Introduction 1 1 2 3 4 5 7 8 9 Changes in BMC of a given ROI are determined not only by the quantity of bone resorbed on the endosteal surfaces but also by the quantity of bone deposited on the periosteal surface. Thus, change in BMC does not reflect the quantity of bone really lost on the endosteal surfaces and, therefore, the decrease in BMC is better qualified as “net bone loss”. Estimation of bone gain due to periosteal apposition would allow assessment of the quantity of bone really lost on the endosteal surfaces. This quantity of bone could be referred as to “endosteal bone loss”. 5 7 10 12 13 Therefore, the aim of this study was to assess age-related apparent net and endosteal bone loss as well as their morphological basis and the age-related changes during a long-term prospective follow-up in a large cohort of elderly men (the MINOS study). Subjects and methods Cohort 14 Measurements aBMD and BMC were measured at the lumbar spine (L2–L4), hip and whole body using pencil-beam DXA (QDR 1500, Hologic Inc., Waltham, MA, USA) and at the distal nondominant forearm using single energy X-ray absorptiometry (Osteometer, DTX-100, Denmark). The OsteoDyne Hip Positioner System (HPS) was used to minimise hip positioning error. ROI of the femoral neck was positioned perpendicularly to the axis of the femoral neck to cover its narrowest part. When necessary, the femoral neck edges were adjusted manually. The QDR 1500 device was calibrated daily using a lumbar spine phantom, yielding a CV for aBMD of 0.33 %. Twice a month, the Hologic hip phantom was measured, yielding a long-term CV of 0.94 % for femoral neck aBMD and 1.05 % for the femoral neck projected area. Also twice a month, a human lumbar spine embedded in methyl methacrylate was measured. Its long-term CV was 1.07 % for BMC of L2–L4, 1.07 % for the projected area of L2–L4, and 0.62 % for aBMD of L2–L4. At the distal forearm, the distal site includes 20 mm of radius situated proximally to the site where the spacing between the medial edge of the radius and the lateral edge of the ulna is 8 mm. Scans with evident error of positioning were excluded. The densitometer was calibrated daily using a calibration standard for DTX 100; its long-term CV was 0.47 % for aBMD and 0.15 % for the projected area. 14 PA 1 3 15 PA EBL EBL PA 1 Fig. 1 PA EBL 0 1 0 1 PA 3 EBL  brighter colour  Statistical methods All calculations were performed by using SAS version 8.2 software (SAS Institute Inc., Cary, NC, USA). Correlation between continuous variables was assessed by Pearson’s simple correlation coefficient. Individual slopes were calculated by using simple linear regression. Comparisons of the individual slopes between age groups were performed by analysis of variance (ANOVA) and adjusted for multiple comparisons by Dunnett-Hsu test. Participants attended two to six exams; those who had few measurements, e.g. two, had them over different periods of follow-up (18–90 months). Individual slopes were calculated by using two to six points distributed over 18–90 months, which could influence accuracy of the calculation. We adjusted for the duration of follow-up or the number of measurements to check whether duration of follow-up and number of measurements influenced the results. Each of them entered significant in the majority of models, although they influenced the results only to a limited degree. We present data adjusted for the duration of follow-up because this variable attained higher level of significance in the models and contributed more to the final model. Results Characteristics of the investigated cohort 1 Table 1 Comparison of 725 men participating in the prospective study and 65 men lost to follow-up after recruitment Parameter n n p p Age (years) 65 ± 7 70 ± 8 < 0.0001  Body weight (kg) 80 ± 13 79 ± 15 0.29  Body height (cm) 169 ± 6 168 ± 7 0.21  2 27.98 ± 3.64 27.75 ± 4.51 0.63  Fat mass (kg) 22.02 ± 7.48 23.23 ± 9.11 0.28  Lean mass (kg) 54.54 ± 6.66 52.16 ± 7.56 < 0.01 NS Tobacco smoking (%) 11.8 11.6 0.98  Physical activity (h/week) 21.8 ± 12.7 17.2 ± 11.3 < 0.005 NS Prevalent fractures (%) 13.8 19.7 0.18  Diabetes (%) 6.5 15.7 < 0.005  Rhumatoid arthritis (%) 1.4 5.7 < 0.01  Parkinsonism (%) 1.5 5.7 < 0.02  2 1.031 ± 0.184 1.052 ± 0.213 0.21  2 0.845 ± 0.121 0.803 ± 0.127 < 0.01 NS Femoral neck BMC (g) 5.111 ± 0.849 4.883 ± 0.934 < 0.04 NS Femoral neck width (cm) 4.082 ± 0.316 4.123 ± 0.356 0.31  2 0.740 ± 0.109 0.691 ± 0.121 < 0.001 < 0.03 2 0.966 ± 0.127 0.910 ± 0.157 < 0.001 NS Whole-body BMC (g) 2706.6 ± 410.4 2550.3 ± 472.7 < 0.005 NS 2 1.210 ± 0.108 1.167 ± 0.121 < 0.003 < 0.05 Distal forearm BMD 0.524 ± 0.065 0.483 ± 0.070 < 0.0001 < 0.01 Ultradistal radius BMD 0.430 ± 0.064 0.399 ± 0.072 < 0.001 < 0.05 2 0.556 ± 0.068 0.513 ± 0.075 < 0.0001 < 0.01 Radius BMC (g) 2.743 ± 0.403 2.527 ± 0.422 < 0.0001 < 0.01 Radius width (cm) 2.471 ± 0.207 2.468 ± 0.216 0.94  Ulna BMD 0.476 ± 0.066 0.438 ± 0.070 < 0.0001 < 0.01 Ulna BMC (g) 1.502 ± 0.244 1.401 ± 0.246 < 0.0001 < 0.01 Ulna width (cm) 1.659 ± 0.136 1.684 ± 0.152 0.16  BMD  BMC  NS p p Characteristics of bone loss 2 2 p Table 2 Average rate of apparent bone loss [change in areal bone mineral density (aBMD)], net bone loss [change in bone mineral content (BMC)] and of periosteal expansion (increase in bone width or area) as well as the simple correlation coefficients of these variables with age in 725 men aged 50–85 at baseline followed up prospectively for 90 months (the prospective MINOS study) Site of measurement Yearly change Correlation with age Bone mineral density 2 (%/year) r p  Lumbar spine 4.205 ± 14.21 0.495 ± 2.910 −0.069 0.07  Femoral neck −2.463 ± 8.305 −0.282 ± 1.019 −0.154 < 0.0001  Trochanter −1.963 ± 8.031 −0.276 ± 1.123 −0.196 < 0.0001  Total hip −4.714 ± 8.451 −0.496 ± 0.930 −0.213 < 0.0001  Whole body −2.081 ± 8.712 −0.177 ± 0.723 −0.045 0.22  Distal forearm −2.937 ± 4.178 −0.580 ± 0.870 −0.202 < 0.0001  Distal radius −2.986 ± 5.340 −0.561 ± 1.041 −0.180 < 0.0001  Distal ulna −3.353 ± 5.131 −0.730 ± 1.206 −0.119 < 0.002  Ultradistal radius −1.823 ± 5.334 −0.426 ± 1.285 −0.128 < 0.001 Bone mineral content (mg/year)  (%/year)  −0.041 0.27  L3 118.17 ± 552.90 0.658 ± 2.954    Femoral neck 12.67 ± 50.38 0.263 ± 1.010 −0.051 0.18  Total hip −226.1 ± 0.687 −0.504 ± 1.551 −0.120 < 0.002  Whole body −7565.9 ± 22500.9 −0.294 ± 0.877 −0.192 < 0.0001  Distal radius −11.26 ± 25.90 −0.426 ± 0.978 −0.179 < 0.0001  Distal ulna −8.64 ± 16.82 −0.559 ± 1.164 −0.123 < 0.001 Bone size     2 35.64 ± 225.40 0.167 ± 1.184 −0.056 0.10  Femoral neck (μm/year) 133.1 ± 217.2 0.321 ± 0.503 0.070 0.06 2 4.82 ± 42.23 0.112 ± 0.915 0.007 0.85  Distal radius (μm/year) 60.48 ± 362.66 0.257 ± 1.497 0.034 0.36  Distal ulna (μm/year) 58.27 ± 156.61 0.355 ± 0.937 0.013 0.73 External diameter of the femoral neck, distal radius and distal ulna as well as the cross-sectional area of L3 and the projected area of total hip increased significantly during the follow-up. Fractional increase in bone size varied from 0.17% to 0.36% per year across the sites. Characteristics of the rate of bone loss according to age at baseline 2 2 Fig. 2 black bars pointed bars white bars p  3 Fig. 3 black bars pointed bars white bars p  4 Fig. 4 Upper panel Lower panel positive hatched bars negative pointed bars  lower panel  Discussion 5 7 10 16 13 7 12 14 16 18 6 16 17 19 20 19 21 22 15 23 24 25 1 26 27 Our study has limitations. Montceau les Mines is a small town, and its inhabitants may not be representative of the French population. The response rate for the invitation was 23%. Men who abandoned the study after the first examination were older and sicker, although they represent only 8% of the initial cohort. Men who were followed up may have been healthier than the general population, especially in the oldest group. However, this difference would have underestimated the age-related bone loss and its age-related acceleration. A number of men had lower number of DXA scans because they did not attend examinations regularly or abandoned the study before the end of the follow-up. A low number of DXA values and shorter follow-up could influence the accuracy of estimation of slopes. However, adjustment for the follow-up duration or the number of scans did not influence the results. DXA presents limitations in the evaluation of bone width. In very old men, subperiosteal bone mass can be low and not recognised by the edge-detection system. This artefact can underestimate the bone width in elderly men and the age-related increase in bone width mainly in the cross-sectional studies (where the age range is large) but less so in the longitudinal study (where the follow-up period is shorter). The projected area of femoral neck may be overestimated because of calcifications in fibrous tissue. The measured radius site is established by the device. According to the individual anatomy, this site may be more distal (larger and more trabecular) or more proximal (narrower and cortical). Again, this artefact may introduce a bias mainly in cross-sectional studies. Calculation of endosteal bone loss is indirect and based on the assumptions such as uniform bone flattening, constant subperiosteal bone vBMD and proportional periosteal expansion in all axes. By contrast, the advantage of this concept is that we do not make any assumption on the morphological basis underlying the endosteal bone loss (cortical thinning or trabecular bone loss, proportion of cortical to trabecular bone, similar or different rates of trabecular and cortical bone loss, etc.). Finally, our calculation of endosteal bone loss was carried out for the predominantly cortical sites and, although globally consistent with the cross-sectional data obtained by hr-pQCT, may not necessarily apply for the predominantly trabecular sites. In conclusion, in a large cohort of elderly men, age-related apparent bone loss (aBMD) at the hip, distal forearm and whole body was determined by the net bone loss (BMC), except for the femoral neck. Apparent and net bone loss accelerated with age, whereas the periosteal expansion rate (widening of ROI) remained constant. At the distal forearm, age-related acceleration of the apparent bone loss was determined by the higher endosteal bone loss, whereas the periosteal apposition rate (estimated mass of deposited bone) remained constant.