Three-Dimensional MicrocomputedTomography Imaging of Basic

Multicellular Unit-Related ResorptionSpaces in Human Cortical Bone

DAVID M.L. COOPER,1* C. DAVID L. THOMAS,2 JOHN G. CLEMENT,2 AND

BENEDIKT HALLGRIMSSON3

1Department of Orthopaedics, Division of Orthopaedic Engineering Research,University of British Columbia, Vancouver, Canada

2School of Dental Science, University of Melbourne, Melbourne, Australia3Faculty of Medicine, University of Calgary, Calgary, Canada

ABSTRACTThis study employed microcomputed tomography (micro-CT) as a novel

means for visualizing the morphology and quantifying the range (length) ofbasic multicellular unit (BMU)-related resorption spaces in human corticalbone. We tested the hypotheses that the density and range of spaces vary withage and sex. The sample included 82 human (18–92 years) anterior femoralmidshaft samples. The morphology of the spaces (n � 99) was varied, includingunidirectional, bidirectional, branched, and even highly clustered forms. Thedensity of resorption spaces was negatively correlated with age for the com-bined sexes and females, with Spearman’s rho values of �0.355 (P � 0.001)and �0.522 (P � 0.002), respectively. The density of spaces did not differsignificantly between the sexes (P � 0.735). Mean range � SD for the combinedsexes, females, and males was 2,706 � 1,177, 2,681 � 1,247, and 2,718 � 1,150�m, respectively. Numerical imulation of the effect of the 7,000 �m scan fieldof view suggested that the actual mean range of the spaces for the pooledsample was actually on the order of 3,770 �m. Range did not correlate signif-icantly with age for the combined sexes (P � 0.587) or females (P � 0.345) andmales (P � 0.896) considered separately and was not significantly different(P � 0.883) between the sexes. These results suggest that the range of BMUsis not affected by age. The age-dependent decrease in resorption space densityfor the females and pooled sexes was most likely a consequence of corticalrarefaction, leading to difficulty detecting resorption spaces with micro-CT,rather than a decrease in overall remodeling activity. Anat Rec Part A, 288A:806–816, 2006. © 2006 Wiley-Liss, Inc.

Key words: microcomputed tomography; cortical bone; remod-eling; basic multicellular unit; osteon; Haversian

The spatial and temporal coupling of the resorptive andformative phases of bone remodeling was first described inthe 1960s (Johnson, 1964) and the coordinated activity ofdifferent cell populations, referred to as the basic multi-cellular unit (BMU), responsible for this coupling is acentral concept in modern bone biology (Frost, 1969).While tremendous advances have been made in under-standing how this cellular activity is regulated (e.g., therole of RANK ligand in the formation of osteoclasts) (Te-itelbaum, 2000), little is known regarding the 3D morphol-ogy of BMUs and the secondary osteons they produce. Theconventionally described cylindrical BMU geometry, anosteoclastic cutting cone followed by an osteoblastic clos-

ing cone, is highly idealized. Although current theoreticalmodels focus on this geometry (Smit and Burger, 2000;

*Correspondence to: David M.L. Cooper, Department of Ortho-paedics, Division of Orthopaedic Engineering Research, Univer-sity of British Columbia, VGH, Research Pavilion, Room 597, 828West 10th Avenue, Vancouver, BC, V5Z 1L8, Canada. Fax: 604-875-4851. E-mail: cooperdv@interchange.ubc.ca

Received 19 December 2005; Accepted 12 April 2006DOI 10.1002/ar.a.20344Published online 7 June 2006 in Wiley InterScience(www.interscience.wiley.com).

THE ANATOMICAL RECORD PART A 288A:806–816 (2006)

© 2006 WILEY-LISS, INC.

Martin, 2002; Smit et al., 2002; Burger et al., 2003), avariety of BMU-related resorption space forms are knownto exist, including localized dilations of existing canals,branched, and double-ended cutting cones, and even irreg-ular interconnected spaces (Cohen and Harris, 1958; Vas-ciaveo and Bartoli, 1961; Johnson, 1964; Jaworski et al.,1972; Tappen, 1977; Konijn et al., 1996). It has also beenproposed that remodeling may have a higher level of or-ganization, above that of the single osteon. Bell et al.(2001) have coined the term “super-osteon” to describe thecollective structure of spatially and temporally clusteredsecondary osteons. While the 2D evidence for such clustersis compelling, the 3D structure of remodeling clusters(super-osteons) has yet to be visualized and described.

Remarkably few studies have directly examined BMUmorphology in human bone and the majority of research inthis area has focused on canine bone (Cohen and Harris,1958; Marshall et al., 1959; Enneking et al., 1972; Jawor-ski and Lok, 1972; Tappen, 1977). Johnson (1964) de-scribes osteonal columns (BMUs) in human bone, but doesnot describe the composition of his sample, while bothBeddoe’s (1977) and Tappen’s (1977) studies included onlytwo human specimens. Robling and Stout’s (1999) study,which was focused on a specific type (drifting) of osteonand related BMU, included three human specimens. Per-haps the most important examination of resorption spacemorphology in human bone was that by Jaworski et al.(1972), who measured 12 cutting cones in five subjects. Inlight of well-established variation in cortical microstruc-ture among many species (Enlow and Brown, 1956, 1957,1958; Jowsey, 1966; Singh et al., 1974; Cespedes Diaz andRajtova, 1975; de Ricqles, 1993; Harsanyi, 1993; Mulhernand Ubelaker, 2001), extrapolation of BMU/osteon mor-phology from one species to another should be viewed withsome caution. The limited exploration of intraspecies vari-ation is also a concern. Variation in osteon diameteramong the skeletal elements of individuals, such assmaller osteons in ribs versus femora (Pfeiffer, 1998),must fundamentally reflect variable BMU activity. Thelength, or range, of a secondary osteon is determined bythe distance traversed by the BMU that formed it. Thisvalue is significant since it relates to the length and rate ofthe remodeling cycle (lifetime of the BMU) and directlyimpacts 2D sampling of cortical microstructure. For exam-ple, the number of newly forming osteons detected in across-section, used to calculate activation frequency, isrelated to the range traversed by BMUs in 3D (Martin,1994). Similarly, range is a central concern when exam-ining the 3D spatial association between events such asmicrodamage and the initiation of remodeling (Martin,2002). Johnson (1964) has described BMUs in youth as“longer and more extensively branched” than those in theold, and others have reported that osteon length decreaseswith age/remodeling (Filogamo, 1946a, 1946b; Amprino,1948; Filogamo, 1946a, based on translation by Simmons,1985). This evidence suggests BMU, and hence osteon,dimensions may vary with age.

High-resolution microcomputed tomography (micro-CT)is capable of resolving the porous network of canals thatpermeates cortical bone (Wachter et al., 2001a, 2001b;Cooper et al., 2003, 2004; Basillais et al., 2004; Bousson etal., 2004; Jones et al., 2004; Schneider et al., 2004; Gruberet al., 2005; Matsumoto et al., 2005). BMU-related resorp-tion spaces are the largest voids within normal corticalbone and therefore micro-CT holds great promise as a

novel noninvasive means of visualizing and quantifyingthe 3D morphology of these spaces. The specific aims ofthis study were to document and describe morphologicalvariability of BMU-related resorption spaces at the hu-man femoral midshaft using micro-CT; to test the hypoth-esis that BMU range, inferred from related porous struc-tures, varies significantly with age and sex; and to assessif there is evidence for the spatial and temporal clusteringof remodeling events as proposed by the super-osteon hy-pothesis.

MATERIALS AND METHODSThe study sample consisted of 82 (30 female, 52 male)

right femoral midshaft specimens, ranging in age from 18to 92 years, derived from the Melbourne Femur Collection(MFC) (Clement, 2005; http://www.vifm.org/attachments/o397.pdf). The mean age of the sample was 54.6 years.When broken down by sex, the mean ages for females (52.2years) and males (55.9 years) were not significantly dif-ferent (P � 0.086). All the tissue was collected at theVictorian Institute of Forensic Medicine with approval ofthe institute’s ethics committee in accordance with Aus-tralian National Health and Medical Research Councilguidelines and with the informed consent of the next ofkin. The Conjoint Faculties Research Ethics Board of theUniversity of Calgary also granted approval for this workon human subjects (CE101-3681). The samples for thisstudy were initially harvested as mid-diaphyseal rings 8mm thick (along the long axis of the femora). In order tofacilitate high-resolution micro-CT imaging, anteriorblocks (5 mm wide), aligned with the cross-sectional cen-troid and the midpoint of the linea aspera (Fig. 1). werecut using a handheld rotary saw (Dremel, Racine, WI).The samples were scanned with 7 �m isotropic voxel size

Fig. 1. Schematic of the sampling location. Solid line depicts thecentral axis passing through the centroid and midpoint of linea aspera.Dashed lines depict cut locations.

8073D MICRO-CT OF BMU SPACES IN CORTICAL BONE

with a SkyScan 1072 microtomograph (Aartselaar, Bel-gium). The scan protocol included rotation through 180° ata rotation step of 0.23°, X-ray settings standardized to 100kV and 100 �A, and an exposure time of 7.5 sec per frame.Six-frame averaging was used to improve the signal-to-noise ratio. Scan times were approximately 12 hr persample. Following scanning, a 2D reconstruction stageutilizing a cone-beam algorithm (Feldkamp et al., 1984)was used to produce 1,024 serial cross-sectional images.Following this step, the voxel size of the data sets wasreduce by half in all three dimensions (14 �m isotropicvoxels) using Tconv 2.0 (SkyScan). This facilitated effi-cient manipulation of the large data sets while still pro-viding resolution sufficient to visualize BMU-related re-sorption spaces.

The serial image data sets were manually screened forresorptive cavities consistent with active remodeling, suchas those revealed by staining and clearing techniques(Dempster and Enlow, 1959; Marotti and Zallone, 1980).Essentially, dilated spaces that were continuous with thecanal network, representing osteoclastic cutting cones,were targeted. In order to be included in this study, aresorption space had to have to at least one such cuttingcone visible in the scan field of view (approximately 7 mmalong the long axis of the diaphysis). Once detected, thevolumes surrounding the resorption spaces were digitallyexcised, using ImageJ 1.33j, from the larger image datasets and resampled to generate serial longitudinal sec-tions. Due to the cubic (isotropic) voxel composition of thedata sets, orthogonal resampling could be performed with-out loss of image quality. Standard deviation projections(images depicting gray-scale standard deviations throughserial image stacks) were used to highlight the morphol-ogy of the resorption spaces against the background of thesurrounding canal networks. Resorption space density(spaces/mm3) was calculated using the total volume ofbone within the specimen. This value was chosen as it didnot require the subjective designation of the endostealborder (e.g., necessary for cortical thickness and cross-sectional area estimates). BMU range was defined as thelength from the tip of the cutting cone back to origin on thecanal network, an oppositely directed cutting cone (dou-ble-ended), or to the edge of the scan field, whichever camefirst. In the case of branched spaces, the longest continu-ous branch was measured. Three-dimensional renderingswere produced using 3D-Creator 2.2h (SkyScan).

To facilitate statistical analysis, the sample was dividedinto four age categories (18–40, 41–60, 61–80, and 81�years; Fig. 2) in a manner similar to that employed pre-viously by Bell et al. (2001). Kolmogorov-Smirnov testswere employed to determine if resorption space densityand range values were normally distributed for the pooledand separated sexes. Resorption space density was notnormally distributed (P � 0.05) and thus the use of non-parametric statistics was indicated. Spearman correlationwas used to examine relationships with age, whileKruskal-Wallis tests were used to examine differencesbetween the sexes. On the other hand, the BMU rangevalues were found to be normally distributed (P � 0.05)and thus Pearson correlation was used to examine rela-tionships with age, while t-tests were used to examinedifferences between the sexes.

A computer-based numerical simulation (Excel, Mi-crosoft) was developed to estimate the bias in the rangemeasurement introduced by the 7,000 �m field of view

(FOV). An idealized sample of 10,000 normally distributedresorption space ranges were randomly generated for aseries of actual means (2,000–10,000 �m) with a standarddeviation of 1,000 �m. The spaces were randomly posi-tioned (even distribution) within a 32 mm vertical rangeand their apparent ranges were measured within a cen-tered 7,000 �m FOV. To replicate the methodology used toacquire the experimental measures, spaces spanning theentire FOV (apparent range � 7,000 �m) were excluded.The smallest experimentally measured range was 784 �m.This is consistent with the selection criteria, where a clearcutting cone had to be identified and reflected the inabilityto identify very small (e.g., newly formed) resorptionspaces. To compensate for this in the simulation of bias,spaces with apparent ranges below 500 �m were excluded.

RESULTSA total of 99 resorption spaces, detected in 48 of the 80

specimens, fit the inclusion criteria. These possessed avariety of different morphologies, including unidirectional(Fig. 3), bidirectional, branched (Fig. 4), and highly clus-tered (Fig. 5) forms. Closing cone morphology varied con-siderably, with some resorption spaces being constrictedadjacent to the cutting cone, while others showed little orno evidence of a closing cone along many millimeters.Based on the serial 2D images, projection views, and 3Drenderings (Fig. 6), it was evident that resorption spacescan frequently intersect with numerous existing canalsalong their length. Cutting cones were most clearly visiblein the periosteal third of the cortex and were increasinglydifficult to identify toward the endosteal surface, whereporosity was greater and consisted of large interconnectedcanals. Similarly, cutting cones were increasingly difficultto identify with advancing age, particularly in females,due to progressive trabecularization of the cortex throughthe coalescence of ever larger canals (Fig. 7). Bone volume(mm3) was negatively correlated with age for the sexespooled (r2 � �0.602; P � 0.001) and considered separately(female: r2 � �0.743, P � 0.001; male: r2 � �0.602, P �0.001).

Resorption space density was not normally distributedfor the combined sexes, nor for either sex individually (P �

Fig. 2. Sample age and sex distribution.

808 COOPER ET AL.

0.05). The lack of resorption spaces in many specimensresulted in an asymmetric distribution of densities,skewed to the left. The percentiles for resorption spacedensity are presented in Table 1. Resorption space densityand age were negatively correlated for the combined sexesand the females, with Spearman’s rho values of �0.355(P � 0.001) and �0.522 (P � 0.004), respectively. Thisrelation was not significant (P � 0.082) for the males. Sexwas not significant for the entire sample (P � 0.735) orwhen broken down by age groups (P � 0.859, 18–40 years;P � 0.225, 41–60 years; P � 0.201, 61–80 years; P �0.414, 81� years).

Mean BMU range � SD for the combined sexes, fe-males, and males were 2,706 � 1,177, 2,681 � 1,247, and2,718 � 1,150 �m, respectively. The maximum range mea-sured for males and females was 5,180 and 5,404 �m,respectively. Range and age were not significantly corre-lated for the combined sexes (P � 0.587) or for females(P � 0.345) or males (P � 0.896) considered separately.Comparing the sexes, the mean values were not signifi-cantly different (P � 0.883) for the entire sample. A posthoc examination of the power for this sample indicatedthat the differences between the means would have toexceed 765 and 886 �m to reach powers of 80% and 90%,respectively (Mace, 1974). When broken down into agecategories (Fig. 8), sex was significant only for the 61–80year olds (P � 0.001). No BMUs were identified in females

over 81 years of age, and thus a comparison could not bemade against the corresponding males.

The bias simulation revealed a nonlinear relation be-tween the actual ranges and the apparent ranges observedwithin the 7,000 �m FOV. This relation approached alimit around 3,700 �m (Fig. 9). For the region of the curvewith ranges at or below 7,000 �m, a strong logarithmic(r2 � 0.997; P � 0.001) relation was identified. The result-ing regression formulae (y � 1456.7Ln(x) � 9295.6) wasused to predict the actual mean values from the experi-mentally derived mean of 2,700 �m (for the pooled sexes),resulting in a value of 3,771 �m. In turn, this actual rangevalue (rounded to 3,770 �m) was reentered into the sim-ulation and the apparent distribution of ranges was com-pared with the experimentally derived results (Fig. 10).Although skewed to the left, the simulated distributionmatched well with the experimental values.

DISCUSSIONMicro-CT imaging proved highly effective for detecting

and visualizing BMU-related resorption spaces within hu-man cortical bone and represents the first noninvasivemethod for this purpose. The morphology of cutting conesat the human femoral midshaft varied considerably, rang-ing from simple cutting cones with the conventionallydescribed form to highly interconnected and clusteredtypes. Closing cones also varied in form, with some resorp-tion spaces constricted adjacent to the cutting cones, whileothers showed no apparent sign of new bone deposition.Osteoid was not discernable in the micro-CT images andrelatively hypomineralized new bone was only occasion-ally apparent adjacent to a cutting cone. Therefore, theextent and location of closing cones could not be fullyappreciated. This limitation may account for the apparentuncoupling of the spatial and/or temporal association be-tween cutting and closing cones found in many of theresorption spaces in this study. However, previous au-thors have also described BMUs with a lag between theresorptive and formative phases of remodeling, resultingin a quiescent zone between the cutting and closing cones(Johnson, 1964; Parfitt, 1983).

Ideally, it was hoped that 3D micro-CT analysis couldfacilitate the volumetric measurement of the total numberof active remodeling events present as inferred by thedensity of BMU-related resorption spaces. However, dis-tinguishing the extent of remodeling activity based solelyon micro-CT images proved more difficult than antici-pated. As such, the number of resorption spaces measuredper specimen in this study should not be viewed as ex-haustive. The inclusion criterion of at least one visiblecutting cone certainly did not encompass the totality ofremodeling-related activity present. In numerous cases,enlarged canals, consistent with the recent passage of acutting cone, spanned the entire field of view and thuswere excluded (Fig. 6). Additionally, due to the highlycomplex and variable nature of the cortical canal network,it is very likely that not all cutting cones were detectedand measured. This was especially the case for very smallspaces, as already noted. Therefore, the results for thisparameter are best viewed as being approximate ratherthan precise.

The negative relationship between age and resorptionspace density observed in the females and pooled samplesuperficially suggested a dramatic decrease in remodel-ing-related activity with age. However, this is clearly not

Fig. 3. Simple resorptive cutting cone in context of surroundingcortical canals. Scale bar � 1 mm.

8093D MICRO-CT OF BMU SPACES IN CORTICAL BONE

the case, as the cortices of older individuals do continue toremodel at rates comparable to younger individuals (e.g.,Bell et al., 2001). The age-dependent decrease in resorp-tion space density found in this study is, therefore, bestexplained by the loss of cortical bone through the trabe-cularization of the endosteal region. Thinner cortices offerless volume in which remodeling can occur, thereby reduc-ing the probability of detecting a resorption space usingour study criteria. Further, the progressive increase inmean osteonal canal diameter coupled with a decrease incanal separation associated with age in this sample (Coo-per et al., 2005) would eventually lead to a state where along continuous cutting cone simply could not exist since anewly formed resorption space would rapidly perforate anadjacent canal. It is important to note, however, thatthinning of the cortex and an increase in porosity are notindependent events, as the coalescence of large canals, incombination with reduced bone formation, has been pro-posed as the underlying mechanism for cortical bone lossvia the expansion of the medullary cavity (Power et al.,2004). This raises the possibility that the decrease in thenumber of cutting cones observed with advancing age mayhave been in part a result of a transition from osteonal tohemiosteonal remodeling (Parfitt, 1994). In other words,there may be a transition from a predominantly corticalpattern (osteonal) of remodeling to a pattern more consis-tent with that seen in trabecular bone (hemiosteonal). Themore dramatic age-related loss of cortical bone in femalesin this sample (Cooper et al., 2005), therefore, may di-rectly account for the stronger relationship between ageand resorption space density.

Previous studies have reported irregular resorptionspaces that are inconsistent with conventional views on

cortical remodeling. Cohen and Harris (1958) examinedthe 3D arrangement of Haversian systems within caninefemoral samples and reported the existence of irregularand highly interconnected resorption spaces. Johnson(1964) described canals with eccentric and variable pat-terns of resorption in human cortical bone. Bell et al.(2000) reported “composite” osteons in the human femoralneck, which were comprised of multiple incomplete pack-ets of new bone and associated with enlarged osteonalcanals. If the activity of multiple BMUs were involvedwith the creation of these composite osteons, it would be apattern of remodeling very similar to that seen in trabec-ular bone. A caveat that must be considered in light of theproposed transition between remodeling modes is thatosteonal remodeling certainly continues in older individ-uals. Therefore, the above interpretation must not be con-strued as implying that new osteons were not forming inthe older individuals, only that a shift toward a trabecularpattern of remodeling and the presence of enlarged canalsreduced the occurrence and chance of detecting resorptionspaces with the classic cutting cone morphology. Bell et al.(2001) found no relationship between age and the densityof remodeling clusters, indicating that new osteon clusterscontinued to be produced with advanced age. Interest-ingly, though, such clusters were most frequently found inthe periosteal region and there was a negative relation-ship between the occurrence of remodeling clusters andgiant (diameter � 385 �m) canals (Bell et al., 2001). Basedon this relationship, they postulated that a breakdown inthe control of resorption depth led to the merger of adja-cent canals into giant canals, thus precluding the forma-tion of clusters (Bell et al., 2001). Again, such a scenario is

Fig. 4. Complex resorption space morphologies. Resorption space with cutting cones on both ends (left),anterior view (left), and lateral view (right). Branched resorption space (right), anterior view (left), and lateralview (right). Scale bar � 1 mm.

810 COOPER ET AL.

consistent with a transition to a more trabecular patternof remodeling.

BMU-related resorption spaces are dynamic structuresand thus the measurement of range in this study repre-sents only a snapshot at the time of death. The presence ofcutting cones directly implies that the BMUs were still inactive resorption and thus increasing their range. Addi-tionally, in many cases, the spaces extended beyond thefield of view and thus could not be measured along their

entire length. While the maximum range measured in thisstudy was 5,404 �m, as noted above, enlarged canalsfrequently spanned the entire scan field, indicating rangescould be in excess of 7 mm (Fig. 6). As such, the experi-mental range measurements obtained in this study cer-tainly underestimate true mature BMU/osteon range. Thesignificantly shorter range for females vs. males in the 61-to 80-year age group perhaps reflects the fact that in-creased porosity in the female specimens precluded the

Fig. 5. Clustered resorption spaces. Anterior view (left) and lateral view (right). Scale bar � 1 mm.

8113D MICRO-CT OF BMU SPACES IN CORTICAL BONE

Fig. 6. Three-dimensional rendering of the cortical canal network (gray) demonstrating a triple-prongedBMU (red) superimposed over an enlarged canal that traverses the entire field of view (blue). Isolated BMUand canal are displayed again on the right; the two structures were not directly interconnected. Recon-structed volume spans 1.6 mm horizontally.

812 COOPER ET AL.

existence of large continuous resorptive cutting cones (asdescribed above). Notably, no resorption spaces werefound in females 81 years and older, which were generallycharacterized by extensively trabecularized cortices.

The limit in range measurement identified by the biassimulation highlights the inability to measure range ac-curately when the mean value exceeds half the FOV. Inour simulation (7,000 �m FOV), this value was shifted upto approximately 3,700 �m due to the exclusion of small(� 500 �m) values. The fact that our experimental resultsfell below half FOV suggests that calibration is possible.While the simulation provided an estimate of the actualrange of resorption spaces within our sample, a few cave-ats must be noted. First, the mean and standard deviationof total BMU range remains unknown. Additionally, theconfounding factor of active cutting cones, at the time ofdeath, remains. This fact is also important, as a normaldistribution of apparent range values is not necessarilyexpected. Due to the presence of resorption spaces at vary-ing stages of progression, it may be more reasonable toexpect a distribution of ranges that is skewed to the right.Ultimately, changing FOV, mean range, and/or the stan-dard deviation of the ranges would result in changing therelation between the actual and apparent ranges. There-fore, the predicted actual mean range value should beviewed only as an approximation.

The mean experimental ranges reported in the currentstudy are generally consistent with previously reportedvalues. A mean osteon length, based on distance betweencanal intersections, for the human femoral cortex has beenreported to be 2.5 mm (Beddoe, 1977). However, in light ofthe ability of BMUs to anastomose with surrounding canalnetworks, the approach of inferring osteon length frominterbranch distance is clearly problematic. Indeed, Co-hen and Harris (1958) noted that there is “no evidence oflongitudinal demarcation of concentric lamellae at anas-tomoses or bifurcations,” a point that complicates measur-ing osteon length by interbranch canal length. The use ofthis method for quantifying osteon length may account for

the age-dependent decrease in osteon length reported inseveral previous studies (Filogamo, 1946a, 1946b; Am-prino, 1948). It can be assumed that “true” mean osteonlength exceeded the 2.5 mm reported by Beddoe (1977),and Johnson (1964) has stated that osteons in humanbone can be as long as 10 mm. Parfitt (1983) reported verysimilar BMU dimensions for canines, monkeys, and hu-mans, estimating the combined lengths of cutting andclosing cones at 2,813, 2,496, and 2,461 �m, respectively.It should, however, be noted that these lengths do notreflect total range and are based on an assumed rate ofradial canal closure. In canine fibulae, Enneking et al.(1972) found that cylindrical osteon cavities averaged2,500 �m and could be as long as 9,500 �m, while in thecanine humerus Ca45 labeling revealed mean osteonlength of 2.1 mm (Marshall et al., 1959). In Tappen’s(1977) study of dog femoral samples, osteons most fre-quently fell between 2,000 and 3,000 �m in length, withthe longest example noted reaching between 8,000 and9,000 �m. Drawing on the evidence from many of theabove noted studies, Martin (2002) estimated 6,000 �m asa “rough” mean for BMU range. In light of the results ofthe current study (both experimental and simulated), thisvalue may be a little high, but certainly remains reason-able.

It would be ideal if BMU parameters such as rangecould be measured from 3D micro-CT reconstructions ofmature osteons. Unfortunately, nonporous structuressuch as cementing lines are not visible in micro-CT imagesand variations in osteon mineralization are only rarelydetected (Engelke et al., 1993; Cooper et al., 2004). Assuch, a direct 3D assessment of mature osteon size (andhence total BMU range) is currently not possible. Futureadvances may ultimately allow the 3D differentiation ofosteons by their degree of mineralization, but micro-CTanalysis of cortical bone is currently limited to porosity.Thus, osteon dimensions can only be inferred from porousstructures such as BMU-related resorption spaces. A his-tological method employing the administration of a fluo-rochrome label coupled with longitudinal histological sec-tioning (Moshin et al., 2002) may ultimately be required tomeasure BMU range definitively. However, the necessaryperiod of labeling would be protracted. For example, John-son (1964) predicted that a 5 mm long double-ended osteonwould require 8 months to be completed, and Martin(1994) has noted that, at an erosion rate of 39 �m per day(value from Jaworski and Lok, 1972), it would take 256days just to form a 1 mm resorption space. These timeperiods, combined with the low probability of obtainingideal sectional views, have, to date, precluded such anapproach. One promising avenue for addressing these dif-ficulties may be combining micro-CT imaging and histol-ogy. This allows the targeting of sections based on struc-tures detected with micro-CT noninvasively.

The numerous resorption spaces with complex branchedand clustered (Figs. 4 and 5) forms detected in this studyare consistent with the super-osteon hypothesis that pro-poses osteons clustered in 2D can be created by a singlelarger remodeling event. Bone formation within the adja-cent spaces would result in spatially associated osteons.Debate remains whether such clusters represent rela-tively permanent structures within the bone and aremaintained throughout remodeling, or are transitorystructures that arise in response to localized stimuli (Bellet al., 2001). The first model has generally been described

Fig. 7. Two-dimensional micro-CT cross-section (left) and 3D ren-dering of marked region (right) demonstrating the appearance of en-larged endosteal canals in a 73-year-old male. For the 2D cross-section,the periosteal surface faces up. The 3D rendering is rotated such that themore periosteal aspect faces down and to the right. The renderedvolume measured 2.1 mm � 2.1 mm � 4.2 mm.

8133D MICRO-CT OF BMU SPACES IN CORTICAL BONE

as a possible mechanism for the initial replacement ofplexiform bone. Currey (1960) hypothesized that second-ary osteons within primary lamellar (plexiform) bone dis-rupt the local primary vascular supply and thus inducesuccessive cycles of remodeling to remove compromisedadjacent packets of bone. Ultimately, this leads to tightlygrouped collections of secondary osteons. Plexiform bone,however, is rarely found in the mature human skeletonand the osteon banding associated with remodeling of thiscortical bone type has even been proposed as a means ofdifferentiating between human and nonhuman bone (Mul-hern and Ubelaker, 2001). In the second model, clustersarise independent of preexisting structures as a directconsequence of BMU branching or spatial clustering. Inlight of the extensive remodeling that human cortical bone

undergoes with age, and the complex resorption spacemorphologies that have been reported in this and previousstudies, this latter model seems to be the most plausible.The underlying reasons for, and mechanisms that regu-late, BMU branching represent intriguing questions forfurther research.

This study demonstrated that micro-CT is an effectiveand efficient tool for the visualization of resorption spaceswithin cortical bone. Although previous studies have re-ported variation in resorption space morphology, ours isthe first to visualize and quantify these structures nonin-vasively. Additionally, the study sample, derived from theexceptional Melbourne Femur Collection, is the largestand most diverse human sample from which resorptionspace morphology has been analyzed to date. Based on themean measured lengths of the spaces (calibrated by nu-merical simulation of the FOV bias), it appears that theaverage BMU in human bone has a range in excess of3,770 �m. The density of resorption spaces with clearlydefined cutting cones was negatively correlated with agefor females and the pooled sexes, which, coupled withvisual evidence of trabecularization of the cortex, demon-strated that there is likely a transition between corticaland trabecular modes of remodeling with advancing age.Complex branched and clustered resorption spaces indi-cated that osteon clusters, as viewed in 2D, can arise froma single simultaneous remodeling event as proposed bythe super-osteon hypothesis. Understanding the underly-ing mechanisms that regulate aspects of BMU morphol-ogy, including diameter, range, orientation, and branch-ing pattern, will require further investigation. However,the improved detection, visualization, and quantificationoffered by the micro-CT approach will undoubtedly aid inthis pursuit.

TABLE 1. Percentile values for resorption space density (spaces/mm3) for the sexescombined and considered separately

Percentile 5 10 25 50 75 90 95

Pooled Sexes 0.0000 0.0000 0.0000 0.0073 0.0138 0.0228 0.0338Female 0.0000 0.0000 0.0000 0.0067 0.0160 0.0276 0.0439Male 0.0000 0.0000 0.0000 0.0075 0.0138 0.0227 0.0330

Fig. 8. Mean BMU range by age and sex. Error bars representstandard deviation. For age category 81�, only one male and no femaleresorption spaces were measured.

Fig. 9. Relation between simulated actual and apparent mean range.Logarithmic regression applied to the 2,000–7,000 �m ranges (closeddiamonds).

Fig. 10. Histogram of apparent BMU range values comparing theexperimental results with those for the simulated mean range of 3,770�m.

814 COOPER ET AL.

ACKNOWLEDGMENTSThe authors thank Professor Stephen Cordner, director

of the Victorian Institute of Forensic Medicine, and thestaff of the mortuary and Donor Tissue Bank for theirassistance in the collection of this series of bone speci-mens. They are also grateful to the next of kin of thedonors for permission to remove bone for research pur-poses. D.M.L.C. thanks the Alberta Provincial CIHRTraining Program in Bone and Joint Health for supportduring the period this study was conducted. Supported inpart by National Science and Engineering grant 238992-02, Canadian Foundation for Innovation grant 3923, andAlberta Innovation and Science grant URSI-01-103-RI (toB.H).

LITERATURE CITEDAmprino R. 1948. A contribution to the functional meaning of the

substitution of primary by secondary bone tissue. Acta Anat 5:291–300.

Basillais A, Chappard C, Bensamoun S, Brunet-Imbault B, Ho Ba ThoM, Benhamou C. 2004. Three dimensional characterization of cor-tical bone porosity on microcomputed tomography images: a newmethod of evaluation. J Bone Miner Res 19:S112–S113.

Beddoe AH. 1977. Measurements of the microscopic structure of cor-tical bone. Phys Med Biol 22:298–308.

Bell KL, Loveridge N, Jordan GR, Power J, Constant CR, Reeve J.2000. A novel mechanism for induction of increased cortical porosityin cases of intracapsular hip fracture. Bone 27:297–304.

Bell KL, Loveridge N, Reeve J, Thomas CD, Feik SA, Clement JG.2001. Super-osteons (remodeling clusters) in the cortex of the fem-oral shaft: influence of age and gender. Anat Rec 264:378–386.

Bousson V, Peyrin F, Bergot C, Hausard M, Sautet A, Laredo JD.2004. Cortical bone in the human femoral neck: three-dimensionalappearance and porosity using synchrotron radiation. J Bone MinerRes 19:794–801.

Burger EH, Klein-Nulend J, Smit TH. 2003. Strain-derived canalicu-lar fluid flow regulates osteoclast activity in a remodelling osteon: aproposal. J Biomech 36:1453–1459.

Cespedes Diaz CM, Rajtova V. 1975. Comparative study of lamellarbone in some carnivora. Folia Morphol (Praha) 23:221–229.

Clement JG. 2005. The Melbourne Femur Collection: the gift of tissueunderpins important medical and forensic research. VIFM Review3:7–11.

Cohen J, Harris W. 1958. The three-dimensional anatomy of haver-sian systems. J Bone Joint Surg Am 40A:419–434.

Cooper DM, Turinsky AL, Sensen CW, Hallgrimsson B. 2003. Quan-titative 3D analysis of the canal network in cortical bone by micro-computed tomography. Anat Rec 274B:169–179.

Cooper DM, Matyas JR, Katzenberg MA, Hallgrimsson B. 2004. Com-parison of microcomputed tomographic and microradiographic mea-surements of cortical bone porosity. Calcif Tissue Int 74:437–447.

Cooper DML, Thomas CDL, Clement JG, Hallgrimsson B. 2005. Therole of age and body size in predicting the 3d structure of corticalporosity at the human anterior femoral mid-shaft. J Bone MinerRes 20:SU016.

Currey JD. 1960. Differences in the blood-supply of bone of differenthistological types. Q J Microsc Sci 101:351–370.

de Ricqles AJ. 1993. Some remarks on paleohistology from a compar-ative evolutionary point of view. In: Grupe G, Garland AN, editors.Histology of ancient human bone: methods and diagnosis. NewYork: Springer-Verlag. p 37–77.

Dempster WT, Enlow DH. 1959. Patterns of vascular channels in thecortex of the human mandible. Anat Rec 135:189–205.

Engelke K, Graeff W, Meiss L, Hahn M, Delling G. 1993. High spatialresolution imaging of bone mineral using computedmicrotomography: comparison with microradiography and undecal-cified histologic sections. Invest Radiol 28:341–349.

Enlow DH, Brown SO. 1956. A comparative histological study of fossiland recent bone tissues: part 1. Texas J Sci 8:405–443.

Enlow DH, Brown SO. 1957. A comparative histological study of fossiland recent bone tissues: part 2. Texas J Sci 9:186–214.

Enlow DH, Brown SO. 1958. A comparative histological study of fossiland recent bone tissues, part 3. Texas J Sci 10:187–230.

Enneking WF, Burchardt H, Puhl JJ, Thornby J. 1972. Temporal andspatial activity in mirror segments of mature dog fibulae. CalcifTissue Res 9:283–295.

Feldkamp LA, Davis LC, Kress JW. 1984. Practical cone-beam algo-rithm. J Opt Soc Am 1:612–619.

Filogamo G. 1946a. Forma e lunghezza degli osteoni della compactaossa lunghe nell’uomo. Ric Morphol 22:1–6.

Filogamo G. 1946b. La forme et la taille des osteones chez quelquesMammiferes. Arch Biol 57:137–143.

Frost HM. 1969. Tetracycline-based histological analysis of bone re-modeling. Calcif Tissue Res 3:211–237.

Gruber HE, Ashraf N, Kilburn J, Williams C, Norton HJ, Gordon BE,Hanley EN Jr. 2005. Vertebral endplate architecture andvascularization: application of micro-computerized tomography, avascular tracer, and immunocytochemistry in analyses of disc de-generation in the aging sand rat. Spine 30:2593–2600.

Harsanyi L. 1993. Differential diagnosis of human and animal bone.In: Grupe G, Garland AN, editors. Histology of ancient human bone:methods and diagnosis. New York: Springer-Verlag. p 79–94.

Jaworski ZF, Lok E. 1972. The rate of osteoclastic bone erosion inHaversian remodeling sites of adult dog’s rib. Calcif Tissue Res10:103–112.

Jaworski ZF, Meunier P, Frost HM. 1972. Observations on two typesof resorption cavities in human lamellar cortical bone. Clin Orthop83:279–285.

Johnson LC. 1964. Morphologic analysis of pathology. In: Frost HM,editor. Bone biodynamics. Boston: Little, Brown. p 543–654.

Jones AC, Milthorpe B, Averdunk H, Limaye A, Senden TJ, Sakel-lariou A, Sheppard AP, Sok RM, Knackstedt MA, Brandwood A,Rohner D, Hutmacher DW. 2004. Analysis of 3D bone ingrowth intopolymer scaffolds via micro-computed tomography imaging. Bioma-terials 25:4947–4954.

Jowsey J. 1966. Studies of Haversian systems in man and someanimals. J Anat 100:857–864.

Konijn GA, Vardaxis NJ, Boon ME, Kok LP, Rietveld DC, Schut JJ.1996. Four-dimensional confocal microscopy for visualisation ofbone remodelling. Pathol Res Pract 192:566–572.

Mace AE. 1974. Sample-size determination. Huntington, NY: E.Krieger.

Marotti G, Zallone AZ. 1980. Changes in the vascular network duringthe formation of Haversian systems. Acta Anat (Basel) 106:84–100.

Marshall JH, White VK, Cohen J. 1959. Microscopic metabolism ofcalcium in bone: I, three-dimensional deposition of Ca45 in canineosteons. Radiat Res 10:197–212.

Martin RB. 1994. On the histologic measurement of osteonal BMUactivation frequency. Bone 15:547–549.

Martin RB. 2002. Is all cortical bone remodeling initiated by micro-damage? Bone 30:8–13.

Matsumoto T, Yoshino M, Asano T, Uesugi K, Todoh M, Tanaka M.2005. Monochromatic synchrotron radiation muCT reveals disuse-mediated canal network rarefaction in cortical bone of growing rattibiae. J Appl Physiol 100:274–280.

Moshin S, Taylor D, Lee TC. 2002. Three-dimensional reconstructionof haversian systems in ovine compact bone. Eur J Morphol 40:309–315.

Mulhern DM, Ubelaker DH. 2001. Differences in osteon bandingbetween human and nonhuman bone. J Forens Sci 46:220–222.

Parfitt AM. 1983. The physiologic and clinical significance of bonehistomorphometric data. In: Recker RR, editor. Bonehistomorphometry: techniques and interpretation. Boca Raton:CRC Press. p 143–222.

Parfitt AM. 1994. Osteonal and hemi-osteonal remodeling: the spatialand temporal framework for signal traffic in adult human bone.J Cell Biochem 55:273–286.

Pfeiffer S. 1998. Variability in osteon size in recent human popula-tions. Am J Phys Anthropol 106:219–227.

Power J, Loveridge N, Lyon A, Rushton N, Parker M, Reeve J. 2004.Osteoclastic cortical erosion as a determinant of subperiosteal os-

8153D MICRO-CT OF BMU SPACES IN CORTICAL BONE

teoblastic bone formation in the femoral neck’s response to BMUimbalance: effects of stance-related loading and hip fracture. Osteo-poros Int 16:1049–1056.

Robling AG, Stout SD. 1999. Morphology of the drifting osteon. CellsTissues Organs 164:192–204.

Schneider P, Stampanoni M, Donahue L, Muller R. 2004. Bone vas-culature as a new marker for bone quality varies within the femoraldiaphysis of two genetically distinct mouse strains. J Bone MinerRes 19:S241.

Simmons DJ. 1985. Options for bone aging with the microscope.Yearbook Phys Anthropol 1985:249–263.

Singh IJ, Tonna EA, Gandel CP. 1974. A comparative histologicalstudy of mammalian bone. J Morphol 144:421–437.

Smit TH, Burger EH. 2000. Is BMU-coupling a strain-regulated phe-nomenon? a finite element analysis. J Bone Miner Res 15:301–307.

Smit TH, Burger EH, Huyghe JM. 2002. A case for strain-inducedfluid flow as a regulator of BMU-coupling and osteonal alignment.J Bone Miner Res 17:2021–2029.

Tappen NC. 1977. Three-dimensional studies of resorption spaces anddeveloping osteons. Am J Anat 149:301–332.

Teitelbaum SL. 2000. Bone resorption by osteoclasts. Science 289:1504–1508.

Vasciaveo F, Bartoli E. 1961. Vascular channels and resorption cav-ities in the long bone cortex the bovine bone. Acta Anat 47:1–33.

Wachter NJ, Augat P, Krischak GD, Mentzel M, Kinzl L, Claes L.2001a. Prediction of cortical bone porosity in vitro by microcom-puted tomography. Calcif Tissue Int 68:38–42.

Wachter NJ, Augat P, Krischak GD, Sarkar MR, Mentzel M, Kinzl L,Claes L. 2001b. Prediction of strength of cortical bone in vitro bymicrocomputed tomography. Clin Biomech 16:252–256.

816 COOPER ET AL.