ARTICLE IN PRESS

Journal of Biomechanics 42 (2009) 555564

Contents lists available at ScienceDirect

Journal of Biomechanics
journal homepage: www.elsevier.com/locate/jbiomech www.JBiomech.com

Sensitivity of tissue differentiation and bone healing predictions to tissue properties

Hanna Isaksson a,b,c,, Corrinus C van Donkelaar b, Keita Ito a,b
a b c

AO Research Institute, AO Foundation, Clavadelerstrasse 8, 7270 Davos, Switzerland Department of Biomedical Engineering, Eindhoven University of Technology, PO Box 513, 5600 MB Eindhoven, The Netherlands Department of Physics, University of Kuopio, PO Box 1627, 70211 Kuopio, Finland

a r t i c l e in fo
Article history: Accepted 2 January 2009 Keywords: Fracture healing Mechanobiology Material properties Fractional factorial design Design of experiments Orthogonal array

abstract
Computational models are employed as tools to investigate possible mechano-regulation pathways for tissue differentiation and bone healing. However, current models do not account for the uncertainty in input parameters, and often include assumptions about parameter values that are not yet established. The aim was to clarify the importance of the assumed tissue material properties in a computational model of tissue differentiation during bone healing. An established mechano-biological model was employed together with a statistical approach. The model included an adaptive 2D nite element model of a fractured long bone. Four outcome criteria were quantied: (1) ability to predict sequential healing events, (2) amount of bone formation at specic time points, (3) total time until healing, and (4) mechanical stability at specic time points. Statistical analysis based on fractional factorial designs rst involved a screening experiment to identify the most signicant tissue material properties. These seven properties were studied further with response surface methodology in a three-level BoxBehnken design. Generally, the sequential events were not signicantly inuenced by any properties, whereas rate-dependent outcome criteria and mechanical stability were signicantly inuenced by Youngs modulus and permeability. Poissons ratio and porosity had minor effects. The amount of bone formation at early, mid and late phases of healing, the time until complete healing and the mechanical stability were all mostly dependent on three material properties; permeability of granulation tissue, Youngs modulus of cartilage and permeability of immature bone. The consistency between effects of the most inuential parameters was high. To increase accuracy and predictive capacity of computational models of bone healing, the most inuential tissue mechanical properties should be accurately quantied. & 2009 Elsevier Ltd. All rights reserved.

1. Introduction Fracture healing mainly aims to restore bones load-bearing function. It involves sequential differentiation of cells and tissues, which is inuenced by the local mechanical environment (Einhorn, 1998; Gerstenfeld et al., 2003). Computational models of tissue differentiation during bone healing are frequently used to study possible mechano-regulation pathways. Increasing biological knowledge and computational power have pushed recent developments towards focusing on biological aspects of tissue differentiation, such as how to better describe cell processes (Gomez-Benito et al., 2005; Isaksson et al., 2008a), cell dispersal (Perez and Prendergast, 2007), inclusion of growth factors and angiogenesis (Bailon-Plaza and van der Meulen, 2001;

Corresponding author at: Department of Physics, University of Kuopio, PO Box 1627, 70211 Kuopio, Finland. Tel.: +358 1716 2341; fax: +358 1716 3032. E-mail address: hanna.isaksson@uku. (H. Isaksson).

Geris et al., 2008a). In contrast to the wealth of literature on mechanical behavior of fracture callus (Claes et al., 1999; Kenwright and Goodship, 1989; Richardson et al., 1994), insufcient data is available on the tissue material properties of the sequentially developing callus tissues. Therefore during computational modeling, callus tissue material properties are often estimated based on material properties of similar tissue types, but obtained from mature tissue or as educated guesses when no literature data is available. Hence, the accuracy of the assumed material properties may become the limiting factor in the precision of the simulations. Most studies of tissue differentiation today assume identical tissue mechanical properties (Table 1) (Andreykiv et al., 2008; Epari et al., 2006b; Geris et al., 2004, 2008b; Isaksson et al., 2006b, 2008a; Kelly and Prendergast, 2005; Lacroix and Prendergast, 2002; Perez and Prendergast, 2007). Unfortunately, many of these properties are not well established. Lacroix introduced this set of material properties and showed in a parametric study by varying-one-parameter-at-the-time that

0021-9290/$ - see front matter & 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.jbiomech.2009.01.001

ARTICLE IN PRESS
556 H. Isaksson et al. / Journal of Biomechanics 42 (2009) 555564

sequential events during bone healing were not altered by changes in material properties, as long as the tissues were sequentially stiffer (Lacroix, 2001). However, current models aim to evaluate rates of healing, as well as effects of biological and mechanical interventions and potential non-union treatment strategies. For such purposes, qualitative descriptions of sequential spatial events during normal healing are no longer sufcient. The importance of the material properties on the quantitative response needs further clarication. A design of experiments (DOE) approach based on fractional factorial designs was used to computationally evaluate the inuence of each assumed tissue property involved during bone regeneration in a poroelastic FE model of tissue differentiation. Different from varying-one-parameter-at-a time, the DOE approach does not need a baseline model, and can reach a more reliable conclusion about factor effects with fewer simulations (Funkenbusch, 2005; Phadke, 1989). The outcome was assessed as sequential spatial and temporal tissue differentiation events, bone formation rate, time until complete healing and mechanical stability. The investigated properties were Youngs modulus, Poissons ratio, permeability and porosity in each of the tissue types; bone marrow, granulation tissue, brous tissue, cartilage, immature bone and mature bone. The objective was to determine which material parameters are of the greatest inuence to each of

the major processes during tissue differentiation and to the bone healing capacity. We hypothesize that material properties would inuence both the spatial and the temporal progression of sequential tissue transformation during bone healing, since material properties in combination with loading govern the mechano-regulation algorithm.

2. Methods 2.1. Adaptive tissue differentiation model The computational mechano-regulatory model was developed to describe the temporal and spatial distributions of brous tissue, cartilage and bone, regulated through cellular activity (Isaksson et al., 2008a). Dependent on mechanical stimulation, mesenchymal stem cells, broblasts, chondrocytes and osteoblasts responded by proliferation, differentiation, migration and/or apoptosis. Additionally, the cells could produce or degrade extracellular matrix for their respective tissue type (Isaksson et al., 2008a). An axisymmetric FE model of an ovine tibia was adopted from a previous fracture healing study (Isaksson et al., 2006b). The geometry represented a 3 mm transverse fracture gap and an external callus (Fig. 1a). A 1 Hz cyclic load of 300 N was applied proximally on the cortical bone. The magnitudes of deviatoric shear strain and uid velocity were calculated at the peak load (v 6.5 ABAQUS, Simulia, Dassault Systems) and used to predict cell and tissue differentiation behavior (Prendergast et al., 1997). Parameter values for all cell processes remained constant (Table 2).

Table 1 Tissue material properties. Commonly used properties Additional literature review Factor levels High Cortical Cortical Cortical Cortical bone bone bone bone Youngs modulus (MPa) Permeability (m4/N s) Poissons ratio Porosity 15750a 1.0E17d 0.325b 0.04c Not Not Not Not included included included included Low

Granulation Granulation Granulation Granulation

tissue tissue tissue tissue

Youngs modulus (MPa) Permeability (m4/N s) Poissons ratio Porosity

1 1.0E14 0.167 0.8 2e 1.0E14e 0.167 0.8 10g 5.0E15f

0.99l

1.5 1.5E14 0.2004 0.96

0.5 5.0E15 0.1336 0.64

Fibrous Fibrous Fibrous Fibrous

tissue tissue tissue tissue

Youngs modulus (MPa) Permeability (m4/N s) Poissons ratio Porosity

1.9e; 7.8m 0.19m 0.70m 3.10l; 14n; 5.3o ; 7p; 5.8q; 4.511.8r ; 10s 4.7E15f; 2.0E15t; 1.9E157.0E15u; 2.3E15s 0.1740.185h; 0.19u; 0.17v 0.79k; 0.73t; 0.76v 201l; 2250x; 2139y; 540z 10E13ab; 4.7E13y; 0.8E1310E13aa,ae,af 0.32x; 0.23z; 0.24aa 0.79x,z; 0.77ac; 0.750.80ag,ah,ai 8300z; 13000aa 10E13ab; 4.7E13y; 0.8E1310E13aa,ae,af 0.32x; 0.23z; 0.24aa 0.79x,z; 0.77ac; 0.750.80ag,ah,ai

3 1.5E14 0.2004 0.96

1 5.0E15 0.1336 0.64

Cartilage Cartilage

Youngs modulus (MPa) Permeability (m4/N s)

15 7.5E15

5 2.5E15

Cartilage Cartilage

Poissons ratio Porosity

0.167h 0.8k

0.2004 0.96

0.1336 0.64

Immature bone Immature bone Immature bone Immature bone

Youngs modulus (MPa) Permeability (m4/N s) Poissons ratio Porosity

1000 1.0E13 0.325 0.8

1500 1.5E13 0.39 0.96

500 5.0E14 0.26 0.64

Mature bone Mature bone Mature bone Mature bone

Youngs modulus (MPa) Permeability (m4/N s) Poissons ratio Porosity

6000i 3.7E13j 0.325 0.8

9000 5.55E13 0.39 0.96

3000 1.85E13 0.26 0.64

ARTICLE IN PRESS
H. Isaksson et al. / Journal of Biomechanics 42 (2009) 555564 557

Table 1 (continued ) Commonly used properties Additional literature review Factor levels High Marrow Marrow Marrow Marrow Youngs modulus (MPa) Permeability (m4/N s) Poissons ratio Porosity 2 1.0E14 0.167 0.8 2x,ad 3 1.5E14 0.2004 0.96 Low 1 5.0E15 0.1336 0.64

Youngs modulus and permeability were chosen 750%, and Poissons ratio and porosity were chosen 720% of those commonly used. a Smit et al. (2002). b Cowin (1999). c Schafer and Burr (1988). d Johnson et al. (1982). e Hori and Lewis (1982). f Armstrong and Mow (1982). g Lacroix and Prendergast (2002). h Jurvelin et al. (1997). i Claes and Heigele (1999). j Ochoa and Hillberry (1992). k Mow et al. (1980). l Leong and Morgan (2008). m Moussa et al. (2008). n Wei et al. (1998). o Korhonen et al. (2002). p Laasanen et al. (2003). q Akizuki et al. (1986). r Shepherd and Seedhom (1999). s Setton et al. (1997). t Wayne et al. (2003). u Julkunen et al. (2007) v Julkunen et al. (2008). x Hosokawa and Otani (1997). y Kohles and Roberts (2002). z Wear et al. (2005). aa Pakula et al. (2008). ab Arramon and Nauman (2001). ac Fellah et al. (2004). ad Hosokawa and Otani (1998). ae Grimm and Williams (1997). af Nauman et al. (1999). ag Chaffai et al. (2000). ah Lundeen et al. (2000). ai Salome et al. (1999).

2.2. Literature review of tissue material properties An extensive literature review was conducted to determine how well each tissue property is dened. We focused on nding soft tissue properties and on determining the variability in reported literature properties for well-characterized tissues. All tissues were assumed linear poroelastic and required a Youngs modulus, Poissons ratio, permeability and porosity. We assumed that granulation tissue, brous tissue, cartilage, immature and mature bone are involved during sequential tissue differentiation. Also bone marrow in the intramedullary canal and cortical bone were included. The initial tissue material properties were taken identical to those commonly used during computational analyses of tissue differentiation (Table 1) (Isaksson et al., 2006b; Lacroix and Prendergast, 2002). An extensive literature review was conducted to nd additional experimental references for characterization of the tissues involved as well as the variability of the reported tissue properties (Table 1). Cortical bone and bone marrow do not undergo tissue differentiation during bone healing. The material properties of cortical bone are well established and much stiffer than the other tissues. Therefore, it was assumed that variation would not have a signicant effect and it was excluded from the parametric study. For bone marrow, mechanical properties are less well known. They were assumed to be potentially important during the early phases of healing, and were therefore included. Despite variations in constitution between yellow and red bone marrow (Blebea et al., 2007; Hartsock et al., 1965) and with site, age, and species (Meunier et al., 1971; Schnitzler and Mesquita, 1998), the only material properties for bone marrow found are those by Hosokawa and Otani (1997, 1998), who used ultrasound to quantify the modulus to be 2 MPa. Immature bone is less mineralized than mature bone and was therefore assigned a lower Youngs modulus. Recently, a nanoindentation study on fracture callus tissue reported variations in bone modulus between 271010 MPa, depending on degree of mineralization (Leong and Morgan, 2008) the low values

being in the range of those measured for early embryonic mineralized bone (Tanck et al., 2004). Similarly, the reported values for permeability of human cancellous bones range over two orders of magnitude (10141012 m4/N s) and depend strongly on porosity and anatomical site (Arramon and Nauman, 2001; Grimm and Williams, 1997; Lim and Hong, 2000; Nauman et al., 1999; Pakula et al., 2008). Porosity of human and bovine cancellous bone ranges from 7095% depending on the anatomical site and bone status (Table 1) (Chaffai et al., 2000; Fellah et al., 2004; Hosokawa and Otani, 1997, 1998; Kohles and Roberts, 2002; Lundeen et al., 2000; Pakula et al., 2008; Salome et al., 1999; Wear et al., 2005). Literature values for Youngs modulus of cartilage vary greatly, partly because different types of moduli are reported. We collected studies that measured instantaneous modulus, since our mechanical model is evaluated during a load cycle of 1 s. Generally this parameters is reported to be 35 MPa under compression, with variations up to 10 MPa (Elliott et al., 1999; Korhonen et al., 2002; Laasanen et al., 2003; Setton et al., 1993, 1997; Wei et al., 1998). One study assessed the cartilage modulus within a rat fracture callus to be 3.10 MPa (Leong and Morgan, 2008). Permeability, Poissons ratio and porosity are well characterized in young, normal, and aged cartilage, and reported with fairly high consistency (Table 1). Fibrous tissue in ligaments and tendons are well characterized under tension (Anaguchi et al., 2005). However, this tissue in its native environment is vastly different from the quickly formed brous tissue during repair. Hory and Lewis determined brous tissue modulus during repair under compression at a bonecement interface after total joint replacement in a canine model to be 1.9 MPa (Hori and Lewis, 1982). The formed tissue was described as consisting of heavy collagen bers with brocytes interspersed throughout the tissue matrix (Hori and Lewis, 1982). Hence, it is a fair assumption that it is similar to the tissue that develops temporarily during bone healing. Granulation tissue, formed shortly after the trauma, is assumed the softest and least organized tissue. It is also the least characterized tissue. Recently, its modulus was quantied for the rst time in

ARTICLE IN PRESS
558 H. Isaksson et al. / Journal of Biomechanics 42 (2009) 555564

Fig. 1. (a) Geometric FE model. Poroelastic axisymetric FE model (left) used for all analyses. The initial conditions include concentrations of mesenchymal stem cells at the periosteum, at the marrow interface, at the outer boundary, and randomly in the callus tissue. All other cell types and tissue types have zero concentrations initially and the tissue material parameters of 100% granulation tissue. (b) Sketch of the adaptive tissue differentiation model including the cell processes involved (Isaksson et al., 2008a).

Table 2 Normalized cell parameter data that was used for all treatment conditions. Cell Initial cell density Periost MSC FB CC OB Matrix FT C B 0.5 0.0 0.0 0.0 Marrow 0.30 0.0 0.0 0.0 Outer/ external 0.05 0.0 0.0 0.0 Callus 0.005 0.0 0.0 0.0 Initial conc. 0.0 0.0 0.0 0.65 0.50 0.0 0.20 0.60 0.55 0.20 0.30 ProductionfPM (day1) 0.20 0.05 0.10 0.30 0.20 0.10 0.15 0.05 0.05 0.10 0.15 DegradationfDM (day1) 0.05 0.05 0.05 TransportD (mm2day1) ProliferationfPR (day1) DifferentiationfD (day1) ApoptosisfAP (day1)

Parameter values were calculated based on the literature review in Isaksson et al. (2008a). The rates of all processes for mesencymal stem cells (MSC), broblasts (FB), chondrocytes (CC), osteoblasts (OB), brous tissue (FT), cartilage (C), and bone (B) were constant throughout the parametric study.

a rat fracture callus, where it was determined to be 1 MPa (Leong and Morgan, 2008).

2.3. Design of experiment approach to study bone healing A two-step parametric analysis was conducted, similar as the study by Isaksson et al. (2008b). All the material properties were investigated and their inuences were determined using analysis of variance (ANOVA). The investigated material properties were Youngs modulus, Poissons ratio, permeability and porosity for granulation tissue, brous tissue, cartilage, immature bone and mature bone, respectively. The chosen parameter space for Youngs modulus and permeability were 50% and 150%, and for Poissons ratio and porosity it was 80% and 120% of the commonly assumed properties for each tissue type. These parameter spaces covered most reported properties in literature (Table 1). First, a two-level screening experiment was used to identify the most important factors (material properties). It investigated all material properties at two levels, high and low (Table 1), using a L64 resolution IV array (Funkenbusch, 2005; Phadke, 1989), with 24 control factors (material properties) and a total of 64 treatment conditions

(simulations with different factor level combinations). The screening experiment assumed approximately linear factor inuence (Isaksson et al., 2008b; Phadke, 1989). Thereafter, a more detailed examination was carried out using the response surface methodology on the identied most important factors to further evaluate curvature and interactions. A BoxBehnken design was used with 7 factors, each with 3 equally spaced levels, by adding a mid level to the high and low levels from the screening experiment. This design resulted in 62 treatment conditions, and allowed us to independently estimate all factors, quadratic factors and two factor interactions. The arrays were generated and analyzed using JMP software (7.0.1., SAS Institute, Inc., NC). To assess the results obtained from the parametric study, four criteria that characterize the performance of the system for each treatment condition were determined. The rst criterion assessed the ability to predict sequential spatial events observed during normal fracture healing, independent of time. Each event received a score of 0 for non-physiological and 1 for normal event. The events were: (1) brous tissue formation in the gap, (2) initial periosteal-bone formation, (3) growing periosteal callus including endochondral ossication, (4) brous/ cartilage formation in the gap, (5) external bony bridging, (6) bone creeping substitution, and (7) complete callus lled with bone. The second criterion

ARTICLE IN PRESS
H. Isaksson et al. / Journal of Biomechanics 42 (2009) 555564 559

immature bone. Outcome criteria evaluated during early phases of healing were most dependent on modulus and permeability of granulation tissue and modulus of cartilage, and outcome criteria assessed during later phases of healing were more highly dependent on the modulus of cartilage and permeability of immature bone (Table 3). From the results of the screening experiment, the most contributing factors were collected for the BoxBehnken design (Table 3). These were Youngs modulus of bone marrow, granulation tissue, cartilage, immature and mature bone as well as permeability of granulation tissue and immature bone. 3.2. BoxBehnken experiment All expected sequential events scored high, and the contribution to the variance was not informative. For amount of bone formation, the time until complete healing and mechanical stability, the properties that were of highest importance concurred with those identied in the screening experiment (Table 4). The amount of bone formation during the early stages of healing was most inuenced by the permeability of granulation tissue (20%), whereas at mid and late stages of healing it was most sensitive to the Youngs modulus of cartilage (mid 39%, late 20%) and permeability of granulation tissue (mid 28%, late 23%). Time to complete healing was substantially inuenced by parameters related to immature bone (permeability 38%, modulus 15%). Mechanical stability during the early phases was most inuenced by permeability of granulation tissue (interfragmentary movement 35%, stiffness 44%), followed by modulus of cartilage and interaction between modulus and permeability of granulation tissue. During later time points, the mechanical stability was more inuenced by modulus of cartilage (interfragmentary movement 41%), and permeability of immature bone (stiffness 33%) (Table 4). Most material properties had an approximately linear inuence on the outcome criteria (Figs. 3 and 4). The moduli of cartilage and immature bone were the only parameters which showed nonlinear responses for bone formation during the late phases of healing (Fig. 3b). Response surface analysis combined with the ANOVA showed that most interactions were minor. In contrast, few outcome criteria showed signicant interactions, exemplied by the amount of bone formation at late phases where the interaction between modulus of cartilage and permeability of granulation tissue was the second most important parameter (Fig. 4). However, signicant interactions were always related to already identied parameters of high importance for that outcome criterion (Table 4). Finally, the results of the statistical model were conrmed by running single simulations with the most benecial material properties for amount of bone formation and time until complete healing criteria. It conrmed that those simulations resulted in the highest amount of bone formation (15% more than average), the shortest time until complete healing (16 days shorter than average) as well as the lowest interfragmentary movement and highest stiffness (80% lower and 280% higher) (Fig. 5).

Fig. 2. Regions of interests (ROI) that were used for the amount of bone formation outcome criteria. During early stage (day 10) of healing, the amount of bone in periosteal reaction and callus formation were measured and averaged. During midphase (day 25) of healing, the amount of bone in the endosteal (intramedullary canal) callus and the bridging regions were assessed. To assess the amount of bone formation during the late stage, the bridging and gap (complete healing) regions were evaluated.

measured the progression of bone healing, based on the amount of bone formation in regions of interest (Fig. 2), during early (day 10), mid (day 25), and late (day 50) phases of healing. The third criterion measured the total time required until complete fracture healing as the number of days until the whole callus was predicted to be lled with bone, i.e. when each element contained over 75% bone matrix. These three criteria originated from our previous study (Isaksson et al., 2008b). The fourth criterion was added based on mechanical stability assessed by interfragmentary movement and axial stiffness at early and mid phases of healing. ANOVA was used to investigate the signicance and contribution of each factor. The percentage of total sum of square (%TSS) was calculated as the ratio of the sum of square of deviation about the mean for each factor divided by the total sum of square of deviation about the mean (Funkenbusch, 2005). %TSS for each of the outcome criteria were used to determine the contribution of each factor to the variance (Dar et al., 2002).

3. Results 3.1. Screening experiment Former predictions of bone healing were used as the baseline for evaluation of normal healing (Isaksson et al., 2008a). The expected sequential events during normal bone healing were not affected by the material properties. All simulations scored high, and the contribution to the variance was not informative. In contrast, the amount of bone formation, the time until complete healing and the mechanical stability were signicantly affected by the tissue properties. In general, Youngs modulus and permeability had high inuence, whereas Poissons ratios and porosity had little inuence (Table 3). All outcome criteria were mostly inuenced by three parameters; the permeability of granulation tissue, the Youngs modulus of cartilage and the permeability of

4. Discussion This study was motivated by the importance of callus tissue material properties on the mechanical behavior of the fracture callus, and thereby the predictions of tissue differentiation during healing. Similar to what was suggested by Lacroix (2001), material properties did not have a signicant effect on the sequence of predicted events during bone healing. However, they did inuence the rates of healing and the mechanical stability (Tables 3 and 4). Time to complete healing, amount of bone

ARTICLE IN PRESS
560 H. Isaksson et al. / Journal of Biomechanics 42 (2009) 555564

Table 3 ANOVA of each of the outcome criteria for the L64 screening experiment. ANOVA, %TSS Factors Time to complete healing Amount of bone formation Early phase 9.4 12.1 4.3 5.9 1.4 8.4 0.0 14.5 3.5 0.1 9.8 0.4 0.0 0.0 2.0 2.5 0.9 1.0 0.0 0.0 0.0 0.5 0.6 0.0 Mid phase 3.8 0.7 5.2 19.8 3.5 0.5 1.3 28.5 0.0 0.2 4.2 3.2 1.5 0.4 0.8 0.8 1.6 0.2 2.0 0.0 0.3 0.2 0.4 0.6 Late phase 1.6 1.8 0.0 18.9 5.4 11.4 0.9 7.3 0.0 1.6 16.4 0.0 1.3 0.0 0.0 0.0 1.6 2.4 1.3 2.0 0.0 1.9 1.4 0.0 Mechanical characterization IFM early phase 11.1 15.9 3.5 11.5 4.0 0.1 2.1 20.1 0.3 0.2 0.1 1.6 0.4 0.1 1.1 0.6 0.8 0.5 1.4 0.2 0.5 0.7 0.4 1.6 IFM mid phase 5.9 1.4 7.3 13.7 4.5 1.2 0.0 17.3 0.7 1.4 0.1 0.7 0.2 1.0 0.4 1.1 1.6 0.7 0.5 0.6 2.2 1.2 0.3 0.5 Stiffness early phase 1.8 12.7 1.3 10.5 5.1 0.9 0.3 25.5 0.2 0.0 7.8 0.1 0.1 0.0 0.0 0.6 0.4 0.3 0.0 0.3 0.1 0.0 1.0 0.0 Stiffness mid phase 1.7 0.4 4.5 16.9 6.4 2.5 0.7 17.2 0.5 0.7 15.5 2.8 0.8 0.3 0.5 0.4 1.5 0.9 1.2 0.0 0.2 0.0 0.5 0.1 4.7 6.2 3.7 14.1 5.0 3.4 0.7 18.4 0.8 0.7 8.5 1.2 0.6 0.2 0.7 0.7 1.3 0.9 0.9 0.5 0.5 0.8 0.6 0.4 Average

X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24

Youngs modulus Youngs modulus Youngs modulus Youngs modulus Youngs modulus Youngs modulus Permeability Permeability Permeability Permeability Permeability Permeability Poisson ratio Poisson ratio Poisson ratio Poisson ratio Poisson ratio Poisson ratio Porosity Porosity Porosity Porosity Porosity Porosity

Marrow GT FT C IMB MB Marrow GT FT C IMB MB Marrow GT FT C IMB MB Marrow GT FT C IMB MB

2.7 4.6 3.3 15.8 9.7 1.9 0.4 16.6 1.5 1.1 14.1 0.5 0.1 0.1 1.0 0.0 2.2 1.5 0.4 0.5 0.4 1.5 0.6 0.7

The percentages of the total sum of squares (%TSS) are listed. The most inuential parameters are highlighted. The total inuence and average were used to determine the factors in the higher level design. Abbreviations: GTgranulation tissue, FTbrous tissue, Ccartilage, IMBimmature bone, MBmature bone.

Table 4 ANOVA of each of the outcome criteria for the BoxBehnken response surface array. ANOVA, %TSS Main factor effects Time to complete healing Amount of bone formation Early phase Marrow GT C IMB MB GT IMB 2.2 0.8 10.9 15.4 0.7 13.8 37.7 6.6 3.0 7.5 0.0 0.0 20.3 11.7 Mid phase 5.7 1.8 38.6 6.3 0.0 27.7 6.9 Late phase 0.1 0.1 19.8 5.1 0.1 22.6 0.9 Mechanical characterization IFM early phase 14.3 7.0 23.9 6.8 0.0 34.9 1.4 IFM mid phase 4.4 2.4 41.1 1.3 0.0 10.5 1.5 Stiffness early phase 2.7 7.0 2.2 4.0 0.0 44.2 3.5 Stiffness mid phase 3.1 1.4 19.7 9.9 0.6 12.3 32.7 4.9 2.9 20.5 6.1 0.2 23.3 12.0 Average

X1 X2 X3 X4 X5 X6 X7

Youngs modulus Youngs modulus Youngs modulus Youngs modulus Youngs modulus Permeability Permeability

Signicant interactions X3 X4 Modulus C Modulus IMB X2 X6 Modulus GT Permeability GT X3 X6 Modulus C Permeability GT

0.5 0.2 1.2

0.3 18.7 2.6

0.3 0.4 1.1

9.7 0.0 21.5

0.7 0.2 0.4

0.4 0.7 9.7

0.0 10.0 0.7

0.0 1.0 0.7

1.5 3.9 4.7

All main factor effects are given together with three signicant two-factor interactions as percentages of the total sum of squares (%TSS). The most inuential parameters are highlighted. Abbreviations: GTgranulation tissue, FTbrous tissue, Ccartilage, IMBimmature bone, MBmature bone.

formation at specic time points and interfragmentary movement and stiffness were mainly affected by permeability of granulation tissue, Youngs modulus of cartilage, and permeability of immature bone. Clinical and experimental evidence exists for the importance of two of these factors. The character and magnitude of initial mechanical stability is important for success-rate of healing

(Lienau et al., 2005; Schell et al., 2005). Therefore, it could be anticipated that the properties of the haematoma and initial granulation tissue are important. When the relative parameter space was identical, the permeability was more inuential than the Youngs modulus. We speculate that this is because the modulus is too low to largely affect the outcome. However, both these tissue properties are today inadequately characterized.

ARTICLE IN PRESS
H. Isaksson et al. / Journal of Biomechanics 42 (2009) 555564 561

Fig. 3. Proles of the seven parameters that were used in the response surface study based on (a) time until complete healing, (b) amount of bone at early, mid, and late time points, and (c) the mechanical stability at early and mid phases. The dotted lines represent the average response of that outcome criteria based on all 62 treatment conditions, and the prole line the contribution of that parameter at high (1), mid (0) and low (+1) levels of each material property. Abbreviations: ModYoungs modulus, Permpermeability, Mbone marrow, GTgranulation tissue, FTbrous tissue, Ccartilage, IMBimmature bone, MBmature bone.

Also turnover of cartilage to bone is essential for successful bone healing (Colnot and Helms, 2001; Ford et al., 2003), and the duration of that phase is related to mechanical stability (Epari et al., 2006a), i.e. to the mechanical properties of the cartilage. The permeability of cartilage is very low compared to the other tissues involved during bone repair. Therefore, only the modulus of cartilage was highly inuential in this study. Effects of permeability of immature bone have not been studied experimentally in callus tissue, but literature reports large variation of this parameter (10141012 m4/N s) in trabecular bone dependent on anatomical location, porosity of the bone and species (Table 1). However, most of these studies have measured permeability in trabecular bone immersed in water after removing the bone marrow. The applicability of these permeability values to bone healing can therefore be questioned. Most material

properties were derived from mature tissues. Only recently properties of tissues encountered during repair, including granulation tissue were quantied (Leong and Morgan, 2008). Hence, additional experiments to quantify these properties within callus tissues are desirable. The sequential tissue differentiation events were not signicantly affected by the material properties. This implies that as long as no time-progressive events are studied or quantied, the accuracy of the material properties are of minor importance within the investigated parameter space. Modeling has reached far enough that studies aim to draw conclusions that are raterelated or quantitative (Bailon-Plaza and van der Meulen, 2003; Boccaccio et al., 2008; Gomez-Benito et al., 2007; Isaksson et al., 2006a, 2007), and in those cases the quantied material properties are essential.

ARTICLE IN PRESS
562 H. Isaksson et al. / Journal of Biomechanics 42 (2009) 555564

Fig. 4. Surface contour plots of the most inuential parameters for amount of bone formation during early and late phases of healing and the mechanical stability during early and mid phases of healing. The interactions between parameters were generally of minor importance, but for the amount of bone at late phases of healing, the interactions between permeability of granulation tissue and modulus of cartilage were second most inuential parameter. The contribution of the material parameters were calculated at high (1), mid (0) and low (+1) levels.

Despite differences in basic biological assumptions between mechano-biological models, the assumed equations for converting tissue partition into mechanical properties are similar, and since most current studies assume a similar set of material properties the identied most important parameters are likely to be the same. Early developments of computational models of fracture repair focused on improving mechanical models of biological tissues. Today poroelastic mechanical models are standard and recent developments have focused on implementing biological aspects of healing. These characteristics are crucial for bone healing models. However, until the assumptions and descriptions of tissue mechanical properties are better validated, the predictive capacity of these models remains qualitative. The benets and limitations of fractional factorial analysis have been discussed extensively in Isaksson et al. (2008b). When matrix production occurs, several tissue properties will be affected simultaneously. In the current study a 3 level BoxBehnken design was used to additionally be able to evaluate these interactions between parameters. The results indicate that certain interactions are prominent such as combinations of modulus and permeability of cartilage and granulation tissue (Table 4). However, from all two-variable interactions, only three interactions were within the 3 most inuential parameters for any of the outcome analyses. Those 3 interactions all involved the para-

meters that also had the highest main factor inuences. Porosity and Poissons ratio were given a relatively smaller parameter space compared to Youngs modulus and permeability. This was motivated by the higher consistency in literature for these parameters, and by the physical limitations for Poissons ratio and porosity (Table 1). Youngs modulus and permeability of both tissues that are well characterize and those without literature references, were given identical relative parameter spaces to avoid bias related to the parameter space in the statistical model (Isaksson et al., 2008b). Due to the difculty in nding one parameter which can be used to characterize the progression of fracture healing, we chose to use several outcome criteria. Three of them were used before (Isaksson et al., 2008b). Since this study is focusing on material properties, we also quantied the mechanical stability based on interfragmentary movement and stiffness. All together, these four criteria are believed to characterize the system well. For the rst time, this study provides a systematic approach to evaluate the sensitivity of the assumed tissue material properties during computational modeling of bone healing and showed that material properties, especially permeability of granulation tissue, Youngs modulus of cartilage and permeability of immature bone needs better characterization before the full potential of computational mechano-biological models can be achieved.

ARTICLE IN PRESS
H. Isaksson et al. / Journal of Biomechanics 42 (2009) 555564 563

Fig. 5. The statistical model was conrmed by running single simulations with the most benecial parameters. For the outcome criterion based on mechanical stability, the most benecial parameter combination resulted in 80% lower interfragmentary movement and 280% higher stiffness compared to the average response.

Conict of interest statement None of the authors have any conicts of interest.

Acknowledgements We acknowledge CSC, the Finnish IT Center for Science for computational tools, and the European Commission for funding (BONEQUAL-219980). References
Akizuki, S., Mow, V.C., Muller, F., Pita, J.C., Howell, D.S., Manicourt, D.H., 1986. Tensile properties of human knee joint cartilage: I. Inuence of ionic conditions, weight bearing, and brillation on the tensile modulus. J. Orthop. Res. 4, 379392. Anaguchi, Y., Yasuda, K., Majima, T., Tohyama, H., Minami, A., Hayashi, K., 2005. The effect of transforming growth factor-beta on mechanical properties of the brous tissue regenerated in the patellar tendon after resecting the central portion. Clin. Biomech. (Bristol., Avon.) 20, 959965. Andreykiv, A., van Keulen, F., Prendergast, P.J., 2008. Simulation of fracture healing incorporating mechanoregulation of tissue differentiation and dispersal/ proliferation of cells. Biomech. Model Mechanobiol. 7, 443461. Armstrong, C.G., Mow, V.C., 1982. Variations in the intrinsic mechanical properties of human articular cartilage with age, degeneration, and water content. J. Bone Joint Surg. Am. 64, 8894. Arramon, Y.P., Nauman, E.A., 2001. Intrinsic permeability of cancellous bone. In: Cowin, S.C. (Ed.), Bone Mechanics Handbook. CRC Press, Boca Raton, FL, pp. 25.1725.21.

Bailon-Plaza, A., van der Meulen, M.C., 2001. A mathematical framework to study the effects of growth factor inuences on fracture healing. J. Theor. Biol. 212, 191209. Bailon-Plaza, A., van der Meulen, M.C., 2003. Benecial effects of moderate, early loading and adverse effects of delayed or excessive loading on bone healing. J. Biomech. 36, 10691077. Blebea, J.S., Houseni, M., Torigian, D.A., Fan, C., Mavi, A., Zhuge, Y., Iwanaga, T., Mishra, S., Udupa, J., Zhuang, J., Gopal, R., Alavi, A., 2007. Structural and functional imaging of normal bone marrow and evaluation of its age-related changes. Semin. Nucl. Med. 37, 185194. Boccaccio, A., Prendergast, P.J., Pappalettere, C., Kelly, D.J., 2008. Tissue differentiation and bone regeneration in an osteotomized mandible: a computational analysis of the latency period. Med. Biol. Eng. Comput. 46, 283298. Chaffai, S., Padilla, F., Berger, G., Laugier, P., 2000. In vitro measurement of the frequency-dependent attenuation in cancellous bone between 0.2 and 2 MHz. J. Acoust. Soc. Am. 108, 12811289. Claes, L.E., Heigele, C.A., 1999. Magnitudes of local stress and strain along bony surfaces predict the course and type of fracture healing. J. Biomech. 32, 255266. Claes, L., Heitemeyer, U., Krischak, G., Braun, H., Hierholzer, G., 1999. Fixation technique inuences osteogenesis of comminuted fractures. Clin. Orthop. Relat. Res., 221229. Colnot, C.I., Helms, J.A., 2001. A molecular analysis of matrix remodeling and angiogenesis during long bone development. Mech. Dev. 100, 245250. Cowin, S.C., 1999. Bone poroelasticity. J. Biomech. 32, 217238. Dar, F.H., Meakin, J.R., Aspden, R.M., 2002. Statistical methods in nite element analysis. J. Biomech. 35, 11551161. Einhorn, T.A., 1998. The cell and molecular biology of fracture healing. Clin. Orthop. Relat. Res., S7S21. Elliott, D.M., Guilak, F., Vail, T.P., Wang, J.Y., Setton, L.A., 1999. Tensile properties of articular cartilage are altered by meniscectomy in a canine model of osteoarthritis. J. Orthop. Res. 17, 503508. Epari, D.R., Schell, H., Bail, H.J., Duda, G.N., 2006a. Instability prolongs the chondral phase during bone healing in sheep. Bone 38, 864870. Epari, D.R., Taylor, W.R., Heller, M.O., Duda, G.N., 2006b. Mechanical conditions in the initial phase of bone healing. Clin. Biomech. (Bristol., Avon.) 21, 646655. Fellah, Z.E., Chapelon, J.Y., Berger, S., Lauriks, W., Depollier, C., 2004. Ultrasonic wave propagation in human cancellous bone: application of Biot theory. J. Acoust. Soc. Am. 116, 6173. Ford, J.L., Robinson, D.E., Scammell, B.E., 2003. The fate of soft callus chondrocytes during long bone fracture repair. J. Orthop. Res. 21, 5461. Funkenbusch, P. (Ed.), 2005. Practical Guide to Designed Experiments: A Unied Modular Approach. Marcel Dekker, New York. Geris, L., Andreykiv, A., Van Oosterwyck, H., Sloten, J.V., van Keulen, F., Duyck, J., Naert, I., 2004. Numerical simulation of tissue differentiation around loaded titanium implants in a bone chamber. J. Biomech. 37, 763769. Geris, L., Gerisch, A., Sloten, J.V., Weiner, R., Oosterwyck, H.V., 2008a. Angiogenesis in bone fracture healing: a bioregulatory model. J. Theor. Biol. 251, 137158. Geris, L., Vandamme, K., Naert, I., Vander Sloten, J., Duyck, J., Van Oosterwyck, H., 2008b. Application of mechanoregulatory models to simulate peri-implant tissue formation in an in vivo bone chamber. J. Biomech. 41, 145154. Gerstenfeld, L.C., Cullinane, D.M., Barnes, G.L., Graves, D.T., Einhorn, T.A., 2003. Fracture healing as a post-natal developmental process: molecular, spatial, and temporal aspects of its regulation. J. Cell Biochem. 88, 873884. Gomez-Benito, M.J., Garcia-Aznar, J.M., Kuiper, J.H., Doblare, M., 2005. Inuence of fracture gap size on the pattern of long bone healing: a computational study. J. Theor. Biol. 235, 105119. Gomez-Benito, M.J., Fornells, P., Garcia-Aznar, J.M., Seral, B., Seral-Innigo, F., Doblare, M., 2007. Computational comparison of reamed versus unreamed intramedullary tibial nails. J. Orthop. Res. 25, 191200. Grimm, M.J., Williams, J.L., 1997. Measurements of permeability in human calcaneal trabecular bone. J. Biomech. 30, 743745. Hartsock, R.J., Smith, E.B., Petty, C.S., 1965. Normal variations with aging of the amount of hematopoietic tissue in bone marrow from the anterior iliac crest a study made from 177 cases of sudden death examined by necropsy. Am. J. Clin. Pathol. 43, 326331. Hori, R.Y., Lewis, J.L., 1982. Mechanical properties of the brous tissue found at the bonecement interface following total joint replacement. J. Biomed. Mater. Res. 16, 911927. Hosokawa, A., Otani, T., 1997. Ultrasonic wave propagation in bovine cancellous bone. J. Acoust. Soc. Am. 101, 558562. Hosokawa, A., Otani, T., 1998. Acoustic anisotropy in bovine cancellous bone. J. Acoust. Soc. Am. 103, 27182722. Isaksson, H., van Donkelaar, C.C., Huiskes, R., Ito, K., 2006a. Corroboration of mechanoregulatory algorithms for tissue differentiation during fracture healing: comparison with in vivo results. J. Orthop. Res. 24, 898907. Isaksson, H., Wilson, W., van Donkelaar, C.C., Huiskes, R., Ito, K., 2006b. Comparison of biophysical stimuli for mechano-regulation of tissue differentiation during fracture healing. J. Biomech. 39, 15071516. Isaksson, H., Comas, O., van Donkelaar, C.C., Mediavilla, J., Wilson, W., Huiskes, R., Ito, K., 2007. Bone regeneration during distraction osteogenesis: mechanoregulation by shear strain and uid velocity. J. Biomech. 40, 20022011. Isaksson, H., van Donkelaar, C.C., Huiskes, R., Ito, K., 2008a. A mechano-regulatory bone-healing model incorporating cell-phenotype specic activity. J. Theor. Biol. 252, 230246.

ARTICLE IN PRESS
564 H. Isaksson et al. / Journal of Biomechanics 42 (2009) 555564

Isaksson, H., van Donkelaar, C.C., Huiskes, R., Yao, J., Ito, K., 2008b. Determining the most important cellular characteristics for fracture healing using design of experiments methods. J. Theor. Biol. 255, 2639. Johnson, M.W., Chakkalakal, D.A., Harper, R.A., Katz, J.L., Rouhana, S.W., 1982. Fluid ow in bone in vitro. J. Biomech. 15, 881885. Julkunen, P., Kiviranta, P., Wilson, W., Jurvelin, J.S., Korhonen, R.K., 2007. Characterization of articular cartilage by combining microscopic analysis with a bril-reinforced nite-element model. J. Biomech. 40, 18621870. Julkunen, P., Korhonen, R.K., Herzog, W., Jurvelin, J.S., 2008. Uncertainties in indentation testing of articular cartilage: a bril-reinforced poroviscoelastic study. Med. Eng. Phys. 30, 506515. Jurvelin, J.S., Buschmann, M.D., Hunziker, E.B., 1997. Optical and mechanical determination of Poissons ratio of adult bovine humeral articular cartilage. J. Biomech. 30, 235241. Kelly, D.J., Prendergast, P.J., 2005. Mechano-regulation of stem cell differentiation and tissue regeneration in osteochondral defects. J. Biomech. 38, 14131422. Kenwright, J., Goodship, A.E., 1989. Controlled mechanical stimulation in the treatment of tibial fractures. Clin. Orthop. Relat. Res., 3647. Kohles, S.S., Roberts, J.B., 2002. Linear poroelastic cancellous bone anisotropy: trabecular solid elastic and uid transport properties. J. Biomech. Eng. 124, 521526. Korhonen, R.K., Wong, M., Arokoski, J., Lindgren, R., Helminen, H.J., Hunziker, E.B., Jurvelin, J.S., 2002. Importance of the supercial tissue layer for the indentation stiffness of articular cartilage. Med. Eng. Phys. 24, 99108. Laasanen, M.S., Saarakkala, S., Toyras, J., Hirvonen, J., Rieppo, J., Korhonen, R.K., Jurvelin, J.S., 2003. Ultrasound indentation of bovine knee articular cartilage in situ. J. Biomech. 36, 12591267. Lacroix, D. (Ed.), 2001. Simulation of tissue differentiation during fracture healing. University of Dublin. Lacroix, D., Prendergast, P.J., 2002. A mechano-regulation model for tissue differentiation during fracture healing: analysis of gap size and loading. J. Biomech. 35, 11631171. Leong, P.L., Morgan, E.F., 2008. Measurement of fracture callus material properties via nanoindentation. Acta Biomater. 4, 15691575. Lienau, J., Schell, H., Duda, G.N., Seebeck, P., Muchow, S., Bail, H.J., 2005. Initial vascularization and tissue differentiation are inuenced by xation stability. J. Orthop. Res. 23, 639645. Lim, T.H., Hong, J.H., 2000. Poroelastic properties of bovine vertebral trabecular bone. J. Orthop. Res. 18, 671677. Lundeen, G.A., Vajda, E.G., Bloebaum, R.D., 2000. Age-related cancellous bone loss in the proximal femur of caucasian females. Osteoporos. Int. 11, 505511. Meunier, P., Aaron, J., Edouard, C., Vignon, G., 1971. Osteoporosis and the replacement of cell populations of the marrow by adipose tissue. A quantitative study of 84 iliac bone biopsies. Clin. Orthop. Relat. Res. 80, 147154. Moussa, M., Swider, P., Babile, R., Fernandez, X., Remignon, H., 2008. Effects of physical activities on biochemical and biomechanical properties of tendons in two commercial types of chickens. Connect. Tissue Res. 49, 7684. Mow, V.C., Kuei, S.C., Lai, W.M., Armstrong, C.G., 1980. Biphasic creep and stress relaxation of articular cartilage in compression? Theory and experiments. J. Biomech. Eng. 102, 7384. Nauman, E.A., Fong, K.E., Keaveny, T.M., 1999. Dependence of intertrabecular permeability on ow direction and anatomic site. Ann. Biomed. Eng. 27, 517524.

Ochoa, J.A., Hillberry, B.M., 1992. Permeability of bovine cancellous bone. Trans. 38th ORS 162. Pakula, M., Padilla, F., Laugier, P., Kaczmarek, M., 2008. Application of Biots theory to ultrasonic characterization of human cancellous bones: determination of structural, material, and mechanical properties. J. Acoust. Soc. Am. 123, 24152423. Perez, M.A., Prendergast, P.J., 2007. Random-walk models of cell dispersal included in mechanobiological simulations of tissue differentiation. J. Biomech. 40, 22442253. Phadke, M. (Ed.), 1989. Quality Engineering Using Robust Design. PTR Prentice-Hall Inc., Engelwood Cliffs, NJ. Prendergast, P.J., Huiskes, R., Soballe, K., 1997. ESB Research Award 1996 biophysical stimuli on cells during tissue differentiation at implant interfaces. J. Biomech. 30, 539548. Richardson, J.B., Cunningham, J.L., Goodship, A.E., OConnor, B.T., Kenwright, J., 1994. Measuring stiffness can dene healing of tibial fractures. J. Bone Joint Surg. Br. 76, 389394. Salome, M., Peyrin, F., Cloetens, P., Odet, C., Laval-Jeantet, A.M., Baruchel, J., Spanne, P., 1999. A synchrotron radiation microtomography system for the analysis of trabecular bone samples. Med. Phys. 26, 21942204. Schafer, M.B., Burr, D.B., 1988. Stiffness of compact bone: effects of porosity and density. J. Biomech. 21, 1316. Schell, H., Epari, D.R., Kassi, J.P., Bragulla, H., Bail, H.J., Duda, G.N., 2005. The course of bone healing is inuenced by the initial shear xation stability. J. Orthop. Res. 23, 10221028. Schnitzler, C.M., Mesquita, J., 1998. Bone marrow composition and bone microarchitecture and turnover in blacks and whites. J. Bone Miner. Res. 13, 13001307. Setton, L.A., Zhu, W., Mow, V.C., 1993. The biphasic poroviscoelastic behavior of articular cartilage: role of the surface zone in governing the compressive behavior. J. Biomech. 26, 581592. Setton, L.A., Mow, V.C., Muller, F.J., Pita, J.C., Howell, D.S., 1997. Mechanical behavior and biochemical composition of canine knee cartilage following periods of joint disuse and disuse with remobilization. Osteoarthritis Cartilage 5, 116. Shepherd, D.E., Seedhom, B.B., 1999. The instantaneous compressive modulus of human articular cartilage in joints of the lower limb. Rheumatology (Oxford) 38, 124132. Smit, T.H., Huyghe, J.M., Cowin, S.C., 2002. Estimation of the poroelastic parameters of cortical bone. J. Biomech. 35, 829835. Tanck, E., Van Donkelaar, C.C., Jepsen, K.J., Goldstein, S.A., Weinans, H., Burger, E.H., Huiskes, R., 2004. The mechanical consequences of mineralization in embryonic bone. Bone 35, 186190. Wayne, J.S., Kraft, K.A., Shields, K.J., Yin, C., Owen, J.R., Disler, D.G., 2003. MR imaging of normal and matrix-depleted cartilage: correlation with biomechanical function and biochemical composition. Radiology 228, 493499. Wear, K.A., Laib, A., Stuber, A.P., Reynolds, J.C., 2005. Comparison of measurements of phase velocity in human calcaneus to Biot theory. J. Acoust. Soc. Am. 117, 33193324. Wei, X., Rasanen, T., Messner, K., 1998. Maturation-related compressive properties of rabbit knee articular cartilage and volume fraction of subchondral tissue. Osteoarthritis Cartilage 6, 400409.