Portal de Periódicos da CAPES

Bone morphology in response to alendronate as seen by high-resolution computed tomography: Through a glass darkly

2010; Oxford University Press; Volume: 25; Issue: 12 Linguagem: Inglês

10.1002/jbmr.261

1523-4681

Ego Seeman,

Bone health and osteoporosis research

1 Corinthians 13:12 All genetic and environmental factors, diseases, and therapeutic agents influencing the material composition and structure of bone do so via a final common pathway: bone modeling and remodeling.1 This cellular machinery fashions and refashions bone throughout life by adding and removing bone matrix on bone's outer (periosteal) envelope and the three (intracortical, endocortical, and trabecular) components of its inner (endosteal) envelope, altering bone's external size and shape and the proportions of mineralized matrix volume and void volume within the periosteal envelope.2 Beneath the periosteal envelope, the mineralized matrix is more compactly assembled; there is more matrix than void volume (cortical bone). The medullary canal contains more void than mineralized matrix volume; the latter is assembled as thin sheets or plates forming cancellous or spongy trabecular bone. The heterogeneity in material composition of the mineralized matrix and the way it is distributed with void space confers strength—its ability to resist crack initiation and propagation during loading, and lightness—which allows the host to move against gravity. While it has been obvious for several centuries that bone's material composition and structure determine its strength, quantifying bone's material composition in vivo using noninvasive methods is still not feasible. However, quantifying bone's microstructure noninvasively—its cortical area and the pores within it (haversian canals, resorption cavities in longitudinal section), trabecular number, thickness, and connectedness—has become a reality recently.3-5 These measurements discriminate between persons with and without fractures when areal bone mineral density (aBMD) does not differ.6 Although the assessment of microstructure is an important advance, there are limitations that may result in misleading inferences concerning changes in bone morphology during growth, aging, and drug therapy. Burghardt and colleagues present the results of a pilot study of 53 early postmenopausal women with low aBMD randomized to alendronate or placebo for 2 years.7 The investigators report (1) treatment increased distal tibial cortical area and thickness and decreased medullary area relative to baseline, suggesting that periosteal and/or endocortical bone formation occurred, (2) treatment did not increase cortical volumetric bone mineral density (vBMD) relative to baseline, suggesting that there was no reduction in cortical porosity and/or no increase in the degree of matrix mineralization despite suppression of bone remodeling, and (3) trabecular vBMD remained unchanged globally, suggesting that decay of trabecular morphology was prevented. Regionally, in some quadrants, trabecular vBMD increased, purportedly owing to an increase in trabecular number. (4) Bone strength at the radius, estimated using micro–finite-element analysis, decreased in both groups, perhaps less so in the alendronate group, or was maintained in the treated group at the tibia. Correlations with strength and stiffness were observed with regional trabecular, not cortical, morphology. How should the results be interpreted given our understanding of the mechanism of action of antiresorptive agents and methodologic issues concerning the assessment of structure based on imaging produced by X-ray photon attenuation and strength derived from finite-element modeling? Bisphosphonates are not anabolic. There is no evidence that antiresorptives result in deposition of mineralized bone on the periosteal envelope, enlarging its total (external) cross-sectional area or the area of its cortex.8 Nor is there evidence that antiresorptives deposit bone on the endocortical surface, thickening the cortex and reducing medullary area. While the antiresorptives reduce remodeling on the trabecular surfaces and so prevent or minimize trabecular thinning and perforation, there is no evidence based on dynamic histomorphometry for increases in trabecular number, thickness, or connectedness. Four events are likely account for the observations reported in this study. The first is the progression to completion of remodelling cycles initiated before alendronate was started. This is the only event that increases the area of a cross section of the mineralized cortical bone matrix (Fig. 1). Before treatment, the number of resorption sites excavating bone on its three internal surfaces equals the number completing bone formation.9 When an antiresorptive is administered, this steady-state remodeling is perturbed. The number of new basic multicellular units (BMUs) appearing is reduced by about 50% to 60%. Concurrenty, the many BMUs initiated prior to treatment complete remodeling with bone formation refilling the excavated sites almost completely with osteoid, which undergoes primary and then slower secondary mineralization. Consequently, the diameter of intracortical pores (in a cross section, canals in longitudinal section) is reduced. The total cortical area (matrix area plus pore area) is unchanged, but the proportion that is matrix increases and the proportion that is void decreases. This increase in cortical matrix area occurs "from within" cortex with no change in periosteal or endocortical perimeters. The refilling on the intracortical and endocortical surfaces is incomplete because the negative BMU balance is not abolished by bisphosphonates.10 BMU balance is not made positive, so endocortical perimeter cannot decrease. (It is not possible to deposit more bone into an intracortical canal than was removed.) This is not an anabolic effect of treatment; it is the completion of the sequence of bone formation "coupled with" bone resorption by BMUs initiated before alendronate was started. In placebo-treated patients, remodeling is in steady state. Resorption sites fill incompletely with new bone (gray) that undergoes primary and slower secondary mineralization (darker gray). Incomplete refilling leaves pores. The same number of new resorption sites are created concurrently. Net volumetric bone mineral density (vBMD) decreases because of accumulating porosity and a reduction in tissue mineralization density (older, more mineralized osteons are removed and replaced with younger, less densely mineralized osteons). Trabeculae thin or disappear, and intracortical and endocortical remodeling thins the cortex. Antiresorptive treatment perturbs the steady state temporarily. Net cortical vBMD increases because the large number resorption sites excavated before treatment fill incompletely with new bone, reducing porosity, but half the number of newly excavated sites appear. Tissue mineralization density increases in the newly completed osteons and existing osteons that would have been removed had high remodeling continued. When steady state is restored at the new, lower remodeling rate, decay continues from the higher vBMD if the negative balance between the volumes of bone resorbed and formed persists. The tempo of the slowed remodeling rate and decay depends on the potency of the drug in suppressing remodeling and the baseline remodeling rate in the patient. Second is the progression of secondary mineralization of matrix deposited during these currently active remodeling cycles and of preexisting osteons, the fossilized footprints of previously completed remodeling cycles. As remodeling is suppressed, existing osteons are no longer rapidly remodeled and replaced with younger, less densely mineralized bone. They undergo continued secondary mineralization (which can require several years before their matrix is fully mineralized).11 There is no change in the area of the cortex that is mineralized matrix during secondary mineralization because enlargement of crystals of calcium hydroxyapatite deposited within collagen fibrils and within extrafibrillary matrix of a collagen fiber during primary mineralization displaces water and does not increase fiber or fibril volume, so there is only an increase in the proportion of the matrix volume that is mineral.12 There is neither an increase in the area of cortical matrix nor a decrease in the proportion of the cortex that is void volume. Third is the concurrent inhibition of the appearance of new resorption cavities. This does not increase or even maintain mineralized matrix area; it decreases it because the birth rate of the BMUs is reduced, not abolished. Each of the fewer BMUs continues to erode bone. Remodeling is inhibited by 50% to 60% using alendronate. When steady state is restored at the new, slower remodeling rate after about 12 to 18 months, continued remodeling erodes the benefit derived when steady state was perturbed. There is a decline from the higher cortical vBMD because slow remodeling increases porosity by removing matrix and reduces tissue mineralization density as older osteons are replaced by younger, less densely mineralized osteons. This continued structural decay may explain the third observation reported by the investigators: Cortical vBMD was higher relative to placebo-treated patients (in whom vBMD continued to decrease as the high remodeling increased intracortical porosity and perhaps decreased tissue mineral density) but was not increased relative to baseline in the alendronate group. This continued decay is not "seen" using bone densitometry. Areal bone mineral density increases for years because of continued secondary mineralization,13 but this occurs in a progressively slowly diminishing bone matrix volume. The continued secondary mineralization of the large mineralized matrix volume not currently being remodeled increases areal BMD overwhelmingly and so obscures decay because this larger and increasingly mineralized matrix volume attenuates many more photons than the increase in photons transmitted by the slow loss of matrix with its mineral. Remodeling suppressants such as denosumab or zoledronate reduce remodeling by 80%, at least initially, and result in the appearance of few new remodeling sites, so continued structure decay during treatment is minimal because there are so few new BMUs initiated to erode bone.14, 15 Drugs that suppress remodeling modestly, such as calcium supplements and raloxifene, allow most of the remodeling present before treatment to resume after the transient perturbation of remodeling. The decline in areal BMD during continued treatment can be seen graphically in the third and subsequent years after steady state is restored at a remodeling rate that is only modestly slower than prior therapy.16, 17 If baseline remodeling rate is low, it makes sense to use less potent remodeling suppressants, lower doses of more potent ones, or perhaps to review the need for treatment at all. It remains to be determined whether there is a net benefit in bone strength achieved in preventing structural decay by suppressing remodeling at the price of reduced material strength produced by reduced microdamage removal, an increase in the proportion of the mineralized matrix that becomes fully mineralized, and so more homogeneously mineralized (allowing crack propagation), while collagen cross-linking by advanced glycation products also predisposes to reduced resistance to crack initiation and propagation.18, 19 Little is known about targeting treatment to the material, to structural properties of a bone in an individual, or to the mechanisms producing fragility in that individual; contemporary therapeutics ignores individuals and trait variances. We treat all patients, with or without fractures, essentially the same way, ignoring the heterogeneity in bone's material composition, structure, remodeling intensity, and BMU imbalance from person to person—a heterogeneity that is more than likely to be important deciding when to treat, what drug to choose, and how to monitor outcomes that may define treatment success or failure.20-23 Fourth is the concurrent reduction in endocortical resorption, preventing cortical thinning or enlargement of the medullary area. While endocortical resorption contributes to cortical thinning, the main cause is likely to be the vigorous intracortical remodeling in cortex adjacent to the medullary canal, leaving cortical remnants that look like trabeculae.24 Almost complete refilling or partial filling of these existing cavities and the reduced (not abolished) appearance of new porosity in this location reduce the rate of cortical thinning from the "inside" and gives the false impression of cortical thickening by endocortical apposition. Actual reversal of thinning by deposition of bone on the endocortical surface contracting the medullary canal, as reported using anabolic therapy and proposed by Burghardt and colleagues,7 is unlikely for reasons discussed earlier; the filling phase of the remodeling cycle will at best replace the excavated volume with a similar volume, closing the hemiosteon and so maintaining, but not reducing, the endocortical perimeter. Thus cortical area has increased, but only by an amount produced by filling of the transient remodeling deficit that is a consequence of the normal delay in initiation and slower progression of the formation phase of actively remodeling BMUs; the higher the baseline remodeling rate, the higher is the transient remodeling deficit, and the greater is the increase in cortical area achieved as the diameter of the larger numbers of existing pores decreases reducing the area that is void space without altering periosteal and endocortical perimeters. Concurrent secondary mineralization, part of the remodeling transient,9 increases the density, not the area, of the existing matrix, increasing edge detection as discarded voxels "seen" as "nonbone" are now included, leading to an apparent increase in cortical area, an apparent increase in cortical thickness (derived by dividing the cortical area by the fixed bone perimeter), and an apparent decrease in medullary area. Continued remodeling then slowly erodes earlier benefits to an extent that will vary based on the individual's baseline remodeling rate and the potency of the drug in suppressing remodeling. The fifth and final observation is null: There was no detected change in trabecular morphology. Antiresorptives reduce remodeling on trabecular surfaces, preventing trabecular thinning, perforation, and loss of connectivity so that no change in trabecular vBMD relative to baseline is expected. Trabecular vBMD and morphology may be higher than in placebo-treated patients because trabecular decay continues in these controls, not because trabecular vBMD increased in the treated group. Despite the absence of detectable differences in trabecular morphology globally, increases in trabecular vBMD were reported in several quadrants or subregions. Correlations were reported between trabecular morphology in some quadrants and estimates of bone strength determined using micro–finite-element analysis. The changes in trabecular morphology—number or thickness—are likely to reflect either the presence of cortical remnants that look like trabeculae or an increase in the degree of tissue mineralization density of trabeculae so that they now can be "seen" as they attenuate photons rather than any anabolic effect of therapy; any measured increase in trabecular thickness or number or cortical thickness has no plausible physiologic basis. Moreover, any restoration in trabecular thickness produced by almost complete refilling of resorption cavities on a trabecular plate of about 100 µm is difficult to detect with the resolution of 82 µm of high-resolution peripheral quantitative computed tomography. Bone is no longer an impenetrable fortress. Densitometry was a beginning, a very good one because it gave voice to a silent epidemic. It also was a bad beginning because areal BMD, the shadow cast by photons as they were attenuated by this 3D masterpiece of biomechanical engineering, obscured and continues to obscure the material and structural basis of bone strength. This is not a trivial limitation; it misguides our thinking or ensures that none occurs.25 We have emerged from that shadow to higher levels of resolution that reveal the microstructural configuration of bone and its surfaces, where the action is. This technology is a real contribution, but it is not without limitations imposed by the resolution achievable within acceptable levels of radiation exposure given the dimensions of the changes in microstructure we would like to quantify.26, 27 These limitations were discussed by the authors, leaders in exploration and refinement of this technology.28, 29 The use of fixed thresholds to segment (separate) bone from soft tissue and cortical from trabecular bone is a limitation of current image processing methods. For example, newly deposited bone during remodeling has a lower tissue density and may not be included in the quantification of bone's material composition and structure; it is not "seen" because it does not have the threshold characteristics imposed by the threshold algorithm to qualify as bone or as cortical rather than trabecular bone. The in-plane resolution of around 82 µm cannot "see" changes in trabecular morphology that are of similar magnitude and cannot "see" intracortical porosity below this resolution; these smaller pores account for most intracortical porosity, and it is these smaller pores that are partly reduced by antiresorptives.30 These limitations are being overcome. New approaches are being developed that bring the resolution down to around 40 µm, and segmentation without using fixed thresholds is also being explored.31 Are we there yet? No. Is there progress? Yes. The author serves on several advisory board committees and occasionally lectures at meetings sponsored by Amgen, Merck Sharpe & Dohme, Sanofi Aventis, Proctor & Gamble, Servier, Novartis, and Eli Lilly.