Impact microindentation assesses subperiosteal bone material properties in humans

Journal Pre-proof Impact Microindentation Assesses Subperiosteal Bone Material Properties in Humans Stamatia Rokidi, Natalie Bravenboer, Sonja Gamsjaeger, Barbara ´ Misof, Stephane Blouin, Pascale Chavassieux, Klaus Klaushofer, Eleftherios Paschalis, Socrates Papapoulos, Natasha Appelman-Dijkstra

PII:

S8756-3282(19)30403-X

DOI:

https://doi.org/10.1016/j.bone.2019.115110

Reference:

BON 115110

To appear in: Received Date:

7 August 2019

Revised Date:

14 September 2019

Accepted Date:

14 October 2019

Please cite this article as: Rokidi S, Bravenboer N, Gamsjaeger S, Misof B, Blouin S, Chavassieux P, Klaushofer K, Paschalis E, Papapoulos S, Appelman-Dijkstra N, Impact Microindentation Assesses Subperiosteal Bone Material Properties in Humans, Bone (2019), doi: https://doi.org/10.1016/j.bone.2019.115110

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier.

Impact Microindentation Assesses Subperiosteal Bone Material Properties in Humans Stamatia Rokidi1, Natalie Bravenboer2, Sonja Gamsjaeger1, Barbara Misof1, Stéphane Blouin1, Pascale Chavassieux3, Klaus Klaushofer1, Eleftherios Paschalis1*, Socrates Papapoulos2, Natasha Appelman-Dijkstra2. 1

Ludwig Boltzmann Institute of Osteology at Hanusch Hospital of Viennese sickness insurance

funds (WGKK) and Research funds of the Austrian workers compensation board (AUVA)

ro of

Trauma Centre Meidling, 1st Medical Department, Hanusch Hospital Vienna, Austria. 2 Leiden Center for Bone Quality, Leiden University Medical Center, Leiden, The Netherlands. 3

INSERM UMR 1033, University of Lyon, Lyon, France

-p

Short Title: BMSi and bone material properties *Address for correspondence: Eleftherios P. Paschalis, PhD

re

Ludwig Boltzmann Institute of Osteology

lP

at the Hanusch Hospital of WGKK and AUVA Trauma Centre Meidling,

na

1st Medical Department, Hanusch Hospital,

Jo

ur

Heinrich Collin Str. 30, A-1140 Vienna Austria [email protected]

Disclosures: The study was not supported by external funding. SEP and NA-D are unpaid members of the Scientific Advisory Board of Life Science Scientific Inc.

Highlights    

Impact microindentation measures tissue-level properties of cortical bone in vivo The nature of the measured properties is incompletely understood We analyzed iliac crest bone biopsies obtained concurrently with IMI measurements Nanoporosity and Pyridinoline content are important determinants of BMSi

Abstract

ro of

Impact microindentation (IMI) is a Reference Point Indentation technique measuring tissuelevel properties of cortical bone in humans in vivo. The nature, however, of the properties that can affect bone strength is incompletely understood. In the present study we examined bone material properties in transiliac bone biopsies obtained concurrently with measurements of

-p

Bone Material Strength index (BMSi) by IMI in 12 patients with different skeletal disorders

re

and a wide range of BMD, with or without fractures (8 males, 4 females, mean age 48±12.2 (SD) years, range 15-60 years). IMI was performed in the mid-shaft of the right tibia with a

lP

hand‐ held microindenter (OsteoProbe). Cancellous and cortical bone mineralization density distributions (BMDD) were measured in the entire biopsy bone area by quantitative

na

backscattered electron imaging. Raman measurements were obtained right at the outer edge of the cortex, and 5, 50, 100, 500 µm inwards. The calculated parameters were: i) Mineral and

ur

organic matrix content as well as the mineral / matrix ratio. ii) Nanoporosity. iii) Glycosaminoglycan content. iv) Pyridinoline content. v) Maturity/crystallinity of the apatite

Jo

crystallites. There was no relationship between BMSi values with any measurement of mineral content of whole bone tissue (BMD, BMDD) or maturity/crystallinity of bone mineral. On the other hand, a positive correlation between BMSi and local mineral content, and an inverse correlation between BMSi and nanoporosity at the mineralized subperiosteal edge of the sample and at 5 μm inwards was found. A positive correlation was also observed between BMSi and pyridinoline content at the same locations. These results indicate that local mineral

content, nanoporosity and pyridinoline content at the subperiosteal site in the transiliac bone biopsy are linked to the BMSi values measured in the tibia. As both high porosity at the nano level and low pyridinoline content of the bone matrix can negatively impact bone strength, our

Jo

ur

na

lP

re

-p

ro of

findings suggest that BMSi most likely assesses subperiosteal bone material properties.

Introduction Impact microindentation (IMI) is a minimally invasive Reference Point Indentation technique that measures tissue-level properties of cortical bone in humans in vivo at the mid-shaft of the tibia [1]. It is performed by a hand-held device (OsteoProbe) that imparts a single impact load to the bone surface. By driving the probe into the bone surface, the resistance of bone tissue to a given mechanical challenge can be measured as Bone Material Strength index (BMSi). The methodology and clinical utility of IMI have been reviewed [2-4]. Clinical studies have shown

ro of

that the method can generally differentiate patient groups at increased risk for fracture from control groups independently of Bone Mineral Density (BMD) values [5-8]. Moreover, significant decreases in BMSi were reported after seven weeks treatment with glucocorticoids,

-p

that lead to rapid deterioration of bone strength with a subsequent increase in fracture risk, without any concurrent changes in BMD [9]. These findings strongly suggest that IMI

re

measures properties of cortical bone that are related to bone strength but are not captured by

lP

BMD [10-12]. Such properties may include bone structure and composition. A recent study showed that BMSi is not associated with parameters of cortical bone microstructure measured by HR-pQCT, including cortical porosity and cortical tissue mineral density, in

na

postmenopausal women with or without femoral fractures [13]. The relationship between BMSi and bone tissue composition has not yet been investigated in vivo but a significant correlation

ur

between BMSi and advanced glycation endproducts (non-enzymatic collagen cross-links;

Jo

AGEs) accumulation was reported in cadaveric female tibiae [14]. In the present study we examined bone material properties that influence bone strength in iliac bone biopsies obtained concurrently with measurements of BMSi by IMI in patients with different skeletal disorders and a wide range of BMD with or without fractures. Patients and Methods Patients

We studied patients with different bone disorders, attending the Out-Patient Clinic of the Center for Bone Quality of the Leiden University Medical Center (LUMC), who were planned for a diagnostic transiliac bone biopsy and consented to also have IMI of the tibia in vivo. There were no other selection criteria and the only exclusion criteria were localized tibia infection and allergy to local anaesthetic precluding the performance of IMI. The study was approved by the Medical Ethics Committee of LUMC and all participants provided written informed consent.

ro of

Bone mineral density

Bone mineral density was measured at the lumbar spine (L1–L4) and at the right hip by dual-

-p

energy X-ray absorptiometry (DXA) using Hologic QDR 4500 (Hologic, Bedford, MA, USA) and T-scores were calculated using NHANES III reference values compatible with reference

re

values for the Dutch population.

lP

Impact microindentation (IMI)

IMI was performed in all subjects in the mid-shaft of the right tibia with a hand‐ held

na

microindenter (OsteoProbe; Active Life Scientific, CA, USA) by a single operator as previously described and recommended by a group of experts [5, 15]. In brief, the patient is

ur

placed in a decubitus supine position with the tibia in external rotation to orient the flat surface of the medial tibia diaphysis in a horizontal position. The measurement site is located at the

Jo

mean distance between the distal apex of the patella and medial malleolus. Following disinfection of the area and local anaesthesia of the skin and periosteum with Lidocaine 1%, the test probe is gently inserted in the skin until the bone surface was reached. The operator ensures that the test probe is placed perpendicularly to the bone surface and classifies the measurements as “well performed,” “adequate,” or “poorly performed” after the indentation and before checking the computer display of the result according to the following criteria :

“well performed” when the operator judged that the test probe was exactly perpendicular to the bone surface; “adequate” when the test probe was within acceptable deviation from the bone surface; “poorly performed” when the operator judged that the test probe was not appropriately placed. “Poorly performed” measurements are usually due to slipping of the test probe, moving of the subject’s leg or failure to place the device perpendicularly to the bone surface and are excluded from the analysis. Five additional measurements are performed on a polymethylmethacrylate (PMMA) calibration phantom firmly secured in a holder and placed

ro of

on a stable surface. BMSi is calculated as 100 times the harmonic mean of the indentation distance increase from impact into the PMMA material divided by the average indentation distance increase from impact into bone. As the probe indents the surface of the cortical bone

-p

of the tibia, it induces a microfracture. The more easily this occurs, the deeper the probe indents

re

the bone, and thus, the lower the BMSi.

The intra-observer coefficient of variation (CV) of the technique was calculated to be 2.2% in

lP

a previous study of 10 patients with low bone mass with or without fractures by measuring BMSi twice in the right leg according to the above-mentioned protocol with the second

na

investigation performed 2 centimetres below the first measurement site [6]. In addition, in a different group of 11 subjects no differences in BMSi were observed after five or 10 adequate

ur

measurements or between the dominant and non-dominant leg [15].

Jo

Bone biopsies

Transiliac bone biopsies were performed with a Bordier trephine (8.0 mm internal diameter) following double fluorochrome labelling with 200 mg tetracycline HCL qid according to a schedule of 2 days on, 10 days off and 2 days on. Five to seven days after the last tetracycline dose, the biopsy was taken and stored in 70% ethanol and after dehydration it was embedded in methylmethacrylate as previously described [16].

Quantitative backscattered electron imaging (qBEI) Cancellous and cortical bone mineralization density distributions (Cn. and Ct.BMDD) were analysed using quantitative backscattered electron imaging (qBEI). Full details, including the validation, precision of this technique and its application to cancellous and cortical bone have been published previously [17, 18]. The signal intensity of the backscattered electrons is proportional to the local calcium concentration of the analysed bone section. The acquired digital images are employed for the evaluation of the grey-level histograms, which are

ro of

subsequently transformed to weight percent calcium histograms (bone mineralization density distributions, BMDD). Five variables were considered from the BMDD: CaMean, the weighted mean Ca-concentration of the bone area; CaPeak, the mode calcium concentration (the peak

-p

position of the histogram), representing the most frequently occurring calcium concentration

re

of the analysed bone area; CaWidth, the full width at half maximum of the distribution, describing the variation in mineralization density; CaLow, the percentage of mineralized bone

lP

with a calcium concentration less than the 5th percentile of the reference BMDD (less than 17.68 weight percent calcium) signifying the amount of bone area undergoing primary

na

mineralization; and CaHigh, the portion of bone areas with a calcium concentration higher than the 95th percentile (higher than 25.30 wt% Ca) of the reference BMDD (predominantly fully

ur

mineralized interstitial bone) [17, 18].

The whole surface of the longitudinally sectioned biopsy cylinder containing the trabecular and

Jo

cortical compartments was analysed for Cn.BMDD and Ct. BMDD. In biopsy samples with two cortices, the Ct.BMDD parameters represent the mean of both cortices, in cases where only one cortex was available, Ct.BMDD represents the outcomes of this cortex. Raman microspectroscopy

A Senterra (Bruker Optik GmbH) instrument, with a spatial resolution of 1 x 1 µm2 was used for the Raman microspectroscopic analysis. A continuous laser beam was focused onto the sample through a Raman fluorescence microscope (Olympus BX51, objective 50x) with an excitation of 785 nm (100 mW) and a lateral resolution of ~0.6 µm. The Raman spectra were acquired from the surface of the bone biopsy, using a thermo-electric–cooled charge-coupled device (CCD) (Bruker Optik GmbH). All data analysis was done with the Opus Ident software package (OPUS 7.2, Bruker Optik GmbH). Once acquired, the Raman spectra were baseline

ro of

corrected (rubber band, 5 iterations) to account for fluorescence, and the following parameters were calculated:

i. The mineral and organic matrix content as well as the mineral/matrix ratio from the

-p

integrated areas of the v2PO4 (410–460 cm−1) and the amide III (1215–1300 cm−1) bands [19-

re

21].

ii. Nanoporosity, approximated by the ratio of the integrated areas of the spectral slice

lP

494-509 cm-1 (PMMA) to Amide III band [22-24]. This metric in embedded bone tissue is a

na

surrogate for tissue water in fresh bone tissue.

iii. The relative glycosaminoglycan (GAG) content was expressed as the GAG / matrix ratio (the ratio of the integrated areas of the proteoglycan/CH3 [1365–1390 cm−1] band

ur

[representative of mucopolysaccharides] to the Amide III [1215–1300 cm−1] band) [25-27].

Jo

iv. The relative pyridinoline content was calculated as the absorbance height at 1660

cm-1 / area of the amide I (1620 – 1700 cm-1) [28-31]. v. The maturity/crystallinity (MMC) of the bone mineral apatite crystallites was approximated from the full width at half height (FWHH) of the v1PO4 (930–980 cm−1) band,

which inversely correlates with crystallite length (002 crystallographic reflection) [20, 21, 24, 30]. By design, IMI tests the outer cortical bone. Thus, Raman measurements were obtained right at the outer subperiosteal edge of the cortex (first 1 x 1 µm² of mineralized tissue as evidenced by the presence of both Amide I and phosphate peaks [32]), and 5, 50, 100, and 500µm inwards. For each patient, 3 such lines (one in the upper third, the other in the middle third, and the last

and the resulting mean value treated as a single statistical unit. Statistical analysis

ro of

in the bottom third of the cortex) were analyzed, the results averaged at equivalent distances,

BMD, qBEI and Raman spectroscopic outcomes were subjected to correlation analysis with

-p

the BMSi values. Kolmogorov-Smirnov test was used to test the normality of distribution of

re

continuous variables. As all data (except Ct.CaHigh) were normally distributed, Pearson coefficient of correlation was used for bivariate correlations between continuous variables; for

lP

Ct.CaHigh Spearman rank order correlation analysis was used. In all instances statistical significance was assigned to p<0.05.

na

Results

Twelve patients (8M/4F) mean age 48.1 years (range 15-60) were studied (Table 1). Four

ur

patients (pt. 1,4,7,9) had osteoporosis (BMD T-score ≤-2.5) and were investigated for causes

Jo

of secondary osteoporosis; five patients (pt. 2,3,5,11,12) had osteopenia and were investigated for causes of fractures and three patients (pt. 6,8,10) had rare bone diseases. In two of the latter patients (pt 6,10) diagnoses were confirmed by genetic testing while in patient 8 with high bone mass and no fractures no genetic diagnosis has been made yet (no mutations in LRP5/6, SOST or LRP4 were found). In all these patients there was no histological evidence of osteomalacia, bone marrow abnormalities or woven bone.

The differences in pathophysiology of the disorders of studied patients provided a wide range of BMD values (T-scores -3.6 to +5.8 at the spine); similarly, BMSi values ranged between 63.3 and 93.4 (mean 79.9 ± 10.1). There was no significant correlation between BMSi and BMD measured either at the spine or the hip (spine: r=-0.137, p=0.672; right hip: r=-0.127, p=0.71; femoral neck r=-0.134, p=0.678). Bone Mineralization Density Distribution (BMDD)

ro of

BMDD (measured by qBEI) outcomes varied also considerably among studied patients but none of the 5 measured outcomes (CaMean, CaHigh, CaLow, CaWidth, and CaPeak) from the whole cancellous or cortical area was significantly correlated with BMSi values. An example

-p

of the obtained results is depicted in Figure 1 that illustrates the lack of relationship between BMSi and Cn.CaMean or Ct.CaMean. Importantly, Cn.CaMean was significantly correlated

re

with Ct.CaMean (r=0.87, p<0.001) as previously described in healthy subjects and patient

Raman microspectroscopy

lP

cohorts [33-35].

na

Figure 2 (left panel) shows a picture of the cortical compartment in one of the subjects analysed, while the right panel shows where the Raman measurements were obtained. Analysis of the

ur

bone samples from the outer cortical surface up to 500 μm by Raman microspectroscopy revealed no relationship between either organic matrix content or mineral/matrix ratio and

Jo

BMSi values. Furthermore, there were no significant correlations between mineral maturity/crystallinity with BMSi (Table 2). On the other hand, there was a weak but significant correlation (r=0.576, p=0.0499) between mineral content by Raman and BMSi at the mineralized subperiosteal edge, and a stronger one (r=0.661, p=0.019) at 5 µm from the edge (Figure 3a; Table 2), while no significant correlations were evident further into the tissue.

When the relationship between nanoporosity and BMSi was examined, significant inverse correlations at the mineralized outer edge of the sample and at 5 μm (Figure 3b) were observed. From 50 μm onwards the strength of this relationship diminished and was not anymore significant (50 μm r=-0.128 p=0.692, 100 μm r=0.249 p=0.435, 500 μm r=0.0638 p=0.8439; Table 2); notably nanoporosity values measured beyond 100 μm were close to the detection limit of the assay and, potentially, not suitable to capture possible differences among very low values. The pattern of changes of nanoporosity values with increasing the measurement

ro of

distance is illustrated in sequential measurements of individual patients with high or low BMSi values (Figure 4A).

A converse pattern, namely a positive significant correlation, was observed in the relationship

-p

between BMSi and pyridinoline content at the mineralized subperiosteal edge and at 5 μm with

re

no further relationship from 50 μm onwards (Figure 3c; Table 2). Sequential measurements in individual patients illustrate the pattern of changes with increasing distance of the measurement

lP

(Figure 4B). Significant positive correlations were also evident between GAG content at 5 µm from the outer mineralized cortical surface (Table 2).

na

The relationships between nanoporosity and pyridinoline (upper panel) and nanoporosity and GAGs content (lower panel) measured at the mineralized subperiosteal edge and at 5 μm

Jo

ur

distance from the edge are shown in Figure 5.

Discussion

In recent years Impact Microindentation has been introduced as a method to assess bone strength in vivo in individuals at risk of fractures. Results of clinical studies combined with the generally reported lack of an association with BMD values suggest that IMI might be used to complement DXA in the evaluation of fracture risk in clinical practice. Apart from certain

methodological issues that still need to be clarified [36], the nature of IMI-measured bone properties that can affect bone strength is incompletely understood. Elucidation of intrinsic bone parameters that affect measurements may help to better define the potential value of IMI in clinical practice. The major measurable determinants of whole bone strength - the mechanical resistance to fracture - in animals and humans in vivo are the amount of bone mineral, bone size and architecture, and bone tissue/material characteristics. The last one is the least understood and

ro of

hardest to evaluate because it requires investigation of bone biopsy specimens by sophisticated methods that are not widely available [36]. In the present study we studied bone composition and material properties in patients with different bone disorders and variable fracture risk in

-p

whom IMI measurements were performed concurrently with transiliac bone biopsies.

re

We found no association of BMSi values with lumbar spine or hip BMD. This is not surprising as one important determinant of the BMD measured by DXA is the amount of bone present in

lP

the spine or hip which does not play a role in the small bone volume tested by IMI in the tibia. Additionally, no relationship between BMSi and BMDD measured by qBEI was observed. The

na

lack of relationship between BMDD and BMSi further suggests that the BMSi might depend on the mineral content of the local bone packet actually tested by the indenter (which comprises

ur

a bone volume from tibia surface to about 150 microns depth [4]) and does not reflect the

Jo

BMDD measured in entire, longitudinally sectioned, bone area of the biopsy cylinder. We further analyzed bone material locally at the mineralized subperiosteal surface of the cortices of the biopsy samples by Raman microspectroscopy, as we hypothesized that the this site in the tibia resembles that of the iliac crest cortices in material properties. It has previously been reported that bone tissue composition varies across anatomic sites in the proximal femur and the iliac crest [37], but the areas of analysis were randomly chosen, thus most likely

reflecting variations of bone turnover among different anatomical sites [38, 39]. In the present study, Raman analysis was performed at well-defined positions in the cortical bone of iliac crest biopsies. The observed correlations were evident within the first 5 µm from the mineralized subperiosteal edge, an area that is unlikely to be influenced by turnover as the mineralized subperiosteal edge undergoes modelling rather than remodeling, a process more pronounced in young children. In the present study all but one of the analyzed patients were adults. Moreover, none of the bone biopsy surfaces analyzed had fluorescent labels evident

ro of

close to the mineralized subperiosteal edge suggesting that the subperiosteal edges were quiescent; this may be a contributing factor to the lack of any correlations deeper into the cortical bone. The observed correlations of Raman outcomes with BMSi support this notion.

-p

Local mineral content at the iliac crest mineralized subperiosteal surface positively correlates with BMSi, unlike the case with either BMD or BMDD. These findings further highlight the

re

necessity to complement outcomes describing mineral content of the whole bone tissue with

lP

outcomes from confined subperiosteal regions in relation to results of microindentation. Bone material consists of mineral, organic matrix and water [40]. BMSi values did not correlate with the organic matrix content (estimated from the integrated area of the Amide III band) at any

na

measured distance from the bone mineralized subperiosteal surface. It should be kept in mind that the Amide III band involves C=O, C-N, and N-H groups of a peptide bond, thus its

ur

integrated area is a measure of the total organic matrix content (collagen and non-collagenous

Jo

moieties). In addition, values did not correlate with mineral quality measured as mineral maturity / crystallinity, although right at the mineralized subperiosteal edge this metric was almost negatively correlated with BMSi (p=0.0501), and we do know that larger crystallites are more brittle [40]. In contrast, BMSi values were significantly and inversely correlated with nanoporosity. As we have discussed elsewhere [23, 41], nanoporosity calculated from the ratio of PMMA / organic matrix content in fixed and embedded bone tissue is a surrogate for tissue

water content in native bone tissue because PMMA infiltrates spaces not occupied by either mineral or organic matrix. Tissue water can be present in pores such as osteons (in the present work, measurements were made away from osteons), in osteocyte lacunae (in the present work, measurements were performed away from visible lacunae), in the canalicular network, as well as within and between collagen fibers [42]. It is, therefore, expected that the IMI probe will move through water easier, increasing the measured penetration. Combined, the results of all measurements -DXA, BMDD by qBEI, Raman Microspectroscopy- indicate that IMI does not

ro of

assess organic matrix content, but rather localized mineral and tissue water (nanoporosity) content. Moreover, the opposite correlations between mineral and tissue water content at the same micro-locations would agree with the notion that mineral displaces water in the organic

-p

matrix.

re

Apart from mineral, tissue water, and organic matrix content, bone mechanical properties are also determined by the properties of organic matrix [40]. To obtain more insight in the latter

lP

we examined pyridinoline and glycosaminoglycan (GAG) content of the bone biopsies with Raman microspectroscopy. Pyridinoline is an enzymatic trivalent collagen cross-link. Collagen

na

cross-links are responsible for the stability of the collagen fibers, that contribute to the viscoelastic properties of bone and influence both stiffness and toughness [43, 44]. They affect

ur

the mechanical properties of bone even if altered at microanatomical locations and in the absence of any concomitant mineral quantity or quality changes [45, 46]. In animal models

Jo

pyridinoline has been shown to be positively correlated with bending strength and modulus of cortical bone, as well as with compressive energy to failure and bone strength at the distal femur [47]. Our findings in the studied patients are in line with those of the animal studies and suggest that BMSi is affected by bone matrix properties. Finally, pyridinoline affects interfibrillar-spacing (inversely) and tissue mechanical properties amongst others [48]. Osteoporotic bone has been reported to have decreased mean fibril diameter and spacing [49],

and elevated pyridinoline content compared to healthy bone [24]. While the study of mineral and material properties of the bone biopsies of our patients indicate that disturbed bone matrix properties adversely affect BMSi measurements it is the finding of the inverse relationship between nanoporosity (tissue water) and BMSi values that helps to formulate a plausible explanation of the overall BMSi findings. Nanoporosity was inversely related to both pyridinoline and GAGs content (Figure 5). The relationship between pyridinoline and nanoporosity may be explained by the decreasing interfibrillar spacing with increasing

ro of

pyridinoline content. GAGs are part of the proteoglycan molecules which have several key roles in bone. At actively forming bone surfaces they have been shown to modulate bone turnover rates and they are also responsible for keeping the osteocyte lacunae and the

-p

canalicular network free of mineral [50, 51]. Proteoglycans also participate in fibrilogenesis, and in general are in a direct relationship with pyridinoline content [52], a relationship observed

re

in the present study (Figure 5). In the present analysis, GAGs content was significantly and

lP

directly correlated with BMSi only 5 m in from the mineralized subperiosteal edge. This may reflect, in part, a relationship between canalicular network size and density, and BMSi. The

na

content of this network is a viscous fluid that might reduce the penetration of the IMI tip, which in addition to the dependency on local mineral content may result in greater BMSi values. Published reports have shown that there is a positive correlation between local mineral content

ur

and canalicular network density [53, 54]. Proteoglycans undergo extensive posttranslational

Jo

modifications including the addition or shedding of GAGs [55], which have been shown to be inhibitors of hydroxyapatite nucleation and growth [56]. Based on these reports, the picture that emerges is that increased canalicular network density, thus increased proteoglycan and GAG content, is associated with increased mineral content, and consequently decreased nanoporosity (due to the increased mineralization), which is what we observed in the analyzed patients regarding the negative correlation between GAG content and nanoporosity. The lack

of any correlation right at the edge may be due to the fact that the canalicular network is not yet dense enough to influence the initial penetration, while the absence of correlations deeper into the mineralized subperiosteal surface is most likely due to the interference of the induced damage due to material compression during advancing penetration of the OsteoProbe tip. Thus, high nanoporosity and disturbed bone matrix quality may result in mechanically compromised mineralized tissue, facilitating the easier penetration of the OsteoProbe tip in cortical bone, an action that is reflected in the lower BMSi values measure. Conversely, a low

ro of

nanoporosity (i.e. higher local mineral content) with normal bone material properties resists the penetration resulting in higher BMSi values. These observations explain the currently proposed, but not proven, principle of the IDI methodology that assumes that the more porous

-p

(at any length scale) bone facilitates an easier penetration of the OsteoProbe tip and hence the

re

lower the BMSi which is inversely related to the penetration depth. The absence of this relationship at greater distances from the mineralized subperiosteal edge may be due to the fact

lP

that once the conical tip breaches the mineralized edge it compresses bone material as it penetrates deeper into the mineralized tissue; it may, thus, be hypothesized that the depth of

na

penetration during RPI measurements depends on the material properties of the outer mineralized cortical surface plus the induced physical damage due to compression.

ur

Unfortunately, Raman microspectroscopy is not capable of either identifying different proteoglycan species, or positively identifying whether they are located within the canalicular

Jo

network, or within or between the fibrils due to the lateral resolution of the technique which is ~ 1 µm. Nevertheless, the data of the present study indicate that nanoporosity and bone matrix composition at the entrance of the OsteoProbe are important determinants of the measured BMSi values, a finding in general agreement with the results reported by Abraham et al [14]. As both high porosity at the micro-nano level and low pyridinoline content of the bone matrix

can negatively impact bone strength our findings strongly suggest that low BMSi values are indeed consistent with reduced bone strength. The study has several limitations, including the low number of patients studied, and the fact that biopsy results were obtained in the iliac crest whereas BMSi measurements were performed in the tibia. The study provides, however, for the first time a plausible hypothesis leading to an explanation of IMI results based on physical/compositional properties of bone

ro of

that warrants further evaluation in larger groups of patients. In summary, the present study demonstrates that BMSi values are independent of degree and distribution of mineral measured in the total cortical cross-section of the transiliac biopsy, but depends on local mineral content. Moreover, while BMSi values are also independent of matrix

-p

content as well as of mineral quality at mineralized subperiosteal sites, they strongly associate

na

Acknowledgements

lP

re

with organic matrix quality at this site.

The authors thank Petra Keplinger, Sonja Lueger and Phaedra Messmer for qBEI

ur

measurements at the Bone Material Laboratory of the Ludwig Boltzmann Institute of

Jo

Osteology, Vienna, Austria.

References

[1] D. Bridges, C. Randall, P.K. Hansma, A new device for performing reference point indentation without a reference probe, Rev Sci Instrum 83(4) (2012) 044301. [2] M.R. Allen, E.M. McNerny, J.M. Organ, J.M. Wallace, True Gold or Pyrite: A Review of Reference Point Indentation for Assessing Bone Mechanical Properties In Vivo, J Bone Miner Res 30(9) (2015) 1539-50.

ro of

[3] S. Herrera, A. Diez-Perez, Clinical experience with microindentation in vivo in humans, Bone 95 (2017) 175-182.

[4] A. Diez-Perez, M.L. Bouxsein, E.F. Eriksen, S. Khosla, J.S. Nyman, S. Papapoulos, S.Y.

-p

Tang, Technical note: Recommendations for a standard procedure to assess cortical bone at the tissue-level in vivo using impact microindentation, Bone Rep 5 (2016) 181-185.

re

[5] F. Malgo, N.A. Hamdy, T.J. Rabelink, H.M. Kroon, K.M. Claessen, A.M. Pereira, N.R.

lP

Biermasz, N.M. Appelman-Dijkstra, Bone material strength index as measured by impact microindentation is altered in patients with acromegaly, Eur J Endocrinol 176(3) (2017) 339347.

na

[6] F. Malgo, N.A.T. Hamdy, S.E. Papapoulos, N.M. Appelman-Dijkstra, Bone material strength index as measured by impact microindentation is low in patients with fractures

ur

irrespective of fracture site, Osteoporos Int 28(8) (2017) 2433-2437.

Jo

[7] D.D. Sosa, E.F. Eriksen, Reduced Bone Material Strength is Associated with Increased Risk and Severity of Osteoporotic Fractures. An Impact Microindentation Study, Calcif Tissue Int 101(1) (2017) 34-42. [8] T.D. Rozental, K.C. Walley, S. Demissie, S. Caksa, A. Martinez-Betancourt, A.M. Parker, J.N. Tsai, E.W. Yu, M.L. Bouxsein, Bone Material Strength Index as Measured by

Impact Microindentation in Postmenopausal Women With Distal Radius and Hip Fractures, J Bone Miner Res 33(4) (2018) 621-626. [9] L. Mellibovsky, D. Prieto-Alhambra, F. Mellibovsky, R. Guerri-Fernandez, X. Nogues, C. Randall, P.K. Hansma, A. Diez-Perez, Bone Tissue Properties Measurement by Reference Point Indentation in Glucocorticoid-Induced Osteoporosis, J Bone Miner Res 30(9) (2015) 1651-6. [10] J.N. Farr, M.T. Drake, S. Amin, L.J. Melton, 3rd, L.K. McCready, S. Khosla, In vivo

ro of

assessment of bone quality in postmenopausal women with type 2 diabetes, J Bone Miner Res 29(4) (2014) 787-95.

[11] A.G. Nilsson, D. Sundh, L. Johansson, M. Nilsson, D. Mellstrom, R. Rudang, M.

-p

Zoulakis, M. Wallander, A. Darelid, M. Lorentzon, Type 2 Diabetes Mellitus Is Associated

With Better Bone Microarchitecture But Lower Bone Material Strength and Poorer Physical

re

Function in Elderly Women: A Population-Based Study, J Bone Miner Res 32(5) (2017)

lP

1062-1071.

[12] R. Guerri-Fernandez, D. Molina, J. Villar-Garcia, D. Prieto-Alhambra, L. Mellibovsky, X. Nogues, A. Gonzalez-Mena, A. Guelar, M. Trenchs-Rodriguez, S. Herrera-Fernandez, J.P.

na

Horcajada, A. Diez-Perez, H. Knobel, Brief Report: HIV Infection Is Associated With Worse Bone Material Properties, Independently of Bone Mineral Density, J Acquir Immune Defic

ur

Syndr 72(3) (2016) 314-8.

Jo

[13] K.L. Popp, S. Caksa, A. Martinez-Betancourt, A. Yuan, J. Tsai, E.W. Yu, M.L. Bouxsein, Cortical Bone Material Strength Index and Bone Microarchitecture in Postmenopausal Women With Atypical Femoral Fractures, J Bone Miner Res 34(1) (2019) 75-82.

[14] A.C. Abraham, A. Agarwalla, A. Yadavalli, J.Y. Liu, S.Y. Tang, Microstructural and compositional contributions towards the mechanical behavior of aging human bone measured by cyclic and impact reference point indentation, Bone 87 (2016) 37-43. [15] F. Malgo, N.A. Hamdy, S.E. Papapoulos, N.M. Appelman-Dijkstra, Impact Microindentation: Consistency of Serial Measurements and Alterations in Patients With Paget's Disease of the Tibia, J Bone Miner Res 32(12) (2017) 2375-2380. [16] P. Chavassieux, N. Portero-Muzy, J.P. Roux, P. Garnero, R. Chapurlat, Are Biochemical

ro of

Markers of Bone Turnover Representative of Bone Histomorphometry in 370 Postmenopausal Women, J Clin Endocrinol Metab 100(12) (2015) 4662-8.

[17] P. Roschger, P. Fratzl, J. Eschberger, K. Klaushofer, Validation of quantitative

-p

backscattered electron imaging for the measurement of mineral density distribution in human bone biopsies, Bone 23(4) (1998) 319-26.

re

[18] P. Roschger, E.P. Paschalis, P. Fratzl, K. Klaushofer, Bone mineralization density

lP

distribution in health and disease, Bone 42(3) (2008) 456-66.

[19] S. Gamsjaeger, A. Masic, P. Roschger, M. Kazanci, J.W. Dunlop, K. Klaushofer, E.P. Paschalis, P. Fratzl, Cortical bone composition and orientation as a function of animal and

na

tissue age in mice by Raman spectroscopy, Bone 47 (2010) 392-9. [20] S. Gamsjaeger, R. Mendelsohn, A.L. Boskey, S. Gourion-Arsiquaud, K. Klaushofer,

ur

E.P. Paschalis, Vibrational spectroscopic imaging for the evaluation of matrix and mineral

Jo

chemistry, Curr Osteoporos Rep 12(4) (2014) 454-64. [21] E. Paschalis, S. Gamsjaeger, K. Klaushofer, Vibrational spectroscopic techniques to assess bone quality., Osteoporos. Int. 28(8) (2017) 2275-2291. [22] S. Gamsjaeger, W. Brozek, R. Recker, K. Klaushofer, E.P. Paschalis, Transmenopausal changes in trabecular bone quality, J Bone Miner Res 29(3) (2014) 608-17.

[23] S. Gamsjaeger, B. Hofstetter, N. Fratzl-Zelman, P. Roschger, A. Roschger, P. Fratzl, W. Brozek, A. Masic, B.M. Misof, F.H. Glorieux, K. Klaushofer, F. Rauch, E.P. Paschalis, Pediatric reference Raman data for material characteristics of iliac trabecular bone, Bone 69 (2014) 89-97. [24] E.P. Paschalis, P. Fratzl, S. Gamsjaeger, N. Hassler, W. Brozek, E.F. Eriksen, F. Rauch, F.H. Glorieux, E. Shane, D. Dempster, A. Cohen, R. Recker, K. Klaushofer, Aging Versus Postmenopausal Osteoporosis: Bone Composition and Maturation Kinetics at Actively-

ro of

Forming Trabecular Surfaces of Female Subjects Aged 1 to 84 Years, J Bone Miner Res 31(2) (2016) 347-57.

[25] R. Ellis, E. Green, C.P. Winlove, Structural analysis of glycosaminoglycans and

-p

proteoglycans by means of Raman microspectrometry, Connect Tissue Res 50(1) (2009) 2936.

re

[26] S. Gamsjaeger, K. Klaushofer, E. Paschalis, Raman analysis of proteoglycans

lP

simultaneously in bone and cartilage. , Journal of Raman Spectroscopy 45(9) (2014) 794-800. [27] F. Quilès, J.-Y. Balandier, S. Capizzi-Banas, In situ characterisation of a microorganism

(2006) 249-255.

na

surface by Raman microspectroscopy: the shell of Ascaris eggs, Anal Bioanal Chem 386

[28] A. Carden, M.D. Morris, Application of vibrational spectroscopy to the study of

ur

mineralized tissues (review), Journal of biomedical optics 5(3) (2000) 259-68.

Jo

[29] S. Gamsjaeger, S.P. Robins, D.N. Tatakis, K. Klaushofer, E.P. Paschalis, Identification of Pyridinoline Trivalent Collagen Cross-Links by Raman Microspectroscopy, Calcif Tissue Int 100(6) (2017) 565-574. [30] M.D. Morris, G.S. Mandair, Raman Assessment of Bone Quality, Clin Orthop Relat Res 469(8) (2011) 2160-9.

[31] E.P. Paschalis, K. Verdelis, S.B. Doty, A.L. Boskey, R. Mendelsohn, M. Yamauchi, Spectroscopic characterization of collagen cross-links in bone, J Bone Miner Res 16(10) (2001) 1821-8. [32] E.P. Paschalis, S. Gamsjaeger, K. Condon, K. Klaushofer, D. Burr, Estrogen depletion alters mineralization regulation mechanisms in an ovariectomized monkey animal model, Bone 120 (2019) 279-284. [33] B.M. Misof, D.W. Dempster, H. Zhou, P. Roschger, N. Fratzl-Zelman, P. Fratzl, S.J.

ro of

Silverberg, E. Shane, A. Cohen, E. Stein, T.L. Nickolas, R.R. Recker, J. Lappe, J.P. Bilezikian, K. Klaushofer, Relationship of bone mineralization density distribution (BMDD) in cortical and cancellous bone within the iliac crest of healthy premenopausal women, Calcif

-p

Tissue Int 95(4) (2014) 332-9.

[34] B.M. Misof, P. Roschger, S. Blouin, R. Recker, K. Klaushofer, Bone matrix

re

mineralization is preserved during early perimenopausal stage in healthy women: a paired

lP

biopsy study, Osteoporos Int 27(5) (2016) 1795-803.

[35] B.M. Misof, S. Blouin, S. Lueger, E.P. Paschalis, R.R. Recker, R. Phipps, K. Klaushofer, P. Roschger, Baseline mineralizing surface determines the magnitude of the

na

bisphosphonate effect on cortical bone mineralization in postmenopausal osteoporotic patients, J Musculoskelet Neuronal Interact 17(3) (2017) 183-191.

ur

[36] A. Diez Perez, J. Farr, Reference Point Indentation, Primer on the Metabolic Bone

Jo

Diseases and Disorders of Mineral Metabolism, Wiley2018, pp. 823-837. [37] E. Donnelly, D.S. Meredith, J.T. Nguyen, A.L. Boskey, Bone tissue composition varies across anatomic sites in the proximal femur and the iliac crest, J Orthop Res 30(5) (2012) 700-6.

[38] C.M. Schnitzler, S.L. Biddulph, J.M. Mesquita, K.A. Gear, Bone structure and turnover in the distal radius and iliac crest: a histomorphometric study, J Bone Miner Res 11(11) (1996) 1761-8. [39] C.H. Chesnut, 3rd, C.J. Rosen, G. Bone Quality Discussion, Reconsidering the effects of antiresorptive therapies in reducing osteoporotic fracture, J Bone Miner Res 16(12) (2001) 2163-72. [40] P. Fratzl, H. Gupta, E. Paschalis, P. Roschger, Structure and mechanical quality of the

ro of

collagen–mineral nano-composite in bone, J. Mater. Chem. 14 (2004) 2115-2123. [41] E.P. Paschalis, S. Gamsjaeger, N. Fratzl-Zelman, P. Roschger, A. Masic, W. Brozek, N. Hassler, F.H. Glorieux, F. Rauch, K. Klaushofer, P. Fratzl, Evidence for a Role for

-p

Nanoporosity and Pyridinoline Content in Human Mild Osteogenesis Imperfecta, J Bone Miner Res 31(5) (2016) 1050-9.

re

[42] M. Granke, M.D. Does, J.S. Nyman, The Role of Water Compartments in the Material

lP

Properties of Cortical Bone, Calcif Tissue Int 97(3) (2015) 292-307. [43] S.P. Robins, Biochemistry and functional significance of collagen cross-linking, Biochem Soc Trans 35(Pt 5) (2007) 849-52.

na

[44] M. Yamauchi, Collagen: The major matrix molecule in mineralized tissues., in: J. Anderson, S. Garner (Eds.), Calcium and Phosphorus in Health and Disease. , CRC Press,

ur

New York, 1996, pp. 127-141.

Jo

[45] E.M. McNerny, B. Gong, M.D. Morris, D.H. Kohn, Bone fracture toughness and strength correlate with collagen cross-link maturity in a dose-controlled lathyrism mouse model, J Bone Miner Res 30(3) (2015) 455-64. [46] E.P. Paschalis, D.N. Tatakis, S. Robins, P. Fratzl, I. Manjubala, R. Zoehrer, S. Gamsjaeger, B. Buchinger, A. Roschger, R. Phipps, A.L. Boskey, E. Dall'Ara, P. Varga, P. Zysset, K. Klaushofer, P. Roschger, Lathyrism-induced alterations in collagen cross-links

influence the mechanical properties of bone material without affecting the mineral, Bone 49(6) (2011) 1232-41. [47] P. Garnero, The contribution of collagen crosslinks to bone strength, Bonekey Rep 1 (2012) 182. [48] K. Yoshida, H. Jiang, M. Kim, J. Vink, S. Cremers, D. Paik, R. Wapner, M. Mahendroo, K. Myers, Quantitative evaluation of collagen crosslinks and corresponding tensile mechanical properties in mouse cervical tissue during normal pregnancy, PLoS One 9(11)

ro of

(2014) e112391. [49] D. Burr, O. Akkus, Bone morphology and organization, in: D. Burr, M. Allen (Eds.), Basic and Applied Bone Biology, Academic Press, New York, 2013, pp. 3-25.

-p

[50] W.R. Thompson, S. Modla, B.J. Grindel, K.J. Czymmek, C.B. Kirn-Safran, L. Wang,

R.L. Duncan, M.C. Farach-Carson, Perlecan/Hspg2 deficiency alters the pericellular space of

re

the lacunocanalicular system surrounding osteocytic processes in cortical bone, J Bone Miner

lP

Res 26(3) (2011) 618-29.

[51] B. Wang, X. Lai, C. Price, W.R. Thompson, W. Li, T.R. Quabili, W.J. Tseng, X.S. Liu, H. Zhang, J. Pan, C.B. Kirn-Safran, M.C. Farach-Carson, L. Wang, Perlecan-containing

na

pericellular matrix regulates solute transport and mechanosensing within the osteocyte lacunar-canalicular system, J Bone Miner Res 29(4) (2014) 878-91.

ur

[52] T. Nakano, J.R. Thompson, F.X. Aherne, Distribution of glycosaminoglycans and the

Jo

nonreducible collagen crosslink, pyridinoline in porcine menisci, Can J Vet Res 50(4) (1986) 532-6.

[53] B. Hesse, P. Varga, M. Langer, A. Pacureanu, S. Schrof, N. Mannicke, H. Suhonen, P. Maurer, P. Cloetens, F. Peyrin, K. Raum, Canalicular network morphology is the major determinant of the spatial distribution of mass density in human bone tissue: evidence by

means of synchrotron radiation phase-contrast nano-CT, J Bone Miner Res 30(2) (2015) 34656. [54] A. Roschger, P. Roschger, W. Wagermaier, J. Chen, A.F. van Tol, F. Repp, S. Blouin, A. Berzlanovich, G.M. Gruber, K. Klaushofer, P. Fratzl, R. Weinkamer, The contribution of the pericanalicular matrix to mineral content in human osteonal bone, Bone 123 (2019) 7685. [55] W.J. Grzesik, C.R. Frazier, J.R. Shapiro, P.D. Sponseller, P.G. Robey, N.S. Fedarko,

ro of

Age-related changes in human bone proteoglycan structure. Impact of osteogenesis imperfecta, J Biol Chem 277(46) (2002) 43638-47.

[56] A.L. Boskey, L. Spevak, S.B. Doty, L. Rosenberg, Effects of bone CS-proteoglycans,

-p

DS-decorin, and DS-biglycan on hydroxyapatite formation in a gelatin gel, Calcif Tissue Int

Jo

ur

na

lP

re

61(4) (1997) 298-305.

Legends to Figures

Figure 1. Relation between BMSi and Ca.CaMean (A) and Ct.CaMean (B) obtained from total surface of the longitudinally sectioned transiliac bone samples. Note the lack of any relationship between BMSi and the two quantitative mineral parameters. Relationship of Cn.CaMean with

ro of

Ct.CaMean (C); r= 0.87, p<0.001. Figure 2.

-p

Figure 2 (left panel) shows a picture of the cortical compartment in one of the subjects analysed, while the right panel shows where the Raman measurements were obtained.

re

Figure 3.

(a) Relationship between BMSi and mineral content measured at 3 different distances from the

lP

mineralized subperiosteal edge of bone biopsies by Raman microspectroscopy. (b) Relationship between BMSi and nanoporosity measured at 3 different distances from the

na

mineralized subperiosteal edge of bone biopsies by Raman microspectroscopy. (c) Relationship between BMSi and pyridinoline content measured at 3 different distances from

Jo

ur

the mineralized subperiosteal edge of bone biopsies by Raman microspectroscopy.

Figure 4.

A. Nanoporosity at mineralized subperiosteal edge and at 5, 50, 100 and 500 μm inside the bone cortex of 4 patients, two with low and two with high BMSi values (shown in the box). B. Pyridinoline content at mineralized subperiosteal edge and at 5, 50, 100 and 500 μm inside the bone cortex in 4 patients, two with low and two with high BMSi values (shown in the box).

Figure 5. Relationship between nanoporosity and pyridinoline (upper panel), and nanoporosity and GAGs content (lower panel) measured at the mineralized subperiosteal edge and at 5 μm

Jo

ur

na

lP

re

-p

ro of

distance from the periosteal edge.

ro of

-p

re

lP

na

ur

Jo

ro of

-p

re

lP

na

ur

Jo

ro of

-p

re

lP

na

ur

Jo

ro of

-p

re

lP

na

ur

Jo

ro of

-p

re

lP

na

ur

Jo

Gender

Age (yrs)

LS-BMD

TH-BMD

FN-BMD

BMSi

1

M

56

-3.6

-2.2

-1.8

85.60

2

M

36

-1.9

3

F

55

-1.6

-1.0

-0.9

4

M

45

-2.6

-1.9

5

M

56

-2.2

-0.6

6

M

60

-1.0

-0.7

7

M

49

-2.9

8

F

46

9

M

53

10

F

15

11

F

52

12

M

Osteoporosis, no Fx Fx L hip and wrist

71.5

VFx (multiple), Fx Wrist

88.6

Osteoporosis, Fx rib

-0.7

67.6

VFx (multiple)

-1.1

63.3

Erdheim Chester disease

-1.9

-2.1

93.4

Osteoporosis, VFx (multiple)

5.8

4

4.2

79.3

High Bone Mass, unknown etiology

-2.7

-1.9

-1.5

88.2

Osteoporosis (fam) no Fx

3.9

2.7

2.2

85.6

Osteogenesis Imperfecta, Fx (multiple)

-0.1

-1.6

-0.1

70.8

Fx tibia, multiple osteolytic lesions femur

-2.4

-1.4

-1.4

90.6

VFx, Fx Humerus , Thyrotoxicosis

na l

Pr

e-

74.8

-1.9

Jo ur 54

-1.7

Diagnosis

pr

Nr

oo

f

Table 1. Characteristics of Studied Patients

Fx=Fracture, VFx=Vertebral Fracture

Table 2: Correlations between Raman-derived parameters and BMSi at the 5 anatomical positions analyzed in the cortical compartment. BMSi vs Mineralized subperiosteal Edge

Mineral content Organic Matrix content

r p r

0.5761

p

Mineral / Matrix Nanoporosity

BMSi vs 5 μm from mineralized subperiosteal edge

BMSi vs 50 μm from mineralized subperiosteal edge

0.6614

0.1355

r p r

0.6745

p

0.0499

0.4975

0.4073

0.2212

r p r

0.4895

p

0.0192

0.5288

BMSi vs 100 μm from mineralized subperiosteal edge

BMSi vs 500 μm from mineralized subperiosteal edge

0.4558

0.2110

r p r

0.2053

r p r

0.5104

p

0.5221

p

0.1888

0.1179

0.1365

0.1475

0.1944 0.5450 0.3274 0.2989 -0.0850

Jo

ur

na

lP

re

-p

ro of

r r r r r 0.0998 0.0771 0.7152 0.6473 0.7928 p p p p p -0.7425 -0.6832 -0.1278 0.2493 0.0638 r r r r r 0.0057 0.0143 0.6922 0.4346 0.8439 p p p p p -0.5757 -0.4348 -0.0780 -0.3084 -0.3580 r r r r r MMC 0.0501 0.1578 0.8095 0.3294 0.2532 p p p p p 0.6593 0.5379 -0.2031 -0.1457 -0.3047 r r r r r GAG content 0.0197 0.0712 0.5267 0.6514 0.3355 p P p p p 0.6935 0.7159 -0.1826 -0.0908 -0.1536 r r r r r Pyridinoline 0.0124 0.0088 0.5700 0.7790 0.6337 p p p p p content MMC = Mineral Matrix/Crystallinity, GAG = Glycosaminoglycan; r = Pearson correlation coefficient. Significant correlations (p<0.05) are listed in bold typeface.