Fourier transform infrared imaging of femoral neck bone: Reduced heterogeneity of mineral-to-matrix and carbonate-to-phosphate and more variable crystallinity in treatment-naive fracture cases compared with fracture-free controls



After the age of 60 years, hip fracture risk strongly increases, but only a fifth of this increase is attributable to reduced bone mineral density (BMD, measured clinically). Changes in bone quality, specifically bone composition as measured by Fourier transform infrared spectroscopic imaging (FTIRI), also contribute to fracture risk. Here, FTIRI was applied to study the femoral neck and provide spatially derived information on its mineral and matrix properties in age-matched fractured and nonfractured bones. Whole femoral neck cross sections, divided into quadrants along the neck's axis, from 10 women with hip fracture and 10 cadaveric controls were studied using FTIRI and micro-computed tomography. Although 3-dimensional micro-CT bone mineral densities were similar, the mineral-to-matrix ratio was reduced in the cases of hip fracture, confirming previous reports. New findings were that the FTIRI microscopic variation (heterogeneity) of the mineral-to-matrix ratio was substantially reduced in the fracture group as was the heterogeneity of the carbonate-to-phosphate ratio. Conversely, the heterogeneity of crystallinity was increased. Increased variation of crystallinity was statistically associated with reduced variation of the carbonate-to-phosphate ratio. Anatomical variation in these properties between the different femoral neck quadrants was reduced in the fracture group compared with controls. Although our treatment-naive patients had reduced rather than increased bending resistance, these changes in heterogeneity associated with hip fracture are in another way comparable to the effects of experimental bisphosphonate therapy, which decreases heterogeneity and other indicators of bone's toughness as a material. © 2013 American Society for Bone and Mineral Research


Femoral fragility fractures are a significant cause of morbidity and mortality in the elderly and represent a major social and economic burden.1–3 Reduction in bone mineral density (BMD), measured clinically at the hip in g/cm−2, is moderately predictive of who will experience a hip fracture,4 yet 50% of cases have a BMD above the diagnostic threshold for osteoporosis.1 The risk of hip fracture increases as much as 10-fold from age 60 to 80 years, an increase of which decreasing BMD accounts for only 20%.1 Other important characteristics than BMD, loosely termed “bone quality,” may contribute to bone strength and be independently associated with fracture.5, 6 The parameters usually considered to contribute to bone quality are: geometry;7, 8 micro-architecture and micro-cracks (their volumetric density and geometric characteristics);9 cortical bone porosity;10 connectivity of trabeculae;11 and the material properties of the mineralized matrix.12 Material properties depend on mineral and matrix components and the interactions between them.12, 13

Fourier transform infrared spectroscopic imaging (FTIRI) has been successfully used to describe the changes in iliac cortical and trabecular bone as functions of age, disease, fracture, and osteoporosis treatment.5, 14–17 Although average FTIRI properties in the iliac crest are comparable to those in the trochanter and hip, their distributions differ from one tissue to the other.15 FTIRI uses a spectrometer coupled with a light microscope to examine undecalcified sections of bone at ∼7-µm pixel resolution to provide spatially detailed information concerning mineral and matrix composition. Typically, four key indices are reported: mineral-to-matrix ratio; carbonate-to-phosphate ratio; crystallinity (which is related to the size and perfection of mineral crystals); and cross-link ratio (which is related to the collagen cross-link maturity).5, 14

Past reports of the effects of diseases and treatments on these FTIRI indices have tended to focus on changes in mean values, but sometimes important observations on heterogeneity were made; for example, in osteoporosis fluoride treatment is accompanied by increased size of crystals associated with loss of integration with the organic matrix, restricted to newly formed bone,18 whereas both fracture and bisphosphonate therapy resulted in an increased mineral-to-matrix ratio16, 19 and decreasing heterogeneity.15, 20 Furthermore, in an assessment of the degree of matrix mineralization based on backscatter electron imaging of iliac crest biopsies from cases of vertebral fractures and controls, it was the subject variability rather than the mean mineralization level that was lower in the fracture cases.21

By assessing the distribution of these parameters across microscopic fields, the level of heterogeneity for selected parameters in the bone tissue can be analyzed. Healthy bone is heterogeneous, and there is now evidence that as hypothesized by Keaveny and Hayes,22 this heterogeneity improves bone's mechanical properties, perhaps by redistributing mechanical strain levels from the surface, where cracks can initiate comparatively easily, to the less vulnerable interior of trabecular bone.23 Conversely, increasing the homogeneity of these properties might increase bone's brittleness.

The purpose of this study was to test the hypotheses that adverse changes in bone mineral and matrix properties may characterize femoral neck bone in cases of hip fracture. We also wished to examine variations between the anatomical quadrants of the femoral neck because of their different experiences of mechanical loading in the ambulant elderly.

Materials and Methods

Samples and analyses

Complete, unbroken, femoral neck cross sections from 10 women aged 65 to 91 years were obtained at hemi-arthroplasty after an intracapsular hip fracture resulting from a fall from standing height or less. Fracture fixation was performed within 48 hours of the fracture. Written informed consent to participate in the study was obtained from all patients. Equivalent material from control necropsies, matched for age (74 to 89 years) and sex, with no history of disease such as carcinoma or use of drugs known to affect bone metabolism were used as controls. These control subjects died within 7 days of admission to hospital; exclusions were cancer, local bone disease, and prior immobility. All patients were naive to bisphosphonate therapy. The quasi-ellipse–shaped femoral neck cross sections (Fig. 1) had minor diameters of 30 ± 5 mm and major diameters of 35 ± 5 mm. The thickness (along the femoral neck axis) of the bone samples ranged from 2 to 5 mm.

Figure 1.

(A) Sketch of the cross section of the femoral neck at mid-cortex. For the purposes of our study, the sections were divided into four quadrants: anterior, superior, posterior, and inferior (from Bell and colleagues10). (B) Histogram of pixel values from a typical FTIRI image to show the derivation of the full width at half maximum (FWHM) parameter used to characterize FTIRI heterogeneity (see text).

Each biopsy was fixed with 80% ethanol (EtOH) and slowly dehydrated through a series of increasing concentrations of EtOH, cleared with xylene, and finally infiltrated and embedded in polymethylmethacrylate (PMMA) using the Erben method.24 All but one biopsy was evaluated histomorphometrically, after sectioning of the whole specimen. Cortical thickness25 and porosity10 was measured in 20-µm sections, and cortical turnover was evaluated by counting cortical canals that were being eroded, were covered in a layer of osteoid, or were quiescent.26 Turnover rates as measured by histomorphometry differed in the anterior portions of the fracture cases and controls.26 The blocks were later orientated and divided into four quadrants as indicated in Fig. 1A to provide four blocks per biopsy.

For the FTIRI analysis, 2- to 3-µm-thick sections were cut from each block, normal to the femoral neck axis with a Leica microtome (Leica SM 2500, Leica, Germany) and mounted on barium fluoride infrared windows (SpectraTek, Hopewell Junction, NY, USA). Each biopsy section was examined using a Perkin Elmer Spotlight 300 Infrared Imaging system (Perkin Elmer Instruments, Waltham, MA, USA) at a spectral resolution of 4 cm−1. Images from cortical and trabecular regions of each quadrant as defined in Fig. 1 were analyzed separately. Background (BaF2 window only) and PMMA spectra were collected for each section analyzed, and these spectra were used for correction of the sample spectral data. Spectra were baseline corrected and the PMMA spectral contribution subtracted using ISYS software (Spectral Dimensions, Olney, MD, USA) as previously described.27 The following FTIR parameters, reviewed in detail elsewhere,14 were calculated using ISYS software. Mineral-to-matrix ratio, which measures bone mineral content (correlated to ash weight) is calculated from the integrated area of phosphate (916 to 1180 cm−1)/Amide I (1592 to 1712 cm−1) peaks. The carbonate-to-mineral ratio (C/P), which reflects the level of carbonate substitution in the hydroxyapatite (HA) crystal, is calculated from the ratio of the integrated area of the ν2 carbonate peak (840 to 892 cm−1) and that of the phosphate. Crystallinity, which is related to mineral crystal size and perfection as determined by X-ray diffraction, was calculated as the 1030/1020 cm−1 peak intensity ratio. The collagen cross-link network maturity (cross-link ratio) was estimated as the intensity ratio of Amide I sub-bands at 1660 and 1690 cm−1. Heterogeneity in each of these indices was assessed by measuring the full width at half maximum (FWHM) of the pixel histograms. These were calculated from distribution of each parameter in each image for trabecular and cortical bone.15, 20

The residual bone blocks were then used for the micro-computed tomography (micro-CT). Each quadrant was analyzed both for trabecular and cortical bone by micro-CT using an eXplore Locus SP micro-CT scanner (GE Healthcare, London, Ontario, Canada). Eighty KVp, 0.4 degrees rotation steps (200 degrees angular range), and 2000 ms exposure per view were used for the scans of PMMA-embedded samples, giving a 28-µm voxel resolution. Aluminum 0.02-mm foil was used for reducing the beam hardening effect. MicroView GEHC 2.2 software (GE Healthcare) was used for 3D reconstruction and viewing of images as detailed elsewhere.28 After reconstruction, volumes with thickness of 100 µm adjacent to the surface used for FTIRI analysis were segmented using a global threshold of 0.4 g/cm3. Directly measured tissue volume (TV), bone volume (BV), bone volume fraction (BV/TV), total mineral content (TMC), and tissue mineral density (TMD) along with other parameters were calculated for both cortical and trabecular bone, using instrument-provided software.

Statistical analysis

Statistical normality of the distributions of residuals (required for valid linear regression analysis) was assessed by quantile-quantile (Q-Q) probability plots and the Shapiro-Wilk W test. Untransformed distributions were acceptably normal except for the heterogeneities; log transformation was needed to correct the skewness of the distributions (Shapiro Wilk W test p values before log transformation all <0.004). After transformation, only the cross-link heterogeneities were still significantly (p < 0.05) positively skewed. Two-way ANOVA was used to assess the effects of anatomical quadrant and disease status (hip fracture case versus control) on the outcome variables of interest. The micro-CT data were analyzed using SigmaStat (Systat Software Inc., San Jose, CA, USA). Differences between means of the analyzed variables were assessed for each quadrant and between fracture cases and controls. The data were also assessed between trabecular and cortical bone (all quadrants) and between fracture cases and controls. The FTIRI outcome variables (four quadrants each for mean values and heterogeneities: mineral-to-matrix ratio, crystallinity, carbonate-to-phosphate ratio, and collagen cross-link ratio) were evaluated separately for cortical and trabecular bone. Each subject was nested within his or her appropriate fracture or control group, and within-group variation between subjects was assumed to be statistically random, drawn from either a normal or a log-normal distribution as appropriate. Multiple linear regression models were also constructed, making the outcome variable a function of hip fracture status, anatomical quadrant, and an interaction between fracture status and quadrant. The reproducibility of the results across the three measured regions of interest per quadrant was assessed from the root mean square (RMS) error. When log transformations were needed, geometric means are reported, whereas variability is reported as the relative standard error (calculated as % geometric mean) and differences between means are reported as a percentage increase or reduction compared with the control group. Because of multiple testing (four outcome measures), we set our significance level at p < 0.012 for each of the FTIRI results (ie, p < 0.05/4; 2-tailed test29).

Indirect evidence (see, eg, Verdelis and colleagues30) suggests a relationship between carbonate-to-phosphate ratio and crystallinity in which carbonate may contribute to limiting crystal size. To examine the statistical dependence of mean crystallinity on carbonate-to-phosphate ratio, regression analysis was used with crystallinity as the outcome variable, adjusting for case versus control status and bone quadrant.


The fracture cases studied and their controls ranged in age from 65 to 91 years. The histomorphometric analyses of the properties of cortical bone in these femoral neck biopsies are shown in Table 1. These subjects represent a subset of subjects reported on previously using the same methods.10, 25, 26

Table 1. 2D Morphometry of Femoral Neck Specimens

Cortical thickness25


Cortical porosity10


Osteoid-bearing canals26


Eroded canals26

  1. Controls: mean age 80.4 years, range 68 to 89 years; fracture cases: mean age 77.5, range 65 to 91 years. One control was not analyzed. When tested across all four quadrants (superior, posterior, anterior, and inferior), mean differences between cases and controls in cortical thickness, cortical porosity, osteoid bearing canals, and eroded canals were not significant at p < 0.05 because of substantial within-group variability.


Micro-CT analyses

Based on micro-CT, the tissue mineral density (TMD) was similar across the four bone quadrants for both the fracture cases and the control (Fig. 2A). However, total mineral content (TMC) in the fracture cases was significantly lower than in controls in both cortical and trabecular bone (p < 0.05) with the inferior region showing a reduction of 26% (Fig. 2B). In controls, the inferior quadrant presented higher TMC than the other three quadrants, whereas in the fractures these differences were not significant. No differences were observed between fractures and controls for bone volume fraction (BVF) (Fig. 2C). Bone volume (BV) was lower in the fracture group compared with the controls, which was largely attributable to the inferior quadrant that had a 25% reduction in the fracture cases compared with the controls (Fig. 2D). All other micro-CT parameters calculated using ASBMR guidelines are summarized in Table 2 and show some significant differences between quadrants in the controls that were not observed in the fracture cases. No significant differences were observed between fracture cases and controls for most of the morphometric bone parameters. Only Tt.Ar in the superior quadrant was significantly lower in fracture cases (–50%) compared with controls. Both fracture cases and controls presented variation with anatomical quadrant especially at the level of the inferior quadrant. For the cortical bone morphology, the lower mean values for Tt.Ar was observed for the anterior quadrant in the control cases, and this anatomical specificity was not seen in fractures cases. Ct.Ar was highest in the inferior quadrant in both control and fracture cases. Inferior and anterior quadrant showed higher Ct.Ar/Tt.Ar values in controls. In both control and fracture cases, the highest values were observed in the inferior quadrant. For the trabecular bone morphology, the inferior quadrant had the highest values for BV/TV in the controls, and there were no differences in BV/TV in different quadrants in fracture cases. The inferior quadrant also showed the highest values for Tb.Th in both control and fracture cases.

Figure 2.

Box plot summaries of bone properties analyzed by micro-CT. For both control (C) and fracture (F) groups, (A) tissue mineral density (TMD), (B) bone volume fraction (BVF), (C) tissue mineral content (TMC), and (D) bone volume (BV) were recorded in trabecular (Tb) and cortical (Cort) bone for each quadrant (sup = superior; ant = anterior; inf = inferior; and post = posterior) of the femoral neck biopsies.

Table 2. Micro-Computed Tomography of Fractured Femoral Necks and Age-Matched Control Femoral Necks (Mean ± SD)
ParametersControl cases (n = 10)Fracture cases (n = 10)
Cortical bone morphology
  • Tt.Ar = anterior quadrant lower than the other quadrants in control (p < 0.05), not observed in fracture cases; Ct.Ar = inferior quadrant higher than the other quadrants in control (p < 0.05), in fracture cases inferior higher than superior and anterior quadrants only; Ct.Ar/Tt.Ar = inferior and anterior are significantly different than the superior and posterior quadrants in control, in fracture cases the anterior quadrant does not present any difference compared with the other quadrants; Ct.Th = inferior quadrant higher than the other quadrants in control and fracture cases (p < 0.05); BV/TV = inferior quadrant higher than the other quadrants in control (p < 0.05), not observed in fracture cases; Tb.Th = inferior quadrant higher than the other quadrants in control and fracture cases (p < 0.05).

  • a

    p < 0.05 (inf ≠ sup).

  • b

    p < 0.05 (inf ≠ ant).

  • c

    p < 0.05 (inf ≠ post).

  • d

    p < 0.05 (ant ≠ sup).

  • e

    p < 0.05 (ant ≠ post).

  • f

    p < 0.05 (control ≠ fracture cases).

Tt.Ar77 ( ± 69)86 ( ± 80)108 ( ± 56)17 ( ± 27)b,d,e96 ( ± 71)56 ( ± 61)53 ( ± 62)49 ( ± 36)f40 ( ± 69)80 ( ± 36)
Ct.Ar15 ( ± 14)31 ( ± 16)a,b,c10 ( ± 7)5.5 ( ± 4)15 ( ± 13)12 ( ± 14)22 ( ± 19)a,b7 ( ± 3)7 ( ± 9)13 ( ± 14)
Ct.Ar/Tt.Ar0.43 ( ± 0.3)0.63 ( ± 0.3)a,c0.107 ( ± 0.06)0.73 ( ± 0.3)d,e0.26 ( ± 0.27)0.44 ( ± 0.3)0.72 ( ± 0.3)a,c0.208 ( ± 0.13)0.509 ( ± 0.31)0.32 ( ± 0.36)
Ct.Th0.096 ( ± 0.6)1.9 ( ± 0.72)a,b,c0.54 ( ± 0.14)0.77 ( ± 0.3)0.63 ( ± 0.13)0.84 ( ± 0.5)1.7 ( ± 0.6)a,b,c0.56 ( ± 0.14)0.59 ( ± 0.16)0.57 ( ± 0.12)
 Trabecular bone morphology
BV/TV0.12 ( ± 0.07)0.17 ( ± 0.09)a,b,c0.104 ( ± 0.04)0.102 ( ± 0.04)0.105 ( ± 0.06)0.1 ( ± 0.04)0.12 ( ± 0.06)0.1 ( ± 0.03)0.096 ( ± 0.04)0.09 ( ± 0.02)
Tb.N1.04 ( ± 0.35)1.14 ( ± 0.44)1.08 ( ± 0.33)0.93 ( ± 0.31)0.99 ( ± 0.3)1.04 ( ± 0.35)0.98 ( ± 0.35)0.99 ( ± 0.3)0.906 ( ± 0.44)0.97 ( ± 0.27)
Tb.Th0.11 ( ± 0.03)0.14 ( ± 0.03)a,b,c0.094 ( ± 0.01)0.106 ( ± 0.01)0.101 ( ± 0.02)0.11 ( ± 0.03)0.13 ( ± 0.04)a,b,c0.102 ( ± 0.01)0.11 ( ± 0.04)0.094 ( ± 0.01)
Tb.Sp0.97 ( ± 0.4)0.88 ( ± 0.48)0.9 ( ± 0.29)1.08 ( ± 0.43)1.02 ( ± 0.42)0.97 ( ± 0.4)1.06 ( ± 0.54)0.98 ( ± 0.26)1.2 ( ± 0.51)1.01 ( ± 0.33)

Effects of hip fracture status and anatomical quadrant on mean FTIRI parameters

Typical FTIRI images recorded in the inferior cortical quadrant are illustrated for one patient from the control group (Fig. 3A, a) and one patient from the fracture group (Fig. 3A, b). These two subjects had comparable TMD and similar ages. The coefficients of variation between regions of the bone images as assessed by RMS errors derived from the best-fit models for cortical bone were 5.3%, 14.0%, 4.4%, and 1.9% for mineral-to-matrix ratio, carbonate-to-phosphate ratio, collagen maturity, and crystallinity, respectively. The equivalent percentages for trabecular bone were 6.7%, 16.7%, 4.6%, and 2.9%. These coefficients of variation were adjusted for mean differences between quadrants.

Figure 3.

(A) Typical infrared images for the IR parameters recorded in inferior cortical quadrant from two patients, one with fractures and one without. Patient A had no fractures and was 85 years old at time of biopsy. The micro-CT data for the cortical bone of this patient were: TMD = 867, TMC = 12.8, BVF = 0.2, and Ct.Ar/Tt.Ar = 0.42. Patient B had a hip fracture and was 86 years old at time of biopsy. The micro-CT data for the cortical bone of this patient were: TMD = 858, TMC = 9.08, BVF = 0.14, and Ct.Ar/Tt.Ar = 0.55. Numerical values above the images are the means ± SEM for the parameter in the figure, and indicate the range of data for the pixels shown. 1 pixel = 6.25 microns. (B) Variations in mean FTIRI indices in trabecular and cortical bone of fracture cases and control subjects, averaged across all subjects in each group. Individual points refer to means for each quadrant and lines connect means averaged across all quadrants for fracture cases and controls separately. For statistical analyses of differences, see text. (C) Variations in the logarithms of the FWHMs (heterogeneities) of the four FTIRI indices, displayed as for (B). For statistical analysis, see text.

In cortical bone, the mean mineral-to-matrix ratio was 5.7% lower in the fracture cases compared with the controls (p = 0.009). In trabecular bone, the mean mineral-to-matrix ratio was 10.1% lower in the fracture cases compared with the controls (p = 0.003; Fig. 3B). In cortical as well as trabecular bone, in both fracture cases and controls, mineral-to-matrix ratio varied significantly with anatomical quadrant, the highest mean values occurring inferiorly and the lowest anteriorly (6.2% and 5.8% reductions in anterior compared with inferior, respectively). Mean overall carbonate-to-phosphate ratios and mean overall collagen maturity ratios (Fig. 3B) did not differ between fracture cases and controls or between quadrants. Although there were apparent differences between fracture cases and controls in collagen maturity ratios, these were masked by the substantially larger differences between individual subjects within each group, resulting in a large overlap between fracture cases and controls. Collagen maturity ratios were larger in the anterior than in the posterior quadrants of trabecular bone, and this difference was 4.5% larger in magnitude in the fracture cases compared with the controls (p = 0.007).

Mean crystallinity (Fig. 3B) showed no significant overall differences between fracture cases and controls in cortical or trabecular bone. However, mean crystallinity varied by anatomical quadrant. In trabecular bone, the inferior quadrant had a significantly higher mean crystallinity (2.25% above the overall mean; p < 0.0001) and the same quadrant had higher crystallinity than average by 1.44% in cortical bone (p < 0.0001). The posterior quadrant's mean crystallinity values were similar to the inferior quadrant's in the controls but lower than the inferior quadrant's by ∼3% in the fracture cases (p for differences: <0.011 for trabecular bone; <0.0001 for cortical bone).

Effects of hip fracture status and anatomical quadrant on heterogeneity (FWHM)

Heterogeneity of infrared parameters differed when fracture cases and controls were compared. The within-quadrant heterogeneity of mineral-to-matrix ratio was more variable than the mean mineral-to-matrix ratio, with coefficients of variation of 10.2% and 13.5% in cortical and trabecular bone, respectively. Overall, the heterogeneity of the mineral-to-matrix ratio was significantly reduced (p < 0.001) in the fracture group compared with the controls in both trabecular and cortical bone (by 19% and 16%, respectively) (Fig. 3C). Between quadrants, there was also significant variation in the heterogeneity of the mineral-to-matrix ratio that was comparable in both fracture cases and controls and in trabecular as well as in cortical bone (p = 0.008 for both). Posteriorly, heterogeneity was 7% to 10% greater than in the superior or anterior quadrants.

The carbonate-to-phosphate ratio heterogeneity showed posterior quadrant-specific reductions in hip fracture compared with control values (27% and 23% in trabecular and cortical bone, respectively). There were no comparable differences in carbonate-to-phosphate ratio heterogeneity in the three other quadrants. Neither were significant differences in heterogeneity seen between fracture cases and controls or between quadrants in collagen maturity ratios (p > 0.05) (Fig. 3C).

The crystallinity parameter (1030/1020 cm−1 peak height ratio) showed between-region differences as well as differences between fracture cases and controls. In both trabecular and cortical bone, the heterogeneity of crystallinity was lowest in the inferior quadrant (which also had higher mean values), heterogeneity being only 81% of that seen superiorly in the trabecular bone and 88% of that seen superiorly in the cortical bone. The heterogeneity of crystallinity in the cortical bone also was highest posteriorly (12% higher than superiorly). In cortical bone, but not trabecular bone, there was an additional contrast between fracture cases and controls: Fracture cases had a 30% higher heterogeneity of crystallinity posteriorly and a 17% higher heterogeneity inferiorly than controls.

The statistical variations of anatomical quadrant with crystallinity (p < 0.0001) and carbonate-to-phosphate ratio (p = 0.0009) in controls were significant. Hip fracture had a weaker effect, increasing the overall crystallinity by 8% (p = 0.015) through its interactions with anatomical quadrant and carbonate-to-phosphate ratio. Overall, 58% of the heterogeneity in crystallinity was accounted for by these independent variables. Fig. 4 shows the associations between mean values for crystallinity and carbonate-to-phosphate ratio in each quadrant for fracture cases and controls. These results suggested the possibility that the heterogeneities of crystallinity and carbonate-to-phosphate ratio might be causally associated.

Figure 4.

Scatter plots of relationships between the means of the logarithms of the carbonate: phosphate ratio and crystallinity for trabecular (dotted line) and cortical bone (solid line), respectively. Each point represents the mean result for one quadrant in one diagnostic group. The regression lines shown, with their 95% confidence intervals, are unadjusted for the effects of diagnosis, subject, or anatomical quadrant. Triangles = cases of hip fracture; circles = controls.

Because anatomical quadrant has pronounced statistical effects on such properties as cortical thickness, porosity and remodeling, we questioned whether each quadrant was acting as a surrogate for structural strength or remodeling activity. We also asked whether the differences between cortical and trabecular bone reflected the much greater “porosity” of trabecular bone. The effects of the structural and remodeling indices listed in Table 1 on the means of the four FTIRI outcome measures were not statistically significant (p > 0.05). However, increasing cortical porosity did have a marginally significant statistical effect to reduce the heterogeneity of the mineral-to-matrix ratio (p = 0.049), which was independent of the reduction in heterogeneity also associated with hip fracture.

In a model of the determinants of the log-transformed heterogeneities that included quadrant and fracture status, trabecular bone had generally higher heterogeneities of all the FTIRI-derived measurements compared with cortical bone. There was overall a 32% higher heterogeneity of the mineral-to-matrix ratio (p < 0.001) and a 17% higher heterogeneity of the carbonate-to-phosphate ratio (p < 0.001), an 11% higher heterogeneity of the collagen cross-links ratios (p < 0.001), and a 32% higher heterogeneity of the crystallinity (p < 0.0001) compared with cortical bone. These differences between cortical and trabecular bone were not significantly affected by fracture history or by anatomical quadrant (p > 0.05).


In this study, we found several bone FTIR spectroscopic imaging compositional properties associated with femoral neck fracture. Mineral-to-matrix ratio, previously reported,31, 32 was generally reduced in fracture cases compared with controls. Otherwise, the differences found were in the heterogeneity of these FTIRI properties. In the herein studied fracture cases, there was reduced heterogeneity (ie, microscopic variation) in the mineral-to-matrix and carbonate-to-phosphate ratios amounting to more than 15% and 20%, respectively, in all quadrants, measured in cortical or trabecular bone.

Cracks in bone may initiate at locations of increased stress, eg, where there is discontinuity such as an osteocyte lacuna or a Howship's lacuna.33 The growth of cracks is resisted by processes that increase energy absorption during fracture (toughening mechanisms). The effect of length scale in determining the contribution of toughening mechanisms to fracture resistance was investigated by Zimmermann and colleagues,34 who found that increased mean mineralization levels are associated with reduced toughness at the micrometer scale; however, submicrometer scale loss of toughness was associated with nonenzymatic collagen cross-linking. The importance of length scale to the effects of toughening mechanisms may also be considered in relation to the development of cracks, which in a tough material like normal bone are generally restricted in length. As a crack develops, it tends to accelerate, a process inhibited by toughening mechanisms that absorb the energy of the crack. Failure of “crack capture” by submicrometer mechanisms require that other mechanisms, capable of absorbing considerably more energy at micrometer scales, be present to prevent the crack from developing into a clinical fracture. At the scale level of the whole cancellous bony trabeculae, there is uncertainty as to how inhomogeneity in calcium content influences overall toughness. Inhomogeneity was associated with increased shear stresses.35 Similarly, a finite element modeling study by Renders and colleagues23 suggests that inhomogeneity also internalizes these stresses. This process is likely to help protect the trabeculae from mechanical failure, by diverting crack-derived stresses away from surface stress-enhancers such as osteoclastic resorption lacunae.

Although FTIRI technology enables us to assess FTIR properties of an individual pixel with an area of 39 × 10−6 mm2, heterogeneity was measured across areas slightly less than 0.2 mm2, considerably larger than the cross-sectional area of normal osteons, typically 20-fold smaller. Reduced mean mineral density in hip fracture cases, found in our previous studies, was based on different individual subjects from the same cohorts.31, 32 There were no comparable reductions in modulus,32 which might be relevant to the toughness of bone at multiple scale lengths. Our results indicating FTIRI heterogeneity are principal to the study of mechanisms opposing the largest cracks that escape capture at the submicrometer scale.

The rate of normal remodeling in the femoral neck is roughly 10-fold lower than that in iliac trabecular bone.36 An explanation for decreased heterogeneity of mineral content in fracture cases is a less actively remodeled tissue compared with controls, possibly associated with reduced osteocyte density adjacent to the canal.37 The proportion of cortical osteons undergoing remodeling is not reduced in cases of hip fracture. This implies that there might be a longer formative phase in each remodeling osteon, potentially driven by increased late-phase sclerostin expression.38 An alternative explanation is the physio-chemical differences (ie, lower levels of calcium and/or phosphate) in the bone micro-environment. The much higher turnover in the iliac crest than in the femoral neck bone might explain why Ciarelli and colleagues found increased mineralization heterogeneity in iliac bone from osteoporotic vertebral fracture cases21 in contrast to our results. Ciarelli and colleagues later found that this increased iliac heterogeneity was attributable to variations in mineralization between different BMUs according to whether they were recently formed or were older.39 Ciarelli and colleagues similarly found reductions in differences between the mineralization levels in adjacent light and dark lamellae in some categories of osteoporosis.40 Present resolution of FTIRI technology (∼7 µm) is unable to detect differing mineral-to-matrix ratios between adjacent lamellae. We cannot, as yet, explain the discrepancy between our results and those of Bousson and colleagues,41 who found increased heterogeneity of mineralization in fractured inferior femoral neck bone using microradiography.

Micro-CT data showed a slight but not significantly reduced mineral density in the fracture group. Relative differences in mineral density between fracture and nonfracture groups were smaller than estimated by FTIRI in this study and by BSE imaging in previous studies.31, 32 These differences were likely owing, in part, to the micro-CT partial volume effect, which can be substantial in analysis of the outermost 100 µm of a cortex42 (as little as 300 µm thick43). Moreover, with FTIR spectroscopy, the mineral content is assessed by the mineral-to-matrix ratio. This includes all tissues present and cannot be directly compared with TMD or BMD, which only consider those regions where the chosen density threshold defines the presence of mineral. These results highlight the importance of both the mineral and the organic matrix in contributing to bone strength.44 One should be cognizant of the limitations of micro-CT methods when variance in organic matrix composition may be involved.

Crystallinity as well as mineral-to-matrix ratio was most homogeneous in the inferior quadrant, whereas the heterogeneity of crystallinity was increased in the posterior quadrant in the hip fracture cases. It is generally accepted that in standing and walking, the inferior quadrant undergoes compression loading, whereas the posterior quadrant lies across the neutral axis,45 where the bone is neither compressed nor under tension during stance. However, in a sideways fall, compression loading is maximal postero-superiorly, a quadrant that is lightly loaded during walking.45

Biological hydroxyapatite is a carbonate-substituted apatite of low crystallinity.44 Carbonate substitution influences bone properties, increases bone solubility,44 increases with aging,5, 46, 47 and is locally regulated by osteocalcin.48 High carbonate substitution in hydroxyapatite has been linked to increased bone dissolution49 and osteoclastic resorption50 and is related to fracture events in iliac crest biopsies.5 Interest in crystallinity has been long-standing because of previously reported tendency for hip fracture cases to develop some very large crystals51 and the likelihood that such crystals are brittle.52, 53 The inverse association between crystallinity and the carbonate-to-phosphate ratio present in the current study suggests a role for carbonate substitution in determining crystallinity because the larger, more perfect crystals, being more stable, should have a lower carbonate substitution.44 We expected higher carbonate content in the fracture group because this would reflect the presence of a less mature bone. We found instead that carbonate substitution was significantly more homogeneous in the fracture group, whereas mean levels were similar to controls. In contrast to previous FTIR imaging studies of osteoporotic iliac crest biopsies,54, 55 no increase in mean bone crystal size was observed in the hip fracture cases, which is in agreement with recent results obtained by scanning small-angle X-ray scattering.32 Crystallinity, however, reflecting crystal size and perfection, was significantly more heterogeneous, suggesting the possibility that the wide size range seen may indicate an excess of brittle crystals in hip fractures with dimensions exceeding the postulated critical length of 30 nm.53

Some clinical treatments may increase the variance of normal values for heterogeneity of FTIRI parameters (mineral-to-matrix ratio, collagen maturity, carbonate-to-phosphate ratio, and crystallinity) measured in treatment-naive cases of hip fracture. Osteoporotic patients treated with bisphosphonates are likely to be affected by further reductions in heterogeneity. This decrease in bone heterogeneity, observed in dog tibias after 1 year's bisphosphonate treatment,15, 20 was used to explain the significant accumulation of microdamage observed in vertebrae of dogs given similar treatments.56, 57 This association raises the possibility that reduced heterogeneity per se, in treatment or disease, is detrimental to bone's toughness. Healthy bone tissue is highly heterogeneous, and there is evidence that this heterogeneity affects the mechanical properties of the bone material by preventing the proliferation of micro-cracks.15, 22, 23 Reduced heterogeneity at the nano- and macro-level is found in elderly patients,34 and reduced heterogeneity in modulus and hardness was also found in bisphosphonate-treated patients with subtrochanteric fractures,15 although the heterogeneity of plastic deformation resistance was increased in the same study. The beneficial effect of bisphosphonates in reducing remodeling by slowing the excessive removal of bone in osteoporotic patients58–60) might be counterbalanced by the detrimental effects of limiting repair of micro-cracks and causing the development of excessive homogeneity, each of which would lead the bone to become more brittle61 and hence more likely to fracture. Therefore, bisphosphonates protect against fracture by maintaining bone mass. A loss of bone mass under the same amount of force would decrease bending resistance or increase the applied stress, leading to fracture. Changes in homogeneity and the decrease in repair of micro-cracks62 might increase fracture risk despite a similar bone mass.

In conclusion, we have confirmed that hip fracture cases naive to bisphosphonate therapy show reduced mineralization of their bone matrix compared with controls. This may contribute to their increased risk43 of mechanically loading parts of the femoral neck beyond the so-called “yield point” at which damage-inducing plastic deformation takes place. In addition, hip fracture cases showed less heterogeneity of their mineral-to-matrix and carbonate-to-phosphate ratios, raising concerns that this could contribute to lower bone toughness, leading to fracture. Also, there was a greater heterogeneity of crystallinity, which might increase the proportion of fragile crystals in fracture cases compared with the controls. Thus, the mineral phase of femoral neck bone in hip fracture is different from control bone in several ways that are not captured by the clinical technique of bone densitometry. With aging, in contrast to osteoporosis, both mineralization and the carbonate-to-phosphate ratio increase progressively, alongside indentation modulus and hardness, at least in baboons.63 Further studies are needed to ascertain the effects of these differences on whole bone fragility and on how current and future treatments affect these abnormal properties of femoral neck bone in those at risk of hip fracture.


All authors state that they have no conflicts of interest.


The samples used in the study were part of our library of femoral neck biopsies obtained under previous local ethics permission (REC 07/H0305/61 & P97/256) and stored under HTA regulations (license no. 12318). We acknowledge the help of our orthopaedic colleagues (principally Mr Martyn Parker and Prof Neil Rushton) for obtaining the biopsies and collaboration with previously published aspects of this work. We appreciate the editorial assistance of Dr J Gerstein. This study was supported by NIH grants R01AR041325 and AR046121; MRC (UK) grant G9321536; NIHR (UK) Cambridge Biomedical Research Centre Grant; and EU FP7 Integrated Project grant TALOS 201099.

Authors' roles: SG-A performed the IR analysis and wrote and edited the first draft of the manuscript. LL performed and interpreted the micro-CT analysis and edited the manuscript. JP prepared the embedded biopsy material, performed the light microscope histomorphometry, and edited the manuscript. NL advised on the interpretation of the data, edited the manuscript, and with JR, wrote the non-NIH grants that supported collection and initial processing of the biopsies and light microscopy data. JR also suggested the study and was responsible for the statistics. AB wrote the grant that supported the study, coordinated the study performance, developed the study design, and edited the final version of the study.