Spinal Compressive Forces in Adolescent Idiopathic Scoliosis With and Without Carrying Loads: A Musculoskeletal Modeling Study
Stefan Schmid, Katelyn A. Burkhart, Brett T. Allaire, Daniel Grindle, Tito Bassani, Fabio Galbusera, Dennis E. Anderson
ffbioe-08-00159 February 28, 2020 Time: 19:23
ORIGINAL RESEARCH published: 03 March 2020doi: 10.3389/fbioe.2020.00159
Edited by:
Alexandros E. Tsouknidas,University of Western Macedonia,Greece
Reviewed by:
Rizwan Arshad,Royal Military College of Canada,CanadaMohammad Nikkhoo,Islamic Azad University, Iran *Correspondence:
Stefan [email protected]
Specialty section:
This article was submitted toBiomechanics,a section of the journalFrontiers in Bioengineering andBiotechnology
Received:
16 December 2019
Accepted:
17 February 2020
Published:
03 March 2020
Citation:
Schmid S, Burkhart KA, Allaire BT,Grindle D, Bassani T, Galbusera F andAnderson DE (2020) SpinalCompressive Forces in AdolescentIdiopathic Scoliosis With and WithoutCarrying Loads: A MusculoskeletalModeling Study.Front. Bioeng. Biotechnol. 8:159.doi: 10.3389/fbioe.2020.00159
Spinal Compressive Forces inAdolescent Idiopathic Scoliosis Withand Without Carrying Loads: AMusculoskeletal Modeling Study
Stefan Schmid * , Katelyn A. Burkhart , Brett T. Allaire , Daniel Grindle ,Tito Bassani , Fabio Galbusera and Dennis E. Anderson Center for Advanced Orthopaedic Studies, Beth Israel Deaconess Medical Center, Boston, MA, United States, Department of Orthopaedic Surgery, Harvard Medical School, Boston, MA, United States, Spinal MovementBiomechanics Group, Division of Physiotherapy, Department of Health Professions, Bern University of Applied Sciences,Bern, Switzerland, Division of Engineering Mechanics, Department of Biomedical Engineering and Mechanics, VirginiaPolytechnic Institute and State University, Blacksburg, VA, United States, Laboratory of Biological Structures Mechanics(LABS), IRCCS Istituto Ortopedico Galeazzi, Milan, Italy
The pathomechanisms of curve progression in adolescent idiopathic scoliosis (AIS)remain poorly understood and biomechanical data are limited. A deeper insightinto spinal loading could provide valuable information toward the improvement ofcurrent treatment strategies. This work therefore aimed at using subject-specificmusculoskeletal full-body models of patients with AIS to predict segmental compressiveforces around the curve apex and to investigate how these forces are affected bysimulated load carrying. Models were created based on spatially calibrated biplanarradiographic images from 24 patients with mild to moderate AIS and validated bycomparing predictions of paravertebral muscle activity with reported values from in vivo studies. Spinal compressive forces were predicted during unloaded upright standing aswell as standing with external loads of 10, 15, and 20% of body weight (BW) applied tothe scapulae to simulate carrying a backpack in the regular way on the back as well as infront of the body and over the shoulder on the concave and convex sides of the scolioticcurve. The predicted muscle activities around the curve apex were higher on the convexside for the erector spinae (ES) and multifidi (MF) muscles, which was comparable tothe EMG-based in vivo measurements from the literature. In terms of spinal loading,the implementation of spinal deformity resulted in a 10% increase of compressive forceat the curve apex during unloaded upright standing. Apical compressive forces furtherincreased by 50–62% for a simulated 10% BW load and by 77–94% and 103–128%for 15% and 20% BW loads, respectively. Moreover, load-dependent compressive forceincreases were the lowest in the regular backpack and the highest in the frontpackand convex conditions, with concave side-carrying forces in between. The predictionsindicated increased segmental compressive forces during unloaded upright standing,which could be ascribed to the scoliotic deformation. When carrying loads, compressive March 2020 | Volume 8 | Article 159 bioe-08-00159 February 28, 2020 Time: 19:23
Schmid et al. Spinal Loading in AIS forces further increased depending on the carrying mode and the weight of the load.These results can be used as a basis for further studies investigating segmental loadingin AIS patients during functional activities. Models can thereby be created using thesame approach as proposed in this study.
Keywords: OpenSim, inverse dynamics, validation, spinal loading, AIS, muscle asymmetry, subject-specificmodeling, backpack carrying
INTRODUCTION
Adolescent idiopathic scoliosis (AIS) is a complex three-dimensional deformity of the spine, which affects as manyas 4 out of 100 adolescents and occurs in early puberty(Cheng et al., 2015). Among congenital, neuromuscular, andother types of scoliosis, AIS is by far the most commonform and is characterized by a poorly understood etiologyand pathogenesis (de Seze and Cugy, 2012; Schlosser et al.,2014). Once the diagnosis of AIS is made, adolescents aregenerally treated conservatively using scoliosis-specific exercises(SSEs) and other forms of physiotherapy in order to minimizecurve progression (Romano et al., 2012). For 10% of theinitially diagnosed AIS patients, however, scoliosis exceeds alateral deformation of 20–30 ◦ and brace treatment is indicated(Negrini et al., 2015). And finally, in about one fourth of thepatients treated with exercises and braces, the progression ofdeformation cannot be contained and surgical intervention isrequired (Dolan and Weinstein, 2007).In order to avoid bracing, surgery, and any associated healthproblems (e.g., chronic back pain), stopping curve progression inan early stage by means of SSEs is highly important. However,the effects of SSEs are not evident and further research is neededto clearly define the best types of SSEs as well as the frequencyand intensity with which they should be administered (Romanoet al., 2012). Since the pathomechanics of the AIS spine are notwell understood, SSE protocols that affect spinal loading in atargeted, and scientifically sound manner cannot be developed.The literature currently lacks any studies reporting on spinalloading during functional activities or exercises in AIS patients.Furthermore, it has been suggested that schoolbag carryingmight play a role in the progression of scoliotic deformity andcontribute to the development of back pain in AIS patients (Chowet al., 2006; Sahli et al., 2013). Sahli et al. (2013) recommendedlimiting backpack loads to 10% of body weight (BW) and carryingthe load equally over both shoulders or over the shoulder on theconcave rather than the convex side. However, these statementsshould be considered with caution since none of these studiesinvestigated the effects of load carrying on spinal loading ortrunk muscle forces.Due to recent advancements in radiography-based geometric3D reconstruction (Bassani et al., 2017) and musculoskeletalmodeling (Bruno et al., 2015; Schmid et al., 2019), suchparameters can be studies non-invasively and do not requireinvasive procedures such as intradiscal pressure or implant-basedvertebral load measurements. The aim of this study was twofold:(1) To create subject-specific musculoskeletal full-body models ofpatients with mild to moderate AIS and validate predicted muscle activities with EMG data available in literature, and (2) to predictsegmental compressive forces around the curve apex and howforces are affected by load carrying conditions. MATERIALS AND METHODSDevelopment of Subject-Specific Models
Base Models
The base models for this study were created using our previouslyvalidated OpenSim-based musculoskeletal full-body models forchildren and adolescents aged 6–18 years (Schmid et al., 2019).They include a fully articulated thoracolumbar spine with arib cage and are age- and gender-adjusted for sagittal spinalalignment as well as segmental inertial properties and maximumtrunk muscle force capacity.We enhanced the models with non-linear stiffness propertiesfor flexion-extension and lateral-bending motions in allsegments between T1/2 and L5/S1 using values from a recentmeta-regression analysis over 45 studies involving experimentson adult cadaveric spines (Zhang et al., 2020). The propertieswere implemented using standard built-in linear bushingelements (expression-based bushing forces), which createreaction moments based on the rotational displacements of theadjacent vertebrae (Meng et al., 2015; Senteler et al., 2016). Toavoid stiffness-related reaction moments in the neutral positionof the spinal segments (i.e., position of the spinal segmentswhen standing in an upright position), the bushing frameswere oriented accordingly. Passive moments for segmental axialrotation were modeled using reserve coordinate actuators.
Implementing Spinal Deformity
Using our enhanced base models, we created subject-specificmodels for 24 patients with mild to moderate AIS (
Table 1 ).Spinal deformity was thereby implemented based on existingsimultaneously captured and spatially calibrated anterior–posterior and lateral radiographic images (EOS Imaging, France)that were acquired within a previous study conducted at theIRCCS Istituto Ortopedico Galeazzi in Milan, Italy (Bassani et al.,2017). The protocol for this study was approved by the local ethicscommission and patient assent and parental permission to usethe anonymized radiological data were given by signing a writteninformed consent form.Three-dimensional position and orientation of each vertebrafrom T1 to L5 was extracted from the radiographs usinga custom MATLAB script (Bassani et al., 2017). In brief,this script contains a graphical user interface (GUI), whichallows for the manual identification of nine characteristic March 2020 | Volume 8 | Article 159 bioe-08-00159 February 28, 2020 Time: 19:23
Schmid et al. Spinal Loading in AIS
TABLE 1 |
Demographics of the patients with adolescent idiopathic scoliosis (AIS) from which the biplanar radiographic images were used to create the musculoskeletalmodels for the current study.
Patient Sex Age (years) Height (cm) Mass (kg) Curve type Cobb ( ◦ ) Convexity P1 Female 14 157 47 4B 23.9 LeftP2 Female 13 168 46 5 18.1 LeftP3 Female 12 140 35 1B 14.0 RightP4 Female 14 165 50 3B 24.0 RightP5 Female 15 168 48 5 24.5 LeftP6 Female 16 172 54 1A 25.5 RightP7 Female 14 160 50 5 23.4 RightP8 Male 11 134 39 5 26.5 LeftP9 Female 12 160 45 5 22.1 LeftP10 Female 15 165 45 5 22.4 RightP11 Male 14 180 50 5 16.3 RightP12 Female 13 155 44 5 18.3 LeftP13 Male 15 175 52 1A 28.3 LeftP14 Female 15 170 60 5 19.2 LeftP15 Female 15 157 48 5 27.7 RightP16 Female 9 133 29 5 18.8 RightP17 Female 11 144 35 5 18.3 RightP18 Female 14 165 42 5 27.7 LeftP19 Female 13 167 60 5 21.1 LeftP20 Female 14 167 48 5 27.9 RightP21 Female 16 169 63 1A 25.0 RightP22 Female 15 168 51 1A 13.6 RightP23 Female 17 170 53 5 24.7 RightP24 Female 14 160 45 5 18.3 LeftAverage (SD) 13.8 (1.8) 161.2 (12.3) 47.5 (8.0) – 22.1 (4.4) – Curve type classification according to Lenke et al. (2001). landmarks per vertebra, i.e., upper and lower vertebral cornersin the sagittal and frontal planes as well as the location ofspinous process in the frontal plane. Based on these landmarks,sagittal and frontal vertebral orientations were calculated asthe average of the slopes of the lines connecting upperand lower vertebral corners. The axial vertebral orientationwas obtained by transforming the landmark which identifiesthe spinous process in 3D space through evaluating a fittedreferential anatomic mesh model for the vertebra underassessment (Bassani et al., 2017). To account for individualspine height, we determined the height of each vertebralbody using the geometric centers of the proximal and distalintervertebral disc spaces in the sagittal plane (centroids ofthe lower and upper corners of the proximal and distalvertebrae, respectively) and the vertebral tilt angle in the frontalplane (
Figure 1 ).Subject-specific models were created in four steps: (1) scalingthe base model from the corresponding age- and gender-groupby body height and body mass, (2) implementing 3D spinaldeformity by adjusting the orientation of the vertebral bodies T1to L5 in the flexion/extension, lateral-flexion and axial rotationdirections, (3) scaling intersegmental joint distances by vertebralbody height, and (4) re-adjusting 3D orientation of the lumpedhead and neck segment as well as arms and ribs to a neutralposition (
Figure 2 ). Evaluation of Muscle Geometry
It was previously reported that AIS is associated with side-to-sideasymmetries in erector spinae (ES) and multifidi (MF) musclegeometry (Zoabli et al., 2007; Zapata et al., 2015). To ensureappropriate handling of the muscle geometry in our models,we therefore estimated bilateral CSAs of the modeled ES andMF muscles for each thoracic and lumbar vertebral mid-planeby summing the CSAs of the individual fascicles [calculated bydividing the maximum force generating capacity of the respectivefascicle by an assumed uniform maximal muscle stress (MMS)of 100 N/cm ] crossing the respective mid-plane to compute anequivalent muscle group CSA at that level (Bruno et al., 2015).We refer to this procedure as a “virtual CT scan” of the model,which enables the comparison of model muscle geometry withconventional medical imaging studies. For the comparison withthe literature, an asymmetry ratio was calculated by dividing theCSA of the muscles on the convex side by the CSA of the muscleson the concave side. In accordance with the in vivo studies, thisratio was calculated for ES at the curve apex and for MF at thelevels T8, L1, and L4 as well as at the curve apex. Simulations
All simulations were carried out using OpenSim 3.3 (Delp et al.,2007) and MATLAB R2019a (MathWorks Inc., Natick, MA,United States). Models were solved using an inverse dynamics March 2020 | Volume 8 | Article 159 bioe-08-00159 February 28, 2020 Time: 19:23
Schmid et al. Spinal Loading in AIS
FIGURE 1 |
Determination of vertebral body height ( H Total ) from radiographicimages in two steps: (1) Extracting the distance between the centroids of theupper and lower intervertebral disc spaces in the sagittal plane ( H Sagittal ), and(2) correct it for scoliosis-dependent projection errors using the vertebral tiltangle measured in the frontal plane.
FIGURE 2 |
Model creation in four steps: (1) scaling of base model by bodyheight and body mass, (2) implementing 3D spinal deformity by adjusting theorientation of the vertebral bodies T1 to L5, (3) scaling intersegmental jointdistances by vertebral body height, and (4) re-adjusting 3D orientation of thelumped head and neck segment as well as arms and ribs to a neutral position. based static optimization with a cost function that minimizedthe sum of squared muscle activation (Herzog, 1987). Dueto uncertainties of how the scoliotic deformation might affectmuscle physiology, we solved the models without consideringforce-length relationships.
Model Validation
To comply with the best practice guidelines for verificationand validation of musculoskeletal models (Hicks et al., 2015),we aimed at providing a reasonable validation for our modelsby comparing model predictions to in vivo values availablefrom the literature. Unfortunately, the literature lacks in vivo studies reporting on functional spinal loading (e.g., segmental compressive forces or intradiscal pressure in upright standingconditions) in AIS patients. For this reason, we conductedsimulations of trunk muscle activity in standing and pronepositions to compare them with the results of three reported in vivo studies using surface electromyography (EMG) (Cheunget al., 2005; Kwok et al., 2015; Stetkarova et al., 2016). For each ofthe simulations, we selected the AIS models that matched best therespective in vivo study population in terms of curve location.To evaluate the accuracy of our models to predict ES muscleactivity, we placed the models in a neutral upright standingposition and compared the convex to concave ratios of theaverage activation levels of the ES muscle fascicles in the lumbarand thoracic regions as well as at the curve apex and upperand lower curve limits (two levels above and below the apex) tosurface EMG-based in vivo measurements in AIS patients withmain thoracic and thoracolumbar curves (Kwok et al., 2015)as well as patients with non-progressive AIS (Cheung et al.,2005), respectively. Accuracy of MF muscle activity predictionswas evaluated by placing the models in a prone position andcomparing the convex to concave ratio of the average activationlevels of the MF muscle fascicles at the curve apex to needle EMG-based in vivo measurements in AIS patients with main thoraciccurves (Stetkarova et al., 2016).Muscle fascicles were thereby selected based on the surfaceelectrode placement and needle electrode insertion locationsdescribed in the respective in vivo studies. To simulate the floorand the table where the models were “standing or lying on,”we used residual point actuators with maximum activation ata force of 10 kN, which was shown to be large enough toprovide the required support with minimal expenses in the staticoptimization (Schmid et al., 2019). Model predictions and in vivo measurements were compared qualitatively.
Prediction of Compressive Forces
In order to investigate spinal compressive forces in uprightstanding AIS patients with and without carrying loads, weconducted simulations in five different conditions (
Table 2 ).After solving the models, joint reaction analysis was carriedout to calculate the axial compressive forces acting on thespinal segments at the curve apex as well as one and two levelsabove and below. In addition to the absolute force magnitudes,compressive forces in AIS patients in the unloaded condition (1)were expressed as a percentage of the forces derived from theundeformed models, whereas compressive forces in AIS patientsin the loaded conditions (2–5) were expressed as a percentageof the unloaded condition. To provide a coherent overview ofthese percentages, data were presented using violin plots withsuperimposed boxplots and individual values. Wilcoxon signedrank tests with an alpha-level set to 5% were conducted to test fordifferences from 100%.
RESULTSMuscle Geometry
The predicted mean CSA ratio for the ES muscle at the apexwas 1.05 SD 0.08, which compares to the ratio of 1.02 calculated March 2020 | Volume 8 | Article 159 bioe-08-00159 February 28, 2020 Time: 19:23
Schmid et al. Spinal Loading in AIS
TABLE 2 |
Conditions for simulating spinal compressive forces during uprightstanding with and without carrying loads in patients with adolescent idiopathicscoliosis (AIS).
Condition Models External load
Unloaded AIS and undeformed No external load appliedBackpack Only AIS External loads of 10, 15, and 20% ofbody weight (BW) applied 20 ◦ dorsallyangled to the lateral third of the upperedge of the scapulae (equallydistributed between sides) to simulatea regular backpackFrontpack Only AIS External loads of 10, 15, and 20% ofBW applied 25 ◦ ventrally angled tothe lateral third of the upper edge ofthe scapulae (equally distributedbetween sides) to simulate abackpack carried in front of the bodySidepackconcave Only AIS External loads of 10, 15, and 20% ofBW applied 5 ◦ dorsally and 10 ◦ laterally angled to the lateral third ofthe upper edge of the scapula on theconcave side to simulate a backpackcarried unilaterally over the shoulderon the concave sideSidepackconvex Only AIS External loads of 10, 15, and 20% ofBW applied 5 ◦ dorsally and 10 ◦ laterally angled to the lateral third ofthe upper edge of the scapula on theconvex side to simulate a backpackcarried unilaterally over the shoulderon the convex side The angles for the external loads were derived from standardized photographs ofone of the investigators carrying a backpack in the different modes. Models beforeimplementation of spinal deformity. from MRI-derived ES muscle volumes reported in the literature(Zoabli et al., 2007;
Figure 3 ). For the MF muscle, predicted meanCSA ratios for the levels T8, L1, and L4 were 0.95 SD 0.21, 0.92SD 0.09, and 1.01 SD 0.03, respectively. This corresponded wellto the values reported in the literature for T8 (0.96), L1 (0.95),and L4 (0.98) (Zapata et al., 2015). Predicted mean CSA ratio forthe MF muscle at the curve apex was 0.89 SD 0.11, however, no in vivo values were available for comparison.
Validation Studies
The predicted mean ES muscle activity ratios during uprightstanding indicated higher activity on the convex side of themuscle for thoracic (thoracic region: 1.29 SD 0.71; lumbar region:1.18 SD 0.78) and thoracolumbar curves (thoracic region: 1.05SD 0.19; lumbar region: 1.12 SD 0.31) (
Figure 4 ). Furthermore,ES muscle activity within the curve also indicated higher convexactivity at the apex (1.72 SD 1.06) as well as the upper (1.37 SD0.78) and lower curve ends (1.02 SD 0.83). These ratios comparereasonably well to the literature for the thoracic and lumbarportions of the ES muscle in patients with main thoracic curves(1.87 SD 1.7 and 1.7 SD 0.85, respectively) as well as at thecurve apex (2.1 SD 1.38) and the lower curve end (0.96 SD 0.32)(Cheung et al., 2005; Kwok et al., 2015). For MF muscle activity,models predictions indicated higher activity at the apex on the convex side of the curve (1.21 SD 0.50), which goes along withthe in vivo measured ratio of 1.38 (Stetkarova et al., 2016).
Spinal Compressive Forces
The implementation of spinal deformity resulted in highermedian compressive forces within the scoliotic curve, with forcesconstantly increasing from two levels above the apex (103% IQR13%, p = 0.092), to the apex (110% IQR 13%, p < p < Figure 5 ).All loaded median axial compressive forces were significantlydifferent from 100% (i.e., from an unloaded condition) at alevel of p < Figure 6 ). In the regular backpack as well as both sidepackconditions, compressive force increased the most above the apexand the least below the apex, whereas in the frontpack condition,compressive force increased about equally on all spinal levels.When applying loads corresponding to 15 and 20% of BW,median axial compressive forces increased on average by 77and 103% for the backpack, 93 and 125% for the frontpack, 94and 128% for the convex sidepack, and 85 and 116% for theconcave sidepack conditions, respectively (
Figures 7 , ). The loaddistribution pattern within the spinal levels of the scoliotic curveremained similar as described for the 10% of BW load.A complete set of the absolute force magnitude andrelative force percentage values can be found in the Supplementary Material . DISCUSSION
We created 24 subject-specific musculoskeletal full-body modelsfrom biplanar radiographic images of patients with mildto moderate AIS and validated these models by comparingpredictions of paravertebral muscle activity with reported valuesfrom in vivo studies. Moreover, we predicted apical spinalloads with and without simulated load carrying, i.e., carrying abackpack in the regular way, carrying a backpack in front of thebody and carrying a backpack over the shoulder on the concaveand convex sides of the scoliotic curve.The evaluation of muscle geometry indicated that theimplementation of spinal deformity resulted in side-to-sideasymmetries, which agreed with reports in the literature. Thevalidation studies showed higher convex ES and MF muscleactivity around the curve apex, which was comparable tothe EMG-based in vivo measurements from the literature.Measurements of overall thoracic and lumbar ES muscle activityagreed well for thoracic but less thoracolumbar curves. In termsof spinal loading, the implementation of spinal deformity resultedin a 10% increase of compressive force at the curve apex duringunloaded upright standing. Apical compressive forces furtherincreased by 50–62% for a simulated 10% BW load and by 77–94% and 103–128% for 15 and 20% BW loads, respectively.Moreover, load-dependent compressive force increases were March 2020 | Volume 8 | Article 159 bioe-08-00159 February 28, 2020 Time: 19:23
Schmid et al. Spinal Loading in AIS
FIGURE 3 |
Convex to concave ratios for cross-sectional areas (CSA) of the erector spinae (ES) muscle at the curve apex (left) as well as the multifidi (MF) musclesat the levels T8, L1, L4, and the curve apex (right) . FIGURE 4 |
Convex to concave ratios for erector spinae (ES) muscle activity in the thoracic and lumbar regions (left) as well as at the apex and the upper and lowerends of the curve (middle) during upright standing, and for multifidi (MF) muscle activity at the curve apex (right) in prone position. the lowest in the regular backpack and the highest in thefrontpack and convex conditions, with concave side-carryingforces in between. Even though the evaluation of muscle geometry estimatedfrom our models indicated larger ES muscle CSA on the convexside of the scoliotic deformation, about one third of our models March 2020 | Volume 8 | Article 159 bioe-08-00159 February 28, 2020 Time: 19:23
Schmid et al. Spinal Loading in AIS
FIGURE 5 |
Spinal compressive forces at the curve apex as well as two levelsabove and below in patients with adolescent idiopathic scoliosis (AIS)expressed as percentages of the forces in the same individuals but withoutspinal deformity (% of undeformed). Data are presented in form of violin plotswith superimposed boxplots and individual values. showed in fact larger ES muscle CSA on the concave side.This agrees with the study of Zoabli et al. (2007), reportingthat even though there was an overall tendency for largerES muscle volume on the convex side, some patients alsopresented larger volumes on the concave side. For the MFmuscle at the curve apex, a larger CSA on the concave sidewas found in about 90% of the models, whereas the remaining10% showed a larger CSA on the convex side. Despite theseuncertainties, however, the muscle geometry estimated from ourAIS models was comparable to what was measured in vivo . Thisraises the question of whether the previously reported musclevolume/thickness asymmetries in AIS patients represent an actualchange in muscle, or is just a result of measurement of musclewith different spinal geometry.The partial disagreements between our model predictionsand the ES muscle activity ratios derived from the resultsreported by Kwok et al. (2015) might be related to differencesin curve characteristics (i.e., location of the apex and severityof deformation) between the respective patient populations.Especially when evaluating overall ES muscle activity in thethoracic and lumbar regions with the same electrode placementfor all patients, different curve characteristics could have asignificant effect on muscle activation. Furthermore, muscle fiberredistribution with a higher proportion of type I fibers on theconvex side of the AIS curve (Gonyea et al., 1985; Meier et al.,1997; Mannion et al., 1998; Stetkarova et al., 2016) might beanother contributing factor, since muscle fiber type seems to have an influence on the EMG signal (Poosapadi Arjunan et al., 2016).However, due to the lack of appropriate data, the considerationof fiber distribution change in our current models would beassociated with too many assumptions. Finally, it is not knownwhether AIS has an influence on the force-length-relationshipof the paravertebral muscles, which could also have an influenceon EMG activity. Based on our validation studies, however, weconsider the muscle activation patterns predicted by our modelscomparable to the patterns reported in the in vivo studies.This is the first study using inverse dynamics-basedmusculoskeletal full-body modeling to investigate the immediateeffect of AIS-related spinal deformity on axial compressive forceswithin the scoliotic curve. The results suggested that spinaldeformity causes an overall increase in compressive forces, withforces increasing the most at the lower and the least at the upperend of the curve. When considering the different curve types,it appears that the average increase in segmental loading wasnot significantly related to the spinal level of the curve apex orthe curve severity, i.e., the Cobb angles (Pearson correlation: r = 0.28, p = 0.193 and r = 0.22, p = 0.302, respectively). Itshould be considered, however, that all the patients in this studyhad mild to moderate AIS, providing a relatively small rangeof Cobb angles. The relationship between compressive forceincrease and curve severity would probably be more pronouncedwith larger range of Cobb angles in the dataset. It should alsobe noted that a small fraction of the predictions resulted incompressive force decreases. When looking at these cases,especially those predicting compressive forces of less than 90%of the undeformed, it appears that these patients tended to haveparticularly flat thoracic sagittal profiles (i.e., < ◦ of thoracickyphosis), which might have resulted in a significant reductionof muscular effort (specifically the ES muscle) and therefore inreduced compressive forces.The simulation of load carrying in AIS patients indicatedcompressive force increments that were dependent on thecarrying mode as well as the weight of the load. Carryingthe load in front of the body resulted in considerably highercompressive forces compared to regular backpack carrying. Thisis not surprising since the front carrying mode would be assumedto cause higher muscular effort to prevent increased flexion.Interestingly, carrying the backpack on the concave side yieldedcompressive forces that were slightly higher than the ones forregular backpack carrying, but lower compared to carrying theload in front. This indicates that shifting the load from the backto the concave side does result in increased compressive forces –most likely due to the higher muscular effort on the contralateralside – but not as much as carrying the load symmetrically infront. In addition, carrying the load on the convex side causedcompressive forces that were comparable to the front carryingmode, most likely because the load was acting more directlyon the spinal segments. The different patterns of compressiveforce increments between the frontpack condition (constant overall levels) and the other carrying conditions (decrease in forceincrements from two levels above to two levels below the apex)might be related to the direction of the applied external load inthe sagittal plane. In the frontpack condition, the external loadwas directed anteriorly, which resulted in increased paraspinal March 2020 | Volume 8 | Article 159 bioe-08-00159 February 28, 2020 Time: 19:23
Schmid et al. Spinal Loading in AIS
FIGURE 6 |
Spinal compressive forces at the curve apex as well as two levels above and below in patients with adolescent idiopathic scoliosis (AIS) when carrying abackpack with a weight corresponding to 10% of body weight in the regular way (top left) , in front of the body (top right) as well as over the shoulder on theconcave (bottom left) , and convex sides (bottom right) of the scoliotic curve. Forces are expressed as percentages of unloaded upright standing and presented inform of violin plots with superimposed boxplots and individual values. muscle activity. In the other carrying conditions, however, theexternal load is directed posteriorly, causing an increased activitypredominantly of the abdominal muscles. Considering that most of the abdominal muscles are only indirectly connected to thespine, i.e., over the rib cage, it seems plausible that this affectedspinal loading differently than in the frontpack condition. In March 2020 | Volume 8 | Article 159 bioe-08-00159 February 28, 2020 Time: 19:23
Schmid et al. Spinal Loading in AIS
FIGURE 7 |
Spinal compressive forces at the curve apex as well as two levels above and below in patients with adolescent idiopathic scoliosis (AIS) when carrying abackpack with a weight corresponding to 15% of body weight in the regular way (top left) , in front of the body (top right) as well as over the shoulder on theconcave (bottom left) , and convex sides (bottom right) of the scoliotic curve. Forces are expressed as percentages of unloaded upright standing and presented inform of violin plots with superimposed boxplots and individual values. any way, these results do not allow any conclusions on whethercarrying a load in front, on either side or regularly on theback is advantageous to minimize curve progression or preventcomplications such as joint degeneration or back pain. In fact, it is possible that carrying the load on the convex side mightput the patients at a higher risk for back pain, but at the sametime slow down curve progression by modulating the segmentalload in a way that vertebral growth is positively affected due to March 2020 | Volume 8 | Article 159 bioe-08-00159 February 28, 2020 Time: 19:23
Schmid et al. Spinal Loading in AIS
FIGURE 8 |
Spinal compressive forces at the curve apex as well as two levels above and below in patients with adolescent idiopathic scoliosis (AIS) when carrying abackpack with a weight corresponding to 20% of body weight in the regular way (top left) , in front of the body (top right) as well as over the shoulder on theconcave (bottom left) , and convex sides (bottom right) of the scoliotic curve. Forces are expressed as percentages of unloaded upright standing and presented inform of violin plots with superimposed boxplots and individual values. the Hueter-Volkmann law, where bone growth is slowed withcompression and accelerated with distraction (de Seze and Cugy,2012). Future studies should therefore address this issue by usinga combination of motion capture-driven musculoskeletal andfinite element models. This study has some important limitations that should bediscussed. First of all, due to a lack of appropriate data forhealthy children and adolescents as well as patients with AIS,the passive segmental stiffness properties that were implementedin the current base models were derived from healthy adult March 2020 | Volume 8 | Article 159 bioe-08-00159 February 28, 2020 Time: 19:23
Schmid et al. Spinal Loading in AIS cadaveric spines. This issue could be addressed in the futureby conducting clinical studies in AIS patients using approachessuch as the intraoperative determination of load-displacementbehavior proposed by Reutlinger et al. (2012). However, thelack of appropriate stiffness properties did not affect the currentpredictions since all simulations were conducted in a neutralposition, i.e., with the spinal segments assumed to be inthe neutral zone (Smit et al., 2011). Stiffness-related reactionmoments would only have occurred with induced segmentalrotations. When using the models for future investigationsinvolving simulations beyond the neutral position of the spine,on the other hand, these limitations will have to be considered.Secondly, we did only consider compressive forces in thisstudy, but forces in other directions (i.e., anterior–posteriorand medial–lateral shear forces) might also be strongly affectedby scoliotic deformities. Furthermore, the patient populationsof the in vivo studies used for the validation of our modelsdid not exactly match the population from which they werecreated. It is therefore advised that future studies investigatingspinal loading in AIS include EMG measurements of theparaspinal muscles for more specific validations of the respectivemodels. Lastly, the current simulations were not based on real-life kinematics, i.e., they were not driven by motion capturedata. Especially for the load carrying investigations, it canbe assumed that real subjects would have slightly adaptedtheir posture based on the applied load, such as previouslyobserved for regular backpack carrying in healthy young adults(Neuschwander et al., 2010).In conclusion, this study used validated subject-specificOpenSim-based musculoskeletal full-body models to providean insight into spinal loading in patients with AIS with andwithout carrying loads. The predictions indicated increasedsegmental compressive forces of about 10% around thecurve apex during unloaded upright standing. When carryingloads, compressive forces further increased depending on thecarrying mode and the weight of the load. These resultscan be used as a basis for further studies investigatingsegmental loading in AIS patients during functional activities.Models can thereby be created using the same approach asproposed in this study.
DATA AVAILABILITY STATEMENT
The datasets generated for this study are available on request tothe corresponding author.
AUTHOR CONTRIBUTIONS
SS substantially contributed to the conception and design ofthe study, created the subject-specific models and performedthe simulations, analyzed the data, interpreted the results, andwrote the manuscript. KB, BA, and DG provided substantialassistance to the creation of the models and the simulations. TBand FG provided the EOS images and MATLAB scripts for theextraction of the spinal deformity parameters. DA substantiallycontributed to the conception and design of the study, andprovided substantial assistance to the creation of the models andthe simulations. All authors critically revised the manuscript andapproved the version to be published.
FUNDING
This work was funded by the Swiss National Science Foundation(SNSF, grant no. 178427) and the National Center for Simulationin Rehabilitation Research (NCSRR, sub-award of NIH grantno. 5P2CHD065690).
ACKNOWLEDGMENTS
The authors thank Dr. Bram Verhofste for assistance in theclinical interpretation of the radiographs.
SUPPLEMENTARY MATERIAL
REFERENCES
Bassani, T., Ottardi, C., Costa, F., Brayda-Bruno, M., Wilke, H. J., and Galbusera,F. (2017). Semiautomated 3D spine reconstruction from biplanar radiographicimages: prediction of intervertebral loading in scoliotic subjects.
Front. Bioeng.Biotechnol.
J. Biomech. Eng.
Nat. Rev. Dis. Primers
Eur. Spine J.
14, 130–137.doi: 10.1007/s00586-004-0780-7Chow, D. H., Kwok, M. L., Cheng, J. C., Lao, M. L., Holmes, A. D., Au-Yang, A.,et al. (2006). The effect of backpack weight on the standing posture and balance of schoolgirls with adolescent idiopathic scoliosis and normal controls.
Gait.Posture
24, 173–181. doi: 10.1016/j.gaitpost.2005.08.007de Seze, M., and Cugy, E. (2012). Pathogenesis of idiopathic scoliosis: a review.
Ann. Phys. Rehabil. Med.
55, 128–138. doi: 10.1016/j.rehab.2012.01.003Delp, S. L., Anderson, F. C., Arnold, A. S., Loan, P., Habib, A., John, C. T.,et al. (2007). OpenSim: open-source software to create and analyze dynamicsimulations of movement.
IEEE Trans. Biomed. Eng.
54, 1940–1950. doi: 10.1109/tbme.2007.901024Dolan, L. A., and Weinstein, S. L. (2007). Surgical rates after observation andbracing for adolescent idiopathic scoliosis: an evidence-based review.
Spine
J. Pediatr.Orthop.
5, 323–329. doi: 10.1097/01241398-198505000-00013Herzog, W. (1987). Individual muscle force estimations using a non-linear optimaldesign.
J. Neurosci. Methods
21, 167–179. doi: 10.1016/0165-0270(87)90114-2Hicks, J. L., Uchida, T. K., Seth, A., Rajagopal, A., and Delp, S. L. (2015).Is my model good enough? Best practices for verification and validation March 2020 | Volume 8 | Article 159 bioe-08-00159 February 28, 2020 Time: 19:23
Schmid et al. Spinal Loading in AIS of musculoskeletal models and simulations of movement.
J. Biomech. Eng.
Biomed. Res. Int.
J. Bone Joint Surg. Am.
83, 1169–1181.doi: 10.2106/00004623-200108000-00006Mannion, A. F., Meier, M., Grob, D., and Muntener, M. (1998). Paraspinalmuscle fibre type alterations associated with scoliosis: an old problemrevisited with new evidence.
Eur. Spine J.
7, 289–293. doi: 10.1007/s005860050077Meier, M. P., Klein, M. P., Krebs, D., Grob, D., and Muntener, M. (1997).Fiber transformations in multifidus muscle of young patients with idiopathicscoliosis.
Spine
22, 2357–2364. doi: 10.1097/00007632-199710150-00008Meng, X., Bruno, A. G., Cheng, B., Wang, W., Bouxsein, M. L., and Anderson,D. E. (2015). incorporating six degree-of-freedom intervertebral joint stiffnessin a lumbar spine musculoskeletal model-method and performance in flexedpostures.
J. Biomech. Eng.
Cochrane Database Syst. Rev.
6, Cd006850.Neuschwander, T. B., Cutrone, J., Macias, B. R., Cutrone, S., Murthy, G., Chambers,H., et al. (2010). The effect of backpacks on the lumbar spine in children: astanding magnetic resonance imaging study.
Spine
35, 83–88. doi: 10.1097/brs.0b013e3181b21a5dPoosapadi Arjunan, S., Kumar, D. K., Wheeler, K., Shimada, H., and Siddiqi, A.(2016). Effect of number of motor units and muscle fibre type on surfaceelectromyogram.
Med. Biol. Eng. Comput.
54, 575–582. doi: 10.1007/s11517-015-1344-1Reutlinger, C., Hasler, C., Scheffler, K., and Buchler, P. (2012). Intraoperativedetermination of the load-displacement behavior of scoliotic spinalmotion segments: preliminary clinical results.
Eur. Spine J.
Cochrane Database Syst. Rev.
8, Cd007837.Sahli, S., Rebai, H., Ghroubi, S., Yahia, A., Guermazi, M., and Elleuch, M. H. (2013).The effects of backpack load and carrying method on the balance of adolescentidiopathic scoliosis subjects.
Spine J.
13, 1835–1842. doi: 10.1016/j.spinee.2013.06.023 Schlosser, T. P., Van Der Heijden, G. J., Versteeg, A. L., and Castelein, R. M.(2014). How ’idiopathic’ is adolescent idiopathic scoliosis? A systematic reviewon associated abnormalities.
PLoS One
J. Biomech. doi: 10.1016/j.jbiomech.2019.07.049Senteler, M., Weisse, B., Rothenfluh, D. A., and Snedeker, J. G. (2016).Intervertebral reaction force prediction using an enhanced assembly ofOpenSim models.
Comput. Methods Biomech. Biomed. Engin.
19, 538–548.doi: 10.1080/10255842.2015.1043906Smit, T. H., Van Tunen, M. S., Van Der Veen, A. J., Kingma, I., and Van Dieen,J. H. (2011). Quantifying intervertebral disc mechanics: a new definition of theneutral zone.
BMC Musculoskelet. Disord.
Eur Spine J.
25, 3146–3153. doi: 10.1007/s00586-016-4628-8Zapata, K. A., Wang-Price, S. S., Sucato, D. J., and Dempsey-Robertson, M. (2015).Ultrasonographic measurements of paraspinal muscle thickness in adolescentidiopathic scoliosis: a comparison and reliability study.
Pediatr. Phys. Ther.
J. Biomech.
Spine J.
7, 338–344. doi: 10.1016/j.spinee.2006.04.001
Conflict of Interest:
The authors declare that the research was conducted in theabsence of any commercial or financial relationships that could be construed as apotential conflict of interest.
Copyright © 2020 Schmid, Burkhart, Allaire, Grindle, Bassani, Galbusera andAnderson. This is an open-access article distributed under the terms of the CreativeCommons Attribution License (CC BY). The use, distribution or reproduction inother forums is permitted, provided the original author(s) and the copyright owner(s)are credited and that the original publication in this journal is cited, in accordancewith accepted academic practice. No use, distribution or reproduction is permittedwhich does not comply with these terms.12