Abnormal bone remodeling in patients with spontaneous painful vertebral fracture


  • Presented at the Sixth International Symposium on Clinical Disorders of Bone Metabolism, Venice, Italy, 1999; proceedings published by Centro Italiano Congressi, Rome.


The application of tetracycline-based iliac bone histomorphometry to the study of the pathogenesis of osteoporosis has given conflicting results. Accordingly, we performed this procedure in 78 postmenopausal white women with one or more vertebral fractures identified according to rigorous criteria that excluded other causes of vertebral deformity and 66 healthy postmenopausal white women recruited from the same geographic region; the groups did not differ in age or weight. In each subject, measurements were made separately on the cancellous (Cn), endocortical (Ec), and intracortical (Ct) subdivisions of the endosteal envelope. In the fracture patients, osteoblast surface was reduced substantially on each subdivision, most markedly on the Cn surface, where about 25% of the deficit was in cuboidal (type II) osteoblasts, suggesting impaired recruitment; the remaining 75% of the deficit was in intermediate (type III) cells, suggesting earlier transition from type III to type IV (flat) cells. On the Ec and Ct surfaces, the deficit was exclusively in type III cells. Mean bone formation rate was reduced by about 18% on the Cn but not on the Ec or Ct surfaces. The deficit was more significant in subjects matched for Cn BV/TV when adjusted for the inverse regression on osteocyte density and after logarithmic transformation. The difference in bone formation rate resulted from a corresponding reduction in wall thickness without a change in activation frequency. The frequency distribution of bone formation rate was more skewed to the left in the fracture patients than in the controls. Osteoclast surface was significantly lower on each subdivision. The variation in osteoblast surface, bone formation rate, and osteoclast surface was significantly greater in the fracture patients than in the controls, with more abnormally low and abnormally high values. The data suggest the following conclusions: (1) The histologic heterogeneity of postmenopausal osteoporosis is reaffirmed; (2) the different subdivisions of the endosteal envelope, although in continuity, behave differently in health and disease; (3) a combination of defective osteoblast recruitment and premature osteoblast apoptosis would account for the deficit in type II and III cells and the reductions in wall thickness and bone formation rate on the Cn surface and the previously reported osteocyte deficiency in Cn bone; (4) premature disaggregation of multinuclear to mononuclear resorbing cells could account for the osteoclast deficit; and (5) some patients with vertebral fracture have one or another disorder of bone remodeling that at present cannot be identified by noninvasive means. © 2011 American Society for Bone and Mineral Research.


Iliac bone histomorphometry after double tetracycline labeling has contributed substantially to understanding the cellular basis of age-related bone loss,1 the disorders of bone remodeling in hyperparathyroidism,2 and the mechanisms of osteoid accumulation in osteomalacia.3 However, in patients with clinically overt osteoporosis, the findings have been inconclusive. Early reports emphasized histologic heterogeneity, with subsets of patients having both higher and lower bone remodeling than normal,4–6 but in a recent large multiauthored book on osteoporosis, the emphasis was on the usual normality of remodeling.7 Nevertheless, there has been some progress in understanding the cellular basis of continuing bone loss. The postmenopausal increase in cancellous erosion depth8 that leads to loss of trabeculae9 is short-lived in normal subjects and no longer present in osteoporotic patients,10 but increased erosion depth persists on the endocortical surface.1 In cancellous bone there is reduced wall thickness, an index of the amount of bone made by previous teams of osteoblasts, probably the result of decreased recruitment of osteoblasts at each remodeling site,11 accounting for continued thinning of trabeculae. Wall thickness is not so reduced on the endocortical and intracortical surfaces,1 so in the ilium continued cortical bone loss is due mainly to increased erosion depth and continued cancellous bone loss to reduced wall thickness.10, 11

Double tetracycline labeling was first applied to the study of bone remodeling using biopsies of the eleventh rib.12 In patients with clinically overt osteoporosis, the most distinctive finding was an extreme prolongation of sigma, the term coined by Frost to denote the time taken to complete one cycle of remodeling at a single cross-sectional location,13 comprising the resorption, quiescent, and formation periods.7 As later found in the ilium, rates of bone formation covered a wider than normal range, although the mean value was below normal.14 These studies were performed in patients seeking relief from back pain, most of whom had vertebral compression fractures. This experience led Frost to distinguish between “true” osteoporosis, by which he meant a disease causing disability, and asymptomatic or “physiologic” osteoporosis.15 This distinction is repudiated by most current observers, who diagnose osteoporosis solely by bone densitometry regardless of fracture history,16 but the distinction has never been refuted.

Our report of substantial osteocyte deficiency in vertebral fracture patients17 demonstrated that subjects with and without fracture differ not only in the severity of their bone deficit, providing some support for Frost's concept. We now present additional data from these patients that confirm in iliac biopsies some but not all of the conclusions drawn 40 years ago from rib biopsies13, 14 and disclose other abnormalities. The data extend a previous report confined to the cancellous surface11 with results from other surfaces and a larger and more representative set of control subjects. The earlier article focused on the cellular mechanisms of cancellous bone loss; this article is more concerned with the process of bone remodeling. We demonstrate more completely than before that some patients with compression fractures have one or other pathophysiologic abnormality of unknown etiology that is not revealed by any currently available diagnostic procedures.

Materials and Methods


We studied 78 patients with vertebral fracture, all white women aged 45 to 75 years who were at least 1 year postmenopause.18 They all presented with back pain, either spontaneous or after no more than trivial trauma, and were found to have at least one compression fracture with loss of posterior height or at least two noncontiguous wedge fractures.19 These criteria established the presence of significant vertebral bone fragility and ruled out other causes of vertebral body deformation.20 The mean number of fractures was 4 (range 1 to 9). All patients were studied prior to randomization for the sodium fluoride trial between August 1981 and December 1987. Twenty patients were taking a calcium supplement and 14 patients were taking thyroid replacement, but no patient was taking estrogen replacement. One patient, previously included,11 was found to have had probable hyperparathyroidism.

Healthy white postmenopausal control subjects were recruited in one of two ways, both in accordance with institutional review board approval. Twenty-eight institutional employees or their friends or relatives who responded directly or indirectly to notices in circulars or newsletters were enrolled between 1981 and 1986. Thirty-eight members of Health Alliance Plan, the institutional health maintenance organization (HMO), who were among 362 who volunteered for a prospective study relating baseline indices of bone turnover to rates of bone loss21 were enrolled between 1989 and 1993. The two groups did not differ significantly in height or weight, but the HMO group was about 4 years older, and the dates of the biopsy differed by approximately 9 years. All 66 subjects were skeletally healthy according to standard criteria.18, 21 Demographic data are summarized in Table 1.

Table 1. Demographic Data in Study Subjects
 Normal subjectsVertebral fracture
  • Note: All subjects were postmenopausal white women.

  • a

    Subjects recruited 1981–1986.

  • b

    Subjects recruited 1989–1993. See text for further details.

Age (years)62.5 (6.0)60.6 (6.7)64.2 (5.2)66.2 (5.9)
Height (cm)160.8 (5.9)162 (6.0)160.1 (5.8)156.1 (7.2)
Weight (kg)67.8 (8.6)67.5 (8.1)68.0 (8.0)62.4 (14.5)

In each subject, between 2 and 4 months after the most recent vertebral fracture, a cylindrical bone biopsy core was obtained22 and processed, embedded, stained, sectioned, and examined by previously reported methods.23 The internal diameter of the biopsy trephine was 7 mm, but the diameter of the core was closer to 6 mm. All biopsy cores were unfragmented with intact cortices at both ends. Each subject underwent in vivo double tetracycline labeling with an interlabel time of 14 days. The schedule was oxytetracycline 250 mg every 8 hours for 3 days, followed by an 11-day interval and then demethylchlortetracycline 150 mg every 8 hours for 3 days, finishing 4 days before the biopsy.23 To keep the latter interval constant, the biopsy invariably was scheduled before the dates of tetracycline administration were determined.

Histologic Methods

All measurements based on tetracycline labels were made on sections 5 µm thick taken from the Villanueva-stained block without further treatment; the methods of fluorescence microscopy have been described in detail previously.24 The length of the first (oxytetracycline) label is systematically shorter than the length of the second (demethylchlortetracycline) label24 and was multiplied by 1.18 before calculation of the length of the mineralizing surface (MS) as the mean of the two label lengths,25 which is referred either to the bone surface (MS/BS) or the osteoid surface (MS/OS). The mineral apposition rate (MAR) was calculated from the distance between the midpoint of the labels measured directly at multiple locations about 100 µm apart and corrected for section obliquity by multiplication by π/4.25 Adjusted apposition rate (Aj.AR) was calculated as MAR × MS/OS; in the absence of osteomalacia, Aj.AR represents the mean rate of matrix apposition averaged throughout the osteoid seam lifespan.11, 26 The surface-based bone formation rate (BFR/BS) was calculated as MS/BS × MAR. If only one label was present, the minimum value for MAR of 0.236 µm/day (0.3 × π/4) was used. If neither label was present, MAR was treated as a missing value, and MS/BS, Aj.AR, and BFR/BS were reported as 0.27 At least one label was present on at least one surface in every subject.

The lengths of osteoid surface (OS/BS) and osteoblast surface (Ob.S/BS), as defined previously,26 were measured in sections stained by a modified toluidine blue method. We have previously classified cells adjacent to osteoid seams into four types: I (pre-Ob), II (cuboidal Ob), III (intermediate Ob), and IV (flat Ob).26, 28 Only type II and III cells were designated as osteoblasts for this study, the combined total corresponding approximately to the definitions of osteoblasts by previous investigators.26 About half the osteoid seam surface is adjacent to cells lacking visible cytoplasm and indistinguishable from the cells that line quiescent surfaces or to no identifiable cell. Since individual cells were not counted, changes in Ob.S/BS could result from changes in cell number, cell size, or both. OS/BS and Ob.S/BS, together with MS/BS and BFR/BS, constitute class 1 bone-formation indices, which depend mainly on total osteoblast number and hence activation frequency (Ac.f), calculated as BFR/BS/wall thickness (W.Th).11

MS/OS, Ob.S/OS, MAR, Aj.AR, and W.Th constitute class 2 bone-formation indices, which depend mainly on osteoblast team performance.11 Osteoid thickness (O.Th) was measured directly in the same manner as label separation, including only values of 2 µm or greater.11 W.Th was measured in the same manner using the gallocyanine-stained cement line to demarcate the boundary between the surface packet and older bone.11, 29 Osteoid volume/bone volume (OV/BV) was calculated indirectly using osteoid thickness and surface measurements.25 Also measured in the same section were three indices of bone resorption: the eroded surface as a fraction of bone surface (ES/BS) and osteoclast surface as a fraction of bone surface (Oc.S/BS) and of mineralized surface (Oc.S/MdS). The rationale for the latter expression is that osteoclasts do not normally begin resorbing on osteoid.30 Because the values are much smaller, indices of bone resorption have greater sampling variation and lower precision than indices of bone formation. Since individual cells were not counted, changes in Oc.S/BS and Oc.S/NOS could result from changes in cell number, cell size, or both.

Several other indices were derived, according to the following expressions: Formation period (FP) = W.Th/Aj.AR; FP is the time taken to complete bone formation at a single location, allowing for the extent of osteoid where no mineralization is occurring1, 11; two different explanations for unlabeled osteoid will be discussed later. Formation period (FP) is the current term corresponding to Frost's definition of sigma; FP is a 2D concept, but sigma is now redefined in 3D terms.25 Active formation period (FPa) = W.Th/MAR; FPa is the time it would take to complete bone formation at a single location if matrix apposition and mineralization proceeded continuously at the average rate given by MAR.31 Osteoblast vigor (Ob.Vg) = reciprocal of FP. Ob.Vg provides the same information as FP, but the values have a lower limit of 0 instead of an upper limit of infinity.27

All these variables were measured separately on the cancellous (Cn), endocortical (Ec), and intracortical (Ct) subdivisions of the endosteal envelope, as defined previously,25 which encloses the bone marrow and its extensions, demarcated as described previously.32 Because the endocortical subdivision is much the smallest, differences are less likely to attain statistical significance. Since vertebral fracture risk depends as much on loss of cortical bone as on loss of cancellous bone, the usual exclusive attention to cancellous bone is inappropriate.33 The BFR is also expressed as percent per year in relation to bone volume (BV), tissue volume (TV), and core volume (CV) referents using the appropriate surface-to-volume ratios.25 Mean bone age (years) was calculated as the reciprocal of BFR/BV. Since the core volume referent is the same for each surface subdivision, the individual values for BFR/CV can be summed to obtain the total bone formation rate for the entire biopsy core.25 This is the most representative expression available from a biopsy of the BFR for the whole iliac bone and hence for the entire central skeleton.34

To maintain consistency of microscopic and metric methodology and histologic feature recognition over an extended period of time, we kept a set of learning slides from representative biopsies unrelated to this study that were originally measured by AR Villanueva. These were used with a three-headed teaching microscope with movable pointer to instruct each new laboratory staff member, who then remeasured the learning slide set until reasonable agreement with the original results was obtained (from within 5% to within 30% for different measurements and locations)35; if necessary new sections were cut from the original block.

Data are reported as mean (SD). For each variable, mean values for the normal and vertebral fracture groups were compared by two-tailed Student's t test.36 Subsets were compared by two-way analysis of variance (ANOVA37). Adjustment of mean values for dependence on another variable was performed by linear regression.37 Adjusted means were compared by analysis of covariance.37 Frequency distributions were compared by the Kolmogorov-Smirnov test.37 All calculations were performed using Sigmastat 2.03 (Chicago, IL, USA) or SPSS 11.5 (SPSS, Inc., Chicago, IL, USA) software.


There was no difference in any histologic variable between patients taking calcium, thyroxin, or neither and no difference between normal volunteers recruited earlier or later, so the results were pooled.

Class I formation indices in both groups are given in Table 2. The most striking difference was a substantial reduction in osteoblast surface (as a percent of bone surface) on all three subdivisions of the endosteal envelope (but of greater magnitude on the cancellous surface) and on the combined total surface. On all three surfaces, the osteoblast deficit was predominantly (Cn) or exclusively (Ec and Ct) in intermediate (type III) rather than cuboidal (type II) osteoblasts (Table 3). However, there also was a significant, although smaller, deficit in cuboidal osteoblasts on the Cn surface and small but significant deficits in preosteoblasts and flat osteoblasts on all three surfaces (Table 3). In addition, there were modest reductions in osteoid surface and BFR that were confined to the cancellous surface, with no difference in activation frequency. The difference in BFR/BS was greater in subgroups (n = 38) matched for cancellous BV/TV [11.9(11.9) versus 16.7(9.9) µm3/µm2 per year; p < .01]. In 34 patients in whom osteocyte density was measured,17 the mean value for BFR/BS predicted by the inverse relationship to osteocyte density38 was 18.6, but the observed mean value was 14.6 (p < .001). Although volume-based formation rates and their reciprocals (mean bone age) did not differ between the groups, tissue-based formation rate was reduced in cancellous bone (2.33 versus3.71 µm3/µm2 per year) mainly because of reduced BV/TV (14.9% versus 20.9%). Core-based formation rates also were reduced, mainly because of less bone; cancellous bone contributed 51% of the total bone formation in normal subjects and 44% of the total in fracture patients.

Table 2. Class I Formation Indices in 66 Normal Subjects and 78 Patients With Vertebral Fracture
Variable (unit)CnEcCtCombined
  1. Note: Data given as mean (SD). N = normal; VX = vertebral fracture; other abbreviations as specified in text.24

  2. Significance values *p < .5; **p < .01; ***p < .001.

OS/BS (%)20.3 (11.2)15.7** (11.1)24.2 (14.1)21.7 (14.0)18.5 (8.0)17.1 (10.9)20.1 (9.8)16.9* (10.3)
Ob.S/BS (%)5.82 (4.08)2.58*** (2.58)6.77 (7.08)4.30* (5.19)5.68 (3.32)4.41** (4.65)5.86 (3.53)3.39*** (3.01)
MS/BS (%)7.9 (5.1)7. 1 (6.3)12.7 (9.8)13.5 (11.0)9.3 (4.8)10.5 (8.5)8.8 (4.5)9.0 (6.5)
BFR/BS (µm3/µm2/y)15.9 (10.5)13.0* (12.3)25.9 (23.2)26.4 (26.5)23.1 (14.7)24.4 (22.1)18.7 (10.0)17.8 (13.9)
Ac.f (/y)0.416 (0.289)0.428 (0.413)0.598 (0.581)0.589 (0.587)0.509 (0.341)0.503 (0.486)0.455 (0.249)0.473 (0.363)
BFR/CV (%/y)2.56 (1.89)1.75** (1.65)0.734 (0.654)0.874 (0.826)1.69 (1.22)1.40* (1.60)4.90 (2.72)3.98* (3.01)
Table 3. Types of Cells Adjacent to Osteoid Seams in 66 Normal Subjects and 78 Patients With Vertebral Fracture
Cell tTypesCnEcCtCombined
  • Note: Data given as mean (SD). N = normal; VX = vertebral fracture; other abbreviations as specified in text.24

  • *

    p < .05;

  • **

    p < .01;

  • ***

    p < .001.

Pre - Ob (I)0.512 (0.473)0.456* (0.660)0.534 (0.742)0.400** (0.828)0.278 (0.402)0.126*** (0.313)0.438 (0.387)0.368* (0.476)
Cuboidal Ob (II)2.39 (2.02)1.61** (1.67)2.74 (3.74)2.83 (3.84)3.00 (2.40)3.80 (4.20)2.57 (1.80)2.40 (2.20)
I + II combined2.9 (2.29)2.06** (2.06)3.27 (4.06)3.23 (3.98)3.28 (2.51)3.92 (4.24)3.01 (2.02)2.77 (2.37)
Intermediate Ob (III)3.43 (3.45)0.971*** (1.56)4.03 (5.12)1.48*** (3.24)2.68 (2.56)0.611*** (1.62)3.29 (3.10)0.985*** (1.56)
Flat Ob (IV)6.43 (5.94)3.32** (2.76)8.51 (7.83)6.31* (8.38)3.66 (3.12)2.66** (3.62)5.91 (5.00)3.48** (2.95)
III + IV combined9.85 (8.98)4.29*** (3.61)12.5 (11.4)7.78** (9.29)6.34 (5.27)3.27*** (4.64)9.20 (7.73)4.46*** (3.94)
Unidentified7.55 (5.69)9.28 (8.54)8.30 (5.80)10.6 (8.28)8.86 (5.59)10.0 (7.59)7.93 (4.92)9.67 (7.41)
Total (= OS/BS)20.3 (11.2)15.6** (11.2)24.1 (14.0)21.7 (14.0)18.5 (7.96)17.2 (10.9)20.1 (9.76)16.9* (10.4)

Class II formation indices in both groups are given in Table 4. There were significant reductions in osteoblast surface (as a percent of osteoid surface) on all three subdivisions and on the combined surface. MAR was reduced by about 10% at all locations, although this was not significant on the endocortical surface. However, adjusted apposition rates did not differ between the groups at any location. Wall thickness was reduced by about 15% on the cancellous surface but did not differ on the other surfaces. Formation period and its reciprocal, osteoblast vigor, did not differ between the groups on any surface (data not shown). Active formation period was slightly shorter on the cancellous surface because of the reduced W.Th but somewhat prolonged on the endocortical and intracortical surfaces because of the reduced MAR without significant change in W.Th. Osteoid thickness was reduced by about 10% on all surfaces, but this was not significant on the endocortical surface.

Table 4. Class II Formation Indices in 66 Normal Subjects and 78 Patients With Vertebral Fracture
Variable (Unit)CnEcCtCombined
  1. Note: Data given as mean (SD). N = normal; VX = vertebral fracture; other abbreviations as specified in text.24

  2. Significance values: *p < .5; **p < .01; ***p < .001.

MS/OS (%)42.4 (22.7)44.2 (23.5)50.3 (27.8)58.0 (28.7)53.2 (23.9)58.1 (26.5)47.0 (18.7)52.3 (21.1)
Ob.S/OS (%)28.6 (13.7)18.9*** (16.2)24.9 (17.4)18.8* (19.2)30.8 (15.8)24.0* (19.9)28.7 (11.5)20.5*** (14.0)
MAR v0.531 (0.100)0.483* (0.141)0.534 (0.137)0.496 (0.157)0.651 (0.173)0.595* (0.184)0.569 (0.100)*0.514** (0.141)
Aj.AR (µm/d)0.230 (0.135)0.217 (0.133)0.281 (0.192)0.297 (0.189)0.353 (0.195)0.358 (0.219)0.271 (0.121)0.277 (0.142)
W.Th (µm)38.6 (6.4)31.0*** (6.5)45.0 (7.2)45.7 (10.2)46.9 (9.3)50.0 (11.2)41.6 (5.2)37.6*** (7.0)
FPa (d)75.9 (24.6)68.5 (21.2)89.2 (24.1)102.1* (37.3)76.8 (26.1)91.7** (30.6)75.6 (17.8)77.7 (21.8)
O.Th (µm)8.49 (2.10)7.63* (2.31)8.52 (2.69)7.92 (2.75)9.15 (2.28)8.32* (2.34)8.74 (1.79)8.06* (2.02)

There was a significant positive relationship between Ob.S/OS and Ob.Vg on the endocortical and intracortical surfaces (Table 5) but not on the cancellous surface. The regression slopes did not differ significantly, but the intercepts and adjusted mean values were significantly lower in the fracture patients than in normal subjects.

Table 5. Relationships Between Ob/OS and Ob.Vg
  • Note: Regression coefficients of Ob/OS (%) as dependent variable and Ob.Vg (%/day) as independent variable. For both Ec and Ct surfaces, the slopes did not differ significantly between the groups, but the intercepts (p < .05) and adjusted means were significantly lower in the 78 vertebral fracture (VX) patients than in the 66 normal (N) subjects.

  • a

    By analysis of covariance.

p < .001 < .001 < .001 < .001
Adj. meana25.318.230.324.4
p.008 .024 

Indices of bone resorption are given in Table 6. Eroded surface did not differ between the groups at any location. Osteoclast surface was reduced significantly by about 35% in the fracture patients at all locations and with either referent. Because total surface is smaller in the fracture patients,39 osteoclast surface/core volume was even more reduced.

Table 6. Resorption Indices in 66 Normal Subjects and in 78 Patients With Vertebral Fracture
Variable (unit)CnEcCtCombined
  1. Note: Data given as mean (SD). N = normal; VX = vertebral fracture; other abbreviations as specified in text.24

  2. Significance values: *p < .5; **p < .01; ***pi< .001.

ES/BS (%)7.56 (3.26)6.42 (4.09)9.64 (5.22)10.61 (8.37)6.41 (5.12)7.65 (6.14)7.47 (2.92)7.34 (4.09)
Oc.S/BS (%)0.938 (0.812)0.566*** (0.589)1.38 (1.67)0.89* (1.02)1.01 (0.79)0.71** (0.88)0.988 (0.65)0.646*** (0.536)
Oc.S/NOS (%)1.24 (1.13)0.72*** (0.82)1.89 (2.29)1.19* (1.46)1.29 (1.07)0.87** (1.05)1.29 (0.87)0.82*** (0.73)

A general survey of the data indicated that many variables had a wider range and larger standard deviation in fracture patients than in normal subjects. The coefficient of variation (CV, =SD/mean × 100) was significantly (p < .001) greater in the fracture patients for all class I formation variables except OS/BS, for Ob.S/OS but for no other class II formation variable, and for both osteoclast variables. The normalized frequency distribution of cancellous BFR/BS in both groups is shown in Fig. 1. Both distributions are skewed to the left, but more so in the fracture patients, and the distributions are significantly different (p = .03). The geometric means calculated after logarithmic transformation differed more strongly (7.0 versus 11.0 µm3/µm2 per year, p < .02) than the arithmetic means. Because of these observations and the previous suggestions of histologic heterogeneity,4–6 we divided both groups into three subgroups based on BFR/BS (Table 7). The differences in W.Th and O.Th on the cancellous surface were somewhat greater in the low-turnover groups and somewhat smaller in the high-turnover groups, but two-way ANOVA indicated that the differences in O.Th and MAR did not persist when adjusted for BFR/BS, but the differences in Ob.S/OS and W.Th between the normal and fracture groups were unrelated to BFR/BS (Table 8). The differences in resorption indices were in the same direction and of similar absolute magnitude in all three turnover groups but were not significant in the high-turnover group (Table 7).

Figure 1.

Normalized frequency distribution of cancellous bone-formation rates expressed as BFR/BS in normal subjects (light shading) and vertebral fracture patients (dark shading). Both distributions are skewed to the left, but more so in the fracture patients.

Table 7. Selected Cancellous Bone Histomorphometric Variables in 66 Normal Subjects and 78 Patients With Vertebral Fracture Classified According to Cancellous Bone Formation Rate
BFR groupVariableNormalVertebral fracture
  1. Note: Data given as mean (SD). Units, abbreviations, and significance values as in Tables 1 through 3.

 n (%)11 (15.7)21 (26.9)
 BFR/BS2.22 (1.67)2.25 (1.59)
Low (<5)Ob.S/OS28.3 (13.0)18.8 (20.3)
 W.Th39.5 (4.2)31.4 (6.5)***
 MAR0.400 (0.095)0.388 (0.151)
 O.Th7.40 (1.49)5.86 (1.58)*
 Oc.S/BS0.672 (0.527)0.341 (0.374)**
 n (%)48 (72.7)50 (64.1)
 BFR/BS16.1 (7.0)13.2 (5.0)*
Medium (5–30)Ob.S/OS28.5 (14.7)18.1 (13.8)***
 W.Th38.1 (6.9)30.3 (5.6)***
 MAR0.551 (0.0780.500 (0.116)**
 O.Th8.57 (2.07)8.01 (2.02)
 Oc.S/BS0.955 (0.867)0.618 (0.639)*
 n (%)7 (10.6)7 (9.0)
 BFR/BS35.5 (3.9)43.8 (14.6)
High (>30)Ob.S/OS29.3 (7.57)25.4 (19.3)
 W.Th39.4 (4.6)35.0 (11.1)
 MAR0.660 (0.076)0.626 (0.122)
 O.Th9.16 (2.58)10.29 (2.48)
 Oc.S/BS1.244 (0.750)0.932 (0.527)
Table 8. Two-Way Analysis of Variance Concerning Table 7
VariableDiagnostic group (normal versus fracture)BFR group (low, medium, high)Interaction
  • a

    MAR is a component of BFR/BS.

W.Th < 0.001NSNS
MARNS < 0.001aNS


We have previously reported important differences in several histologic indices among the three subdivisions of the endosteal envelope in healthy women.1, 40 BFR/BS, Ac.f, and W.Th all were significantly higher on the endocortical (Ec) and intracortical (Ct) surfaces than on the cancellous (Cn) surface,1 and MS/OS and Aj.AR were both higher on the Ct surface than on the Ec or Cn surfaces.40 We now report for the first time corresponding differences in patients with vertebral fracture, but we focus mainly on the differences in the comparison between fracture patients and normal controls among the three surfaces. Several patterns were evident (Table 5). For MS/OS, MS/BS, Ac.f, and Aj.AR, there were no differences between the groups on any surface, in contrast to the findings in the rib mentioned previously.14 For O.Th, MAR, Oc.S/BS, and Oc.S/NOS, there were differences of similar magnitude on all three surfaces, and for Ob.S/BS and Ob.S/OS, there were differences on all surfaces, but the magnitude was substantially greater on the Cn surface. Furthermore, there was no significant relationship between Ob.S/OS and Ob.Vg on the Cn surface but a highly significant relationship on the Ec and Ct surfaces (Tables 2 through 5). For FPa, there were differences on the Ec and Ct surfaces but not on the Cn surface. Finally for OS/BS, BFR/BS, and W.Th, there were differences on the Cn surface but no differences on the other two surfaces.

In the only previous studies in postmenopausal women in which more than one surface was measured, several indices of resorption and formation were higher on the Ec than on the Cn surface in both normal subjects41, 42 and vertebral fracture patients,41 but the only significant difference between the groups was a shorter FPa on the Cn surface and a greater osteoclast number on the Ec surface in the vertebral fracture group.41 However, in premenopausal women with idiopathic osteoporosis, there were significant reductions in W.Th and BFR/BS on the Cn and Ec surfaces but not on the Ct surface,43 and in idiopathic juvenile osteoporosis there was much lower BFR/BS on the Cn surface but not on the Ct surface.44 Why surfaces that are in continuity should behave so differently is unknown, but a complete understanding of vertebral fracture pathogenesis must take this phenomenon into account.

In both groups and on all surfaces, about half the osteoid surface lacked a tetracycline label (Table 4). Some osteoid seams could have been initiated in the 4-day interval between second-label administration and biopsy, but this would account for only a small proportion of unlabeled surface. Some investigators have used the existence of unlabeled osteoid as evidence that osteoblast function is intermittent and from its extent calculate so-called off time or down time.31 We think that this is very unlikely for several reasons. First, in subjects given three labels of different color, in no case was the second label missing with the first and third labels present.45 Second, the proportion of osteoid labeled is directly related to the dose of tetracycline given.46 Third, almost all unlabeled osteoid is adjacent to type IV or unidentified cells and so occurs late in the osteoblast life history.26, 27 We believe it much more likely that when matrix and mineral apposition have become very slow, too little mineral is deposited to prevent escape of tetracycline as the blood level falls to zero.40 In this circumstance, the extent of tetracycline fixation and the indices calculated from it, such as BFR/BS, will be systematically underestimated on average by about 10%.40

In the vertebral fracture patients, the extent of osteoblast surface (whether expressed as Ob.S/BS or as Ob.S/OS) was significantly reduced not only on the cancellous surface, as we have reported previously,11 but also on the endocortical and intracortical surfaces, albeit to a lesser extent (Tables 2 through 5). By contrast, in normal postmenopausal women, Ob.S/BS was significantly higher and Ob.S/OS essentially the same as in premenopausal women1, 35 (and unpublished data). This finding must be related to the normal osteoblast life history.26 When osteoblasts assemble on the cement surface, they are tall and columnar in shape, and their secretory territory47 is small. As they move further from the cement surface, they become progressively flatter in shape, their secretory territory expands, and eventually, they become indistinguishable from lining cells. At the same time, osteoblast function, expressed as MAR and Aj.AR, declines progressively.26 To indicate this morphologic and functional progression, we have classified cells adjacent to osteoid seams into four types, as described previously.26, 28 The demarcation between types is somewhat arbitrary, and it is more accurate, although more time-consuming, to express the evolution as nuclear height, which progressively declines with increasing distance from the cement surface.48

An osteoblast deficit must be due either to reduced birth rate or reduced lifespan or some combination thereof. Because preosteoblasts may be difficult to see,26 we believe that the sum of types I and II is a more accurate reflection of osteoblast birth rate and was reduced significantly only on the Cn surface (Table 3). We have discussed previously the importance of defective osteoblast recruitment in the pathogenesis of vertebral fracture,11 a notion that subsequently received substantial experimental support.49 However, this cannot be the whole story. The deficit in type III and IV cells was substantially greater than the deficit in type I and II cells on the Cn surface and was the only deficit on the Ec and Ct surfaces (Table 3). Since the extent of mineralizing surface (MS/BS and MS/OS) did not differ between the groups, it is evident that a higher proportion of cells classified as type IV (flat) still were able to make bone in the vertebral fracture patients than in normal control individuals. Even the lining-like cells lacking visible cytoplasm may show two tetracycline labels in vertebral fracture patients.50

Another way of expressing this conclusion is that the spreading and flattening that indicate the transition from type III to type IV osteoblasts occur earlier in vertebral fracture patients than in normal subjects. From the beginning and throughout its life history, each osteoblast has to “choose” between three options: death by apoptosis,51 transformation to an osteocyte,26 or continuation of matrix synthesis. The first two options both contribute to progressive reduction in the number of surface cells, which inevitably leads to the morphologic changes described previously in the cells that remain. If apoptosis occurred earlier in the osteoblast life history in vertebral fracture patients than in controls, this would explain the earlier transition from type III to type IV and, when combined with defective osteoblast recruitment, would explain the reduction in wall thickness and BFR (Table 3) (because there will be fewer osteoblasts initially and their average lifespan would be shorter) and the reduction in osteocyte density,17 because fewer cells would be available to undergo the transition. W.Th was not reduced on the Ec and Ct surfaces, but there was a lesser reduction in Ob.S/OS (44% on Cn, 24% on Ec, and 22% on Ct). It may be that both reduced birth rate and shorter lifespan are required to substantially impair osteoblast function.

In only a few previous studies was tetracycline-based BFR measured in patients with postmenopausal osteoporosis and in appropriate normal control subjects, but the measurements were restricted to the Cn surface. The results have shown either no significant difference between groups31 or a higher mean value in the osteoporotic group.52 By contrast, we found a significantly lower mean value (Table 1); the difference was larger and more significant in groups matched for BV/TV when adjusted for the inverse regression on osteocyte density and after logarithmic transformation. Furthermore, the frequency distribution was more skewed to the left (Fig. 1), the CV was significantly greater, and there were more very low and very high values (Tables 2, 4, and 5 and Fig. 1). It is useful to recall that BFR/BS (µm3/µm2 per year) = W.Th (µm) × Ac.f (years). There was no difference in Ac.f, so the difference in mean BFR/BS was entirely due to the difference in W.Th discussed previously.

Although the data are incomplete, it seems unlikely that the different results can be explained by differences in the severity of estrogen deficiency, in vitamin D nutrition, or in parathyroid hormone secretion between Detroit, MI, Rochester, MN, and Omaha, NE. A possible explanation for our results is that we used more stringent radiographic criteria to identify vertebral fragility so that some of our patients may have differed in some other respect from women with low bone density but no fracture. In the Omaha study, the mean number of vertebral fractures was fewer than in our study (2.7 versus 4.0), but the mean height loss was greater (6.0 versus 4.7 cm), so it is unlikely that the severity of vertebral deformity was much different.31 The difference in W.Th was greater in our study (7.6 versus 4.1) but smaller than in the Rochester study (7.6 versus 8.1).52 A substantial reduction in W.Th was found in all three studies, so regional differences in BFR/BS are due mainly to regional differences in Ac.f, for which we are unable to suggest an explanation.

Cessation of bone remodeling, as in patients with radiation necrosis of bone or in dogs given a very high dose of a bisphosphonate, leads to spontaneous long bone fractures.53, 54 But it is still unclear to what extent low cancellous bone turnover contributes to vertebral fragility—the senior author has equivocated on this question for more than 20 years.54 Increased bone age could lead to accumulation of fatigue microdamage in several ways.54 Mean bone age calculated as the reciprocal of BFR/BV did not differ between the groups, but when distance from the surface is taken into account, the proportion of bone more than 20 years old and the proportion of subjects in whom such bone constituted more than 10% of the total were both higher in the vertebral fracture group.55 Increased susceptibility to fatigue microdamage has not been demonstrated, but because this phenomenon is more difficult to study in cancellous than in cortical bone, neither has it been ruled out.54 Some vertebral fracture patients have abnormally high levels of bone mineralization, which would increase brittleness and reduce fracture toughness, but there was no relationship to BFR/BS.56 Osteocyte death becomes more likely with increasing bone age,54 but it is not known whether the osteocyte deficiency in vertebral fracture patients17 is related to the greater extent of very old bone.

The relationship between bone fragility and bone turnover is U-shaped—both very low and very high values increase fracture risk.57, 58 The adverse effects of unnecessarily high bone turnover are mediated in cortical bone by increased cortical porosity59 and by confluence of intracortical cutting cones.60 In cancellous bone, each resorption cavity acts as a stress concentrator, and when horizontal trabeculae have been removed, each focal weak point increases the risk of buckling.54 None of the usual causes of high bone turnover were present in our study subjects, but some patients with vertebral fracture appear to have increased sensitivity to the effects of parathyroid hormone on osteoclast recruitment.61 It is also possible that in some patients the effects of estrogen deficiency on cytokine production62 or on reactive oxygen species63, 64 are potentiated by some unknown factor. Whatever the explanation, it seems reasonable to regard vertebral fracture patients with abnormally high cancellous bone turnover as a pathophysiologically distinct subset.38

The most puzzling finding was that in the fracture group, osteoclasts were substantially either fewer or smaller or both than in the normal subjects, even though bone-resorption rates could not have been very different. A similar finding has been reported previously, but without comment.31, 65 Osteoclasts are multinucleated cells and are usually identified in histologic sections as cell profiles enclosing more than one nuclear profile. Osteoclast cytoplasm may be tinctorially distinctive, but absent a stain specific for acid phosphatase, cell profiles with no or only one nuclear profile cannot be identified confidently as osteoclasts. According to Eriksen's kinetic model of the resorptive site in normal subjects,66 multinucleated osteoclasts are responsible only for the first 8 days of excavation, which is completed over the next 34 days by mononuclear cells that presumably are disaggregated osteoclasts but which we would not have recognized. Our findings suggest that in the fracture patients, the transition from multinuclear to mononuclear resorbing cells occurred substantially earlier than in normal subjects; if so, this abnormality would be present on each subdivision of the endosteal envelope.

In conclusion, our data enable us to reaffirm the histologic heterogeneity among postmenopausal women with vertebral fracture first noted in rib biopsies by Frost and colleagues in 196613 and later confirmed in iliac biopsies by research groups in Lyon,4 St Louis,5 and Augusta6 but generally ignored by those who diagnose osteoporosis solely by bone densitometry or investigate its pathogenesis solely by studies in animal models. But the situation is more complex than simply recognizing subgroups with abnormally low or abnormally high bone turnover, important as these are. We have identified several abnormalities of osteoblast function that can be accounted for by diminished recruitment and premature death of these cells and an abnormality of osteoclast function that may be accounted for by their premature disaggregation into mononuclear cells. Furthermore, several of the abnormalities were present on every subdivision of the endosteal envelope, not just the cancellous surface. Together with our previous finding of osteocyte deficiency in interstitial bone formed many years earlier,17 our data demonstrate that in some patients vertebral fragility cannot be explained solely by the effects of estrogen deficiency, aging, or other established risk factors but is accompanied by one or more disorders of bone remodeling of unknown etiology that at present cannot be identified by noninvasive means.


All the authors state that they have no conflicts of interest.


We thank Michael Kleerekoper for insisting on the distinction between bone density and bone fragility; AR Villanueva for establishing the laboratory and devising most of the methods; J Stanciu, MS Shih, J Foldes, ZH Han, and AR Villanueva for their contributions to the database used in the study; and Paula Roberson for assistance with statistical analysis.