In the first phase of the selection 624 references identified, of which 315 were duplicates, leaving 309 unique references for screening phase (Table 4). Full texts of seventeen articles were retrieved being considered potentially to possess the inclusion criteria. However, after full text review, 7 of these articles were excluded for not meeting determined inclusion criteria. Remaining ten studies matching inclusion criteria included in this systematic review for final analysis (Figure 1).
3.1. Study characteristics
All ten studies included in this systematic review investigated the capacity of dental CBCT imaging to accurately diagnose osteoporosis. Studies included female participants only and used observational cross-sectional methodologies. Three studies were conducted in Syria(1,3,17), three in Brazil(6,9,18), one in Iran(19), one in Egypt(20), one in Korea(2) and one in the Unites States(21). All studies were published in English language.
All ten studies reporting on our outcomes of interest enrolled a total of 510 participants. Case-control was the most common method of sampling and was used in 6 of 10 studies(1–3,6,17,20) while cohort sampling was used in remaining four studies(9,18,19,21). Summary of included studies is reported in Table 2.
3.2. Participants’ characteristics
All studies enrolled female participants only; 3 of 10 studies enrolled menopausal and post-menopausal women(1,3,17) while 7 of 10 studies enrolled post-menopausal women only(2,6,9,18–21). Participants’ ages ranged from 46 to 83 years across the studies.
3.3. Intervention (CBCT) characteristics
There was extensive variation among studies in the characteristics of CBCT devices used to assess low BMD. Three studies used WhiteFox® version 3(1,3,17), two used KODAK 9000 C 3-D system(9,18) while remaining five studies used different devices. Field of view (FOV), voxel size and viewer software varied across included studies. Characteristics of CBCT devices are reported in Table 2.
3.3. Quality Assessment
Quality assessment and applicability concerns for each of the included articles were outlined in Figure 2 and Figure 3 and represent review of authors' judgements regarding each section across studies included in this systematic review. Overall, there were large variations among studies in cases and controls classifications, CBCT devices and viewer software. Small sample sizes and insufficient data ruled out the meta-analysis. Only one study fulfilled all of the methodological quality criteria (9).
3.3.1. Patient Selection
Patient selection posed a high risk of bias (ROB) in six of ten because of case-control sampling. Two studies were placed into the low risk of bias category because of the use of cohort sampling technique and random sampling strategy. ROB was unclear for two of ten studies; Shokri et al. (2019)(19) and Tadinada et al. (2013)(21). Shokri et al. (2019)(19) avoided case-control design for patients’ selection but did not report the sampling method, whereas Tadinada et al. (2013)(21) did not present sufficient information to assess selection bias. Patients and settings did coincide with the review question; hence applicability concern was low for all ten studies.
3.3.2. Index Test
ROB was unclear in 7 of 10 studies, primarily due to the lack of information about masking for conducting and interpreting the index test results. Moreover, the authors in these studies provided no information on the use and pre-specification of thresholds. However, the remaining three studies had a low risk of bias(6,9,19). Index test applicability concerns were low in all studies except one which had unclear applicability concern due to lack of information.
3.3.3. Reference Test
ROB was low in 6 of 10 studies. However, the remaining four studies had unclear ROB due to lack of blinding information, and It was not clear whether or not findings were evaluated without knowing the results of the study. Reference test applicability concerns were low in all studies except one which had unclear applicability concern due to lack of information.
3.3.4. Flow and timing
ROB was high in the study conducted by De Castro et al. (2020)(6). Although the time interval between index test and reference test was not a matter of concern, the authors excluded osteopenia patients, which raised the concern of false increase in sensitivity; hence we assessed it high ROB. ROB was unclear in Koh and Kim (2011)(2) study as the authors did not report timing between two tests. All other included studies had low flow and timing ROB.
Reporting of outcome variables varied across included studies. Six of ten studies reported sensitivity and specificity(1,3,6,9,17,19) while accuracy was reported in half of the included studies , , , , . Four of the included studies, measured qualitative radio-morphometric indices and linear measurements on the CBCT scans(2,9,18,20).
Brasileiro et al. (2017) performed measurements on CBCT derived cross-sectional images to determine mean values of computed tomography mandibular index, computed tomography index (inferior) and computed tomography index (superior). The authors concluded that the mean values of all indices were significantly lower in osteoporosis group compared to osteopenic and healthy patients (p < 0.02)(18).
Mostafa et al. (2016) conducted a study to assess the capability of using box-counting fractal dimension (FD) and mandibular CBCT radio-morphometric indices to detect osteoporosis in 25 postmenopausal women in comparison to 25 healthy females in control group. The authors found significant differences for the CT mental index, CT cortical index scores, and CT mandibular index between the cases and control. The osteoporotic group showed lower mean values compared to the healthy group. However, FD values of control group were not found to be significantly higher in comparison to osteoporotic group. Authors suggested that CBCT is a useful secondary diagnostic tool to implement in patients at risk of insufficient bone density for further diagnostic assessment.(20)
Koh and Kim (2011) studied the possible use of the radio-morphometric indices derived from CBCT images for diagnosis of osteoporosis in the sample of 21 postmenopausal osteoporotic women and 21 postmenopausal healthy women. The authors found that there were statistically meaningful differences between two groups in the values of CTI(S), CTI(I), and CTCI while no statistically significant difference was observed in the values of computed tomography mental index (CTMI)(2).
Barra et al. (2020) conducted the study to evaluate four new radio-morphometric indices: symphysis (S), anterior (A), molar (M) and posterior (P). The authors found that there was no statistically significant difference in mean values of the S index and A index between normal, osteopenia and osteoporosis groups. However, mean M index showed noticeable lower values in osteopenia (p < 0.001) and osteoporosis (p = 0.001) compared to healthy individuals. Mean P index was markedly lower in osteoporosis than in healthy patients (P = 0.008) while showing no significant difference between osteopenia and healthy group (p=0.031). These findings demonstrate that M and P indices may become useful tools in dental CBCT images of the mandible for the evaluation of bone density and diagnosis of osteoporosis in postmenopausal women(9).
Barngkgei et al. (2014, 2015) studied lumbar spine and femoral neck T score and its correlation with radiographic density (RD) derived from dental CBCT for evaluation of osteoporosis in postmenopausal females (1,3). Authors demonstrated that RD values derived from the CBCT of the mandibular slice that passes through the lower borders of the mental foramina shows high accuracy in predicting osteoporosis and correlation with femoral neck and lumbar T-scores.
Barngkgei et al. (2015) also assessed RD from the cervical vertebrae. RD values derived from CBCT of
the left part of the atlas (C1) and the dens (odontoid process of the second cervical vertebra) showed the strongest correlation coefficients (r = 0.6, 0.7; P < .001) and the highest sensitivity (70%, 76.9%), specificity (92.9%, 92%), and accuracy (86.4%, 90.8%) in predicting osteoporosis in the femoral neck and the lumbar vertebrae, respectively(3). The dens and the trabecular bone structure of jawbones in osteoporotic and non-osteoporotic women was assessed by using CBCT in a separate study. Barngkgei et al. (2016) found that Jawbone-derived measures showed minor differences between osteoporotic and non-osteoporotic groups. Correlation of Jawbone-derived measures with femoral neck and lumber vertebrae T scores (between osteoporotic and non-osteoporotic groups) was ≤ 0.4 (P > 0.05). However, the correlation between Dens-derived measures and femoral neck and lumber neck was 0.34 to 0.38 (p = 0.02–0.036) and 0.48 to 0.61 (p ≤ 0.003), respectively(17).
Shokri et al. (2019) assessed the correlation between BMD determined by CBCT gray values and BMD determined by DEXA. The authors found that 61 asymptomatic patients (47.5%) had abnormal BMD based on the T-scores of the femoral neck while 55.7% of the patients had abnormal BMD based on the T-scores of lumbar spine. The correlations between the T-scores and the gray values of the maxillary incisor and tuberosity areas were significant(19).
De Castro et al. (2020) used cross-sectional and panoramic reconstructed images to evaluate mandibular cortical width (MCW) and cortical quality. The authors found significantly lower MCW values in osteoporotic women compared to women with normal BMD according to DEXA at all three bone sites; lumbar spine, femoral neck and total hip(6). New three-dimensional morphometric index on panoramic reconstructed images derived from CBCT has been found to be helpful to assess the osteoporosis status in postmenopausal women.
Tadinada et al. (2013) investigated the accuracy of pixel intensity values and width of the mandibular inferior cortical border measured on images derived from a CBCT in comparison to T scores on 32 postmenopausal women. The authors found a strong correlation between the right and left mental indices while no statistically significant correlation found between the T scores of hip, lumbar vertebra, and mental index values on the right or left sides. A Strong correlation was found between thickness of the mandibular cortical border, pixel intensity values, and age. There was no correlation between pixel intensity values and T-scores.