecular bone obtained from CBCT could be used for BMD estimation [26]. Similarly, Barngkgei et al. suggested that the bone density measurement of the mandible with CBCT can predict subjects with osteoporosis [20]. However, they did not use GS values to predict osteoporosis during their investigations, and they studied the mandible, not the maxilla [21]. Other researches also indicated that osteoporosis could be predicted with high accuracy from the radiographic density value of the mandibular body performed by CBCT [21].
The outcomes of the present study showed that GS values at the posterior maxilla and anterior mandible were able to predict T value. This is in agreement with previous investigations that used different methods and different research settings in this regard [45]. Marquezan et al. compared the BMD values obtained by CBCT and DXA, and found a positive correlation between the two modalities [45]. Moreover, Mostafa et al. used the fractal dimension (FD) and radio-morphometric indices of jawbone CBCT images and correlated them with BMD accessed via DXA [46]. They observed high correlations between the radio-morphometric indices and BMD values. In addition, Hsu et al. compared CBCT and DXA to evaluate the bone strength of the cortical bone of rats' femur and tibia bones. Their observations suggested that CBCT could provide better fracture load predictions than DXA [47]. In contrast, other researchers reported no correlation between the BMD obtained from DXA and the CBCT for the mandible [48]. This could be attributed to utilizing different CBCT devices, exposure protocols and radiation dose voxel sizes [42,49]. Moreover, other studies concluded that the radiographic densities may be inaccurate and depend on the voxel size [50]. It should be emphasized that the aforementioned studies use different CBCT machines and different voxel sizes [51][52][53]. The findings of the present study are restricted to the current model of CBCT used in this study, unless experimental verification and calibration are done, since some manufacturers do not calibrate the gray value scale along with a Hounsfield unit [51].
The findings of this study demonstrated the significant relationships between CBCT GS values on one side and DXA scores and presence of osteoporosis on the other side. CBCT GS values at certain jaw regions were able to predict osteoporosis. Therefore, it would be useful to use CBCT scans obtained for dental implant purposes for predicting osteoporosis and refer suspected patients for further medical investigations. This would help potential patients, because osteoporosis is a silent condition with no symptoms, and patients usually find out that they are osteoporotic, either after fracture or after random screening. In this context, it is suggested to refer patients for the DXA screening of osteoporosis if they showed CBCT GS of 630 or less at anterior mandible and/or 233 or less at posterior maxilla.
Limitations: Considering the sample size, the wide variation of CBCT machines, the different nature between DXA and CBCT and the fact that CBCT is used mostly for dental purposes, we call for future research on this topic.

Conclusions
The CBCT GS values of bone mineral density at different jaw bone regions are significantly related to the DXA scores, as well as the presence of osteoporosis. CBCT GS values at posterior maxilla were able to predict the presence of osteoporosis and DXA T score values. CBCT GS values at the anterior mandible were able to predict DXA T score values. Therefore, it would be useful to use CBCT scans obtained for dental implant purposes for predicting osteoporosis and refer suspected patients for further medical investigations if they showed a CBCT GS of 233 or less at the posterior maxilla and/or a GS of 630 or less at the anterior mandible.