Changing trabecular patterns in panoramic radiographs of Swedish women during 25 years of follow-up

Abstract

Objectives:

The radiographic trabecular pattern on dental radiographs may be used to predict fractures. The aim of this study was to analyze longitudinal changes in the mandibles of 145 females between 1980 and 2005.

Methods:

Panoramic radiographs were obtained in 1980 and 2005. On 290 radiographs, regions of interest (ROIs) were selected in the ramus, angle and body. In all ROIs, the orientation was measured in 36 directions with the line frequency deviation method. The effects of ageing were analyzed for the fracture and the non-fracture groups separately.

Results:

During the follow-up, 61 females suffered fractures of the hip, wrist, spine, leg or arm. The fracture and non-fracture groups displayed dissimilar age changes in each investigated ROI. All significant changes pertained to increasing values of line frequency deviation. With increasing age, the trabecular network in the mandible lost details and the trabeculae became more aligned in their main direction. In the “ramus”, the alignment was to the 110–120˚ axis, parallel to the posterior and anterior ramus border. In the “angle”, the alignment was to the 135–150˚ axis, parallel to the oblique line, and in the “body” ROI to the 150–175˚ direction, approximately parallel to the occlusal plane and inferior cortex.

Conclusion:

Most changes were consistent with the notion that the bone aged less severely in the non-fracture group. In the fracture group, the findings indicate that bone loss leads to redistribution of the remaining bone tissue in such a way that the trabeculae are accentuated perpendicular to the principal loading.

Introduction

Osteoporosis has been defined as a disease characterized by low bone mineral density (BMD), deteriorated bone structure, and increased fracture risk.¹ Many studies focus on osteoporosis and BMD rather than on microstructure and fractures. Low BMD in the spine, femur and radius indicates a high risk of fractures^2–6 but, in general, most fracture patients are not osteoporotic, and BMD is only one of many determinants of the fracture risk.^4,5,7–9

Radiographs offer a window to the inside of the bones. The access to the internal bone structure that they offer may be only partial, but the widespread use of radiographs makes it desirable to explore and increase their diagnostic yield. The microstructure of bone has dimensions between 0.01 and 0.5 mm.¹⁰ The pores in trabecular bone are up to 1 mm, which is in the realm of macrostructure rather than microstructure.¹¹ Considering that dental panoramic devices may resolve details down to 0.1 mm, it is evident that standard panoramic radiographs depict the microstructure of bone only to a limited extent.¹² Nevertheless, it seems that panoramic radiographs are relevant with respect to osteoporosis^13–16 and fracture prediction.^17–20

On dental radiographs, the structure of the mandibular trabecular bone is projected as the radiographic trabecular pattern (Figure 1). Methods have been developed to describe its morphological image features and directional parameters.^21–26 Geraets et al developed the line frequency deviation (LFD) method to measure orientation in two- and three-dimension.^23–26 Changes in the mandible are of interest to dentists, but it has become clear that the mandible and its radiographs are relevant to endocrinologists as well. The aim of the present study is to quantify longitudinal changes in the mandibular trabecular bone of aging females.

Figure 1.

Panoramic radiograph depicting the radiographic trabecular pattern. Three square ROIs with fixed size were selected in each radiograph: “ramus”, “angle” and “body”. ROI, region of interest.

Methods and materials

Participants

The present study is based on an ongoing a longitudinal study of perimenopausal females, the Prospective Population Study of Females in Gothenburg, Sweden. The prospective study was initiated in 1968. The sample was obtained from the Revenue Office Register. To ensure a representative sample, all females living in Gothenburg in 1968 who were born in 1908, 1914, 1918, 1922 and 1930, on dates of the month divisible by six, were invited for a combined medical, psychiatric and dental examination. The initial purpose was to study anemia and health factors related to the menopause. Gothenburg had approximately 44,5000 inhabitants in 1968. The same females were re-examined in 1980 and in 1992. In 2005, the 1930 and 1922 age-cohorts were re-examined. Since then, only 38-year-olds and 50-year-olds have been invited for investigation of health-related trends. The dental examination was performed by dentists and included a clinical inspection of the teeth, gums and oral mucosa, color photography of the dentition, a questionnaire, and a panoramic radiograph.

The participants provided informed consent in accordance with the Helsinki Declaration. The study was approved by the Regional Ethical Review Board in Gothenburg (T453-04 and T075-09).

Of the females who joined the first part of the study in 1968, 73% underwent the medical and dental re-examinations of the second part starting in 1980. An extensive non-participation analysis was performed at the 1992 follow-up. Non-participants were interviewed by means of a telephone call or a letter, and additional information was obtained from national registers and in-patient and out-patient records. The females who declined participation in the first study did not differ significantly from the participants, except that they had shorter long-term survival.²⁷ In 1992, the survivor participation rate was 69% for the medical examination and 64% for the dental examination.²⁸

The occurrence of fractures between 1980 and 2005 was self-reported and hospital-verified using the County Patient Register.^18,19Spine fractures were only included if the patient had clinical symptoms, such as back pain. No fractures of the fingers or toes were recorded. No attempt was made to separate fragility fractures from other fractures. Females who sustained more than one fracture were included only once. Patients with jaw fractures were excluded to avoid interference with the trabecular pattern. Jaw fractures are not osteoporotic but caused by violence or radiation treatment.

The present study focuses on a random selection of 145 females who participated both in 1980 and 2005. Between 1980 and Between, 2005 and Between, 61 of them suffered a fracture: 47%, an arm or wrist fracture; 29%, a leg or hip fracture; 15%, a spine fracture, and 9%, other. The remaining 84 females were fracture-free until 2005. In 1980, the clinical variables age, weight, height, and BMI were recorded. Table 1 gives an overview of those variables for the fracture and the non-fracture groups. None of the variables differed significantly between the groups.

Table 1.

Clinical variables (mean ± SD) recorded in 1980

	N	Age (years)	Weight (kg)	Height (cm)	BMI (kg/m2)
Fracture patients	61	52.8 ± 3.8	65.0 ± 7.9	165.2 ± 5.3	23.8 ± 3.2
Non-fracture	84	52.0 ± 3.2	66.8 ± 11.1	163.2 ± 6.0	25.1 ± 3.9
Total	145	52.4 ± 3.5	66.0 ± 9.9	164.1 ± 5.7	24.5 ± 3.6

BMI, body mass index; SD, standard deviation.

Differences between fracture and non-fracture patients were not significant (p > 0.0125).

Radiographs

Analog panoramic radiographs were made in 1980 and in 2005 to assess the number of teeth, endodontic treatment, and the distance from the cementoenamel junction to the bone crest. The radiographs were made with Scanora (Orion Corp., Soredex, Helsinki, Finland) with 66–70 kV and 20 mA. The Scanora equipment used in 1968 and 2005 provided analog panoramic radiographs with a fixed magnification of 1.3. This made it possible to compare the radiographs, despite the fact that they were taken with two different machines 25 years apart.

For the present study, the radiographs were scanned with a flatbed scanner (Microtek Medi-2200 plus) at a resolution of 236 pixels per centimeter (600 dpi). The panoramic radiographs from 1968, 1980, 1992, and 2005 were of very high quality, but during the digitizing process, the radiographs from 1992 darkened considerably and large “burn-out” areas were found, especially around the jaw angle. This was probably a result of insufficient fixation in the early automatic processing machines. In 1968, there were few fractures and it was therefore decided to use only panoramic radiographs from 1980 and 2005 of the two age-cohorts born in 1930 (n = 102) and in 1922 (n = 43).

Regions of interest

One observer (WG) manually selected three regions of interest (ROIs) on the right side-of the mandible (Figure 1). Afterwards, the ROIs were adjusted automatically to fixed sizes. The “ramus” ROI comprised a large part of the ramus. It consisted of 1000 × 1000 pixels, corresponding to 4.23 × 4.23 cm². The “body” ROI amounted to 800 × 800 pixels located near the first molar and the second premolar. The “angle” ROI was located between the two previous ROIs and amounted to 800 × 800 pixels. Selecting the ROIs is not critical and does not require extensive training,²⁹ at least not when it comes to cross-sectional studies. In the present investigation, the magnification was 1.3 both in 1980 and in 2005. Small deviations in the total size of the panoramic radiographs cannot be excluded but are not important for assessing the directions of the trabeculae. In contrast, if strut analyses had been performed (calculations of the number of termini and nodes per unit, and the number and lengths of strut segments), size deviations would have been more crucial.

The ROIs were made binary (white or black) in three steps. First, a median filter adjusted isolated pixels with deviating grey values (Figure 2a). Next, an unsharp self-masking filter calculated large-scale variations in gray value, caused by varying thickness of the cortex and soft tissues (Figure 2b). These variations were removed from the ROI (Figure 2c). The most common gray value was used as a threshold. Pixels whose gray value exceeded the threshold were made white (gray value 255), whereas the remaining pixels were made black (gray value 0) (Figure 2d).

Figure 2.

(a) ROI from the ramus. (b) blurred version; (c) filtered version (enhanced); (d:) binarized version. ROI, region of interest.

In the binarized ROI, the trabecular orientation was measured along 36 directions: 0˚, 5˚, ..., 175˚. A virtual grid was projected at the center of the ROI (Figure 3). Algebraic equations simulated the grid rotating around the center. The size of the grid was made √two smaller than the sample so that it overlapped the ROI and never protruded, not even when rotated 45˚. For the “ramus” ROI, the grid measured 707 × 707 pixels, and for the “angle” and “body” ROIs, it measured 565 × 565 pixels. When the grid was in position, the fraction of white pixels was calculated for each row of the grid. The standard deviation (SD) of those fractions was defined as the LFD value for that position (Figure 3). The LFD values for 180˚ to 355˚ were obtained by assuming that LFD(180˚)=LFD(0˚), LFD(185˚)=LFD(5˚), ..., LFD(355˚)=LFD(175˚) and included in the polar plots (Figures 4–7).

Figure 3.

Explaining the measurement of LFD orientation with a ROI of 16 × 16 pixels and a measuring grid of 10 × 10 pixels. The grid is rotated 15˚ with respect to the horizon. For each row in the grid, the fraction of white pixels is calculated. Note that the fraction value of 1.0 occurs once and fraction 0.0 twice. The standard deviation of the ten fractions amounts to 0.34, abbreviated LFD(15˚)=0.34. LFD, line frequency deviation.

Figure 4.

(a) The 1980 “ramus” ROI from a fracture patient. (b) Same patient in 2005. In 2005, the trabeculae, depicted as white pixels, are more distinct and the image contains fewer details than 25 years earlier. This favors larger LFD values. (c) Combined polar plots, LFD 1980 (solid line) and LFD 2005 (dotted). The LFD values are plotted as the distance from the circle center in all 36 assessed directions (the upper half of the circle, from 0˚ to 175˚). Similarly, the 36 “assumed” means from the directions 180˚−355˚ are plotted in the lower half of the circle. The diameters in the polar plots are 1. The mean LFD value in the 110˚ direction for 2005 is approximately 0.5, which means an almost perfect alignment of the trabeculae in that direction. Also, the plot for 1980 demonstrates alignment to the 110˚ – 120˚ direction but with a lower LFD value. Large increases in LFD are visible along 100˚ and 115˚ as white gaps between the solid and the dotted lines. Along 50˚, a small decrease in LFD is barely visible. Full scale: 0.12. LFD, line frequency deviation.

Figure 5.

Directional changes in the “ramus” ROI of the fracture and non-fracture groups. The polar diagram for 1980 is drawn as a solid line, whereas the line for 2005 is dotted. The means of the LFD values are plotted as the distances from the center for each direction. The small bars represent the SDs of the means of the LFD values of 61 fractured females to the left and 84 non-fractured females to the right. The end of each bar is plotted where the means of the LFD -values are plotted. The bars are directed inward for the 1980 polar diagrams and outward for the 2005 diagrams. In both the fracture and the non-fracture group, the directional changes after 25 years consist of trabecular alignment to the 110˚−120˚ axis. The “white gaps” between the dotted and the solid diagrams illustrate the 25 year directional changes. Significant changes indicated by * (p < 0.00023) and ** (p < 0.00005). Full scale: 0.12. LFD, line frequency deviation; ROI, region of interest; SD, standard deviation.

Figure 6.

Directional changes in the “angle” ROI of the fracture and non-fracture groups. In both the fracture and the non-fracture group, the directional changes after 25 years consist of trabecular alignment to the 135˚−150˚ direction. Increasing values of orientation show up as “white gaps” between the solid and the dotted diagrams. The smaller the longitudinal changes, the smaller the SD bars. Significant change indicated by * (p < 0.00023) and ** (p < 0.00005). Full scale: 0.12. ROI, region of interest; SD, standard deviation.

Figure 7.

Longitudinal changes in the “body” ROI of the fracture and the non-fracture groups. The changes after 25 years are significant only in the fracture group and consist of trabecular alignment to the 150˚−175˚ direction; i.e. approximately parallel to the occlusal plane and the inferior cortex. Increasing values of orientation show up as “white gaps” between the solid and the dotted diagrams. Very small changes were found in the non-fracture group. Significant changes indicated by * (p < 0.00023) and ** (p < 0.00005). Full scale: 0.12. ROI, region of interest.

The method is illustrated in Figure 3. The trabeculae are visualized as white pixels. All the pixels are white in row 2, and therefore the fraction value is 1. Row 5 and Row 6 have only black pixels, and therefore they have a fraction value of 0. The mean fraction sum in Figure 3 is 0.51. SD is 0.34, which is the LFD value.

The maximum LFD value is 0.5, when the trabeculae are well aligned to the selected direction (Figure 4, right). In contrast, if the trabeculae in the bone are aligned along random directions, then the rows in the grid contain roughly the same fractions of white and black pixels, resulting in a smaller SD (Figure 4, left). The minimum LFD value is 0.0 and occurs when all rows in the grid have the same fraction value.²⁴

Polar plots

In Figure 4, the LFD values are plotted for one fractured female in a two-dimensional polar plot. Each LFD value on the plot surface (circle area) has a specific distance and direction in relation to the center.²⁶ A maximum LFD value of 0.5 corresponds to the radius in polar plots with the diameter 1. Thus, when the trabeculae are well aligned to the selected direction, the LFD value approaches the periphery of the circle (Figure 4, right). In Figures 5–7, the means of the LFD values (calculated in 36 directions and assumed in 36 directions) in the “body”, “angle” and “ramus” ROIs are plotted for the fracture group (n = 61) and the non-fracture group (n = 84). In each of the 36 calculated means, a small bar indicates the SD for the mean LFD values (i.e., the standard deviation of the standard deviations). Due to the 36 small bars, it may be difficult to visualize the dotted line for 2005.

Reliability

Cronbach's α, a measure of the reliability of LFD measurements, was assessed in an earlier study on ROIs in the ramus and body on panoramic radiographs. It yielded values of 0.85 or more.³⁰ Thus, it seems that the selection of the regions of interest does not introduce large amounts of noise and that the LFD measurements are highly reproducible.

Statistics

When comparing the fracture and the non-fracture groups with respect to height, weight, BMI and age, the Bonferroni correction was applied to compensate for the fact that four t tests were performed simultaneously. It meant that for a single t test, the significance level of 0.05/4 = 0.0125 was used.

Similarly, the Bonferroni correction was used to test whether the LFD orientation had changed significantly between 1980 and 2005. This was done for 2 patient groups, 3 ROIs and 36 directions; i.e. 2 × 3×36=216 t tests were performed simultaneously, and the significance level of 0.05/216 = 0.00023 (*) was therefore used. Similarly, the significance level of 0.00005 (**) was applied to find the significant directions corresponding with a Type 1 error of 1% (0.01/216). The statistical calculations were made using the SPSS package (v. 21; SPSS Inc., Chicago, IL).

Results

The LFD values of 36 directions were measured in 870 ROIs from 290 radiographs of 145 females. The measurements from 1980 were compared with those from 2005. Figures 5–7 present the polar diagrams for the fracture group and the non-fracture group in 1980 and 2005. The polar plots for 1980 are drawn as a solid line and the plots for 2005 as a dotted line.

As a compromise between completeness and clarity, only half of the small bars representing the standard deviation for each plotted point (the mean LFD-value) are shown. For the 1980 diagrams, only the inward-directed halves of the bars are shown, and for the 2005 diagrams, only the outward-directed halves. In that way, increasing values of orientation show up as “white gaps” between the solid and the dotted plots.

Figure 5 gives an overview of the findings for the “ramus” ROI. In both the fracture and the non-fracture group, the largest mean LFD value in terms of distance from the center was found along the 110–120˚ axis; i.e. the main direction of the trabeculae. Also, the largest change after 25 years has occurred in that direction; i.e. a trabecular alignment to the 110–120˚ axis. In Figure 5, it is remarkable that the dotted diagrams (2005) encircle the solid line diagrams (1980) completely, especially in the fracture group. It means that in the “ramus” ROI, the LFD values, representing the trabecular orientation, increased in all directions between 1980 and 2005. Significant changes are marked with * or **. They were found in the directions from 120˚ to 175˚, both in the fracture group and the non-fracture group. Along 70˚ and 75˚, the fracture group changed significantly, but the non-fracture group did not. In contrast, along 15˚, 20˚, 25˚ and 30˚, the fracture group did not change significantly, whereas the non-fracture group did.

Figure 6 gives an overview of the findings for the “angle” ROI. In both the fracture and the non-fracture group, the longitudinal changes after 25 years consist of trabecular alignment to the 135–150˚ direction. The plots clearly differ from the ones in Figure 5; they are bigger, and they appear to be rotated about 20˚ anti clockwise. Between 30˚ and 60˚, the dotted diagrams intersect the continuous diagrams, and no significant changes are found.

Figure 7 presents the result for the “body” ROI. The longitudinal changes after 25 years are significant only in the fracture group and consist of trabecular alignment to the 150–175˚ direction; i.e. approximately parallel to the occlusal plane, the mandibular canal and the inferior cortex. Again, the plots appear to be rotated anti clockwise with respect to Figure 6. Between 30˚ and 60˚, small decreases in LFD values are seen both in the fracture group and the non-fracture group, but none of them are significant. In the fracture group, seven significant increases are found, whereas the “body” ROI of the non-fracture group remains stable.

Discussion

The polar plots from the ramus show that the trabeculae are mostly aligned to the 110˚−120˚ direction, approximately parallel to the posterior and anterior border of the ramus. In the “body” ROI, the trabecular alignment is to the 150˚−175˚ direction; i.e. almost parallel to the occlusal plane, the mandibular canal and the inferior cortex, whereas the main directions in the “angle” ROI are 135˚–150˚. The polar plots also show the longitudinal changes after 25 years as white gaps between the dotted and the solid lines. They demonstrate that these changes are greater in the fracture group than in the non-fracture group. In a previous study of the same females, it was found that with aging, the trabecular network becomes sparser, the intertrabecular spaces increase, and the trabeculae seem to be less mineralized.¹⁹ Furthermore, a statistically significantly increased risk of fracture was found in females with a large distance between adjacent trabeculae; i.e. sparse trabeculation.¹⁹ Other studies have also found a reduction in the complexity of the trabecular pattern in osteoporotic individuals.^31,32 An increased distance between trabeculae accounts for more than twice the age-related bone loss compared with the decrease in trabecular thickness.³¹ Furthermore, advanced image analyses have shown that changes in radiographic trabecular patterns are predictive of hip fractures in elderly females.²⁰

Longitudinal fracture studies using trabecular jaw bone are rare.^17–20,28 Many research groups develop software algorithms for artificial intelligence, mostly to identify individuals with low bone density.^{13–16,32–41} Some of them use periapical radiographs,^32–37 whereas others use panoramic radiographs.^38–41 CT, MR and cone beam CT have also been used, mostly for cortical parameters, but at present, the cost and complexity of these methods limit their utility in the clinic.⁴²

According to Wolff's law, the bone of healthy persons adjusts to resist mechanical forces. Bones have the capacity to adapt their architecture to changes in habitual loading. Reduced loading, e.g. due to bed rest induces significant bone loss and mineral changes. In the “body” ROI, the side region of the mandible, the main forces come from tooth clenching related to mastication and bruxism. To resist these forces, the bone mass is redistributed, accentuating the trabeculae perpendicular to the loading; i.e. parallel to the occlusal plane. This idea is supported by investigations showing that the mandibular stiffness and strength is largest in the longitudinal direction.^43–46 The impact of mechanical forces on the human mandible is extremely complicated to investigate.^43–45 The contact area during biting and mastication varies and the “cantilever” construction of the mandible complicates the situation. The mandible is attached only distally to the rest of the skull. During jaw movement, it can deform around the midline in three directions.⁴⁷ Deformation also occurs in the side-regions, where muscular contractions may result in a narrowing of the arch during opening and protrusion, and an arch increase during mandibular retrusion.⁴⁸

The macrostructure of the mandibular bone is related to the microstructure, which, in turn, is related to even smaller levels of organization. The mandibular trabecular orientations in our study are like patterns of the direction of maximum stiffness found in the overlying mandibular cortical bone in the study of Schwartz-Dabney & Dechow.⁴⁵ The directions of the maximum LFD in our Figures 5–7 are similar to the orientation of the facial and lingual cortical plates presented in their Figure 3A.⁴⁵ They found that anterior to the second molar, corresponding to our “body” ROI, the mean orientation was almost parallel to the occlusal plane, whereas in the facial ramus, the orientation was more vertical.⁴⁵There are also some similarities between Figures 5–7 and the orientation of apatite crystals.⁴⁴ Apatite crystals are much too small to be visible on panoramic radiographs, but their orientation confirms that various levels of bone structure are arranged to withstand mechanical loadings.

Panoramic radiographs admit access to the microstructure of the mandible, although only partially. Their widespread use makes them suitable for population studies. Previous studies have shown that the trabecular pattern on panoramic radiographs can be used to predict the occurrence of fractures of the hip, wrist, spine, leg or arm.^17–20 Based solely on the radiographic trabecular pattern on panoramic radiographs, the occurrence of fractures could be predicted with an 0.80 area under the ROC curve.¹⁷ In the present study, significant directional changes were found in the mandibular trabecular bone after 25 years. All changes pertained to significant LFD increases. From the 216 t values that were obtained, only 32 corresponded to decreasing LFD values but none were significant. The significant changes can be explained by loss of trabecular complexity. It is reported that the radiographic trabecular pattern of osteoporotic patients becomes less complex and less detailed than the trabecular patterns of healthy subjects.^19,31–33 This is supported by Figure 4, showing the “ramus” ROIs of a fracture patient. The 1980 ROI appears more detailed than the 2005 ROI, where the trabeculae are more aligned in the 110˚–120˚ direction. Looking at Figure 3, it can be seen that the abundance of detail in the 1980 ROI makes completely black rows as well as completely white rows less likely; the fraction values tend to approach 0.5 and the resulting LFD tends to be less. Thus, the loss of details may explain why the LFD values increase more in all directions in the fracture group than in the non-fracture group, where there are fewer directions with increasing LFD. Moreover, the changes tend to be larger in the fracture group. It appears that the fracture group aged more severely than the non-fracture group. It remains unexplained why, in the “ramus” ROI, significant increases are found along 15˚, 20˚, 25˚ and 30˚ in the non-fracture group but not in the fracture group. It can be hypothesized that the heterogeneity in bone thickness and bone mass around the mandibular foramen, the mylohyoid and the oblique lines play a role.

The fact that the fracture group and the non-fracture group show dissimilar changes supports the notion that the increasing LFD values depict the aging of the mandibular bone rather than an artefact that has been overlooked. The possibility of unwanted changes to the technique, causing apparent longitudinal changes in the radiographs, had to be considered. Our approach to differentiating artifacts from genuine aging relied on the assumption that artefacts would affect the radiographs of fracture patients and non-fracture patients in similar ways, which seems a reasonable assumption. Various authors have reported differences between the radiographs of groups of osteoporotic and normal subjects.^{14–16,32–35} In the present study, the main outcome was not prediction of osteoporosis or future fracture, but the focus was on age changes, which involve a different comparison. If aging affected both groups in similar ways, it would be difficult to rule out all artefacts. However, if the two groups aged in dissimilar ways, this could afford some insight into the changing microstructure.

With respect to the loss of detail explaining the increasing LFD values, one might speculate that in 1980, the panoramic films were of poorer quality and recorded more noise than in 2005. That would invalidate all conclusions on anatomical changes between 1980 and 2005. However, the LFD values in the “body” ROI of the fracture group increased significantly along several directions, whereas the “body” ROI of the non-fracture group remained unaffected. This disproves lower levels of noise in the 2005 radiographs. Obviously the two groups have aged in different ways.

Figures 5–7 might be explained by the preservation of strength of the mandible along the main trabecular direction. The greatest changes in LFD values were found in the fracture group, where the direction of the maximum orientation became more prominent with aging, especially in the “body” ROI. The trabecular alignment in the “body” ROI is in accordance with previous observations that with aging and bone loss, the trabeculae parallel to the occlusal plane seem to be more distinct than the perpendicular ones, which is probably a compensatory biological reaction to avoid jaw fracture.¹⁹

The present study demonstrates changes in the trabecular pattern of the mandible of aging females during a 25 year follow-up period. A strong point of the study is its long-term fracture follow-up. Such studies are rare because they require huge investments of time and effort. Some limitations must be considered. Initially, the LDF method was developed using the trabecular pattern of the distal radius,²⁴ but the method has been used without problems on jawbones as well.^14,17 The placement of the ROI’s is a potential problem. They were set manually by one author (WG) and the sizes were fixed for each region. In a previous study, the effect of the size and location of the ROI on the correlation between the radiographic trabecular structure and BMD was investigated.²⁹ It was found that selecting the ROIs is not very critical; a ROI including parts of the roots performed as well as a smaller ROI excluding dental parts and no extensive training was needed.²⁹ A Cronbach’s α of 0.85 or more for the mandibular LFD values indicates high reproducibility.³⁰ Since a follow-up time of 25 years is long, the panoramic device, the film properties, and the film handling procedures are probably not exactly the same, but at least the magnification factor was 1.3 on both occasions. Obtaining exactly the same head position is not possible, but errors in the positioning of the patients are probably distributed randomly over various subgroups. Considering the very significant t values, it seems that the sources of variation did not matter very much.

The present study shows that panoramic radiographs can be used to quantify long-term longitudinal changes in the microstructure of the mandible. It is shown that the trabecular pattern of aging females becomes more prominently oriented, especially in individuals with postcranial fractures. These findings indicate that bone loss leads to redistribution of the remaining bone mass in such a way that the trabeculae are accentuated perpendicular to the principal forces. Clearly, the mandibles of fracture patients and non-fracture patients age in different ways. Most of the changes were consistent with the notion that the bone aged less severely in the non-fracture group.