IDIOS: An innovative index for evaluating dental imaging-based osteoporosis screening indices.

PURPOSE: The goal of this study was to develop a new index as an objective reference for evaluating current and newly developed indices used for osteoporosis screening based on dental images. Its name; IDIOS, stands for Index of Dental-imaging Indices of Osteoporosis Screening.
MATERIALS AND METHODS: A comprehensive PubMed search was conducted to retrieve studies on dental imaging-based indices for osteoporosis screening. The results of the eligible studies, along with other relevant criteria, were used to develop IDIOS, which has scores ranging from 0 (0%) to 15 (100%). The indices presented in the studies we included were then evaluated using IDIOS.
RESULTS: The 104 studies that were included utilized 24, 4, and 9 indices derived from panoramic, periapical, and computed tomographic/cone-beam computed tomographic techniques, respectively. The IDIOS scores for these indices ranged from 0 (0%) to 11.75 (78.32%).
CONCLUSION: IDIOS is a valuable reference index that facilitates the evaluation of other dental imaging-based osteoporosis screening indices. Furthermore, IDIOS can be utilized to evaluate the accuracy of newly developed indices.

Introduction

Osteoporosis is a skeletal disease characterized by low bone mass, deterioration of the bone structure, and bone fragility, leading to an increased risk of fracture;1 a less substantial decrease in bone mass is termed osteopenia.2 Bone fragility is evaluated using dual-energy X-ray absorptiometry (DXA), which measures bone mineral density (BMD). The measured BMD of an individual being screened for osteoporosis is compared with that of a young healthy adult as a reference. In this context, the standard deviation values of young healthy adults' BMD are known as T-scores, based on the World Health Organization (WHO) classification published in 1994.2 Using this classification, normal individuals are those with a T-score value of at least -1, individuals with osteopenia have T-scores less than -1 but greater than -2.5, and osteoporosis is diagnosed in individuals with T-scores of -2.5 or less. Due to the asymptomatic nature of osteoporosis over time, it may not be diagnosed until it has progressed to its late stage, which corresponds to an elevated risk of associated morbidity and even mortality.13 Thus, early diagnosis can improve the prognosis and the quality of life of individuals with osteoporosis.

Detecting osteoporosis from dental images is a promising diagnostic possibility, as a variety of dental imaging techniques are commonly indicated for the diagnosis of conditions affecting head and neck structures.4 The most commonly used techniques are periapical and panoramic radiography. Cone-beam computed tomography (CBCT), introduced in 1998,5 has become widely accepted in dentistry, but due to concerns about the dose of radiation, it is not part of routine dental imaging. Since they are routine and feasible, various imaging techniques in dentistry have been evaluated as methods of opportunistic screening for osteoporosis. Numerous dental image-derived indices have been suggested and evaluated for this purpose. Although the results of many such studies are comparable, their conclusions are somewhat contradictory.678 This can be attributed to the absence of clear criteria for describing the values resulting from these indices.

Accordingly, this study aimed to develop a new index as an objective reference for evaluating currently used and newly developed indices, and as a tool that allows the objective comparison of different indices. IDIOS, the suggested name of the index, stands for Index of Dental-imaging Indices of Osteoporosis Screening.

The definitions of the indices used to screen for osteoporosis using dental imaging techniques are included in Appendix 1. However, it should be noted that in some papers some deviations from these definitions may have occurred, and such deviations were ignored in this study. Moreover, some indices were only used in a single study, and the reader may therefore refer to the article in question to obtain the details of such indices.

Materials and Methods

A preliminary PubMed search was conducted in October 2014. Relevant terms and combinations of terms were chosen to retrieve all studies on the diagnosis and screening of osteoporosis using dental imaging techniques (Table 1). A second search, using the same terms and combinations of terms, was conducted on April 13, 2015, in order to include all papers published by that time and to confirm the results of the preliminary search.

Table 1

Search terms used and the number of results

Studies were included if they were in English and described the use of any dental imaging technique to detect or predict osteoporosis in live humans. All retrieved studies were reviewed by one author, and all criteria were applied strictly. Each study was reviewed to extract the imaging technique that was used, the index or indices that were evaluated, and the statistical results regarding the relationship between bone fragility status and these indices. The following statistical parameters were evaluated: sensitivity (SN), specificity (SP), positive predictive value (PPV), negative predictive value (NPV), accuracy, area under the receiver operating characteristic (ROC) curve, and correlations. Statistical parameters that are less commonly used in this context, such as odds ratios and positive and negative likelihood ratios, were not included in the calculation of IDIOS or presented in the results.

IDIOS is based on 5 criteria; 3 criteria were derived from the statistical results of the indices described in the included studies, and the other 2 are highly relevant for evaluating the indices in question. IDIOS has a single output value, ranging from 0 to 15 (equivalent to 0% to 100%). The higher the value of IDIOS, the greater the power and the validity of the evaluated index. The formula for calculating IDIOS scores is presented after the following explanation of the criteria.

The power of the index

This refers to the reported ability of an index to determine positive and negative cases accurately. This criterion is based on the values of the SN and SP of the index, which are the most commonly presented statistical parameters in studies of diagnostic and screening methods. Each value of these 2 parameters was given a score as follows: 6 for values between 90% and 100%, 5 for values between 80% and 89.9%, 4 for values between 70% and 79.9%, 3 for values between 60% and 69.9%, 2 for values between 50% 59.9%, 1 for values between 40% and 49.9% and 0 for values less than 40%.

It is worth mentioning that the authors of some studies presented SN and SP values for the lumbar vertebrae and the hip separately, or presented more than 1 value. In such cases, the maximum and the minimum values were recorded in Tables 2, 3, 4.

Table 2

Panoramic radiography-based osteoporosis screening indices included in this study (sorted alphabetically)

SN: sensitivity, SP: specificity, Acc: accuracy, PPV: positive predictive value, NPV: negative predictive value, ROC: area under the receiver operating characteristic (ROC) curve

*The subscores refer to the following criteria used to calculate the IDIOS score: the power of the index, its reproducibility, objectivity, the presence of software, and differentiation between all bone fragility groups.

Table 3

Periapical radiography-based osteoporosis screening indices included in this study (sorted alphabetically)

SN: sensitivity, SP: specificity, Acc: accuracy, PPV: positive predictive value, NPV: negative predictive value, ROC: area under the receiver operating characteristic (ROC) curve

*The subscores refer to the following criteria used to calculate the IDIOS score: the power of the index, its reproducibility, objectivity, the presence of software, and differentiation between all bone fragility groups.

Table 4

CT-based and CBCT-based osteoporosis screening indices included in this study (sorted alphabetically)

SN: sensitivity, SP: specificity, Acc: accuracy, PPV: positive predictive value, NPV: negative predictive value, ROC: area under the receiver operating characteristic (ROC) curve, CT: computed tomography, CBCT: cone-beam computed tomography

*The subscores refer to the following criteria used to calculate the IDIOS score: the power of the index, its reproducibility, objectivity, the presence of software, and differentiation between all bone fragility groups.

For each index that was evaluated, the corresponding scores for the SN and SP values were recorded separately for individual studies. The means of these scores were then calculated across all studies employing that index. Finally, the mean of these two mean values was defined as the power of that index (i.e., the maximum value was 6).

Reproducibility of the index (interobserver and intraobserver agreement)

This parameter refers to the extent to which the observers reported the same scores for the same subjects on two different occasions, or agreed with each other in reporting scores for the same subjects. In this context, kappa statistics, interclass correlations, Pearson correlations, and/or agreement were considered for each index in each study. Scores of 6, 5, 4, 3, 2, 1, and 0 corresponded to reproducibility values of 0.90-1 (90%-100%), 0.8-0.899 (80%-89.9%), 0.7-0.799 (70%-79.9%), 0.6%-0.699 (60%-69.9%), 0.5%-0.599 (50%-59.9%), 0.40-0.499 (40%-49.9%), and 0.39 (39%) or less, respectively.

The mean score was then calculated for all studies in which a given index was used. This mean was considered the reproducibility score of the index.

No effort was made to differentiate between interobserver and intraobserver agreement in this study. When more than reproducibility test was included in a study (e.g., intraobserver and interobserver agreement or if reproducibility was tested among 3 or more observers), the maximum and the minimum values were included in Table 5.

Table 5

Reproducibility of the indices analyzed in this study

BMD: bone mineral density, DXA: dual-energy X-ray absorptiometry, M/M ratio: maxillary-mandibular ratio, CT: computed tomography, CBCT: cone-beam computed tomography

Objectivity

If the index was based on measurements and/or calculations, it was considered an objective index, and a score of 1 was given. Otherwise, the index was considered a subjective index, such as indices based on visual assessment, and a score of 0 was given.

Software

A score of 1 was given if software was used to perform the analytical process for the index under evaluation. It should be clarified that some indices are natively software-dependent, but the user must perform some preliminary steps. A score of 1 for this criterion means that the main analytical steps, including any preliminary steps, are done by the software.

Differentiation between bone fragility groups

A score of 1 was given to an index if at least one of the studies that used this index included all bone fragility groups (normal, osteopenia, and osteoporosis according to the WHO criteria2 presented above), either in the hip or the lumbar vertebrae, and found at least one of the following: (A) statistical significance between each pair of groups, (B) area under the ROC curve (accuracy) greater than 0.8, (C) a Pearson (or Spearman) correlation coefficient greater than 0.6. Otherwise, a score of 0 was given. Studies that included all bone fragility groups are presented in Tables 2, 3, 4.

To calculate the IDIOS score for an index, the above criteria were applied and the sum of the 5 scores for each criterion was calculated. Accordingly, the maximum IDIOS score is 15 (6+6+1+1+1). The IDIOS scores can be presented as percentages by dividing the IDIOS score by 15 and multiplying the result by 100 ((IDIOS score/15)×100).

Results

The PubMed search yielded 676 studies (Table 1). Only 104 studies (15.4%) were found to be eligible for the development of IDIOS according to the inclusion criteria. In these studies, 24 panoramic (Table 2), 4 periapical (Table 3), and 9 CT/CBCT-based indices (Table 4) were used.

The IDIOS scores for the panoramic imaging-derived indices ranged from 0 (0%), for trabecular bone visual assessment, to 11.75 (78.32%) for the mandibular cortical width index (MCW) (Table 2). The IDIOS scores for the periapical imaging-based indices ranged from as low as 1 (6.67%), for width of the lamina dura, to 11 (73.33%) for bone structure analyses (Table 3). Similarly, the IDIOS scores of the CT/CBCT-derived indices ranged from 1 (6.7%) to 10.25 (68.33%), for radiographic density in gray values (CBCT) (Table 4).

Table 5 shows the reproducibility of each index as reported in the studies that we analyzed.

Discussion

Osteoporosis imposes a significant burden on public health.39 Osteoporosis has been extensively studied in the context of dentistry, because in dental practice, the quality of the jawbone is paramount, as it is the supportive structure for the teeth and for dental implants. It is important, therefore, to investigate the effects, if any, of osteoporosis on the jawbone. Additionally, dental imaging techniques may serve as opportunistic screening tools for osteoporosis, considering the large number of individuals who receive dental services.4

In the current paper, 65% of all indices were derived from panoramic radiographs. As its name implies, panoramic radiographs provide a comprehensive view of all teeth and the jaws, including the temporomandibular joint, with a reasonably low radiation dose. This advantage might be the reason for its more widespread use as method of osteoporosis screening than periapical radiographs, which are more frequently indicated in general.40 Furthermore, the wide variation of intraoral radiographs in the imaged area may be the reason that they are less popular for osteoporosis screening.40 Moreover, occlusal radiography has not been used for diagnosing osteoporosis. It was used in some old studies,2641 to measure the buccal-lingual thickness of the mandibular bone to adjust the densitometer calculated using panoramic and periapical radiographs. On the other hand, the inaccessibility, expense, and high radiation dose of advanced dental imaging techniques reduce their utility as screening methods.

IDIOS is suggested as a useful comparative index due to the emergence of a large number of currently used indices, and in light of ongoing debates regarding their validity and reproducibility, along with the continuous evolution of newly developed imaging techniques, which then leads to the development of new indices. IDIOS acts as a reference against which the current indices can be objectively evaluated. As a preliminary suggestion, IDIOS scores of 7.5-12 (50%-80%) and above 12 (>80%) may be considered good and very good indicators, respectively, of the usefulness of the tested index as a screening tool for osteoporosis.

While developing the IDIOS criteria, the power of the index was first calculated using the SN, SP, PPV, and NPV. However, most studies only reported the SN and SP of the indices they evaluated. Furthermore, PPV and NPV are statistical parameters derived from the SN and SP (i.e., they are SN- and SP-dependent values), and the power of various indices was nearly identical regardless of whether the PPV and NPV were included. Hence, the power criterion of IDIOS was limited to SN and SP.

In this study, the validity of the osteoporosis indices was tested in a collective manner based on the results reported in the literature. In the future, when a new study is performed on the detection of osteoporosis based on dental images, it would be helpful to provide the IDIOS score of the index or indices in question. This would aid readers in assessing the validity of each index and allow them to compare indices with each other.

IDIOS evaluations of the retrieved indices

The highest IDIOS scores were reported for MCW, bone structure analyses (on periapical radiographs), radiographic density in gray values (CBCT), and the mandibular cortical index (MCI), respectively. Although the MCI is subjective in nature and requires repeated training, which explains its lower reproducibility, both indices, MCI and MCW, have been studied more extensively and in more depth than the other indices.

When the MCI was introduced by Klemetti et al.,22 they were not enthusiastic about its potential, so they did not recommend it for identifying osteoporosis in women. It is possible that Klemetti et al. expected a stronger correlation of jawbone measurements with bone fragility status. Ongoing research focuses on the development of the MCW and MCI indices. In 2007, Devlin et al.42 introduced a fully automated computer program for measuring the MCW. It was not as useful as they thought, but they improved it and proved its usefulness in a later study.43 Although software for identifying MCI categories has been developed,44 it was not found to be as useful as the MCW software.45 Similarly, some studies used complicated analyses25464748 to improve the practicability of some indices regarding differentiation between bone fragility groups; however, these analyses unfortunately proved impractical.

Measuring the cortical width in other areas of the mandibular cortex (e.g. gonial and antegonial indices)424950 may achieve more precise values than the MCW, which is performed in the mental foramen region, but these measurements face some limitations. For example, it is unclear where exactly they should be performed, which poses challenges for reproducibility. For the MCW, however, the presence of an obvious characteristic radiolucent landmark (the mental foramen) facilitated the development of a program to perform the relevant calculations automatically. One drawback of the MCW (and indices that depend on length measurements) is the fact that this index must be corrected for magnification.5152 This means that calibrations of the measurements should be made, which might complicate the analysis process.

In contrast to the MCW, the panoramic mandibular index and alveolar bone resorption degree (maxillary/mandibular ratio) are indices that proved to be ineffective (IDIOS power scores of 2.83 and 1.5, respectively), despite their high reproducibility scores. The final IDIOS score cannot be high unless all components are high, due to the fact that the nature of IDIOS depends on the summation of multiple values (especially its major components of the power and the reproducibility). This consideration may add to the utility of IDIOS.

Although the results of the relevant studies remain contradictory, measuring radiographic density appears to be a promising approach for predicting osteoporosis. Indices using CBCT35 and periapical radiographs5354 for this purpose had the third-highest and the fifth-highest IDIOS scores (10.25 [68.33%] and 9.02 [60.11%], respectively). This approach was not thoroughly evaluated in the literature for panoramic radiography (no study reported the SN and SP of radiographic density assessed using panoramic radiography), hence its IDIOS score was low (1 [6.7%]). The radiographic density measurements made using multidetector CT analyzed by Naitoh et al.38 revealed a weak correlation between the radiographic density of the mandibular trabecular bone and the lumbar spine BMD; thus, its IDIOS score was low. The inclusion of the cortical bone in the measurement process in the CBCT-based study performed by Barngkgei et al.35 may be the reason for the improved correlation between radiographic density and BMD.

Changes in the trabecular bone of the femoral neck during osteoporosis were confirmed many years ago.55 However, many studies found that trabecular bone measurements of the mandible were useless for osteoporosis prediction,3536385657 in contrast to the correlation between cortical measurements (thickness or density measures), and the BMD of the femoral neck and lumbar vertebrae.355758 A recent study found statistically significant correlations between DXA and the values of Hounsfield units acquired by CT and micro-CT of the mandibular cortex, but not of the mandibular trabecular bone.59 In a 24-year follow-up study, changes of the trabecular bone (toward sparse trabeculation) were not as obvious as the changes of the cortical bone (toward an eroded cortex).60 Two intraoral imaging-based studies2546 found alternations of the trabecular bone pattern; however, these studies had small sample sizes and significant effects were only found in combination with clinical variables.25 Advanced high-resolution imaging or functional imaging techniques (e.g., isotopes) may be future possibilities for exploring the actual influence of osteoporosis on trabecular bone in the jaws.

Simple visual assessment of the cortices on panoramic radiographs may be a useful tool, somewhat similar to the MCI, with an IDIOS score of 8.4 (55.7%). However, such indices are not considered objective and require considerable amount of training, which results in poor overall agreement.61 This also applies to visual assessments of the trabecular bone. In any case, trabecular bone sparseness was found to increase with age.60

Many panoramic radiography-derived indices other than those discussed above have been suggested (e.g., styloid process length, alveolar bone resorption, hierarchic segmentation analysis, antegonial measurements [angles and depth], mandibular angle, and calcified carotid artery plaques),7111628 but they have not been thoroughly evaluated in well-designed studies.

The appropriate and standardized reporting of results will facilitate comparisons among different studies. The power of IDIOS is based on SN and SP, which are the most useful biostatistical parameters in the context of diagnosis/screening studies. Many studies, however, inappropriately used simple correlation/relationship statistics to evaluate the relationships between measurements of dental images and DXA results, instead of using SN and SP. This did not allow a thorough evaluation of the screening power of the index in question, which is why many indices received a low IDIOS score. Ultimately, further evaluation is needed before making the final judgment of such indices.

Repeatability of measurements

A diagnostic tool is worthless if repeatability is not guaranteed, which is why reproducibility is a built-in criterion in IDIOS. Hence, calculating and reporting statistical parameters relating to reproducibility is paramount, and reporting even low values of these parameters is better than confusing readers by not reporting them. Many indices have not been evaluated for reproducibility, and such indices cannot be considered trustworthy tools. However, in one study of the OSTEODENT project,62 the results of one of the observers suggested that the MCI was not useful in osteoporosis screening, unlike the other observers. On the whole, such a study is more reliable than studies that do not report reproducibility parameters at all. Many studies636465 were designed to test the extent to which trained or untrained dentists could apply certain indices. In a study evaluating the reproducibility of the MCI, it was found that this index exhibited poor intraobserver and interobserver agreement. Moreover, minimal training in assessing the MCI (such as might be given in a lecture format) was found to be ineffective, and was associated with poor interobserver agreement and limited diagnostic validity in identifying the signs of osteoporosis.66 Thus, one might ask whether it is possible to generalize the use of these indices among untrained dentists or whether the results were valid. This is why objectivity was used as an additional criterion when calculating IDIOS. It should be emphasized, however, that in most of the studies analyzed in this study, experienced oral radiologists assessed the radiological measurements.

Furthermore, a study comparing digital and conventional panoramic radiographs found that digital panoramic radiographs were better than analog radiographs for measuring panoramic indices.67 Using software developed for specific measurements would resolve these reproducibility-related issues. This is why the presence of specialized software was included as one of the IDIOS criteria. In addition, the software criterion will encourage those interested in programming to create software to perform such analyses automatically.

Limitations of the included studies and suggestions for future research

Many of the studies we analyzed did not contain cases of osteopenia. The main aim of osteoporosis screening indices is the opportunistic screening of osteoporosis-that is, to detect the early stages of the disease, which may be classified as osteopenia. The accuracy of a diagnostic tool cannot be established without determining the capability of this tool to distinguish all individuals with a given condition from those with closely related conditions (such as osteopenia in this case). If a tool does not detect osteopenia, its efficacy as a diagnostic tool cannot be established. This is why the inclusion of all bone fragility groups was one of the IDIOS criteria.

The validity of IDIOS is a major question. However, in addition to the retrospective application of IDIOS to the current indices, its validity can be evaluated by conducting future studies on osteoporosis screening using dental images. Regardless of whether the indices are new or old, the IDIOS score will show if it is valid or not. For example, if an index proven to be effective in osteoporosis screening has a high IDIOS score, that finding would support the validity of IDIOS, and vice versa.

Finally, fracture risk assessment has become an important topic in research related to osteoporosis. The IDIOS criteria may be applied to test the validity of fracture risk assessment rather than detecting osteoporosis. In addition, recent advances in some techniques (e.g., quantitative CT) that assess more than one aspect of bone quality, in comparison to DXA which measures BMD alone, may be used to assess an individual's future risk of bone fracture. Attempts to find correlations or relationships between the measurements derived from advanced techniques and dental imaging indices may lead to completely different results (i.e., higher IDIOS scores). This may be a topic for future research.

In conclusion, the findings of the current study can be summarized as follows:

Numerous indices have been suggested for osteoporosis screening based on dental images. These indices differ in their power to detect osteoporosis. It is expected that new indices will be developed as a result of the continuous development of new imaging techniques, new analytical procedures, and new assessment programs.

IDIOS is an objective reference index to evaluate the osteoporosis detection indices in current use. In this study, this index was applied in a collective manner, assessing the results reported in all papers that were included.

The MCW, bone structure analyses on periapical radiographs, radiographic density in gray values (CBCT), and the MCI had the highest IDIOS scores (11.75 [78.32%], 11 [73.33%], 10.25 [68.33%], and 9.72 [64.77%], respectively).

When performing a new study on the detection of osteoporosis based on dental images, it would be helpful to calculate the IDIOS scores of the relevant indices. This would aid readers in assessing the validity of each index and allow the comparison of indices with each other and with the results reported in other papers.

As the maximum IDIOS score was 11.75 (78.32%), it is clear that dental images are not yet an adequate substitute for conventional methods of diagnosing osteoporosis.