Independent and Combined Contributions of Cancellous and Cortical Bone Deficits to Vertebral Fracture Risk in Postmenopausal Women


  • The authors state that they have no conflicts of interest.


Using iliac bone histomorphometry on 78 patients with vertebral fracture and 66 healthy postmenopausal women, cortical thickness discriminated at least as well as any cancellous bone structural index between the two groups. Subjects with a deficit in both cortical and cancellous bone had much greater likelihood of fracture.

Introduction: Vertebral fracture is often attributed to disproportional loss of cancellous bone, but fracture patients may have deficits in cortical and cancellous bone. Accordingly, we examined the contribution of cortical and cancellous bone deficits, separately and together, to the likelihood of vertebral fracture.

Materials and Methods: Iliac bone histomorphometry was performed in 78 white woman with clinically apparent vertebral fracture, 66 healthy postmenopausal women, and 38 healthy premenopausal women. We measured cancellous bone volume (Cn.BV/TV), trabecular number (Tb.N), trabecular thickness (Tb.Th), cortical bone volume (Ct.BV/TV), and cortical thickness (Ct.Th). For each variable, a value of >1 SD below the mean in premenopausal women was treated as a putative risk factor, and its association with the presence or absence of fracture was determined by OR calculated by logistic regression and by receiver operating characteristic (ROC) curve analysis. Subsets of fracture and control subjects were separately matched for Cn.BV/TV and Ct.Th.

Results: All structural indices differed between fracture patients and controls except Ct.BV/TV. There was a weak but highly significant correlation between Cn.BV/TV and Ct.Th in the entire group (r = 0.389, r2 = 0.151 p < 0.001). Many control subjects had a high value for one of these variables and a low value for the other. Ct.Th., Cn.BV/TV, and Tb.N were all significantly associated with vertebral fracture (ORs, 4.4–5.8; ROC area under the curve [AUC], 0.74–0.85). In subjects matched for Cn.BV/TV, Ct.Th was reduced by 29% (OR, 5.0), and in subjects matched for Ct.Th, Cn.BV/TV was reduced by 27% (OR, 5.0). In patients with deficits in both cortical and cancellous bone, the ORs (28–35) were much higher.

Conclusions: Deficits in cortical bone (reduced value for Ct.Th) and in cancellous bone (reduced values for Cn.BV/TV or Tb.N) were equally effective in discriminating between subjects with and without vertebral fracture. With a deficit in both cortical and cancellous bone, the association with vertebral fracture was much stronger. Vertebral fracture is not the result of disproportionate loss of cancellous bone in the patients as a whole, although individual patients may have relatively greater deficits in either cancellous or cortical bone.


Osteoporosis has been defined as a systemic skeletal disease characterized by low bone mass and deterioration of bone microarchitecture, with a consequent increase in bone fragility and fracture.(1) Low BMD is regarded as a critical factor responsible for fragility fractures, but not all subjects with fragility fractures have low BMD.(2,3) The likelihood of such fractures depends also on architectural deterioration of cancellous bone(4,5) and on a variety of other nonmass factors.(6)

Vertebral fracture is often stated to be the result of disproportionate loss of cancellous bone relative to cortical bone,(7) although dissent from this view has been expressed.(8) Despite assertions to the contrary,(9) cortical bone contributes substantially to vertebral compressive strength,(10,11) and patients with vertebral fracture have a substantial deficit in cortical bone as well as cancellous bone, both in the vertebrae, determined by several noninvasive methods,(12,13) and in the ilium, determined by biopsy.(5,14) Nevertheless, the independent and combined contributions of these deficits to vertebral fracture risk have not been determined, which was the object of this study.


Patients with osteoporotic fractures fulfilled the entry criteria for a controlled trial of sodium fluoride treatment.(15) They were white women, 45–75 years of age, were at least 1 year postmenopausal, and had at least one compression fracture (reduced posterior as well as anterior height) or at least two non-contiguous wedge fractures (reduced anterior height only)‥ The abnormal vertebral body shape was readily apparent without measurement on a lateral spine radiograph, taken for evaluation of back pain that was spontaneous or after no more than trivial trauma. These criteria served to establish the existence of significant vertebral bone fragility and to rule out other causes of vertebral body deformation(16,17); the mean number of vertebral fractures was 4.3 ± 2.5 (SD). No patient had previously received sodium fluoride or was currently taking estrogen or any other treatment. Bone loss from causes other than age or menopause was excluded. Among 663 potential candidates who were evaluated, 84 were enrolled in the study. Of these, 78 (93%) underwent transiliac bone biopsy before randomization between August 1981 and December 1987.(15)

Healthy control subjects were recruited in one of two ways. Between 1980 and 1992, institutional employees and their friends and relatives responded to notices in circulars and newsletters. Between 1990 and 1993, members of Health Alliance Plan, the institutional HMO, who volunteered for a prospective study relating baseline indices of bone turnover to rate of bone loss, were approached.(18) Individuals in both groups volunteered for a study of normal bone histology using transiliac biopsy, which was approved by the Institutional Review Board for studies on human subjects at Henry Ford Hospital. All subjects were skeletally healthy according to standard criteria. Of the 109 volunteers, 43 were premenopausal and 66 were postmenopausal white women. For easy description, the postmenopausal women with and without fracture were termed fracture and control subjects.

In each subject, a cylindrical bone biopsy core was obtained using a trephine with internal diameter 7.5 mm and processed, embedded, stained, sectioned, and examined by previously reported methods. Both cortices were intact in all biopsies. We measured cancellous bone volume (Cn.BV/TV), trabecular thickness (Tb.Th), trabecular number (Tb.N), cortical bone volume (Ct.BV/TV; the converse of porosity), and cortical thickness (Ct.Th) as previously described.(14,18) Cortical and cancellous bone were demarcated according to published criteria.(19)

One-way ANOVA was used to compare each variable among the three groups (premenopausal women, postmenopausal women with vertebral fracture, and postmenopausal women without vertebral fracture). The relationships between Ct.Th and the three cancellous variables were examined by simple regression. The value of >1 SD below the mean from premenopausal women was used as the cut-off for each variable, for determining its relationship as a risk factor for the presence or absence of vertebral fracture. ORs were calculated by logistic regression analysis(20,21) for each variable, adjusted for age. Interaction between Ct.Th and Cn.BV/TV was calculated because they are major determinants of vertebral bone strength.(5,14) The ability of the measurements to discriminate between fracture patients and normal subjects was also examined by construction of receiver operating characteristic (ROC) curves,(22) using the MEDCALC program (MedCalc Software, Mariakerke, Belgium). Subsets of fracture and control subjects were matched separately for Cn.BV/TV and Ct.Th by pairing individual values within <2% and the same analysis performed as on the complete groups.


Mean values of the structural indices and the −1 SD cut-off values are given in Table 1. The postmenopausal women with fracture were significantly older (and as expected, shorter) than those without fracture, but they did not differ in body weight or BMI. All values except Ct.BV/TV were significantly lower in the fracture patients than in the controls, and all values except Tb.Th were significantly lower in post than in premenopausal women. There were no significant differences between the fracture and normal subjects in slope or intercept for the regressions, so the data were pooled. There was a modest but highly significant correlation between Cn.BV/TV and Ct.Th. (Fig. 1) and lesser but still significant correlations between Tb.Th. and Tb.N with Ct.Th. (data not shown). In all three instances, there was substantial overlap between the fracture and control subjects.

Table Table 1.. Comparison of Variables Among Women With Vertebral Fracture and Normal Premenopausal and Postmenopausal Women
original image
Figure Figure 1.

Relationship between principal indices of cancellous and cortical bone structure. •, premenopausal women; ○, postmenopausal women; Δ, fracture subjects. Interrupted lines denote cut-off values for data in Table 1. Solid line is regression line: Cn.BV/TV = 11.2 + 0.0073 × Ct.Th, p < 0.001, r = 0.389, r2 = 0.151.

The ORs showed highly significant associations with vertebral fracture for Cn.BV/TV, Tb.N, and Ct.Th (Table 2). There was significant interaction between Cn.BV/TV and Ct.Th (p < 0.05). The ROC curves showed significant discrimination between fracture cases and controls for the same three variables (Fig. 2). The area under the curve was greatest for Ct.Th. (Table 3) but was not significantly different from that for Cn.BV/TV and Tb.N. None of the analyses were appreciably affected using cut-off values of −0.5 or −1.5 SD instead of −1.0 SD (data not shown).

Table Table 2.. Proportion of Groups Below Cut-Off Values for Each Variable
original image
Table Table 3.. Results of ROC Analysis
original image
Figure Figure 2.

ROC curves for discrimination between subjects with and without vertebral fracture, based on different structural indices as indicated. Numerical values in Table 3. The three largest areas were all significantly different from the smallest one but not significantly different from each other.(20)

In the subsets matched for Cn.BV/TV, Ct.Th was 29% lower in the subjects with than in those without fracture (Table 4). In the subsets matched for Ct.Th, there were significant decreases in Cn.BV/TV and Tb.N in the fracture subjects (Table 5). The ORs were 5.04 and 4.99, respectively (Table 6).

Table Table 4.. Comparison of Variables Between Fracture and Control Subjects Matched for Cn.BV/TV
original image
Table Table 5.. Comparison of Variables Between Fracture and Control Subjects Matched for Ct.Th
original image
Table Table 6.. Association With Vertebral Fracture Estimated for Matched Variables
original image

Each subject was assigned to one of four groups according to the structural changes in cancellous and cortical bone (Table 7). With deficits of both cortical and cancellous bone, the likelihood of fracture was much higher than with deficits in either cortical or cancellous bone alone. About 50% of the fracture patients had a combined bone deficit, ∼15% had a deficit in cortical but not cancellous bone, ∼25% had a deficit in cancellous but not cortical bone, and ∼10% had normal values for both cortical and cancellous bone.

Table Table 7.. Association With Vertebral Fracture for Cancellous and Cortical Bone Deficits Separately or Together
original image


Unlike the vertebral bodies, the iliac crest is not load bearing, and its trabeculae are more isotropic,(23,24) but vertebral biopsy is impracticable, and iliac biopsy is the only available surrogate. Fortunately, the ilium has several important features in common with the vertebral bodies, making it a useful site for studying vertebral fracture pathogenesis. In both sites, cancellous bone tissue and hematopoietic marrow fill the space enclosed by the cortex, differing in this respect from the long bone metaphyses.(25) In both, a substantial proportion of cancellous bone is derived from cortical bone by subendocortical tunneling and so was originally made at the periosteum.(26) In contrast, all metaphyseal cancellous bone is made by endocortical ossification. At both sites, there is hematopoietic marrow so that both blood flow and bone turnover are much higher than in the metaphyses.(27) Although there are differences in the absolute values of bone structural indices (trabeculae are both thinner and fewer in the vertebrae than in the ilium) and bone turnover, which is generally lower in the vertebrae based on static histologic data, there is reasonable correlation between the two sites, and they show similar proportional changes with aging.(23,24)

In traditional medical terms, a fracture is a painful and often disabling event in a patient's life, but to facilitate recruitment for clinical trials(28) a vertebral fracture is now usually defined in the osteoporosis literature as a radiographically detectable deformity without regard to pain or other symptoms(29) and is often discovered fortuitously. Such a deformity may be a reasonable surrogate for a genuine fracture,(17) but the widespread assumption that their pathogenesis is identical(29,30) has never been verified. Patients with asymptomatic vertebral deformation do not differ in several indices of bone remodeling from healthy subjects,(31) and they often have higher than normal bone turnover.(32) However, patients with clinically evident vertebral fracture tend to have lower than normal bone turnover and several abnormalities of osteoblast function.(33,34) All the patients in this study had genuine clinically significant fractures, so that any conclusions we draw concerning their structural basis do not necessarily apply to asymptomatic deformations.

There was a highly significant correlation between the indices of cortical and cancellous bone structure (Fig. 1), presumably because similar factors influence the growth and decay of the skeleton at different locations, but the magnitude of the correlation was remarkably small, variation at one site accounting for only 15% of the variation at the other site (r2 = 0.151). In normal subjects, at the same value for Ct.Th., values for Cn.BV/TV varied over almost a 4-fold range, and at the same value for Cn.BV/TV, Ct.Th varied over almost a 4-fold range. Consequently, some healthy women have a high value at one site and a low value at the other in either direction. Evidently there are differences and similarities in the regulation of cortical and cancellous bone growth. In the fracture patients, the relative variability was even greater (5- and 6-fold), but the absolute variability was smaller. Division of the fracture patients into four groups, although useful for exposition of the results (Table 7), is arbitrary, because the data show no tendency to cluster. Age-related bone loss displays significant long-term tracking,(35,36) so it seems likely that differences between the severity of cortical and cancellous bone deficits mainly reflect differences established by the time of skeletal maturity.

ORs are used in cross-sectional studies to provide similar information as relative risk values in longitudinal studies.(20) The ORs in Tables 2, 6, and 7 are the likelihood of having a low value for some measurement in the fracture group divided by the likelihood of a low value in the control group. These ratios are valid indices of the association of a putative risk factor (such as low Ct.Th) with the event of vertebral fracture within the study population, but this population was chosen to have approximately the same number of subjects with and without fracture. Obviously in a representative sample of the whole population there will be many more patients without than with fracture. It would be inaccurate to extrapolate our results in a small nonrandom sample to the entire population, so we have refrained from referring to the results as indicating individual risk. However, the prospective study needed to determine the individual fracture risk conferred by particular histologic findings will never be performed.

The most important implication of our results is that a deficit in cortical bone makes just as great a contribution as a deficit in cancellous bone to increased vertebral fragility, manifested as fracture. Reduced cortical thickness is at least as good as any index of cancellous bone at discriminating between subjects with and without fracture (Fig. 2; Table 3). If either cortical or cancellous bone mass is within the age-related reference range, a substantial deficit in the other is needed to increase fracture risk (Tables 4–6). If both cortical and cancellous bone mass are reduced, the effect on fracture risk is multiplicative or even synergistic (Table 7). The absolute loss of iliac cortical bone in patients with vertebral fracture is more than twice as great as the absolute loss of cancellous bone,(14) and patients with vertebral fracture have substantial deficits in cortical bone in the peripheral skeleton(8,14) so that the absolute deficit in cortical bone mass in patients with vertebral fracture is much greater than the absolute deficit in cancellous bone mass. Furthermore, loss of both cancellous and cortical bone is initiated by increased depth of resorption.(37) It is no longer reasonable to regard vertebral fracture as a manifestation of selective or disproportionate loss of cancellous bone. Other factors that preferentially affect cancellous bone, such as differences in mineral density distribution(38) or deficiency of osteocytes,(34) may determine the site of the first osteoporotic fracture.


This work was supported by grants from the National Institutes of Health to The Henry Ford Hospital, Detroit, MI, (DK43858, R03AR 48641), and to the University of Arkansas for Medical Sciences, Little Rock, AK (P01AG13198).