Deciphering the "Art" in Modeling and Simulation of the Knee Joint: Assessing Model Calibration Workflows and Outcomes.

Model reproducibility is a point of emphasis for the National Institutes of Health (NIH) and in science, broadly. As the use of computational modeling in biomechanics and orthopedics grows, so does the need to assess the reproducibility of modeling workflows and simulation predictions. The long-term goal of the KneeHub project is to understand the influence of potentially subjective decisions, thus the modeler's "art", on the reproducibility and predictive uncertainty of computational knee joint models. In this paper, we report on the model calibration phase of this project, during which five teams calibrated computational knee joint models of the same specimens from the same specimen-specific joint mechanics dataset. We investigated model calibration approaches and decisions, and compared calibration workflows and model outcomes among the teams. The selection of the calibration targets used in the calibration workflow differed greatly between the teams and was influenced by modeling decisions related to the representation of structures, and considerations for computational cost and implementation of optimization. While calibration improved model performance, differences in the postcalibration ligament properties and predicted kinematics were quantified and discussed in the context of modeling decisions. Even for teams with demonstrated expertise, model calibration is difficult to foresee and plan in detail, and the results of this study underscore the importance of identification and standardization of best practices for data sharing and calibration.

Reproducibility in modeling and simulation of the knee: Academic, industry, and regulatory perspectives.

Stakeholders in the modeling and simulation (M&S) community organized a workshop at the 2019 Annual Meeting of the Orthopaedic Research Society (ORS) entitled "Reproducibility in Modeling and Simulation of the Knee: Academic, Industry, and Regulatory Perspectives." The goal was to discuss efforts among these stakeholders to address irreproducibility in M&S focusing on the knee joint. An academic representative from a leading orthopedic hospital in the United States described a multi-institutional, open effort funded by the National Institutes of Health to assess model reproducibility in computational knee biomechanics. A regulatory representative from the United States Food and Drug Administration indicated the necessity of standards for reproducibility to increase utility of M&S in the regulatory setting. An industry representative from a major orthopedic implant company emphasized improving reproducibility by addressing indeterminacy in personalized modeling through sensitivity analyses, thereby enhancing preclinical evaluation of joint replacement technology. Thought leaders in the M&S community stressed the importance of data sharing to minimize duplication of efforts. A survey comprised 103 attendees revealed strong support for the workshop and for increasing emphasis on computational modeling at future ORS meetings. Nearly all survey respondents (97%) considered reproducibility to be an important issue. Almost half of respondents (45%) tried and failed to reproduce the work of others. Two-thirds of respondents (67%) declared that individual laboratories are most responsible for ensuring reproducible research whereas 44% thought that journals are most responsible. Thought leaders and survey respondents emphasized that computational models must be reproducible and credible to advance knee M&S.

Polynomial chaos expansion based sensitivity analysis of predicted knee reactions-assessing the influence of the primary ligaments in distraction based models.

Computational knee models have shown that predicted condylar reactions are sensitive to the utilized ligament mechanical parameters. These models, however, are computationally expensive with multiple sources of uncertainty. Traditional uncertainty analysis using Monte-Carlo (MC) inspired methods are costly to perform. The purpose of this study was to use two example calibrated knee models to compare quasi-MC versus polynomial chaos expansion (PCE) sensitivity analyses of predicted condylar reactions that included uncertainty in the mechanical parameters of the ligaments. PCE was practically identical versus quasi-MC with 95% and 98% reductions in model evaluations for analyses with 10 and 6 uncertain variables, respectively.

Assessment of reporting practices and reproducibility potential of a cohort of published studies in computational knee biomechanics.

Reproducible research serves as a pillar of the scientific method and is a foundation for scientific advancement. However, estimates for irreproducibility of preclinical science range from 75% to 90%. The importance of reproducible science has not been assessed in the context of mechanics-based modeling of human joints such as the knee, despite this being an area that has seen dramatic growth. Framed in the context of five experienced teams currently documenting knee modeling procedures, the aim of this study was to evaluate reporting and the perceived potential for reproducibility across studies the teams viewed as important contributions to the literature. A cohort of studies was selected by polling, which resulted in an assessment of nine studies as opposed to a broader analysis across the literature. Using a published checklist for reporting of modeling features, the cohort was evaluated for both "reporting" and their potential to be "reproduced," which was delineated into six major modeling categories and three subcategories. Logistic regression analysis revealed that for individual modeling categories, the proportion of "reported" occurrences ranged from 0.31, 95% confidence interval (CI) [0.23, 0.41] to 0.77, 95% CI: [0.68, 0.86]. The proportion of whether a category was perceived as "reproducible" ranged from 0.22, 95% CI: [0.15, 0.31] to 0.44, 95% CI: [0.35, 0.55]. The relatively low ratios highlight an opportunity to improve reporting and reproducibility of knee modeling studies. Ongoing efforts, including our findings, contribute to a dialogue that facilitates adoption of practices that provide both credibility and translation possibilities.

A Method to Compare Heterogeneous Types of Bone and Cartilage Meshes.

Accurately capturing the bone and cartilage morphology and generating a mesh remains a critical step in the workflow of computational knee joint modeling. Currently, there is no standardized method to compare meshes of different element types and nodal densities, making comparisons across research teams a significant challenge. The aim of this paper is to describe a method to quantify differences in knee joint bone and cartilages meshes, independent of bone and cartilage mesh topology. Bone mesh-to-mesh distances, subchondral bone boundaries, and cartilage thicknesses from meshes of any type of mesh are obtained using a series of steps involving registration, resampling, and radial basis function fitting after which the comparisons are performed. Subchondral bone boundaries and cartilage thicknesses are calculated and visualized in a common frame of reference for comparison. The established method is applied to models developed by five modeling teams. Our approach to obtain bone mesh-to-mesh distances decreased the divergence seen in selecting a reference mesh (i.e., comparing mesh A-to-B versus mesh B-to-A). In general, the bone morphology was similar across teams. The cartilage thicknesses for all models were calculated and the mean absolute cartilage thickness difference was presented, the articulating areas had the best agreement across teams. The teams showed disagreement on the subchondral bone boundaries. The method presented in this paper allows for objective comparisons of bone and cartilage geometry that is agnostic to mesh type and nodal density.

WraptMor: Confirmation of an Approach to Estimate Ligament Fiber Length and Reactions With Knee-Specific Morphology.

Knee ligament length can be used to infer ligament recruitment during functional activities and subject-specific morphology affects the interplay between ligament recruitment and joint motion. This study presents an approach that estimated ligament fiber insertion-to-insertion lengths with wrapping around subject-specific osseous morphology (WraptMor). This represents an advancement over previous work that utilized surrogate geometry to approximate ligament interaction with bone surfaces. Additionally, the reactions each ligament imparted onto bones were calculated by assigning a force-length relationship (kinetic WraptMor model), which assumed that the insertion-to-insertion lengths were independent of the assigned properties. Confirmation of the approach included comparing WraptMor predicted insertion-to-insertion length and reactions with an equivalent displacement-controlled explicit finite element model. Both models evaluated 10 ligament bundles at 16 different joint positions, which were repeated for five different ligament prestrain values for a total of 80 simulations per bundle. The WraptMor and kinetic WraptMor models yielded length and reaction predictions that were similar to the equivalent finite element model. With a few exceptions, predicted ligament lengths and reactions agreed to within 0.1 mm and 2.0 N, respectively, across all tested joint positions and prestrain values. The primary source of discrepancy between the models appeared to be caused by artifacts in the finite element model. The result is a relatively efficient approach to estimate ligament lengths and reactions that include wrapping around knee-specific bone surfaces.

Deciphering the "Art" in Modeling and Simulation of the Knee Joint: Variations in Model Development.

The use of computational modeling to investigate knee joint biomechanics has increased exponentially over the last few decades. Developing computational models is a creative process where decisions have to be made, subject to the modelers' knowledge and previous experiences, resulting in the "art" of modeling. The long-term goal of the KneeHub project is to understand the influence of subjective decisions on the final outcomes and the reproducibility of computational knee joint models. In this paper, we report on the model development phase of this project, investigating model development decisions and deviations from initial modeling plans. Five teams developed computational knee joint models from the same dataset, and we compared each teams' initial uncalibrated models and their model development workflows. Variations in the software tools and modeling approaches were found, resulting in differences such as the representation of the anatomical knee joint structures in the model. The teams consistently defined the boundary conditions and used the same anatomical coordinate system convention. However, deviations in the anatomical landmarks used to define the coordinate systems were present, resulting in a large spread in the kinematic outputs of the uncalibrated models. The reported differences and similarities in model development and simulation presented here illustrate the importance of the "art" of modeling and how subjective decision-making can lead to variation in model outputs. All teams deviated from their initial modeling plans, indicating that model development is a flexible process and difficult to plan in advance, even for experienced teams.

Use of distraction loading to estimate subject-specific knee ligament slack lengths.

Knee ligaments guide and restrain joint motion, and their properties influence joint mechanics. Inverse modeling schemes have been used to estimate specimen-specific ligament properties, where external joint forces are assumed to balance with internal ligament and contact forces. This study simplifies this assumption by adjusting experimental loads to remove internal contact forces. The purpose of this study was to use novel experimental loading in an inverse modeling scheme to estimate ligament slack lengths, perform validation using additional loading scenarios, and evaluate sensitivity to the applied loading. Joint kinematics and kinetics were experimentally measured for a set of load cases. An optimization scheme used a specimen-specific forward kinematics model to estimate ligament slack lengths by minimizing the residual between model and experimentally measured kinetics. The calibrated model was used for a form of validation by evaluating non-optimized load cases. Additionally, uncertainty analysis related kinetic errors to previously reported kinematic errors. The six DOF tibial reactions realized RMS errors less than 23 N and 0.75 Nm for optimized load cases, and 33 N and 2.25 Nm for the non-optimized load cases. The uncertainty analysis, which was performed using the optimized load cases, showed average kinetic RMS errors less than 26 N and 0.45 Nm. The model's recruitment patterns were similar to those found in clinical and cadaveric studies. This study demonstrated that experimental distraction loading can be used in an inverse modeling scheme to estimate ligament slack lengths with a forward kinematics model.

Deciphering the "Art" in Modeling and Simulation of the Knee Joint: Overall Strategy.

Recent explorations of knee biomechanics have benefited from computational modeling, specifically leveraging advancements in finite element analysis and rigid body dynamics of joint and tissue mechanics. A large number of models have emerged with different levels of fidelity in anatomical and mechanical representation. Adapted modeling and simulation processes vary widely, based on justifiable choices in relation to anticipated use of the model. However, there are situations where modelers' decisions seem to be subjective, arbitrary, and difficult to rationalize. Regardless of the basis, these decisions form the "art" of modeling, which impact the conclusions of simulation-based studies on knee function. These decisions may also hinder the reproducibility of models and simulations, impeding their broader use in areas such as clinical decision making and personalized medicine. This document summarizes an ongoing project that aims to capture the modeling and simulation workflow in its entirety-operation procedures, deviations, models, by-products of modeling, simulation results, and comparative evaluations of case studies and applications. The ultimate goal of the project is to delineate the art of a cohort of knee modeling teams through a publicly accessible, transparent approach and begin to unravel the complex array of factors that may lead to a lack of reproducibility. This manuscript outlines our approach along with progress made so far. Potential implications on reproducibility, on science, engineering, and training of modeling and simulation, on modeling standards, and on regulatory affairs are also noted.

The potential for intercellular mechanical interaction: simulations of single chondrocyte versus anatomically based distribution.

Computational studies of chondrocyte mechanics, and cell mechanics in general, have typically been performed using single cell models embedded in an extracellular matrix construct. The assumption of a single cell microstructural model may not capture intercellular interactions or accurately reflect the macroscale mechanics of cartilage when higher cell concentrations are considered, as may be the case in many instances. Hence, the goal of this study was to compare cell-level response of single and eleven cell biphasic finite element models, where the latter provided an anatomically based cellular distribution representative of the actual number of cells for a commonly used [Formula: see text] edge cubic representative volume in the middle zone of cartilage. Single cell representations incorporated a centered single cell model and eleven location-corrected single cell models, the latter to delineate the role of cell placement in the representative volume element. A stress relaxation test at 10% compressive strain was adopted for all simulations. During transient response, volume- averaged chondrocyte mechanics demonstrated marked differences (up to 60% and typically greater than 10%) for the centered single versus the eleven cell models, yet steady-state loading was similar. Cell location played a marked role, due to inhomogeneity of the displacement and fluid pressure fields at the macroscopic scale. When the single cell representation was corrected for cell location, the transient response was consistent, while steady-state differences on the order of 1-4% were realized, which may be attributed to intercellular mechanical interactions. Anatomical representations of the superficial and deep zones, where cells reside in close proximity, may exhibit greater intercellular interactions, but these have yet to be explored.

A general framework for application of prestrain to computational models of biological materials.

It is often important to include prestress in computational models of biological tissues. The prestress can represent residual stresses (stresses that exist after the tissue is excised from the body) or in situ stresses (stresses that exist in vivo, in the absence of loading). A prestressed reference configuration may also be needed when modeling the reference geometry of biological tissues in vivo. This research developed a general framework for representing prestress in finite element models of biological materials. It is assumed that the material is elastic, allowing the prestress to be represented via a prestrain. For prestrain fields that are not compatible with the reference geometry, the computational framework provides an iterative algorithm for updating the prestrain until equilibrium is satisfied. The iterative framework allows for enforcement of two different constraints: elimination of distortion in order to address the incompatibility issue, and enforcing a specified in situ fiber strain field while allowing for distortion. The framework was implemented as a plugin in FEBio (www.febio.org), making it easy to maintain the software and to extend the framework if needed. Several examples illustrate the application and effectiveness of the approach, including the application of in situ strains to ligaments in the Open Knee model (simtk.org/home/openknee). A novel method for recovering the stress-free configuration from the prestrain deformation gradient is also presented. This general purpose theoretical and computational framework for applying prestrain will allow analysts to overcome the challenges in modeling this important aspect of biological tissue mechanics.

Commentary on the integration of model sharing and reproducibility analysis to scholarly publishing workflow in computational biomechanics.

OBJECTIVE: The overall goal of this paper is to demonstrate that dissemination of models and analyses for assessing the reproducibility of simulation results can be incorporated in the scientific review process in biomechanics. METHODS: As part of a special issue on model sharing and reproducibility in the IEEE Transactions on Biomedical Engineering, two manuscripts on computational biomechanics were submitted: Rajagopal et al., IEEE Trans. Biomed. Eng., 2016 and Schmitz and Piovesan, IEEE Trans. Biomed. Eng., 2016. Models used in these studies were shared with the scientific reviewers and the public. In addition to the standard review of the manuscripts, the reviewers downloaded the models and performed simulations that reproduced results reported in the studies. RESULTS: There was general agreement between simulation results of the authors and those of the reviewers. Discrepancies were resolved during the necessary revisions. The manuscripts and instructions for download and simulation were updated in response to the reviewers' feedback; changes that may otherwise have been missed if explicit model sharing and simulation reproducibility analysis was not conducted in the review process. Increased burden on the authors and the reviewers, to facilitate model sharing and to repeat simulations, were noted. CONCLUSION: When the authors of computational biomechanics studies provide access to models and data, the scientific reviewers can download and thoroughly explore the model, perform simulations, and evaluate simulation reproducibility beyond the traditional manuscript-only review process. SIGNIFICANCE: Model sharing and reproducibility analysis in scholarly publishing will result in a more rigorous review process, which will enhance the quality of modeling and simulation studies and inform future users of computational models.

Evaluation of a post-processing approach for multiscale analysis of biphasic mechanics of chondrocytes.

Understanding the mechanical behaviour of chondrocytes as a result of cartilage tissue mechanics has significant implications for both evaluation of mechanobiological function and to elaborate on damage mechanisms. A common procedure for prediction of chondrocyte mechanics (and of cell mechanics in general) relies on a computational post-processing approach where tissue-level deformations drive cell-level models. Potential loss of information in this numerical coupling approach may cause erroneous cellular-scale results, particularly during multiphysics analysis of cartilage. The goal of this study was to evaluate the capacity of first- and second-order data passing to predict chondrocyte mechanics by analysing cartilage deformations obtained for varying complexity of loading scenarios. A tissue-scale model with a sub-region incorporating representation of chondron size and distribution served as control. The post-processing approach first required solution of a homogeneous tissue-level model, results of which were used to drive a separate cell-level model (same characteristics as the sub-region of control model). The first-order data passing appeared to be adequate for simplified loading of the cartilage and for a subset of cell deformation metrics, for example, change in aspect ratio. The second-order data passing scheme was more accurate, particularly when asymmetric permeability of the tissue boundaries was considered. Yet, the method exhibited limitations for predictions of instantaneous metrics related to the fluid phase, for example, mass exchange rate. Nonetheless, employing higher order data exchange schemes may be necessary to understand the biphasic mechanics of cells under lifelike tissue loading states for the whole time history of the simulation.

A three-dimensional inverse finite element analysis of the heel pad.

Quantification of plantar tissue behavior of the heel pad is essential in developing computational models for predictive analysis of preventive treatment options such as footwear for patients with diabetes. Simulation based studies in the past have generally adopted heel pad properties from the literature, in return using heel-specific geometry with material properties of a different heel. In exceptional cases, patient-specific material characterization was performed with simplified two-dimensional models, without further evaluation of a heel-specific response under different loading conditions. The aim of this study was to conduct an inverse finite element analysis of the heel in order to calculate heel-specific material properties in situ. Multidimensional experimental data available from a previous cadaver study by Erdemir et al. ("An Elaborate Data Set Characterizing the Mechanical Response of the Foot," ASME J. Biomech. Eng., 131(9), pp. 094502) was used for model development, optimization, and evaluation of material properties. A specimen-specific three-dimensional finite element representation was developed. Heel pad material properties were determined using inverse finite element analysis by fitting the model behavior to the experimental data. Compression dominant loading, applied using a spherical indenter, was used for optimization of the material properties. The optimized material properties were evaluated through simulations representative of a combined loading scenario (compression and anterior-posterior shear) with a spherical indenter and also of a compression dominant loading applied using an elevated platform. Optimized heel pad material coefficients were 0.001084 MPa (µ), 9.780 (α) (with an effective Poisson's ratio (ν) of 0.475), for a first-order nearly incompressible Ogden material model. The model predicted structural response of the heel pad was in good agreement for both the optimization (<1.05% maximum tool force, 0.9% maximum tool displacement) and validation cases (6.5% maximum tool force, 15% maximum tool displacement). The inverse analysis successfully predicted the material properties for the given specimen-specific heel pad using the experimental data for the specimen. The modeling framework and results can be used for accurate predictions of the three-dimensional interaction of the heel pad with its surroundings.

Adaptive surrogate modeling for expedited estimation of nonlinear tissue properties through inverse finite element analysis.

Simulation-based prediction of specimen-specific biomechanical behavior commonly requires inverse analysis using geometrically consistent finite element (FE) models. Optimization drives such analyses but previous studies have highlighted a large computational cost dictated by iterative use of nonlinear FE models. The goal of this study was to evaluate the performance of a local regression-based adaptive surrogate modeling approach to decrease computational cost for both global and local optimization approaches using an inverse FE application. Nonlinear elastic material parameters for patient-specific heel-pad tissue were found, both with and without the surrogate model. Surrogate prediction replaced a FE simulation using local regression of previous simulations when the corresponding error estimate was less than a given tolerance. Performance depended on optimization type and tolerance value. The surrogate reduced local optimization expense up to 68%, but achieved accurate results for only 1 of 20 initial conditions. Conversely, up to a tolerance value of 20 N(2), global optimization with the surrogate yielded consistent parameter predictions with a concurrent decrease in computational cost (up to 77%). However, the local optimization method without the surrogate, although sensitive to the initial conditions, was still on average seven times faster than the global approach. Our results help establish guidelines for setting acceptable tolerance values while using an adaptive surrogate model for inverse FE analysis. Most important, the study demonstrates the benefits of a surrogate modeling approach for intensive FE-based iterative analysis.

Verification of predicted knee replacement kinematics during simulated gait in the Kansas knee simulator.

Evaluating total knee replacement kinematics and contact pressure distributions is an important element of preclinical assessment of implant designs. Although physical testing is essential in the evaluation process, validated computational models can augment these experiments and efficiently evaluate perturbations of the design or surgical variables. The objective of the present study was to perform an initial kinematic verification of a dynamic finite element model of the Kansas knee simulator by comparing predicted tibio- and patellofemoral kinematics with experimental measurements during force-controlled gait simulation. A current semiconstrained, cruciate-retaining, fixed-bearing implant mounted in aluminum fixtures was utilized. An explicit finite element model of the simulator was developed from measured physical properties of the machine, and loading conditions were created from the measured experimental feedback data. The explicit finite element model allows both rigid body and fully deformable solutions to be chosen based on the application of interest. Six degrees-of-freedom kinematics were compared for both tibio- and patellofemoral joints during gait loading, with an average root mean square (rms) translational error of 1.1 mm and rotational rms error of 1.3 deg. Model sensitivity to interface friction and damping present in the experimental joints was also evaluated and served as a secondary goal of this paper. Modifying the metal-polyethylene coefficient of friction from 0.1 to 0.01 varied the patellar flexion-extension and tibiofemoral anterior-posterior predictions by 7 deg and 2 mm, respectively, while other kinematic outputs were largely insensitive.

Concurrent musculoskeletal dynamics and finite element analysis predicts altered gait patterns to reduce foot tissue loading.

Current computational methods for simulating locomotion have primarily used muscle-driven multibody dynamics, in which neuromuscular control is optimized. Such simulations generally represent joints and soft tissue as simple kinematic or elastic elements for computational efficiency. These assumptions limit application in studies such as ligament injury or osteoarthritis, where local tissue loading must be predicted. Conversely, tissue can be simulated using the finite element method with assumed or measured boundary conditions, but this does not represent the effects of whole body dynamics and neuromuscular control. Coupling the two domains would overcome these limitations and allow prediction of movement strategies guided by tissue stresses. Here we demonstrate this concept in a gait simulation where a musculoskeletal model is coupled to a finite element representation of the foot. Predictive simulations incorporated peak plantar tissue deformation into the objective of the movement optimization, as well as terms to track normative gait data and minimize fatigue. Two optimizations were performed, first without the strain minimization term and second with the term. Convergence to realistic gait patterns was achieved with the second optimization realizing a 44% reduction in peak tissue strain energy density. The study demonstrated that it is possible to alter computationally predicted neuromuscular control to minimize tissue strain while including desired kinematic and muscular behavior. Future work should include experimental validation before application of the methodology to patient care.

An elaborate data set characterizing the mechanical response of the foot.

Mechanical properties of the foot are responsible for its normal function and play a role in various clinical problems. Specifically, we are interested in quantification of foot mechanical properties to assist the development of computational models for movement analysis and detailed simulations of tissue deformation. Current available data are specific to a foot region and the loading scenarios are limited to a single direction. A data set that incorporates regional response, to quantify individual function of foot components, as well as the overall response, to illustrate their combined operation, does not exist. Furthermore, the combined three-dimensional loading scenarios while measuring the complete three-dimensional deformation response are lacking. When combined with an anatomical image data set, development of anatomically realistic and mechanically validated models becomes possible. Therefore, the goal of this study was to record and disseminate the mechanical response of a foot specimen, supported by imaging data. Robotic testing was conducted at the rear foot, forefoot, metatarsal heads, and the foot as a whole. Complex foot deformations were induced by single mode loading, e.g., compression, and combined loading, e.g., compression and shear. Small and large indenters were used for heel and metatarsal head loading, an elevated platform was utilized to isolate the rear foot and forefoot, and a full platform compressed the whole foot. Three-dimensional tool movements and reaction loads were recorded simultaneously. Computed tomography scans of the same specimen were collected for anatomical reconstruction a priori. The three-dimensional mechanical response of the specimen was nonlinear and viscoelastic. A low stiffness region was observed starting with contact between the tool and foot regions, increasing with loading. Loading and unloading responses portrayed hysteresis. Loading range ensured capturing the toe and linear regions of the load deformation curves for the dominant loading direction, with the rates approximating those of walking. A large data set was successfully obtained to characterize the overall and the regional mechanical responses of an intact foot specimen under single and combined loads. Medical imaging complemented the mechanical testing data to establish the potential relationship between the anatomical architecture and mechanical responses and to further develop foot models that are mechanically realistic and anatomically consistent. This combined data set has been documented and disseminated in the public domain to promote future development in foot biomechanics.

Adaptive surrogate modeling for efficient coupling of musculoskeletal control and tissue deformation models.

Finite element (FE) modeling and multibody dynamics have traditionally been applied separately to the domains of tissue mechanics and musculoskeletal movements, respectively. Simultaneous simulation of both domains is needed when interactions between tissue and movement are of interest, but this has remained largely impractical due to the high computational cost. Here we present a method for the concurrent simulation of tissue and movement, in which state of the art methods are used in each domain, and communication occurs via a surrogate modeling system based on locally weighted regression. The surrogate model only performs FE simulations when regression from previous results is not within a user-specified tolerance. For proof of concept and to illustrate feasibility, the methods were demonstrated on an optimization of jumping movement using a planar musculoskeletal model coupled to a FE model of the foot. To test the relative accuracy of the surrogate model outputs against those of the FE model, a single forward dynamics simulation was performed with FE calls at every integration step and compared with a corresponding simulation with the surrogate model included. Neural excitations obtained from the jump height optimization were used for this purpose and root mean square (RMS) difference between surrogate and FE model outputs (ankle force and moment, peak contact pressure and peak von Mises stress) were calculated. Optimization of the jump height required 1800 iterations of the movement simulation, each requiring thousands of time steps. The surrogate modeling system only used the FE model in 5% of time steps, i.e., a 95% reduction in computation time. Errors introduced by the surrogate model were less than 1 mm in jump height and RMS errors of less than 2 N in ground reaction force, 0.25 Nm in ankle moment, and 10 kPa in peak tissue stress. Adaptive surrogate modeling based on local regression allows efficient concurrent simulations of tissue mechanics and musculoskeletal movement.