Downloaded from rsta.royalsocietypublishing.

org on December 2, 2009

Numerical modelling of the shoulder for clinical applications

Philippe Favre, Jess G. Snedeker and Christian Gerber Phil. Trans. R. Soc. A 2009 367, 2095-2118 doi: 10.1098/rsta.2008.0282

References Rapid response Subject collections

This article cites 120 articles, 8 of which can be accessed free

http://rsta.royalsocietypublishing.org/content/367/1895/2095.ful l.html#ref-list-1 Respond to this article http://rsta.royalsocietypublishing.org/letters/submit/roypta;367/ 1895/2095 Articles on similar topics can be found in the following collections biomechanics (26 articles)

Email alerting service

Receive free email alerts when new articles cite this article - sign up in the box at the top right-hand corner of the article or click here

To subscribe to Phil. Trans. R. Soc. A go to: http://rsta.royalsocietypublishing.org/subscriptions

This journal is 2009 The Royal Society

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

Phil. Trans. R. Soc. A (2009) 367, 20952118 doi:10.1098/rsta.2008.0282

REVIEW

Numerical modelling of the shoulder for clinical applications

B Y P HILIPPE F AVRE *, J ESS G. S NEDEKER
AND

C HRISTIAN G ERBER

Laboratory for Orthopaedic Research, Department of Orthopaedics, Balgrist, University of Zurich, Forchstrasse 340, 8008 Zurich, Switzerland
Research activity involving numerical models of the shoulder is dramatically increasing, driven by growing rates of injury and the need to better understand shoulder joint pathologies to develop therapeutic strategies. Based on the type of clinical question they can address, existing models can be broadly categorized into three groups: (i) rigid body models that can simulate kinematics, collisions between entities or wrapping of the muscles over the bones, and which have been used to investigate joint kinematics and ergonomics, and are often coupled with (ii) muscle force estimation techniques, consisting mainly of optimization methods and electromyography-driven models, to simulate muscular action and joint reaction forces to address issues in joint stability, muscular rehabilitation or muscle transfer, and (iii) deformable models that account for stressstrain distributions in the component structures to study articular degeneration, implant failure or muscle/tendon/bone integrity. The state of the art in numerical modelling of the shoulder is reviewed, and the advantages, limitations and potential clinical applications of these modelling approaches are critically discussed. This review concentrates primarily on muscle force estimation modelling, with emphasis on a novel muscle recruitment paradigm, compared with traditionally applied optimization methods. Finally, the necessary benchmarks for validating shoulder models, the emerging technologies that will enable further advances and the future challenges in the eld are described.
Keywords: shoulder biomechanics; computer/numerical modelling; glenohumeral joint; muscle force estimation; rigid body model; nite element

1. Introduction (a ) Increasing interest in shoulder modelling Historically, in orthopaedics, most numerical simulations have focused on the hip and the knee, with fewer on the shoulder. From a clinical perspective, the hip and knee have largely occupied the interests of clinical and industrial researchers because the vast majority of joint replacements are performed at these joints.
* Author for correspondence (pfavre@research.balgrist.ch). One contribution of 15 to a Theme Issue The virtual physiological human: tools and applications I.

2095

This journal is q 2009 The Royal Society

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

2096
70 60 50 no. of publications 40 30 20 10 0 <1900

P. Favre et al.

1900 1979

1980 1989 19901999 publication period

2000mid- 2008

Figure 1. International publications based on numerical models of the shoulder, listed on PubMed. The scale on the horizontal axis is not linear.

From a modelling standpoint, the complexity of the shoulder joint itself may be a disincentive to modellers, who have to cope with intricate active and passive stabilizing mechanisms and an extremely large range of joint motions (Hogfors et al. 1995). Upper extremity motions are by necessity more variable than the locomotive movements of the lower limb (a well-dened cyclic motion), and while a two-dimensional analysis of gait can reasonably characterize leg kinematics, such a simplied treatment of the shoulder is inadequate (Rau et al. 2000). Shoulder problems are gaining research interest as 2030 per cent of the population suffer from shoulder pain and 8.8 per cent suffer functional impairment (Lock et al. 1999; Makela et al. 1999). Furthermore, many major clinical challenges of shoulder orthopaedics have proven to be persistent, and a need for deeper understanding of shoulder pathology and treatment has become clear. On the modelling side, lessons learned from other joints can now be applied to the shoulder, and improvements in software and computational power have facilitated the building and solving of ever more complex models. All these factors have driven the exponential increase in the number of publications involving numerical models of the shoulder (gure 1), the most clinically relevant of which we review here. (b ) Essential clinical issues with potential to be addressed using models Among the many shoulder problems encountered in daily clinical practice, several concerns are of particular interest to the modeller. First, they correspond to some of the most challenging issues in shoulder orthopaedics. Second, numerical simulations offer unique potential to improve their treatment.
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

Review. Numerical modelling of the shoulder

2097

(i) Glenohumeral instability The glenohumeral joint is the most frequently dislocated major joint of the body, affecting at least 1.7 per cent of people in the course of their lifetime (Hovelius 1982). The contribution of the glenohumeral surface shape, capsule and ligaments to stability has been widely studied, although glenohumeral joint stability is primarily ensured by coordination of muscular actions (Lippitt & Matsen 1993; Veeger & van der Helm 2007). This is especially complex, as some muscles may act as stabilizers, while others rather tend to destabilize the joint, depending on joint position (van der Helm 1994a; McMahon & Lee 2002; Werner et al. 2007; Favre et al. 2009). Our present knowledge fails to explain many cases of instability, owing to insufcient understanding of how the muscles simultaneously move the humerus and prevent glenohumeral dislocation (Kronberg et al. 1991; McMahon & Lee 2002; Veeger & van der Helm 2007). Here, muscle force estimation models can be considerably helpful in studying active stabilization mechanisms. Rigid body or deformable models can be used to assess the inuence of passive stabilizers and joint conformity in more depth. (ii) Rotator cuff tears The muscles of the rotator cuff maintain joint stability by pulling the humeral head into the glenoid socket. Tears of the rotator cuff tendons are common and have been shown to be present in over 50 per cent of asymptomatic people above the age of 60 (Sher et al. 1995), reecting accumulated damage over the course of a lifetime. These tears may progress to massive, symptomatic tears, eventually leading to irreversible changes in the physiology of the muscle and tendon (Gerber et al. 2004), making surgical repair difcult. Torn cuff tendons can lead to pain, instability, dysfunction and osteoarthritis, and in the worst cases can degenerate to cuff tear arthropathy, in which the humeral head collapses (Neer et al. 1983). The pathogenesis of rotator cuff tears remains unclear, and the best treatment for this common disorder remains a source of debate (Williams et al. 2004). The evolution to a symptomatic tear is not well understood, and although many other factors are involved (vascular, genetic, etc.), nite-element (FE) analyses can provide rst insight into tendon-loading patterns (Luo et al. 1998; Wakabayashi et al. 2003; Sano et al. 2006) or could give guidelines to optimize the mechanics of tendon repair. Interestingly, the functional decits associated with rotator cuff tears are highly variable among patients, varying from normal range of motion to pseudoparalysis, indicating that some patients are able to cope with the torn muscles, while others cannot. Here, muscle force estimation models could help us to understand how compensation involving the remaining viable muscles occurs and to dene new rehabilitation strategies. (iii) Shoulder arthroplasty The number of partial and total shoulder replacements remains relatively small in comparison with that for the hip or knee, but has tripled in just 10 years in the USA, reaching 29 000 partial and total shoulder replacements in 2004 (Kozak et al. 2006). The two major complications associated with joint replacement are loosening of the glenoid component and joint instability (Franklin et al. 1988; Wirth & Rockwood 1996).
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

2098

P. Favre et al.

Although the aetiology of glenoid loosening is not yet fully understood, it is generally believed to have several potential (and possibly interrelated) sources, notably implant design, implant wear and particle formation, surgical technique, poor bone quality and eccentric implant loading. Eccentric loading can be due to migration of the humeral head (Franklin et al. 1988) or disadvantageous implant positioning (Nyffeler et al. 2006). In their ability to explore mechanical phenomena, FE simulations offer potential to understand these issues. As with the normal glenohumeral joint, instability of the prosthetic joint is also a primary concern. Because stability of the shoulder depends so heavily on the active musculature, muscle force estimation models may improve our understanding of the stability changes induced by an implant. Constrained devices such as the reverse prosthesis (in which the ball is transposed to the scapula and the socket to the humerus) provide increased intrinsic stability, but they have been associated with higher rates of surgical revision (Matsen et al. 2007). Moreover, the geometric and kinematic changes induced by the reverse design present many questions to be investigated (Boileau et al. 2005), and rigid body models may provide a tool for investigation. Generally, models may facilitate our quest for increased implant longevity by helping the surgeon in choosing an appropriate size, type and position of the prosthesis and guide in optimizing soft tissue tension and force balance. (iv) Tendon transfer Muscle transfer represents a viable treatment option to restore the lost function in a variety of difcult shoulder issues. The anatomical insertion of a functioning muscle is detached and transferred to a different location to full a new role and restore a lost function. For instance, the latissimus dorsi transfer can restore external rotation in the case of irreparable rotator cuff tears (Gerber 1992), or be further combined with arthroplasty (Gerber et al. 2007). However, the clinical outcome of muscle transfers is still difcult to predict, and the biomechanical implications of such procedures are only beginning to be understood (Magermans et al. 2004a,b; Favre et al. 2008a). Rigid body models simulating muscle wrapping can provide rst information on the muscle path and moment arm of the transferred muscle. Muscle force estimation models can then help in understanding how the changes induced by the transferred muscle inuence the distribution of muscle forces over the whole joint. Generally, simulations should provide guidelines (choice of muscle to transfer, optimal insertion site and muscle training) to maximize the functional potential. 2. Numerical models of the shoulder Simulations performed with numerical models allow investigation of aspects that are otherwise difcult or impossible to quantify, overcoming technical limits (deterioration of tissues, adequate placement of sensors, etc.) and ethical limits (invasiveness and short supply of specimens) on direct in vivo or in vitro measurements. Depending on the aspect of shoulder function to be investigated, various modelling approaches can be selected. Because the clinical question generally dictates the kind of model to be used, in the current review we categorize the models based on the type of output they generate, and thus the type of clinical question they
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

Review. Numerical modelling of the shoulder

2099

Table 1. Potential of different numerical model types to address the most pressing shoulder clinical issues, as dened in this review. (Each clinical issue is subdivided into respective relevant research topics. A cross means that the output delivered by the numerical model type could be or has been implemented to address them directly. This list is not exhaustive.) shoulder model type muscle force estimation

clinical issue glenohumeral instability rotator cuff tears

research topic passive stabilization muscular stabilization joint conformity/constraint muscle overload pathogenesis of tear compensation/rehabilitation tendon repair

rigid body !

deformable !

! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

shoulder implant type or design arthroplasty xation range of motion implant component positioning cement thickness shoulder function implant intrinsic stability tendon transfer

! !

choice of muscle to transfer ! prediction of functional outcome post-operative muscle training optimize insertion site !

can address (table 1). This classication is not exhaustive, largely allows for overlapping between categories and focuses on the clinical issues described above. Accordingly, the available models can be broadly categorized into three main groups: (i) rigid body models that can simulate kinematics, collisions between entities or wrapping of the muscles over the bones, and which have been used to investigate joint kinematics and ergonomics, and have very often been coupled with (ii) muscle force estimation techniques, consisting mainly of optimization methods and electromyography (EMG)-driven models, to simulate muscular action (inverse dynamics) and joint reaction forces to address issues in joint stability, muscular rehabilitation or muscle transfer, and (iii) deformable models that account for stressstrain distributions in the component structures, and which can be used to study articular degeneration, implant failure or muscle integrity. This review does not intend to provide an exhaustive list of all published shoulder models but points out several key models that have been specically tailored to address clinical questions. (a ) Rigid body models Rigid body models idealize a system as consisting of solid bodies connected by kinematic constraints, and in which deformations of the bodies are neglected.
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

2100
(a)

P. Favre et al.

(b)

Figure 2. (a) Notching of the humerus at the level of the inferior glenoid border in patients with a conventional shoulder implant and (b) rigid body model to investigate whether malpositioning of implant components can cause this notching through impingement, eventually leading to eccentric loading of the glenoid component and its loosening.

In applying this approach to the shoulder, the bones are represented as rigid segments, and the muscles as adjustable length segments that link their origin and insertion. Kinematics, collisions between entities or wrapping of the muscles over the bones can be simulated. Because deformations are not considered, the necessary computational power is considerably less than that required for deformable models, and useful information can often be extracted from relatively simple models that run quickly on low-cost computational hardware. Rigid body models have been used in kinematics studies (i.e. to describe motion of the rigid segments, without consideration of the causes leading to movement) to assess the range of shoulder motions in rehabilitation and ergonomics applications (Johnson & Gill 1987; Engin & Tumer 1989; Klopcar et al. 2007). As an example of a model specically tailored to address a clinical question, we implemented a rigid body model to investigate whether malpositioning of conventional shoulder implant components can cause impingement, leading to the notching observed in patients, nally inducing eccentric loosening and possible glenoid loosening (gure 2; Favre et al. 2008b). In an inverse dynamics approach, such models have been used to estimate joint moments in the whole upper limb during a specic motion (Barker et al. 1997).
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

Review. Numerical modelling of the shoulder

2101

Most often, rigid body models have been used for rigid body dynamics, in combination with muscle force optimization techniques (described in detail in 2b later). Here, the rigid body models provide anatomical data, specically the muscle moment arms and lines of action, as input to a coupled muscle force estimation method. Such models have been shown to predict accurately muscle moment arms of the rotator cuff in comparison with experimentally measured values (Gatti et al. 2007), supporting the use of rigid body models for calculation of moment arms, with the advantage that these can be easily computed for any position of the joint at any time, as opposed to those measured experimentally. In turn, the determined muscle forces may then be used to drive the rigid body kinematics in a forward dynamics approach. Here, the rigid body model serves for visualization of an induced motion, making muscle force estimation software more intuitive and facilitating user interaction with the model (Dickerson et al. 2007). Commercially available models for the whole musculoskeletal system (AnyBody Technology A/S, Aalborg, Denmark; SIMM, MusculoGraphics Inc., Santa Rosa, CA, USA) also incorporate muscle force optimization algorithms. These packages have been customized and used in several shoulder studies (Holzbaur et al. 2005; Langenderfer et al. 2005; Charlton & Johnson 2006). While rigid body models are useful for many applications, they do have limitations. Most published rigid body models of the shoulder are limited by a simplied representation of the glenohumeral joint as an ideal joint, with predened degrees of freedom. It has been approximated as consisting of three hinges with dened axes of rotation (Barker et al. 1997; Klopcar et al. 2007) or by dening it as an ideal ball-and-socket joint (van der Helm 1994b; Garner & Pandy 1999; Maurel & Thalmann 1999; Favre et al. 2008b). These simplications prevent translations at the glenohumeral joint and thus limit studies on joint stability. Furthermore, the problem of muscle wrapping requires some improvement. Accounting for physical contact between muscles and bones, or between muscles and other soft tissues, is a complicated issue that can limit accuracy in modelling muscle actions over large ranges of motions. This is a major problem as it directly affects the muscle moment arm and line of action (Charlton & Johnson 2001). It is sometimes solved articially by dening so-called via points through which the muscle is forced to pass or by replacing the complex bony geometry with simpler shapes (van der Helm et al. 1992; Garner & Pandy 2000; Holzbaur et al. 2005); but these techniques do not mirror the in vivo behaviour. More recent methods model the entire muscle volume using the FE method (Blemker et al. 2007). As a nal limitation, rigid body models neglect tissue deformation and cannot be used to address some very important questions of shoulder biomechanics (such as arthroplasty, where quantifying mechanical bone stresses are important). However, these sacrices in realism do yield a signicant advantage in smaller requirements for computational power, allowing larger models of not only the shoulder but also often the entire upper limb or torso. This is essential in efciently studying bi-articular muscles and the coupling between multiple joints. (b ) Muscle force estimation Estimation of muscle loading is critical in predicting when tendon tears will occur. When tears have already been diagnosed, this information is required to
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

2102

P. Favre et al.

develop strategies for safely reconstructing them, or to optimize rehabilitation by focusing on the remaining viable muscles. It is also necessary to understand how the central nervous system controls the musculoskeletal system during specic tasks in order to best treat neurological or muscular deciencies. Finally, accurately estimating the forces at the joint contact interface is fundamental to prosthesis design in preventing accelerated implant wear and premature failure. In order to simulate how muscles are coordinated to move the arm or to hold it in a certain position, two parameters must be dened: rst, an appropriate set of active muscles (muscle recruitment) and second, the distribution of force among these muscles to reach equilibrium. Early models of shoulder muscle force balance were two-dimensional systems that considered only a few muscles, and could be solved by simple analytical methods (Inman et al. 1944; De Luca & Forrest 1973; Poppen & Walker 1978; Bassett et al. 1990). While these models were limited to a simple arm movement (elevation) and included a restricted set of muscles, they nonetheless delivered important insights into some basic working principles of shoulder function. More importantly, they provided a technical foundation on which later three-dimensional models could be built to address more complex questions, necessitating the use of computers. When comprehensively modelling the muscular forces at the shoulder, the main hurdle is of a mathematical nature. In the simplied case in which the glenohumeral joint is represented as a ball-and-socket joint (thus removing/ constraining the three translational degrees of freedom), the system is governed by three equations describing the equilibrium in each rotational degree of freedom. The muscle forces are treated as unknowns in these equations. This system of three equations would have one unique solution only if no more than three unknowns were present. In actuality, the many muscles that cross the glenohumeral joint present a much less dened system. Muscles with large origins are often subdivided to allow for a differentiated action, as moment arm length and direction of tendon action can vary considerably within an individual muscle (van der Helm & Veenbaas 1991; Favre et al. 2005, 2009), so that many shoulder models regularly include more than 20 muscle segments. This yields a mathematically indeterminate system comprising more unknowns than equations and is characterized by an innite number of possible solutions. The main issue is to nd a method that systematically isolates a suitable solution among these innite number of solutions. Most computational methods tackle the indeterminacy issue using mathematical optimization. The forces are distributed among muscles in such a way that a chosen cost function is minimized, while ensuring that these forces satisfy given physical constraints. The cost function can be based upon muscle stress, force, energy, fatigue, etc., and the constraints usually dictate that the muscle forces can apply tension only, must satisfy the equilibrium equations and fall within reasonable upper (tetanic muscle force) and lower (negligible force) muscle force limits (Karlsson & Peterson 1992). The rst optimization models were developed to simulate muscular load sharing at other joints (Penrod et al. 1974; Seireg & Arvikar 1975), and the shoulder represents a more recent application. The rst three-dimensional model of shoulder muscle force estimation was extensively described by Karlsson & Peterson (1992) and was later followed by the most frequently cited and implemented model to date (van der Helm 1994b).
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

Review. Numerical modelling of the shoulder

2103

One of the major problems in optimization techniques is the choice of an appropriate cost function. Comparisons with EMG activity have shown that some cost functions predict onoff muscle recruitment patterns better than others, but it is not yet clear what principles would dictate how muscles are recruited in vivo (Karlsson & Peterson 1992; van der Helm 1994a; Erdemir et al. 2007). Furthermore, muscle load sharing strategy is likely to differ between subjects, depending to some extent on the type of activity and varying in response to fatigue, mental demands and visual feedback or in the presence of musculoskeletal disorders, such as rotator cuff tears (Kronberg et al. 1991; Jensen et al. 2000; Steenbrink et al. 2006). Finally, co-contraction of antagonistic muscles can be predicted (Jinhaa et al. 2006), but it has been shown to confound conventional optimization-based models (Cholewicki et al. 1995; Dickerson et al. 2008). In an attempt to overcome the drawbacks of analytical optimization techniques, experimentally based, EMG-driven models that were rst developed for the lower back (McGill 1992) have been extended to the shoulder (Koike & Kawato 1995; Laursen et al. 1998; Langenderfer et al. 2005). In such models, the set of active muscles is identied in recordings of EMG activity for isolated shoulder positions. The muscle force is then estimated by assuming a linear relationship between EMG and force under isometric conditions. One advantage is that the recruitment of a particular muscle is directly indicated by the EMG measurements, and realistic muscle recruitment patterns can be immediately implemented in the model. In this way, co-contraction of antagonistic muscles may be directly accounted for and simulated. The main weakness of such models is that reliable EMG recordings of all relevant muscles must be available for the simulated position. This is technically demanding (even if possible) and brings with it the well-known difculties inherent in EMG signal measurement and processing (De Luca 1997). All muscle force estimation models are limited in their ability to be individualized, owing to the variability in anatomy and neuromuscular control, and this restricts the general applicability of a single model. Despite the limitations of these methods, and the difculties in validating such models (to be discussed in 3b later), muscle force estimation models have been applied successfully to investigate a variety of clinical questions. They have been used to evaluate the inuence of a prosthesis on the muscular forces (de Leest et al. 1996), to assess the inuence of scapular neck malunion on shoulder function (Chadwick et al. 2004), to test tendon transfers (Magermans et al. 2004a,b) and to prevent overload injuries in wheelchair design and propulsion (van der Helm & Veeger 1996; Veeger et al. 2002; Lin et al. 2004; van Drongelen et al. 2005, 2006). Because muscle force estimation models must necessarily entail a muscle recruitment strategy, they give explicit insight into the manner in which the central nervous system drives the musculoskeletal system. Changes in the applied cost function or recruitment criteria directly affect the prediction of recruited muscles and can be varied to explore neuromuscular control hypotheses. The effect of a neurological or muscular deciency can be assessed by scaling the allowed maximum relative force of selected muscles (van Drongelen et al. 2006). In a forward dynamics approach, muscle forces can be used to calculate joint torques and trajectory using the dynamics equations (Koike & Kawato 1995), and combined with rigid body models with the previously described advantage of motion visualization. Finally, such models are useful for investigating how joint stability is achieved, by considering
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

2104

P. Favre et al.
(b) moment (N m) 3 2 1 0 1 2 3 4

My Mx Mz

(a) moment (N m)

3 2 1 0 1 2 3 4

group 1 Mx Mz muscles

My

group 2 moment (Nm)

3 2 1 0 1 2 3 4

My Mx Mz

Figure 3. Muscle recruitment paradigm in the Algorithm for Shoulder Force Estimation (ASFE): (a) external moment components and (b) muscle potential moment components. The three external moment components (grey bars in (a), shown as dotted or dashed lines in (b)) are assessed to recruit muscles from two groups. Muscles from group 1 oppose the greatest external moment component only (here, the y-component of the external moment, shown in dashed lines, is counteracted by the y-component of the muscle, shown with black ll), disregarding the other two components (shown with grey ll). Muscles from group 2 concurrently oppose all three external moment components simultaneously (external moment components shown in dashed lines, muscle moment components shown with black ll).

the intersection of the resultant force with the glenoid, with the joint remaining stable as long as the resulting joint force falls within the glenoid boundaries (van der Helm 1994b; Favre et al. 2005). (i) A new paradigm for muscle force estimation In an effort to avoid the previously described limitations that accompany traditional muscle force estimation methods (choice of an appropriate cost function, co-contraction prediction and requirements for EMG measurements), we have developed an algorithm to predict the muscle forces required for shoulder joint equilibrium. This algorithm implements a novel recruitment strategy that focuses on selecting muscles with a relative mechanical advantage, and a corresponding set of muscles that counterbalance secondary joint moments. Because the recruitment process is central to the new method, we describe it here briey. A more comprehensive description of the entire algorithm can be found elsewhere (Favre et al. 2005). In this muscle recruitment paradigm (gure 3), an external moment at the joint centre of rotation is decomposed into three orthogonal components. The Algorithm for Shoulder Force Estimation (ASFE) rst recruits the muscles that
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

Review. Numerical modelling of the shoulder

2105

have the largest potential mechanical advantage in offsetting the greatest of the three external moment components, without any regard for the other two components (group 1). Here, potential mechanical advantage is dened by the muscle lever arms for the current position (Favre et al. 2009) multiplied by muscle cross section. In order to offset the two remaining components of the external moment, muscles that simultaneously oppose all three external components are recruited (group 2). Muscles that do not t within these two groups are not active in the current loop (but may become active in later loops, thus allowing simulation of co-contraction, depending on which component of the remaining external moment becomes the greatest). The equilibrium force equations then attribute forces to the set of recruited muscles in proportion to each muscles cross-sectional area (thereby making the system mathematically determinate). The boundary conditions ensure that the assigned muscle forces are physiological (range in tension from zero to the tetanic muscle force) and the resultant muscle forces from this loop are then arbitrarily scaled down (typically by 95%) to ensure that the model converges and the external moment is not overshot. The whole process is iteratively repeated using the remaining external moment as input to the next loop, until all three components of the external moment are balanced. At the end, the resultant force is determined, and if the joint reaction force points outside of the glenoid (unstable joint), the simulation is restarted, but rst giving all rotator cuff muscles a supplementary force. This method has been shown to deliver realistic muscle recruitment patterns in comparison with EMG measurements (Favre et al. 2005). In order to validate the obtained joint reaction force, the performance of the ASFE and other optimization criteria, in comparison with the instrumented shoulder implant (Bergmann et al. 2007), is described in 2b(ii) below for a classic example. (ii) Abduction in the scapular plane Abduction was simulated with the ASFE at 0, 45, 90 and 1208 with respect to the torso in the scapular plane. The arm weight (35 N) was applied at the middle of the arm (0.3 m from humeral head centre). Because the ASFE considers a xed scapula, the rotation of the scapula was taken into account by rotating the external force vector according to a scapulohumeral rhythm ratio of 1 : 2. Figure 4 shows the joint reaction force predicted by the ASFE and compares it with the reaction force reported from other studies (Poppen & Walker 1978; van der Helm 1994a; Terrier et al. 2008). Deviations in predicted joint resultant forces up to 908 can be explained by differences in model geometry (muscle segmentation, humeral head size, muscle size, muscle origin and insertion site) and recruitment strategy. The large deviation after 908 abduction is due to the ASFE incorporation of muscle co-contraction, which cannot be considered in the other models. Muscles of the rotator cuff and the deltoid, which represent the totality of muscles included by Poppen et al. and Terrier et al., also showed a decreased activity past 908 in the ASFE, except for the posterior deltoid segment. The rise in joint resultant force above 908 is due to other muscles. At 1208 abduction, the majority of abductors externally rotate the humerus (table 2). Only the cranial segments of the subscapularis and pectoralis major can counterbalance external rotation, but they are also strong forward exors. Here, other muscles such as teres major are activated to help balance these
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

2106
1000 joint reaction force (N) 800 600 400 200

P. Favre et al.

50

100 150 abduction (deg.)

200

Figure 4. Joint reaction force during abduction in the scapular plane obtained by different models (triangles, Poppen & Walker 1978; circles, van der Helm 1994a; diamonds, Terrier et al. 2008). For the ASFE (squares, Favre et al. 2005), the input data (muscle moment arms and lines of action) were measured in prior experiments so that only four positions are currently available in the scapular plane (0, 45, 90 and 1208 torso-humeral elevation). The grey arrow shows the tendency of the measurements obtained with the instrumented shoulder implant in two patients able to abduct their arm higher than 908 (correspondence with Prof. G. Bergmann).

components, even if such a muscle is a very strong adductor. Interestingly, in vivo measurements using instrumented shoulder prostheses (Bergmann et al. 2007) recently presented at the meeting of the European Society of Biomechanics have indicated that joint reaction forces increase as abduction exceeds 908 (Nikooyan et al. 2008). Although the ASFE is able to replicate this trend, it does not necessarily mean that the muscle forces found here correspond to an in vivo muscular contribution during abduction, as many physiologically plausible muscle force patterns can lead to a similar joint reaction force. Experimental measurements of muscular activity in postures above 908 of abduction are now required. However, it may indicate that co-contraction is a very important aspect of shoulder biomechanics, which has thus far been widely neglected. (c ) Deformable models The FE method can simulate the deformations of complex systems that are otherwise difcult to assess and has been used to address a broad range of problems in the eld of biomechanics and orthopaedics (Huiskes & Hollister 1993). FE modelling opens up a vast range of possibilities for simulating complex material phenomena such as nonlinear elastic and viscoelastic behaviours, plastic deformation, creep and fatigue failure behaviours. Most FE models of the shoulder have been developed to specically investigate glenoid xation and loosening, generally relating excessive bone or cement stresses to failure. The inuence of implant design parameters, such as peg or keeled glenoid anchorage (Lacroix et al. 2000), shape of implant components (Lacroix & Prendergast 1997; Buchler & Farron 2004; Terrier et al. 2006) or unconventional xation type (Murphy & Prendergast 2005), has been studied using FE models. Finally, the FE method has been implemented in comparisons
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

Review. Numerical modelling of the shoulder

2107

Table 2. Muscle segment moment arms at 1208 abduction. (Abduction (C) or adduction (K) moment arms are given in the rst column, forward (C) or backward (K) change in plane of elevation in the second, and internal (C) or external (K) axial rotation in the third column. The fourth column displays the forces attributed by the ASFE.) moment arm components plane of elevation axial rotation (mm) (mm) 14.21 2.58 22.17 21.28 14.32 K4.20 K11.73 K13.76 K15.76 K14.06 28.65 5.17 K9.79 K25.04 K37.34 K30.23 K14.30 K30.08 K7.16 2.88 2.88 0.00 31.09 50.13 54.43 17.19 31.51 K2.63 K5.24 1.43 5.52 8.75 K10.61 K11.63 K12.23 K10.46 K14.32 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 3.44 7.16 7.16 2.86 4.30 5.73 11.46 K10.02 0.00

elevation (mm) anterior supraspinatus posterior supraspinatus cranial subscapularis middle subscapularis caudal subscapularis cranial infraspinatus middle infraspinatus caudal infraspinatus cranial teres minor caudal teres minor anterior deltoid anterior middle deltoid middle middle deltoid posterior middle deltoid posterior distal deltoid posterior proximal deltoid triceps coracobrachialis teres major cranial latissimus dorsi middle latissimus dorsi caudal latissimus dorsi cranial pectoralis major middle pectoralis major caudal pectoralis major biceps caput longum biceps caput breve 17.19 21.53 2.86 K1.72 K11.23 20.05 13.01 4.77 K4.30 K8.59 22.33 27.16 25.78 13.64 7.16 2.81 K20.06 2.86 K25.78 K31.51 K40.11 K42.97 5.73 K1.43 K37.24 2.91 2.91

force (N) 18.56 18.56 4.42 11.85 11.85 18.28 18.28 18.28 0.00 0.00 187.15 63.11 85.60 85.60 126.92 99.35 0.00 23.93 62.90 7.63 7.63 12.90 58.61 12.11 12.11 22.13 30.63

between cemented and uncemented prosthesis xation (Gupta et al. 2004b,c), and testing the inuence of cement thickness (Couteau et al. 2001; Terrier et al. 2005) or implant positioning on bone and cement stresses (Hopkins et al. 2004; Farron et al. 2006). Shoulder FE studies addressing other clinical shoulder questions are fewer. Interesting two-dimensional studies have considered the stress pattern in the supraspinatus tendon, trying to understand the mechanical origin of tears (Luo et al. 1998; Wakabayashi et al. 2003; Sano et al. 2006). Another study tried to quantify the inuence of the shape of the humeral head on the stress distribution in the scapula with the intent to compare a normal with an osteoarthritic shoulder (Buchler et al. 2002). Finally, relative kinematics of implant components have been simulated using the FE method to study implant intrinsic stability (Hopkins et al. 2006, 2007).
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

2108

P. Favre et al.

The main limitations of deformable models usually lie in the denitions of boundary conditions and material properties that are employed. Measuring and describing the material properties of biological tissues is not at all simple, and most models necessarily implement idealized material properties. With regard to boundary conditions, in vivo forces have been simulated by applying the joint resultant force only (Stone et al. 1999; Couteau et al. 2001; Andreykiv et al. 2005). In more elaborate FE models (Lacroix et al. 2000; Gupta et al. 2004b,c; Hopkins et al. 2005, 2007; Murphy & Prendergast 2005), individual muscle contributions were assigned according to van der Helm (1994b). Axial rotation of the humerus has been simulated by applying a gradually imposed displacement at a chosen muscle insertion and by restricting unwanted motions through nonphysiological boundary constraints at the distal extremity of the humerus (Buchler et al. 2002). In a later extension of the same model, this articial boundary condition was removed, but the model was restricted to a two-dimensional representation and to the deltoid and rotator cuff muscles (Terrier et al. 2008). Finally, investigating glenoid loosening requires a profound understanding of the failure mechanisms. FE models can predict mechanical factors such as boneimplant micromotion, wear or excessive stress, but the induced biological response is extremely complex and remains not yet fully understood. Despite these limitations, results obtained from deformable models can still be directly transferred to clinical practice by dening the relative inuence of surgically adaptable parameters (choice of implant, components positioning, use of cement, etc.) on the investigated consequence (stress shielding, aseptic loosening, etc.). 3. Model validation To efciently address scientic and clinical questions through a modelling approach, the goal is not to build a perfectly realistic model, but to create a sufciently accurate representation of the aspect to be investigated. The complexity of a biological system, such as the shoulder, is such that some simplications must be made in the modelling process, although each assumption introduces a potential source of error. To ensure that the imposed simplications do not diminish the veracity of the simulation, validation is essential. The validation process confronts simulation results with reality, or, when this is not directly possible, with controlled experiments that approximate reality. Simulation results are compared against measurable parameters with the aim to identify inappropriate assumptions or simplications that can then be adjusted to improve the delity of the model. (a ) Validating rigid body models Simulated rigid body motion can be compared with experimentally measured kinematic data. For instance, the simulated reachable workspace of the arm can be validated against actual measurements of range of motions (Klopcar et al. 2007). Rigid body models used for the quantication of moment arms can also be directly compared with in vivo or in vitro measurements (Garner & Pandy 2001; Gatti et al. 2007). Finally, when combined with muscle force estimation, in vivo measurements of maximum isometric muscle torques exerted at the joint and/or
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

Review. Numerical modelling of the shoulder

2109

the resultant kinematics can be used for comparison, although these only validate the global simulation quality and do not yield specic information on individual muscle forces (Garner & Pandy 2001). (b ) Validating muscle force estimation models In an indeterminate system of equations, it falls on the modeller to restrict the number of considered solutions to those that can realistically occur in vivo. As described above, muscle force estimation models deliver two main pieces of information: the set of recruited muscles and the force that these exert. EMG signals can be used to compare the timings of the muscle activities (Happee & van der Helm 1995) and to verify that the model activates a reasonable set of muscles (Karlsson & Peterson 1992; Happee 1994; van der Helm 1994b; Happee & van der Helm 1995; Nieminen et al. 1995; Niemi et al. 1996; Favre et al. 2005), or they could also allow comparison of paralysis experiments (either pathological or neurotoxin induced) against model behaviour when specic muscles are removed from consideration. On the other hand, we are currently limited in our ability to quantify individual muscle force magnitude since no in vivo muscular force measurement devices are currently available (Erdemir et al. 2007), and EMG amplitude is a poor measure for validating the forces obtained by musculoskeletal models (Inman et al. 1952; van der Helm 1994b). Given the lack of reliable muscle force measurement techniques, results have generally been compared with those obtained with previous models. However, this approach cannot be considered a valid muscle force validation, but merely serves as an indication that two models represent reality in a similar way. Since this approach cannot root out problems with widely held assumptions, there is obviously a pressing need for better techniques to quantify in vivo muscle forces. (c ) Validating deformable models In the most straightforward validations of FE models, modelling results (surface strain patterns) have been directly compared with experiments performed under similar loading conditions, using an in vitro strain gauge measurement (Couteau et al. 2001; Gupta et al. 2004a; Maurel et al. 2005) or photoelastic techniques (Murphy & Prendergast 2005). As an example of clinical validation, bone formation around a retrieved specimen conrmed that the bone stress predicted by the model (disregarding other important factors) was within acceptable values (Ahir et al. 2004). For a proper validation of an FE model, strict guidelines and best practices have been suggested (Viceconti et al. 2005). Unfortunately, the validation step is very often neglected in deformable models. (d ) Emerging technologies for model validation In the past, validation often compared model results with experiments performed on cadavers. Emerging technologies now allow direct comparison with selected in vivo measurements, bringing the models to a higher level of accuracy. Their use for validation of shoulder models is nascent (Dubowsky et al. 2008; Terrier et al. 2008), but is destined to increase. Rapid progress in the eld of medical imaging opens new possibilities to measure rotation and translation of the glenohumeral joint during in vivo movements using standard uoroscopic
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

2110

P. Favre et al.

sequences (Prrmann et al. 2002), open magnetic resonance imaging (MRI; Graichen et al. 2000) and biplanar X-rays (Bey et al. 2008). These quantities provide useful global criteria for a partial assessment of model performance of rigid body or contact models. Validation of the forces predicted by numerical musculoskeletal models of the shoulder has now been improved with direct measurements of glenohumeral contact forces using an instrumented implant (Bergmann et al. 2007) as well as enhanced techniques to measure in vivo muscle activity (Praagman et al. 2003, 2006; de Groot et al. 2004). 4. Outlook in numerical shoulder modelling Implementation of more realistic modelling approaches to investigate stabilization of the glenohumeral joint is imperative (Veeger & van der Helm 2007). Since translations may no longer be small in comparison with rotations in the unstable shoulder, modellers may have to deviate from the simplication of the shoulder joint as a perfect ball-and-socket (Hopkins et al. 2006). If a priori (articial) restrictions on translations are to be avoided, muscle force estimation models can play a valuable role in introducing muscle-driven stabilization of the joint. Owing to the great individual variability in anatomy, patient-specic models may be needed to study specic skeletal deformities, and altered bone and soft tissue material properties, or for individual diagnosis and surgical planning. With semi-automatic extraction of anatomy from computed tomography and MRI images, models of individual patients are now plausible (Young et al. 2008). For studies on loosening of the glenoid component in shoulder arthroplasty, a more realistic representation of the underlying bone microarchitecture may improve the representation of the interface mechanics and force transmission in the bone, enhancing failure risk prediction (Ulrich et al. 1997; Pistoia et al. 2002). Also, long-term predictions of implant xation could be improved by including bone mechanical adaptation laws into FE models (Beaupre et al. 1990; Cowin et al. 1992; Prendergast & Taylor 1994; Huiskes et al. 2000). Such FE models for the shoulder do exist but have been limited to two-dimensional analyses of the glenoid (Andreykiv et al. 2005; Sharma et al. 2007). Simulation of brous tissue interposition between the implant and bone could also be added (Weinans et al. 1990). In this paper, shoulder models were treated as belonging to three main groups. Although these categories sometimes overlap already, as seen with the coupling of rigid body and muscle force estimation models, an integrated modelling approach can combine the advantages of the different models, offering new possibilities. For example, when studying the inuence of implant component positioning (table 1), the integrated modelling approach would provide a tool to address simultaneously the interconnected implications in range of motions and anatomy (which could be investigated with a rigid body model), the changes in muscle activity and contact forces (which is studied with muscle force estimation models) and the deformations of the implant interface and implant xation (information typically delivered by an FE model). In an effort to reduce computational expense, FE models rarely represent the shoulder as a whole. When possible, parts of the system are simplied (or neglected) and substituted with approximate boundary conditions. However, this can lead to erroneous
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

Review. Numerical modelling of the shoulder

DEL3 DEL4

2111

(a) muscle moment arms and direction of action

(b)

120 100
DEL1

force (N)

80 60 40 20 0
SSP1 ISP1

SSP2

ISP2 ISP3 TMIN1 TMIN2

SSC1 SSC2 SSC3

DEL2

DEL5 DEL6 TRI COR TMAJ LAT1 LAT2 LAT3 PEC1 PEC2 PEC3

BI CL BI CB

motion/ deformations

muscle force (c)

Figure 5. Integrated modelling approach combining the ASFE with an anatomically precise FE model of the glenohumeral joint: (a) FE (rigid body), (b) ASFE and (c) FE (deformable). Here, only a few muscle segments are represented on the three-dimensional model of the shoulder. The deformable muscles can wrap on the bones, relying on detection of contact in the FE software. The moment arms and lines of action of the muscles are computed in this position and sent to the ASFE. The muscle forces are returned as boundary conditions to the FE model. The innitesimal motion of the arm is simulated within the FE model and the positional error between the obtained and the target positions is determined. This loop is reiterated until the positional error is less than a predened value. Results obtained with this model will be validated against EMG patterns, glenohumeral contact forces for given activities, and clinical cases of muscle insufciency.

conclusions (Gupta & van der Helm 2004; Hopkins et al. 2005). For FE, in general, the benets of relying on fully balanced external load regimes have been demonstrated (Duda et al. 1998; Speirs et al. 2007). Explicit incorporation of tissue deformations and failure further allow investigations into how these factors can affect segment motions and muscle force activity. Thus, through an integrated modelling approach, more realistic conditions can be implemented, yielding more reliable results. To this end, we see the FE method as the most promising framework. Modern commercial FE software packages allow simulation of large motions, and can implement the simplifying aspects of rigid body assumptions in a hybrid fashion, permit simultaneous inclusion of muscle force estimation and recruitment strategies, and allow determination of component deformations in critical locations. To this end, we are currently developing a model that combines the ASFE with an anatomically precise FE model of the glenohumeral joint (gure 5).
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

2112

P. Favre et al.

In conclusion, numerical models of the shoulder have proven to be useful in a wide range of clinical applications. In the future, given our growing understanding of shoulder mechanics, advances in modelling methods and increases in computer power, numerical models should enable researchers to address ever more complex clinical questions and ultimately improve patient care.

References
Ahir, S. P., Walker, P. S., Squire-Taylor, C. J., Blunn, G. W. & Bayley, J. I. 2004 Analysis of glenoid xation for a reversed anatomy xed-fulcrum shoulder replacement. J. Biomech. 37, 16991708. (doi:10.1016/j.jbiomech.2004.01.031) Andreykiv, A., Prendergast, P. J., van Keulen, F., Swieszkowski, W. & Rozing, P. M. 2005 Bone ingrowth simulation for a concept glenoid component design. J. Biomech. 38, 10231033. (doi:10.1016/j.jbiomech.2004.05.044) Barker, T. M., Kirtley, C. & Ratanapinunchai, J. 1997 Calculation of multi-segment rigid body joint dynamics using Matlab. Proc. Inst. Mech. Eng. [H ] 211, 483487. (doi:10.1243/ 0954411981534619) Bassett, R. W., Browne, A. O., Morrey, B. F. & An, K. N. 1990 Glenohumeral muscle force and moment mechanics in a position of shoulder instability. J. Biomech. 23, 405415. (doi:10.1016/ 0021-9290(90)90295-E) Beaupre, G. S., Orr, T. E. & Carter, D. R. 1990 An approach for time-dependent bone modeling and remodelingtheoretical development. J. Orthop. Res. 8, 651661. (doi:10.1002/jor. 1100080506) Bergmann, G., Graichen, F., Bender, A., Kaab, M., Rohlmann, A. & Westerhoff, P. 2007 In vivo glenohumeral contact forcesmeasurements in the rst patient 7 months postoperatively. J. Biomech. 40, 21392149. (doi:10.1016/j.jbiomech.2006.10.037) Bey, M. J., Kline, S. K., Zauel, R., Lock, T. R. & Kolowich, P. A. 2008 Measuring dynamic in-vivo glenohumeral joint kinematics: technique and preliminary results. J. Biomech. 41, 711714. (doi:10.1016/j.jbiomech.2007.09.029) Blemker, S. S., Asakawa, D. S., Gold, G. E. & Delp, S. L. 2007 Image-based musculoskeletal modeling: applications, advances, and future opportunities. J. Magn. Reson. Imaging 25, 441451. (doi:10.1002/jmri.20805) Boileau, P., Watkinson, D. J., Hatzidakis, A. M. & Balg, F. 2005 Grammont reverse prosthesis: design, rationale, and biomechanics. J. Shoulder Elbow Surg. 14, 147S161S. (doi:10.1016/ j.jse.2004.10.006) Buchler, P. & Farron, A. 2004 Benets of an anatomical reconstruction of the humeral head during shoulder arthroplasty: a nite element analysis. Clin. Biomech. (Bristol, Avon) 19, 1623. (doi:10.1016/j.clinbiomech.2003.09.009) Buchler, P., Ramaniraka, N. A., Rakotomanana, L. R., Iannotti, J. P. & Farron, A. 2002 A nite element model of the shoulder: application to the comparison of normal and osteoarthritic joints. Clin. Biomech. (Bristol, Avon) 17, 630639. (doi:10.1016/S0268-0033(02)00106-7) Chadwick, E. K. J., van Noort, A. & van der Helm, F. C. T. 2004 Biomechanical analysis of scapular neck maluniona simulation study. Clin. Biomech. (Bristol, Avon) 19, 906912. (doi:10.1016/j.clinbiomech.2004.06.013) Charlton, I. W. & Johnson, G. R. 2001 Application of spherical and cylindrical wrapping algorithms in a musculoskeletal model of the upper limb. J. Biomech. 34, 12091216. (doi:10. 1016/S0021-9290(01)00074-4) Charlton, I. W. & Johnson, G. R. 2006 A model for the prediction of the forces at the glenohumeral joint. Proc. Inst. Mech. Eng. [H ] 220, 801812. (doi:10.1243/09544119JEIM147) Cholewicki, J., McGill, S. M. & Norman, R. W. 1995 Comparison of muscle forces and joint load from an optimization and EMG assisted lumbar spine model: towards development of a hybrid approach. J. Biomech. 28, 321331. (doi:10.1016/0021-9290(94)00065-C)
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

Review. Numerical modelling of the shoulder

2113

Couteau, B., Mansat, P., Estivalezes, E., Darmana, R., Mansat, M. & Egan, J. 2001 Finite element analysis of the mechanical behavior of a scapula implanted with a glenoid prosthesis. Clin. Biomech. (Bristol, Avon) 16, 566575. (doi:10.1016/S0268-0033(01)00029-8) Cowin, S. C., Sadegh, A. M. & Luo, G. M. 1992 An evolutionary Wolffs law for trabecular architecture. J. Biomech. Eng. 114, 129136. (doi:10.1115/1.2895436) de Groot, J. H., Rozendaal, L. A., Meskers, C. G. M. & Arwert, H. J. 2004 Isometric shoulder muscle activation patterns for 3-D planar forces: a methodology for musculo-skeletal model validation. Clin. Biomech. (Bristol, Avon) 19, 790800. (doi:10.1016/j.clinbiomech.2004.05.013) de Leest, O., Rozing, P. M., Rozendaal, L. A. & van der Helm, F. C. T. 1996 Inuence of glenohumeral prosthesis geometry and placement on shoulder muscle forces. Clin. Orthop. Relat. Res. 330, 222233. (doi:10.1097/00003086-199609000-00030) De Luca, C. J. 1997 The use of surface electromyography in biomechanics. J. Appl. Biomech. 13, 135163. De Luca, C. J. & Forrest, W. J. 1973 Force analysis of individual muscles acting simultaneously on the shoulder joint during isometric abduction. J. Biomech. 6, 385393. (doi:10.1016/00219290(73)90098-5) Dickerson, C. R., Chafn, D. B. & Hughes, R. E. 2007 A mathematical musculoskeletal shoulder model for proactive ergonomic analysis. Comput. Methods Biomech. Biomed. Eng. 10, 389400. (doi:10.1080/10255840701592727) Dickerson, C. R., Hughes, R. E. & Chafn, D. B. 2008 Experimental evaluation of a computational shoulder musculoskeletal model. Clin. Biomech. (Bristol, Avon) 23, 886894. (doi:10.1016/ j.clinbiomech.2008.04.004) Dubowsky, S. R., Rasmussen, J., Sisto, S. A. & Langrana, N. A. 2008 Validation of a musculoskeletal model of wheelchair propulsion and its application to minimizing shoulder joint forces. J. Biomech. 41, 29812988. (doi:10.1016/j.jbiomech.2008.07.032) Duda, G. N., Heller, M., Albinger, J., Schulz, O., Schneider, E. & Claes, L. 1998 Inuence of muscle forces on femoral strain distribution. J. Biomech. 31, 841846. (doi:10.1016/S00219290(98)00080-3) Engin, A. E. & Tumer, S. T. 1989 Three-dimensional kinematic modelling of the human shoulder complexpart I: physical model and determination of joint sinus cones. J. Biomech. Eng. 111, 107112. Erdemir, A., McLean, S., Herzog, W. & van den Bogert, A. J. 2007 Model-based estimation of muscle forces exerted during movements. Clin. Biomech. (Bristol, Avon) 22, 131154. (doi:10. 1016/j.clinbiomech.2006.09.005) Farron, A., Terrier, A. & Buchler, P. 2006 Risks of loosening of a prosthetic glenoid implanted in retroversion. J. Shoulder Elbow Surg. 15, 521526. (doi:10.1016/j.jse.2005.10.003) Favre, P., Sheikh, R., Fucentese, S. F. & Jacob, H. A. 2005 An algorithm for estimation of shoulder muscle forces for clinical use. Clin. Biomech. (Bristol, Avon) 20, 822833. (doi:10.1016/ j.clinbiomech.2005.04.007) Favre, P., Loeb, M. D., Helmy, N. & Gerber, C. 2008a Latissimus dorsi transfer to restore external rotation with reverse shoulder arthroplasty: a biomechanical study. J. Shoulder Elbow Surg. 17, 650658. (doi:10.1016/j.jse.2007.12.010) Favre, P., Moor, B., Snedeker, J. G. & Gerber, C. 2008b Inuence of component positioning on impingement in conventional total shoulder arthroplasty. Clin. Biomech. (Bristol, Avon) 23, 175183. (doi:10.1016/j.clinbiomech.2007.09.009) Favre, P., Jacob, H. A. C. & Gerber, C. 2009 Changes in shoulder muscle function with humeral position: a graphical description. J. Shoulder Elbow Surg. 18, 114121. (doi:10.1016/j.jse.2008.06.010) Franklin, J. L., Barrett, W. P., Jackins, S. E. & Matsen III, F. A. 1988 Glenoid loosening in total shoulder arthroplasty. Association with rotator cuff deciency. J. Arthroplasty 3, 3946. (doi:10. 1016/S0883-5403(88)80051-2) Garner, B. A. & Pandy, M. G. 1999 A kinematic model of the upper limb based on the Visible Human Project (VHP) image dataset. Comput. Methods Biomech. Biomed. Eng. 2, 107124. (doi:10.1080/10255849908907981)
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

2114

P. Favre et al.

Garner, B. A. & Pandy, M. G. 2000 The obstacle-set method for representing muscle paths in musculoskeletal models. Comput. Methods Biomech. Biomed. Eng. 3, 130. (doi:10.1080/ 10255840008915251) Garner, B. A. & Pandy, M. G. 2001 Musculoskeletal model of the upper limb based on the Visible Human male dataset. Comput. Methods Biomech. Biomed. Eng. 4, 93126. (doi:10.1080/ 10255840008908000) Gatti, C. J., Dickerson, C. R., Chadwick, E. K., Mell, A. G. & Hughes, R. E. 2007 Comparison of model-predicted and measured moment arms for the rotator cuff muscles. Clin. Biomech. (Bristol, Avon) 22, 639644. (doi:10.1016/j.clinbiomech.2007.02.001) Gerber, C. 1992 Latissimus dorsi transfer for the treatment of irreparable tears of the rotator cuff. Clin. Orthop. Relat. Res. 275, 152160. (doi:10.1097/00003086-199202000-00022) Gerber, C., Meyer, D. C., Schneeberger, A. G., Hoppeler, H. & von Rechenberg, B. 2004 Effect of tendon release and delayed repair on the structure of the muscles of the rotator cuff: an experimental study in sheep. J. Bone Joint Surg. Am. 86, 19731982. (doi:10.1302/0301-620X. 86B6.14577) Gerber, C., Pennington, S. D., Lingenfelter, E. J. & Sukthankar, A. 2007 Reverse delta-III total shoulder replacement combined with latissimus dorsi transfer. A preliminary report. J. Bone Joint Surg. Am. 89, 940947. (doi:10.2106/JBJS.F.00955) Graichen, H., Stammberger, T., Bonel, H., Englmeier, K.-H., Reiser, M. & Eckstein, F. 2000 Glenohumeral translation during active and passive elevation of the shouldera 3D open-MRI study. J. Biomech. 33, 609613. (doi:10.1016/S0021-9290(99)00209-2) Gupta, S. & van der Helm, F. C. T. 2004 Load transfer across the scapula during humeral abduction. J. Biomech. 37, 10011009. (doi:10.1016/j.jbiomech.2003.11.025) Gupta, S., van der Helm, F. C. T., Sterk, J. C., van Keulen, F. & Kaptein, B. L. 2004a Development and experimental validation of a three-dimensional nite element model of the human scapula. Proc. Inst. Mech. Eng. [H ] 218, 127142. (doi:10.1243/095441104322984022) Gupta, S., van der Helm, F. C. T. & van Keulen, F. 2004b The possibilities of uncemented glenoid componenta nite element study. Clin. Biomech. (Bristol, Avon) 19, 292302. (doi:10.1016/ j.clinbiomech.2003.12.002) Gupta, S., van der Helm, F. C. T. & van Keulen, F. 2004c Stress analysis of cemented glenoid prostheses in total shoulder arthroplasty. J. Biomech. 37, 17771786. (doi:10.1016/j.jbiomech. 2004.02.001) Happee, R. 1994 Inverse dynamic optimization including muscular dynamics, a new simulation method applied to goal directed movements. J. Biomech. 27, 953960. (doi:10.1016/00219290(94)90267-4) Happee, R. & van der Helm, F. C. T. 1995 The control of shoulder muscles during goal directed movements, an inverse dynamic analysis. J. Biomech. 28, 11791191. (doi:10.1016/00219290(94)00181-3) Hogfors, C., Karlsson, D. & Peterson, B. 1995 Structure and internal consistency of a shoulder model. J. Biomech. 28, 767777. (doi:10.1016/0021-9290(94)00134-P) Holzbaur, K. R., Murray, W. M. & Delp, S. L. 2005 A model of the upper extremity for simulating musculoskeletal surgery and analyzing neuromuscular control. Ann. Biomed. Eng. 33, 829840. (doi:10.1007/s10439-005-3320-7) Hopkins, A. R., Hansen, U. N., Amis, A. A. & Emery, R. 2004 The effects of glenoid component alignment variations on cement mantle stresses in total shoulder arthroplasty. J. Shoulder Elbow Surg. 13, 668675. (doi:10.1016/j.jse.2004.04.008) Hopkins, A. R., Hansen, U. N. & Amis, A. A. 2005 Finite element models of total shoulder replacement: application of boundary conditions. Comput. Methods Biomech. Biomed. Eng. 8, 3944. (doi:10.1080/10255840500075205) Hopkins, A. R., Hansen, U. N., Amis, A. A., Taylor, M., Gronau, N. & Anglin, C. 2006 Finite element modelling of glenohumeral kinematics following total shoulder arthroplasty. J. Biomech. 39, 24762483. (doi:10.1016/j.jbiomech.2005.07.031)
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

Review. Numerical modelling of the shoulder

2115

Hopkins, A. R., Hansen, U. N., Amis, A. A., Taylor, M. & Emery, R. J. 2007 Glenohumeral kinematics following total shoulder arthroplasty: a nite element investigation. J. Orthop. Res. 25, 108115. (doi:10.1002/jor.20290) Hovelius, L. 1982 Incidence of shoulder dislocation in Sweden. Clin. Orthop. Relat. Res. 166, 127131. (doi:10.1097/00003086-198206000-00021) Huiskes, R. & Hollister, S. J. 1993 From structure to process, from organ to cell: recent developments of FE-analysis in orthopaedic biomechanics. J. Biomech. Eng. 115, 520527. (doi:10.1115/1.2895534) Huiskes, R., Ruimerman, R., van Lenthe, G. H. & Janssen, J. D. 2000 Effects of mechanical forces on maintenance and adaptation of form in trabecular bone. Nature 405, 704706. (doi:10.1038/ 35015116) Inman, V. T., Saunders, J. B. & Abbott, L. C. 1944 Observations of the function of the shoulder joint. J. Bone Joint Surg. Am. 26A, 131. Inman, V. T., Ralston, H. J., Saunders, J. B., Feinstein, B. & Wright Jr, E. W. 1952 Relation of human electromyogram to muscular tension. Electroencephalogr. Clin. Neurophysiol. 4(Suppl.), 187194. (doi:10.1016/0013-4694(52)90008-4) Jensen, B. R., Laursen, B. & Sjogaard, G. 2000 Aspects of shoulder function in relation to exposure demands and fatiguea mini review. Clin Biomech (Bristol, Avon) 15(Suppl. 1), S17S20. (doi:10.1016/S0268-0033(00)00054-1) Jinhaa, A., Ait-Haddoua, R., Binding, P. & Herzog, W. 2006 Antagonistic activity of one-joint muscles in three-dimensions using non-linear optimisation. Math. Biosci. 202, 5770. (doi:10. 1016/j.mbs.2006.03.018) Johnson, G. R. & Gill, J. M. 1987 Prediction of the range of hand positions available to a patient with movement restrictions at the joints of the upper limba mathematical model. Int. J. Rehabil. Res. 10, 291297. (doi:10.1097/00004356-198700105-00061) Karlsson, D. & Peterson, B. 1992 Towards a model for force predictions in the human shoulder. J. Biomech. 25, 189199. (doi:10.1016/0021-9290(92)90275-6) Klopcar, N., Tomsic, M. & Lenarcic, J. 2007 A kinematic model of the shoulder complex to evaluate the arm-reachable workspace. J. Biomech. 40, 8691. (doi:10.1016/j.jbiomech.2005.11.010) Koike, Y. & Kawato, M. 1995 Estimation of dynamic joint torques and trajectory formation from surface electromyography signals using a neural network model. Biol. Cybern. 73, 291300. (doi:10.1007/BF00199465) Kozak, L. J., Defrances, C. J. & Hall, M. J. 2006 National hospital discharge survey: 2004 annual summary with detailed diagnosis and procedure data. National Center for Health Statistics, Vital and Health Statistics, Ser. 13, No. 162. Hyattsville, MD: US Department of Health and Human Services. Kronberg, M., Brostrom, L. A. & Nemeth, G. 1991 Differences in shoulder muscle activity between patients with generalized joint laxity and normal controls. Clin. Orthop. Relat. Res. 269, 181192. (doi:10.1097/00003086-199108000-00026) Lacroix, D. & Prendergast, P. J. 1997 Stress analysis of glenoid component designs for shoulder arthroplasty. Proc. Inst. Mech. Eng. [H ] 211, 467474. (doi:10.1243/0954411981534583) Lacroix, D., Murphy, L. A. & Prendergast, P. J. 2000 Three-dimensional nite element analysis of glenoid replacement prostheses: a comparison of keeled and pegged anchorage systems. J. Biomech. Eng. 122, 430436. (doi:10.1115/1.1286318) Langenderfer, J., Lascalza, S., Mell, A., Carpenter, J. E., Kuhn, J. E. & Hughes, R. E. 2005 An EMG-driven model of the upper extremity and estimation of long head biceps force. Comput. Biol. Med. 35, 2539. (doi:10.1016/j.compbiomed.2003.12.002) Laursen, B., Jensen, B. R., Nemeth, G. & Sjogaard, G. 1998 A model predicting individual shoulder muscle forces based on relationship between electromyographic and 3D external forces in static position. J. Biomech. 31, 731739. (doi:10.1016/S0021-9290(98)00091-8) Lin, H. T., Su, F. C., Wu, H. W. & An, K. N. 2004 Muscle forces analysis in the shoulder mechanism during wheelchair propulsion. Proc. Inst. Mech. Eng. [H ] 218, 213221. (doi:10. 1243/0954411041561027)
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

2116

P. Favre et al.

Lippitt, S. & Matsen, F. 1993 Mechanisms of glenohumeral joint stability. Clin. Orthop. Relat. Res. 291, 2028. (doi:10.1097/00003086-199306000-00004) Lock, C., Allgar, V., Jones, K., Marples, G., Chandler, C. & Dawson, P. 1999 Prevalence of back, neck and shoulder problems in the inner city: implications for the provision of physiotherapy services in primary healthcare. Physiother. Res. Int. 4, 161169. (doi:10.1002/pri.163) Luo, Z. P., Hsu, H. C., Grabowski, J. J., Morrey, B. F. & An, K. N. 1998 Mechanical environment associated with rotator cuff tears. J. Shoulder Elbow Surg. 7, 616620. (doi:10.1016/S10582746(98)90010-6) Magermans, D. J., Chadwick, E. K. J., Veeger, H. E. J., Rozing, P. M. & van der Helm, F. C. T. 2004a Effectiveness of tendon transfers for massive rotator cuff tears: a simulation study. Clin. Biomech. (Bristol, Avon) 19, 116122. (doi:10.1016/j.clinbiomech.2003.09.008) Magermans, D. J., Chadwick, E. K. J., Veeger, H. E. J., van der Helm, F. C. T. & Rozing, P. M. 2004b Biomechanical analysis of tendon transfers for massive rotator cuff tears. Clin. Biomech. (Bristol, Avon) 19, 350357. (doi:10.1016/j.clinbiomech.2003.11.013) Makela, M., Heliovaara, M., Sainio, P., Knekt, P., Impivaara, O. & Aromaa, A. 1999 Shoulder joint impairment among Finns aged 30 years or over: prevalence, risk factors and co-morbidity. Rheumatology (Oxf.) 38, 656662. (doi:10.1093/rheumatology/38.7.656) Matsen III, F. A., Boileau, P., Walch, G., Gerber, C. & Bicknell, R. T. 2007 The reverse total shoulder arthroplasty. J. Bone Joint Surg. Am. 89, 660667. Maurel, W. & Thalmann, D. 1999 A case study analysis on human upper limb modelling for dynamic simulation. Comput. Methods Biomech. Biomech. Eng. 2, 6582. (doi:10.1080/ 10255849908907979) Maurel, N., Diop, A. & Grimberg, J. 2005 A 3D nite element model of an implanted scapula: importance of a multiparametric validation using experimental data. J. Biomech. 38, 18651872. (doi:10.1016/j.jbiomech.2004.08.019) McGill, S. M. 1992 A myoelectrically based dynamic three-dimensional model to predict loads on lumbar spine tissues during lateral bending. J. Biomech. 25, 395414. (doi:10.1016/00219290(92)90259-4) McMahon, P. J. & Lee, T. Q. 2002 Muscles may contribute to shoulder dislocation and stability. Clin. Orthop. Relat. Res. 403, S18S25. (doi:10.1097/00003086-200210001-00003) Murphy, L. A. & Prendergast, P. J. 2005 Acromion-xation of glenoid components in total shoulder arthroplasty. J. Biomech. 38, 17021711. (doi:10.1016/j.jbiomech.2004.06.030) Neer II, C. S., Craig, E. V. & Fukuda, H. 1983 Cuff-tear arthropathy. J. Bone Joint Surg. Am. 65, 12321244. Niemi, J., Nieminen, H., Takala, E. P. & Viikari-Juntura, E. 1996 A static shoulder model based on a time-dependent criterion for load sharing between synergistic muscles. J. Biomech. 29, 451460. (doi:10.1016/0021-9290(95)00073-9) Nieminen, H., Niemi, J., Takala, E. P. & Viikari-Juntura, E. 1995 Load-sharing patterns in the shoulder during isometric exion tasks. J. Biomech. 28, 555566. (doi:10.1016/00219290(94)00107-F) Nikooyan, A. A., Veeger, H. E. J., van der Helm, F. C. T., Westerhoff, P., Graichen, F. & Bergmann, G. 2008 Comparing model predicted GH-joint contact forces by in-vivo measured forces. In Proc. 16th Congr. European Society of Biomechanics, Lucerne, Switzerland, 69 June. Nyffeler, R. W., Sheikh, R., Atkinson, T. S., Jacob, H. A., Favre, P. & Gerber, C. 2006 Effects of glenoid component version on humeral head displacement and joint reaction forces: an experimental study. J. Shoulder Elbow Surg. 15, 625629. (doi:10.1016/j.jse.2005.09.016) Penrod, D. D., Davy, D. T. & Singh, D. P. 1974 An optimization approach to tendon force analysis. J. Biomech. 7, 123129. (doi:10.1016/0021-9290(74)90050-5) Prrmann, C. W., Huser, M., Szekely, G., Hodler, J. & Gerber, C. 2002 Evaluation of complex joint motion with computer-based analysis of uoroscopic sequences. Invest. Radiol. 37, 7376. (doi:10.1097/00004424-200202000-00004)
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

Review. Numerical modelling of the shoulder

2117

Pistoia, W., van Rietbergen, B., Lochmuller, E. M., Lill, C. A., Eckstein, F. & Ruegsegger, P. 2002 Estimation of distal radius failure load with micro-nite element analysis models based on three-dimensional peripheral quantitative computed tomography images. Bone 30, 842848. (doi:10.1016/S8756-3282(02)00736-6) Poppen, N. K. & Walker, P. S. 1978 Forces at the glenohumeral joint in abduction. Clin. Orthop. Relat. Res. 135, 165170. (doi:10.1097/00003086-197809000-00035) Praagman, M., Veeger, H. E. J., Chadwick, E. K. J., Colier, W. N. J. M. & van der Helm, F. C. T. 2003 Muscle oxygen consumption, determined by NIRS, in relation to external force and EMG. J. Biomech. 36, 905912. (doi:10.1016/S0021-9290(03)00081-2) Praagman, M., Chadwick, E. K. J., van der Helm, F. C. T. & Veeger, H. E. J. 2006 The relationship between two different mechanical cost functions and muscle oxygen consumption. J. Biomech. 39, 758765. (doi:10.1016/j.jbiomech.2004.11.034) Prendergast, P. J. & Taylor, D. 1994 Prediction of bone adaptation using damage accumulation. J. Biomech. 27, 10671076. (doi:10.1016/0021-9290(94)90223-2) Rau, G., Disselhorst-Klug, C. & Schmidt, R. 2000 Movement biomechanics goes upwards: from the leg to the arm. J. Biomech. 33, 12071216. (doi:10.1016/S0021-9290(00)00062-2) Sano, H., Wakabayashi, I. & Itoi, E. 2006 Stress distribution in the supraspinatus tendon with partial-thickness tears: an analysis using two-dimensional nite element model. J. Shoulder Elbow Surg. 15, 100105. (doi:10.1016/j.jse.2005.04.003) Seireg, A. & Arvikar, R. J. 1975 The prediction of muscular load sharing and joint forces in the lower extremities during walking. J. Biomech. 8, 89102. (doi:10.1016/0021-9290(75)90089-5) Sharma, G. B., Robertson, D. D., Smolinski, P. J., Gilbertson, L. G., Debski, R. E. & McMahon, P. J. 2007 Glenoid bone remodeling: numerical simulation and validation. In Meeting of the Orthopaedic Research Society, San Diego, CA, USA. Sher, J. S., Uribe, J. W., Posada, A., Murphy, B. J. & Zlatkin, M. B. 1995 Abnormal ndings on magnetic resonance images of asymptomatic shoulders. J. Bone Joint Surg. Am. 77, 1015. Speirs, A. D., Heller, M. O., Duda, G. N. & Taylor, W. R. 2007 Physiologically based boundary conditions in nite element modelling. J. Biomech. 40, 23182323. (doi:10.1016/j.jbiomech. 2006.10.038) Steenbrink, F., de Groot, J. H., Veeger, H. E. J., Meskers, C. G. M., van de Sande, M. A. J. & Rozing, P. M. 2006 Pathological muscle activation patterns in patients with massive rotator cuff tears, with and without subacromial anaesthetics. Man. Ther. 11, 231237. (doi:10.1016/j.math. 2006.07.004) Stone, K. D., Grabowski, J. J., Coeld, R. H., Morrey, B. F. & An, K. N. 1999 Stress analyses of glenoid components in total shoulder arthroplasty. J. Shoulder Elbow Surg. 8, 151158. (doi:10. 1016/S1058-2746(99)90009-5) Terrier, A., Buchler, P. & Farron, A. 2005 Bonecement interface of the glenoid component: stress analysis for varying cement thickness. Clin. Biomech. (Bristol, Avon) 20, 710717. (doi:10. 1016/j.clinbiomech.2005.03.010) Terrier, A., Buchler, P. & Farron, A. 2006 Inuence of glenohumeral conformity on glenoid stresses after total shoulder arthroplasty. J. Shoulder Elbow Surg. 15, 515520. (doi:10.1016/j.jse.2005. 09.021) Terrier, A., Vogel, A., Capezzali, M. & Farron, A. 2008 An algorithm to allow humerus translation in the indeterminate problem of shoulder abduction. Med. Eng. Phys. 30, 710716. (doi:10.1016/ j.medengphy.2007.07.011) Ulrich, D., Hildebrand, T., van Rietbergen, B., Muller, R. & Ruegsegger, P. 1997 The quality of trabecular bone evaluated with micro-computed tomography, FEA and mechanical testing. Stud. Health Technol. Inform. 40, 97112. van der Helm, F. C. T. 1994a Analysis of the kinematic and dynamic behavior of the shoulder mechanism. J. Biomech. 27, 527550. (doi:10.1016/0021-9290(94)90064-7) van der Helm, F. C. T. 1994b A nite element musculoskeletal model of the shoulder mechanism. J. Biomech. 27, 551569. (doi:10.1016/0021-9290(94)90065-5)
Phil. Trans. R. Soc. A (2009)

Downloaded from rsta.royalsocietypublishing.org on December 2, 2009

2118

P. Favre et al.

van der Helm, F. C. T. & Veeger, H. E. J. 1996 Quasi-static analysis of muscle forces in the shoulder mechanism during wheelchair propulsion. J. Biomech. 29, 3952. (doi:10.1016/00219290(95)00026-7) van der Helm, F. C. T. & Veenbaas, R. 1991 Modelling the mechanical effect of muscles with large attachment sites: application to the shoulder mechanism. J. Biomech. 24, 11511163. (doi:10. 1016/0021-9290(91)90007-A) van der Helm, F. C. T., Veeger, H. E. J., Pronk, G. M., van der Woude, L. H. V. & Rozendal, R. H. 1992 Geometry parameters for musculoskeletal modelling of the shoulder system. J. Biomech. 25, 129144. (doi:10.1016/0021-9290(92)90270-B) van Drongelen, S., van der Woude, L. H., Janssen, T. W., Angenot, E. L., Chadwick, E. K. & Veeger, D. J. H. 2005 Mechanical load on the upper extremity during wheelchair activities. Arch. Phys. Med. Rehabil. 86, 12141220. (doi:10.1016/j.apmr.2004.09.023) van Drongelen, S., van der Woude, L. H. V., Janssen, T. W. J., Angenot, E. L. D., Chadwick, E. K. J. & Veeger, H. E. J. 2006 Glenohumeral joint loading in tetraplegia during weight relief lifting: a simulation study. Clin. Biomech. (Bristol, Avon) 21, 128137. (doi:10. 1016/j.clinbiomech.2005.09.003) Veeger, H. E. J. & van der Helm, F. C. T. 2007 Shoulder function: the perfect compromise between mobility and stability. J. Biomech. 40, 21192129. (doi:10.1016/j.jbiomech.2006.10.016) Veeger, H. E. J., Rozendaal, L. A. & van der Helm, F. C. T. 2002 Load on the shoulder in low intensity wheelchair propulsion. Clin. Biomech. (Bristol, Avon) 17, 211218. (doi:10.1016/ S0268-0033(02)00008-6) Viceconti, M., Olsen, S., Nolte, L. P. & Burton, K. 2005 Extracting clinically relevant data from nite element simulations. Clin. Biomech. (Bristol, Avon) 20, 451454. (doi:10.1016/ j.clinbiomech.2005.01.010) Wakabayashi, I., Itoi, E., Sano, H., Shibuya, Y., Sashi, R., Minagawa, H. & Kobayashi, M. 2003 Mechanical environment of the supraspinatus tendon: a two-dimensional nite element model analysis. J. Shoulder Elbow Surg. 12, 612617. (doi:10.1016/S1058-2746(03)00214-3) Weinans, H., Huiskes, R. & Grootenboer, H. J. 1990 Trends of mechanical consequences and modeling of a brous membrane around femoral hip prostheses. J. Biomech. 23, 9911000. (doi:10.1016/0021-9290(90)90314-S) Werner, C. M., Favre, P. & Gerber, C. 2007 The role of the subscapularis in preventing anterior glenohumeral subluxation in the abducted, externally rotated position of the arm. Clin. Biomech. (Bristol, Avon) 22, 495501. (doi:10.1016/j.clinbiomech.2006.12.007) Williams Jr, G. R., Rockwood Jr, C. A., Bigliani, L. U., Iannotti, J. P. & Stanwood, W. 2004 Rotator cuff tears: why do we repair them? J. Bone Joint Surg. Am. 86, 27642776. Wirth, M. A. & Rockwood Jr, C. A. 1996 Complications of total shoulder-replacement arthroplasty. J. Bone Joint Surg. Am. 78, 603616. Young, P. G., Beresford-West, T. B., Coward, S. R., Notarberardino, B., Walker, B. & AbdulAziz, A. 2008 An efcient approach to converting three-dimensional image data into highly accurate computational models. Phil. Trans. R. Soc. A 366, 31553173. (doi:10.1098/rsta.2008. 0090)

Phil. Trans. R. Soc. A (2009)