Dentomaxillofacial Radiology (1998) 27, 17 ± 21  1998 Stockton Press All rights reserved 0250 ± 832X/98 $12.00

The relationships between two indices of mandibular bone quality and bone mineral density measured by dual energy X-ray absorptiometry K Horner and H Devlin Turner Dental School, University of Manchester, UK.

Objectives: To establish whether a relationship exists between the bone quality index (BQI), the mandibular cortical index (MCI) and bone mineral density (BMD) of the body of the mandible as measured by dual energy X-ray absorptiometry (DXA). Methods: Mandibular DXA and panoramic radiography were performed on 40 edentulous female patients. Mandibular body BMD was calculated by manual analysis of DXA scans. Panoramic radiographic appearances of the mandibles were assessed according to the BQI and MCI. All assessments were performed by two observers. Results: There were signi®cant correlations between BQI and BMD (P50.01 for both observers) and between MCI and BMD (P50.005), but MCI was the more dominant of the two variables. Inter-observer agreement between BQI assessments was superior to that between MCI assessments. There were marked di€erences between intra-observer agreement for the two observers for both indices. Conclusions: Both BQI and MCI are signi®cantly related to BMD. However, particularly in the case of MCI, there were problems in repeatability of assessments which might limit their use in clinical practice. Keywords: bone and bones; bone diseases; mandible; osteoporosis

Introduction Radiographic assessment of `bone quality' has applications in implantology1 and in research assessing the relationship between oral bone loss and osteoporosis.2 A large number of quantitative and qualitative measurements of mandibular bone from radiographs have been devised for this purpose, including densitometry3,4 and morphometry.5 ± 8 Many of these require specialised facilities or are time-consuming and necessitate radiography of the highest standards. The two simplest assessments described in the literature, and possibly the most appropriate for application in general practice, are the Bone Quality Index (BQI)1 and the Mandibular Cortical Index (MCI).7 Four types of bone `quality' are described in the BQI, based on subjective evaluation of the cortical thickness and trabecular pattern. This is a four point

Correspondence to: K Horner, Radiology Department, University Dental Hospital, Higher Cambridge Street, Manchester M15 6FH, United Kingdom. Received 19 March 1997; accepted 31 August 1997

index (I-IV) and was assessed using the following criteria, also illustrated in Figure 1: Type I: homogeneous cortical bone; Type II: thick cortical bone with marrow cavity; Type III: thin cortical bone with dense trabecular bone of good strength; Type IV: very thin cortical bone with low density trabecular bone of poor strength. Both Engquist et al.8 and Jan and Berman9 found a greater failure rate in mandibular and maxillary implants in Type IV bone, suggesting that the BQI might have practical use in patient assessment. The BQI represents a means of rationalising subjective judgement of radiographs and is analogous to the simple methods proposed by Singh et al.10 and Kruse and Kuhlencordt11 for assessment of osteoporosis in the femur and spine respectively. Klemetti et al.7 devised the MCI as a classi®cation of the appearance of the lower border of mandibular cortex distally from the mental foramen, as viewed on

Bone quality indices K Horner and H Devlin

18

Figure 1 Jaw bone quality index.1 Four types of bone `quality' are described, based on simple subjective evaluation of the cortical thickness and trabecular pattern. Figure redrawn from reference 1

panoramic radiographs. This is a three point index (C1-3) and was assessed using the following criteria, also illustrated in Figure 2: C1: the endosteal margin of the cortex was even and sharp on both sides; C2: the endosteal margin showed semilunar defects (lacunar resorption) or seemed to form endosteal cortical residues (one to three layers) on one or both sides; C3: the cortical layer formed heavy endosteal cortical residues and was clearly porous. This classi®cation has been subsequently employed by Taguchi et al.12,13 and Klemetti and Kolmakow14 in studies relating it to osteoporosis and bone mineral density (BMD) of the mandibular cortices. In the latter study the severity of changes in the mandibular cortex observed on panoramic radiographs was signi®cantly correlated with buccal, but not lingual, cortical BMD measured by quantitative computed tomography (QCT). `Bone quality' is a term which is widely used but which includes various factors, including cortical thickness, trabecular thickness and density and, importantly, BMD. A number of in vivo methods are

Figure 2 Mandibular cortical index.7 A classi®cation of the appearance on panoramic radiographs of the lower border of mandible cortex distally from the mental foramen. C1: the endosteal margin of the cortex appears even and sharp; C2: the endosteal margin exhibits semilunar defects (lacunar resorption) or endosteal cortical residues (one to three layers); C3: the cortical layer has heavy endosteal cortical residues and is clearly porous. Figure redrawn from reference 7

available for the measurement of BMD, including single photon absorptiometry, dual photon absorptiometry, QCT, single energy X-ray absorptiometry and dual energy X-ray absorptiometry (DXA).15 Currently, DXA is widely accepted as the `gold standard' method of clinical bone mineral measurement,16 but until recently it had not been applied to measure BMD in the jaws. Corten et al.17 ®rst suggested the use of DXA for the mandible, but demonstrated its use in only four patients. Hildebolt et al.18 described its use in cadaver mandibles but did not attempt to use it clinically. Recently DXA of the mandible was carried out in a sample of 40 patients, giving data which showed signi®cant correlations between mandibular BMDs and those of the lumbar spine, femoral neck and forearm.19 If subjective indices like BQI and MCI are to be used clinically, it is important to know what they mean in terms of BMD. The aim of this study was to establish whether a relationship exists between BQI, MCI and BMD of the body of the mandible as measured by DXA. Materials and methods The participants in this study were 40 edentulous females attending the Prosthodontic Department of a

Bone quality indices K Horner and H Devlin

19

Dental Hospital for complete denture construction. The patients ranged in age from 43 to 79 years, (mean age 64.4 years; s.d. 8.8 years). The project received local ethical approval and informed consent was obtained from the patients to perform bone densitometry and panoramic radiography. DXA examinations of the mandible were performed using a Lunar DPX-L densitometer (Lunar Corporation, Madison, Wisconsin, USA). Full details of the method of mandibular DXA have been reported in detail previously.19 Data on mandibular BMD were derived by placing rectangular customised regions of interest over the superimposed right and left mandibular bodies to conform to the bone images of each patient (Figure 3). Manual analysis of scans was carried out on two occasions, once by each author (hereafter referred to as observers 1 and 2) and the mean mandibular body BMD calculated for each patient. Patients with a BMD less than 1 standard deviation below the mean BMD of all patients were considered as having `low' bone mass in the mandible. Each patient also underwent panoramic radiographic examination using a Cranex dental panoramic unit (Soredex Orion Corporation, Helsinki, Finland) operated at 63, 65 or 67 kVcp and 190 mAs. Images were recorded on HR-L ®lm (Fuji Photo Film Co. UK Ltd., London, UK) using a standard cassette ®tted with G-8 rare-earth intensifying screens (Fuji). Radiographs were processed in a Durr Medezin 430 automatic processor (Durr GmbH, Bietigheim-Bissingen, Germany). The panoramic radiograph of each patient was viewed on a standard dental panoramic X-ray viewer

Figure 3 Mandibular DXA image with a region of interest box placed over the superimposed sides of the mandibular body

in a darkened room on separate occasions by observers 1 and 2. Mandibular bone quality was assessed using the BQI. The appearance of the lower border of mandibular cortex distal to the mental foramina were classi®ed according to the MCI. The relationships between the mandibular body BMD and the radiographic measurements (BQI and MCI) were assessed by calculation of the nonparametric Spearman's rank correlation coecient (Rs) using SPSS PC+,20 with signi®cance set at P50.05 level. In addition, the mean BMDs in each diagnostic category (Type I to IV for BQI, C1 to C3 for MCI) were compared by t-tests using Bonferroni criteria with signi®cance set at P50.05 level. To assess the relative dominance of the two indices, a stepwise multiple regression analysis was undertaken using BMD as the dependent variable and the BQI/MCI data collected by the two observers as independent variables. In the stepwise regression analysis, a variable was entered if the signi®cance level of its `F value ± to enter' was P50.001 and removed if the signi®cance level was P40.01. To quantify inter- and intra-observer agreement in radiographic classi®cation using BQI and MCI, all assessments were repeated by both observers after a minimum intervening period of one month. Agreement was assessed by calculation of the kappa (k) statistic.21 Results The mean mandibular body BMD determined by DXA was 1.12 g.cm72 (s.d., 0.30 g.cm72), with a range of 0.396 to 1.866 g.cm72. Five patients (12.5%) had BMDs less than 1 s.d. below mean BMD. The proportions of patients diagnosed into each of the 4 categories of BQI are shown in Table 1. The mean mandibular BMDs (and s.d.) of patients in each category are shown in Table 2. For assessments by both observers there were signi®cant di€erences in mandibular BMD between those individuals classi®ed as having type II and type III bone (P50.05). BQI assessments, made by both observers, were signi®cantly correlated with BMD (Rs=70.41, P=0.009 and Rs=70.51, P=0.001 for observers 1 and 2). The proportions of patients diagnosed into each of the three categories of MCI are shown in Table 1. The mean mandibular BMDs (and s.d.) of patients in each category are shown in Table 3. For observer 1, there was a signi®cant di€erence between mandibular BMD measurements in patients classi®ed as having C1 and C3 cortical appearance (P50.05). For observer 2, there was a signi®cant di€erence between mandibular BMD measurements in patients classi®ed as having C1 and

Table 1 The numbers and proportions (%) of patients diagnosed into each of the categories of BQI and MCI by the two observers

Observer 1 Observer 2

Type I

Type II

3 (8%) 1 (3%)

25 (63%) 28 (70%)

BQI

Type III

Type IV

C1

MCI C2

C3

10 (25%) 10 (25%)

2 (5%) 1 (3%)

12 (30%) 22 (55%)

19 (48%) 15 (38%)

9 (23%) 3 (8%)

Bone quality indices K Horner and H Devlin

20

C2 cortical appearance and between C1 and C3 (P50.05). MCI assessments, made by both observers, were signi®cantly correlated with mandibular body BMD (Rs=70.50, P=0.001 and Rs=70.48, P=0.002 for observers 1 and 2). There was a signi®cant correlation between BMD and MCI for both observers' data (R2=0.30 and 0.29). As adding BQI to the equation only marginally increased the R2 value (0.30 and 0.31 respectively) it was excluded from the model. There was no evidence of interaction between the two independent variables. The assessments of inter- and intra-observer agreement in the classi®cations of radiographic appearances by BQI and MCI are shown in Tables 4 and 5 respectively. Discussion This study used DXA as a `gold standard' for in vivo measurement of bone mineral density. Of course, DXA itself is subject to some inaccuracy in BMD measurement, although this is small; precision of DXA in our Department of Diagnostic Radiology ranges from 0.5 to 5%.15 Following the suggestion of Corten et al.17, the mean BMD of the two manual analyses of the DXA scans was used in an attempt to increase the validity of the data. The choice of edentulous patients was e€ectively dictated by the perceived need to avoid the shadows of teeth on the DXA scans which might have led to an arti®cially high BMD measurement. Thus the results should be interpreted with some quali®cation and further work would be needed to ascertain if the conclusions are valid in dentate patients. Similarly, only women were examined

Table 2 Mandibular bone mineral densities of patients related to classification by BQI. s.d., standard deviation; n, number of cases

BQI

Observer 1

Observer 2

BQI group

Mean BMD (g.cm72)

s.d. (g.cm72)

n

I II III IV I II III IV

1.147 1.201 0.909 0.818 1.257 1.198 0.820 1.188

0.095 0.282 0.219 0.524 ± 0.260 0.237 ±

3 25 10 2 1 28 10 1

Table 3 Mandibular bone mineral densities of patients related to classification by MCI. s.d., standard deviation; n, number of cases

MCI

Observer 1 Observer 2

MCI group

Mean BMD (g.cm72)

s.d. (g.cm72)

n

C1 C2 C3 C1 C2 C3

1.305 1.094 0.861 1.243 0.964 0.797

0.265 0.230 0.290 0.250 0.242 0.372

12 19 9 22 15 3

because this work formed part of a larger study on post-menopausal osteoporosis; further work would be needed to establish whether the results reported here would be similar in men. Examination of Table 1 shows that although there are four `types' of bone in the BQI classi®cation, only two of these occurred frequently; most patients in this sample had a reasonably dense trabecular pattern with varying thicknesses of cortices (types II and III bone). In practice this made BQI a dichotomous variable. This may have been a characteristic of this group of patients, related to age, sex or edentulousness. It would be valuable to examine a larger population over a wider age range to determine whether this observation is true in general. This is the ®rst study to relate BQI to mandibular BMD. The results showed that the index was signi®cantly correlated with BMD. The strength of the association between MCI and BMD was approximately the same, a ®nding which is in concordance with that of Klemetti and Kolmakow.14 However, the strength of the associations between BQI, MCI and BMD suggests that the two indices are limited indicators of BMD. Taking into account the results of the multiple regression analysis, it can be concluded that MCI was more dominant in determining BMD than BQI and that little advantage would be gained by applying both indices in combination. When originally planning this study, we had hoped that the data would permit an estimatation of the diagnostic validity of BQI and MCI in the prediction of low mandibular BMD. However, the small number of individuals with `low' BMD (®ve patients with a BMD 51 standard deviation below the mean) necessarily meant that the con®dence intervals around sensitivity values were unacceptably large. Ideally it would be necessary to study a group of patients within which there was a larger proportion of osteopenic individuals. In an attempt to achieve this, we are currently studying patients referred to an osteoporosis clinic, where there is likely to be a higher proportion of patients with low BMD. Inter- and intra-observer agreement were assessed using the kappa statistic, in which a value of 40.75 is considered as `excellent' agreement, 50.4 as `poor' and intermediate values as `fair' or `good'.22 There were limitations in agreement between observers (Tables 4 and 5), being only `good' for BQI but `poor' for MCI. Bearing in mind the limitations of the study, these ®ndings suggest that clinical use of the two indices of bone quality, particularly MCI, might be considerably limited by variation in judgements between observers. These results di€er from those reported for MCI in a recent study,14 where a second observer had almost perfect agreement with a ®rst in classi®cation of cortical morphology. It is dicult to reconcile this di€erence, as we carefully followed the guidelines of Klemetti and underwent a preliminary `calibration' viewing of panoramic radiographs. Obviously, the greater experience of Klemetti and Kolmakow14 in using MCI might play a part.

Bone quality indices K Horner and H Devlin

21 Table 4 Agreement between repeated assessments of BQI, quantified by Cohen's kappa (k). s.e.=standard error BQI

Intra-observer agreement (observer 1) Intra-observer agreement (observer 2) Inter-observer agreement

k

s.e.

0.81

0.09

0.53

0.12

0.70

0.11

Table 5 Agreement between repeated assessments of MCI, quantified by Cohen's kappa (k). s.e.=standard error MCI

Intra-observer agreement (observer 1) Intra-observer agreement (observer 2) Inter-observer agreement

k

s.e.

0.54

0.11

0.38

0.12

0.30

0.12

There were also considerable limitations in intraobserver agreement in assessments of the radiographs

(Tables 4 and 5). Repeated assessments of BQI by observer 1 had `excellent' agreement but those by observer 2 were only `moderate'. In the case of MCI, agreement was `moderate' for observer 1 but `poor' for observer 2. This further underlines the practical problems of using subjective radiographic indices, and suggests that careful training and calibration of observers would be advantageous. In conclusion, this study has three signi®cant ®ndings: First, BQI was signi®cantly correlated with Mb BMD and there was moderate to good agreement between repeated assessments. Second, MCI was signi®cantly correlated with Mb BMD and, of the two independent variables (BQI and MCI), was the more dominant in determining BMD. However, there was only poor to moderate agreement between repeated assessments of MCI. Third, while BQI and MCI do re¯ect mandibular BMD, there are limitations in reproducibility of assessments which might considerably diminish their usefulness in clinical practice. Acknowledgements This work was supported by grants from the North Western Regional Health Authority and The Sir Halley Stewart Trust. We thank Professor Judith Adams and the staff of the Department of Diagnostic Radiology and Bone Disease Research Centre, University of Manchester for their help and advice.

References 1. Lekholm U, Zarb GA. Patient selection and preparation. In: Branemark P-I, Zarb GA, Albrektsson T, (eds). Tissueintegrated Prostheses. Osseointegration in Clinical Dentistry. Quintessence Publishing: Chicago, 1985, pp 199 ± 209. 2. Hildebolt CF. Osteoporosis and oral bone loss. Dentomaxillofac Radiol 1997; 26: 3 ± 15. 3. Kribbs PJ, Smith DE, Chesnut CH. Oral ®ndings in osteoporosis. Part I: Measurement of mandibular bone density. J Prosthet Dent 1983; 50: 576 ± 579. 4. Devlin H, Horner K. Measurement of mandibular bone mineral content using the dental panoramic tomogram. J Dent 1991; 19: 116 ± 120. 5. Bras J, van Ooij CP, Abraham-Inpijn L, Kusen GJ, Wilmink JM. Radiographic interpretation of the mandibular cortex: A diagnostic tool in metabolic bone loss. Part I. Normal state. Oral Surg Oral Med Oral Pathol 1982; 53: 541 ± 545. 6. Benson BW, Prihoda TJ, Glass BJ. Variations in adult cortical bone mass as measured by a panoramic mandibular index. Oral Surg Oral Med Oral Pathol 1991; 71: 349 ± 356. 7. Klemetti E, Kolmakov S, Kroger H. Pantomography in assessment of the osteoporosis risk group. Scand J Dent Res 1994; 102: 68 ± 72. 8. Engquist B, Bergendal T, Kallis T. A retrospective multicentered evaluation of osseo-integrated implants supporting overdentures. Int J Oral Maxillofac Implant 1998; 3: 129 ± 134. 9. Jan RA, Berman CL. The excessive loss of BraÊnemark ®xtures in Type IV bone: a 5-year analysis. J Periodontol 1991; 62: 2 ± 4. 10. Singh M, Nagrath AR, Maini PS. Changes in trabecular pattern of the upper end of the femur as an index of osteoporosis. J Bone Joint Surg 1970; 52A: 457 ± 467. 11. Kruse HIP, Kuhlencordt F. Studies in primary osteoporosis. In: Dixon AStJ, (ed). Osteoporosis: a Multidisciplinary Problem. Academic Press: London, 1983, pp 149 ± 152.

12. Taguchi A, Tanimoto K, Suei Y, Otani K, Wada T. Oral signs as indicators of possible osteoporosis in elderly women. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1995; 80: 612 ± 616. 13. Taguchi A, Suei Y, Ohtsuka M, Otani K, Tanimoto K, Ohtaki M. Usefulness of panoramic radiography in the diagnosis of postmenopausal osteoporosis in women. Width and morphology of inferior cortex of the mandible. Dentomaxillofac Radiol 1996; 25: 263 ± 267. 14. Klemetti E, Kolmakow S. Morphology of the mandibular cortex on panoramic radiographs as an indicator of bone quality. Dentomaxillofac Radiol 1997; 26: 22 ± 25. 15. Adams JE. Osteoporosis and bone mineral densitometry. Curr Opin Radiol 1992; 4: 11 ± 20. 16. National Osteoporosis Society Scienti®c Advisory Committee. Priorities for Prevention. Osteoporosis: a decision-making document for diagnosis and prevention. National Osteoporosis Society: Bath, 1994. 17. Corten FGA, van't Hof MA, Buus WCAM, Hoppenbrouwers P, Kalk W, Corstens FHM. Measurement of mandibular bone density ex vivo and in vivo by dual-energy x-ray absorptiometry. Archs Oral Biol 1993; 38: 215 ± 219. 18. Hildebolt CF, Rupich R, Vannier MW et al. Inter-relationships between bone mineral content measures: dual energy radiography (DER) and bitewing radiographs (BWX). J Clin Periodontol 1993; 20: 739 ± 745. 19. Horner K, Devlin H, Alsop CW, Hodgkinson IM, Adams JE. Mandibular bone as a predictor of skeletal osteoporosis. Br J Radiol 1996; 69: 1019 ± 1025. 20. Norusis MJ. SPSS PC+TM Version 2.0 Base Manual. SPSS: Chicago, 1988. 21. Cohen J. A coecient of agreement for nominal scales. Educ Psychol Measurement 1960; 20: 37 ± 46. 22. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33: 159 ± 174.