Journal of Biomechanics 158 (2023) 111766

Contents lists available at ScienceDirect

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

A combined experimental and finite element analysis of the human elbow

under loads of daily living
Stephanie L. Kahmann a, b, *, Amelie Sas b, Nils Große Hokamp c, G. Harry van Lenthe b,
Lars-Peter Müller a, Kilian Wegmann a, d
a
University of Cologne, Faculty of Medicine and University Hospital Cologne, Center for Orthopedic and Trauma Surgery, Kerpener Str. 62, Cologne 50937, Germany
b
Biomechanics Section, Dept. of Mechanical Engineering, KU Leuven, Belgium
c
University of Cologne, Faculty of Medicine and University Hospital Cologne, Institute for Diagnostic and Interventional Radiology, Germany
d
OCM München, Steinerstr. 6, 81369, München, Deutschland

A R T I C L E I N F O A B S T R A C T

Keywords: Elbow trauma is often accompanied by a loss of independence in daily self-care activities, negatively affecting
In-silico model patients’ quality of life. Finite element models can help gaining profound knowledge about native human joint
Human elbow mechanics, which is crucial to adequately restore joint functionality after severe injuries. Therefore, a finite
Upper extremity
element model of the elbow is required that includes both the radio-capitellar and ulno-trochlear joint and is
Finite element analysis
Cadaveric biomechanical testing
subjected to loads realistic for activities of daily living. Since no such model has been published, we aim to fill
this gap.
For comparison, 8 intact cadaveric elbows were subjected to loads of up to 1000 N, after they were placed in
an extended position. At each load step, the displacement of the proximal humerus relative to the distal base
plate was measured with optical tracking markers and the joint pressure was measured with a pressure mapping
sensor.
Analogously, eight finite element models were created based on subject-specific CT scans of the corresponding
elbow specimens. The CT scans were registered to the positions of tantalum beads in the experiment. The
optically measured displacements were applied as boundary conditions.
We demonstrated that the workflow can predict the experimental contact pressure distribution with a mod­
erate correlation, the experimental peak pressures in the correct joints and the experimental stiffness with
moderate to excellent correlation. The predictions of peak pressure magnitude, contact area and load share on
the radius require improvement by precise representation of the cartilage geometry and soft tissues in the model,
and proper initial contact in the experiment.

1. Introduction (Anderson et al., 2008; Li et al., 2021; Osawa et al., 2014) and knee
research (Haut Donahue et al., 2002; Mononen et al., 2013; Peña et al.,
The elbow joint provides crucial mobility and functionality for ac­ 2006); however, the elbow is not yet well represented.
tivities of daily living (Fornalski et al., 2003). Severe injuries, such as the To be able to draw clinically meaningful results from finite element
terrible triad lesion of the elbow, can cause stiffness and severe pain models of the elbow, these must be tested under realistic conditions,
(Bozon et al., 2022; Hotchkiss, 1996; Lindenhovius et al., 2008; Papa­ including realistic loads and adequate representation of all relevant
theodorou et al., 2014), which negatively impact the patients’ inde­ structures. Common loads in the ulno-trochlear and radio-capitellar
pendence and dignity immediately. Therefore, profound knowledge of joints range up to 1600 N and 800 N during occupational activities
the intact state may help improve the treatment (Bryce and Armstrong, (Chadwick and Nicol, 2000). The elbow joint is regularly subjected to
2008) and finite element models (FEM) are an established tool to forces between 0.5 and 3 times body weight in activities of daily living,
improve understanding of joint biomechanics (Stylianou, 2021). This ranging up to 2 times body weight (An et al., 1984; Safran et al., 2005).
computational approach has been applied to a great extent in the hip According to the literature, 54–67% of the pressure is transmitted

* Corresponding author.
E-mail address: [email protected] (S.L. Kahmann).

https://doi.org/10.1016/j.jbiomech.2023.111766
Accepted 11 August 2023
Available online 12 August 2023
0021-9290/© 2023 Elsevier Ltd. All rights reserved.
S.L. Kahmann et al. Journal of Biomechanics 158 (2023) 111766

through the radius in elbow extension (Hackl et al., 2017; Halls and the overall pressure distribution, being relevant parameters for correct
Travill, 1964; Hwang et al., 2018; Smithson et al., 2020). The distribu­ prosthesis design, and the stiffness as a parameter for the deformation
tion of load transfer directly impacts the pressure distribution within the occurring in the joint. Furthermore, we compare the contact areas, peak
joint, which is an important evaluation criterion in the development of pressure magnitudes and percentage of load share on the radius.
prosthesis and treatment options (Langohr et al., 2015).
Thus far, published models of the elbow have been validated up to 2. Materials and method
200 N (Kim and Miller, 2016). In view of the actual loads, this is very
low. Furthermore, this model neglected a detailed description of the 2.1. Preparation
ulna. Willing et al. have presented an elbow model including both
forearm bones (Willing et al., 2013). Their study focused on the cartilage Eight cadaveric arms from 4 donors (1 female and 3 male), with an
contact mechanics, and, for their purpose, 80 N of load sufficed. How­ average age of 78 (range 62–89) years were used. Ethical approval was
ever, this range does not fulfill the wide range of loads in the activities of granted under the number VT 21-1514_1 prior to the study.
daily living. Before in-vitro testing, the humerus was dissected 12 cm proximal of
Hence, among the published elbow models, we identified the lack of the joint line and the forearm bones 7 cm distal of it, which were the
a model that includes both the radio-capitellar and the ulno-trochlear highest common lengths of all samples. The specimens were potted into
joint, and that is subjected to realistic loads. custom-made aluminum cylinders with PMMA (Technovit, Kulzer
We aim to fill this gap with a subject-specific finite element workflow GmbH, Wehrheim, Germany) in full extension and 60◦ pronation. The
with parameters that are optimized for one specimen and subsequently height of the fixation was marked at 3 cm. Each specimen was equipped
tested in a test-cohort of seven specimens of the native elbow joint. We with 6 tantalum beads (0.8 mm, x-medics, Frederiksberg, Denmark) of
compare the experimental and simulated locations of peak pressure and which two were inserted in the proximal shaft, one in the medial

Fig. 1. The set-up of the simulation (a) and experiment (b) are analogous to each other. The proximal fixation of the simulation is derived from the resin height in the
aluminum pot and its displacement, measured with the position markers, is applied as boundary condition to the simulation. In both cases, the humerus presses
axially on the pressure measurement: measured in the cartilage (a) and the pressure mapping sensor (b). In the simulation (a), the ulna and radius are well visible
whereas the soft tissue and the pressure mapping sensor obstruct the view in (b). The distal fixation of the simulation (a) corresponds in height and degrees of
freedom to the resin, fixated to the bottom table with screws (b).

2
S.L. Kahmann et al. Journal of Biomechanics 158 (2023) 111766

epicondyle of the humerus, one in the proximal end of the olecranon and markers, allowing to monitor the stiffness, defined as force over
two in the radial shaft. Then, cross sectional computed tomography (CT) displacement. The whole protocol was repeated three times per
images of the specimens were obtained in the anatomic position of the specimen.
experiment, using a clinical CT system with a modified protocol (Philips
IQon, Philips Healthcare, Best, The Netherlands). Helical CT image
2.3. Simulation
acquisition was performed using 120 kVp, 250 mAs and a pitch of 0.235.
Images were reconstructed using a hybrid-iterative reconstruction al­
All CT scans were segmented semi-automatically using the same
gorithm with a sharp kernel (iDose4 and C, Philips Healthcare, Best, The
scripted Mimics functions (smoothing with a smoothing factor of 0.4 for
Netherlands). Images were reconstructed in a slice thickness of 0.67 mm
three iterations, eroding and dilating with the size of 1 pixel in relation
without increment. Given a matrix of 1024 × 1024 and the field-of-view
to the 26 connected pixels, wrapping with the default values (detail
of 350 × 350 mm2, in plane resolution was 0.34 × 0.34 mm2.
0.3906 and gap 0.195313), and then manual correction of intersections,
A pressure mapping sensor (PMS, Model 5040; Tekscan, Inc., Boston,
Mimics Version 22.0, Materialise, Leuven, Belgium). For correct align­
MA) with spatial and temporal resolutions of 1.03 mm2 and 10 Hz was
ment, the geometries of the bones were registered to their respective
preconditioned and calibrated under experiment-like conditions. The
positions in the experimental set-up by transforming the positions of the
PMS was placed between two silicon layers, lubricated with machine oil,
tantalum beads from the CT scan to the experimentally measured posi­
on a flat table and a stamp, both made of steel, in a material testing
tion in a CAD Software (3-matic Version 12.0, Materialise, Leuven,
machine (error ± 1 N, Modell Z010, Zwick Roell, Ulm, Germany) to
Belgium). This way, the coordinate systems from the measurement in
mimic the cartilage covered bone. During the calibration and the
the experiments matched those in the simulations. A preliminary study
experiment, the PMS was protected from the moist environment by a
demonstrated the necessity of this measure to reflect the joint alignment
Tegaderm Film (3 M Deutschland GmbH, Neuss, Germany) (Jansson
correctly (1◦ axial rotation of the humerus changed the peak pressure by
et al., 2013; Renani et al., 2017). The conditioning loads ranged up to
2%).
110% of the highest expected pressure (1.27 MPa), estimated based on a
Once the bones were aligned, the cartilage geometries on the hu­
previous study (Hackl et al., 2017). A custom-adapted second order
merus, ulna and radius were created based on the joint surfaces. The
polynomial calibration curve was found to show the lowest error (RMSE
constant cartilage thickness was manually increased up to contact with
0.05 MPa), as suggested (Brimacombe et al., 2009). Several parameters
the opposing cartilages and ranged from 0.7 to 1.1 mm among all
were based on preliminary studies in one specimen and the best
specimens.
matching set was applied to the remaining 7 specimens in the compar­
In the next step, the mesh size was assigned. In a mesh sensitivity
ison cohort.
study, an edge length of 1 mm in the region of interest covering 25 mm
proximal and distal of the joint lines proved to be the appropriate bal­
2.2. Experiment ance of precision and runtime (<6 % deviation in peak contact pressure
compared to 0.9 and 1.1 mm edge lengths). The meshes outside the
The potted specimens were fixed in the material testing machine region of interest had an average edge length of 3 mm. Regular tetra­
(Z010, Zwick Roell, Ulm, Germany) to the bottom table with screws and hedral elements (C3D10) were used, due to bending in a preliminary
press fitted to the top pot, which was itself rigidly attached to the study and for accurate peak pressures in penalty-ruled contact (Abaqus
actuator (Fig. 1b). The specimens were placed in full extension and 60◦ Version 2017, Simulia, Dassault Systèmes, Vélizy-Villacoublay, France)
pronation. This position was chosen as a well-controlled set-up. The skin (ABAQUS, 2013).
was incised axially at the ventral center of the elbow over 7 cm. Further Young’s Moduli were assigned dependent on the density, according
blunt dissection, moving the muscles to the sides, giving unobstructed to the linear elastic portion of the Keyak bone fracture model (Keyak
view on the ventral capsule, which was cut transversally along the joint et al., 2005). Therefore, the density was discretized in 10 bins for each
gap, allowed for insertion of the PMS. Through two dorsal puncture the spongious and the cortical bone (Pellikaan et al., 2018). More than
incisions medially and laterally of the olecranon, two clamps grabbed 99% of all elements were assigned a Young’s Modulus of 2–16,389 MPa.
the PMS and prevented it from shifting during force application. Thus, it The cartilage material properties were based on a preliminary analysis.
was ensured that the PMS covered the whole joint. In the resins around The Young’s modulus was iteratively optimized approximating the
the humerus and the forearm, two position markers (Smart Cluster experimentally measured stiffness with a fixed Poisson’s ratio of 0.4
Marker, Optotrak, Northern Digital Inc., Waterloo, Ontario, Canada) (Kim and Miller, 2016). Within the tested range of 4 to 16 MPa, the 6
were mounted. The position markers measured the vertical displace­ MPa case was found to be the best match. It was compared to Neo-
ments with a maximum error of 0.005 mm. Hookean material properties (Willing et al., 2013) in the exemplary
Under a preload of 10 N, the positions of the tantalum beads were load cases of 100 N and 1000 N. This property had only little effect on
measured with an optical tracking device (Four Marker Rigid Body, the stiffness and pressure distribution at a cost of longer computation
Optotrak Certus, Northern Digital Inc., Waterloo, Ontario, Canada). time in the case of 100 N, which is in line with the finding of (Kim and
These positions were the basis for the fine registration in the simulation Miller, 2016), and did not converge at 1000 N. Therefore, the linear
and were measured with an accuracy of 0.03 mm. elastic model was chosen.
The humeri were loaded axially with a rate of 0.6 mm/min. A load The cartilages were tied to their respective bones, and frictionless
cycling 30 times between 10 and 100 N served as the preconditioning. tangential contact was assigned within the joint (Renani et al., 2017). In
This was followed by the test protocol, starting at 10 N, increasing to normal direction, the penalty stiffness increased linearly from Kinitial =
100 N and from there in 100 N increments up to 1000 N. Subsequently, Eelement at an overclosure e = 0.01g0 up to Kfinal = 100*Eelement at d =
the load was decreased in the same manner back to 10 N. At each step of 0.03g0 with g0 being the faceted element length (Willing et al., 2013).
the test protocol, the position was held for 60 s. Meanwhile, the pressure The distal ends of both forearm bones were fixated in all six degrees
and the displacements were recorded with the PMS and the position of freedom up to the individual height of the resin, derived from the CT

3
S.L. Kahmann et al. Journal of Biomechanics 158 (2023) 111766

scans (Fig. 1a). Similarly, the proximal end of the humerus was
restricted in transverse translation and rotations, only allowed to
displace in axial direction. These displacements were governed by the
subject-specifically measured displacements of the experiments using
the optical tracking markers. This was found to be more sensitive than
applying the load in a parameter sensitivity study: changing the assigned
displacement by 0.1 mm (experimental error ± 0.005 mm), the force
varied by around 70 N. When changing the applied load by 1 N
(experimental error ± 1 N), no change in displacement was detectable.
Therefore, the less sensitive displacement was chosen as input and the
more sensitive reaction force as comparison parameter. The model
stiffness is determined by the resulting forces over the applied
displacements.

2.4. Data processing and statistical evaluation

The experimental and simulated contact pressures were compared in

a full-field approach. Therefore, the nodal joint contact pressure of the
finite element analysis and moving average (3 × 3 matrix) filtered
pressure mapping data were registered. That was achieved by projecting
the 3-dimensional finite element geometry of the contact area onto the
plane of the pressure mapping sensor and normalizing both datasets to
an area of the dimensions [0 1], both in the medial–lateral and ventral-
dorsal directions. Using a published validation tool (Abel et al., 2020),
the higher resolved simulation output was averaged around the lower Fig. 2. A subject-specific correlation of the simulated over the measured full-
resolved pressure mapping output. field pressure distribution with 137 local measurements is shown as an
For the comparison of the peak pressure locations, the computational example. The finite element nodal output was averaged around the pressure
and experimental coordinates were registered and normalized in the mapping pixel output. Perfect agreement would be represented by a slope of 1,
an intercept of 0 and a correlation R2 of 1.
same way as for the full-field pressure comparison. But, instead of
comparing the pressure values, the coordinates of the respective peak
pressures were extracted and with them the identification of the contact
joint (radial or ulnar). contact area, the model correlated poorly to moderately with the
The experimentally measured displacements were partially affected experiment (Table 1).
by a non-congruency of the joints at lower loads, apparent in a contact Furthermore, it was predicted correctly for all 7 comparison speci­
area that was unphysiologically low (on average below 50% at the mens, in which joint the peak pressure occurred (Fig. 4). In 6 cases it was
radius). This led to non-linear stiffness in the simulation. To prevent this detected in the radio-humeral joint, one in the humero-ulnar joint. The
error from propagating, we linearized the force–displacement diagram average magnitude of this peak pressure among all specimens was 10.1
based on the load cases higher than 400 N, extrapolated the missing data ± 3.3 MPa (mean ± standard deviation) in the experiments and 13.7 ±
and shifted the inception to the point of origin. In cases 2 and 6, no 7.8 MPa in the FEM (Fig. 5). The predicted and measured peak pressures
contact was detected in the humero-radial joint up to 300 N and 400 N, of both joint contacts separately at 1000 N correlated with Lin’s CCC of
respectively, which strongly impacted the corresponding displacements. 0.333 ± 0.397 (Table 1). The experimental peak pressures before and
Therefore, these values were excluded from evaluation. after the waiting period of 60 s differed <0.04 MPa; the repetition
Statistical evaluation was performed using SPSS (IBM SPSS Statistics showed an average standard deviation of 0.1 MPa with no apparent
Version 28, IBM, Armonk, NY, USA). All tests for normal distribution trend. Further, we found that neither the ascending nor the descending
were carried out using the Shapiro-Wilk test. Correlations over all order of load cases affected the measured parameters.
samples were tested using Lin’s concordance correlation coefficient Apart from the pressure distribution, also the model stiffness was
(Lin’s CCC) and subject-specific correlations using Spearman’s rho. The evaluated. We did that by comparing the total contact forces resulting
significance level for all tests was defined at p < 0.05. The 95 % confi­ from the subject-specifically assigned displacements (Fig. 7), which
dence interval (CI) of the simulated force (Fig. 6) was calculated as a deviated from the experimental ones by 3–9 % up to 700 N and by 10–12
first-order normal approximation (Puth et al., 2015), showing the range, % from 800 to 1000 N (Fig. 6). The subject-specific stiffness correlated
in which a population value would lie with a probability of 95 %, based on average moderately with Lin’s CCC of 0.82 (range 0.65–1.00, 4
on the samples included in this study. moderately, 3 excellent, Cohen, 1977).
When evaluating the force share among the two forearm bones, we
3. Results found that on average, in the experiment 57.2 ± 18.2% of the force were
transferred via the radial column, the simulation predicted 51.7 ±
The computational and experimental pressure distributions of each 15.7% at 1000 N (Fig. 8, Table 1).
subject correlated with an average Spearman’s rho of 0.496 (range In general, no particular correlation between contralateral speci­
0.260–0.888, p < 0.01) (Fig. 2). In both distributions of Fig. 3, the mens was apparent in the experimental parameters or results, therefore
medial side of the radius and the medial and lateral side of the ulna any correlation in the simulations would be incidental and was not
showed high pressure with elliptical contact areas. Regarding the tested.

4
S.L. Kahmann et al. Journal of Biomechanics 158 (2023) 111766

Fig. 3. A comparison of the pressure distributions from simulation (top) and experiment (bottom), showing a similar distribution in a right arm as an example. The
blue dots of the simulation indicate unloaded nodes of the cartilages, similar to the blue background of the pressure mapping. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)

5
S.L. Kahmann et al. Journal of Biomechanics 158 (2023) 111766

Table 1
Parameters are evaluated at 1000 N load, here shown as mean (±standard deviation). Lin’s correlation concordance coefficient (Lins’s CCC ± 95 % confidence in­
terval) indicates the correlation between the experimental and the simulated results.
Parameter Experiment FEA

Peak pressure in MPa

Mean 10.1 (±3.3) 13.7 (±7.8)
Lin’s CCC 0.333 (±0.397)

Load share in % on radius

Mean 57.2 (±18.2) 51.7 (±15.7)
Lin’s CCC 0.655 (±0.467)

Contact area in mm2

Mean 395.4 (±68.8) 225.4 (±63.4)
Lin’s CCC 0.152 (±0.198)

Fig. 4. The peak pressure locations both predicted (yellow) and measured (blue), indicated by crosses and the subject-specifically corresponding ones connected with
a line, projected into a right sample. All peak pressures were found in the correct joint: 6 in the radius and one in the ulna. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5. The peak pressure magnitudes of the simulation are shown over the experimental ones. The averaged correlation is shown in a dotted line whereas the
optimal line, with a slope of 1 and an R2 of 1, is shown in a solid line.

6
S.L. Kahmann et al. Journal of Biomechanics 158 (2023) 111766

Fig. 6. The contact forces of the simulation are guided by the displacements measured in the subject-specific experiment. Shown here is the contact force of the
simulation over the experimentally applied load subject-specifically for the seven test-specimens, their 95 % confidence interval (blue area), their average (short-
dotted line) and the optimal correlation with a slope of 1 (long-dotted line). (For interpretation of the references to colour in this figure legend, the reader is referred
to the web version of this article.)

Fig. 7. Experimentally measured loads over displacements of the seven test-specimens. The seven specimens show different stiffness and non-linear behavior,
especially for lower loads.

7
S.L. Kahmann et al. Journal of Biomechanics 158 (2023) 111766

Fig. 8. The average share of the total load (mean ± standard deviation) transmitted by the radius is below 50 % in the experimental (blue) load cases up to 400 N.
The simulated (yellow) share is overpredicted in the loads up to 500 N and underpredicted over 600 N. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)

4. Discussion magnitudes only correlated weakly, which might be due to the manually
adjusted constant cartilage thickness up to contact. That is also why our
In this study, we developed a finite element workflow for the elbow, model was not able to predict the missing contact in two cases for lower
which includes both the radio-capitellar and the ulno-trochlear joint, loads - a discrepancy to the experiments. Since our CT parameters were
that ranges up to realistic loads and that is compared to cadaver tests. focused on the bone, the cartilage geometry could not be extracted.
The experimental and simulated pressure distributions correlate Therefore, we assigned a constant cartilage thickness like (Kim and
moderately to strongly in 6 out of 7 cases, in one weakly, the stiffness Miller, 2016). When varying the thickness from 1.1 to 0.9 mm, the
moderately to excellent (Cohen, 1977). Further, the peak pressures were maximum contact stress changed by 7.5%. In our study, the cartilage
predicted in the correct joint in all specimens: 6 in the radio-capitellar thickness was adapted to the joint gap after registration of the digital to
and one in the ulno-trochlear joint. This outcome was expected when the experimental bone positions and ranged from 0.7 to 1.1 mm among
considering the peak pressure an indicator for risk of fracture and radial the different specimens. We could not detect a correlation or a trend
head fractures being the most common fracture of the elbow (Kaas et al., between cartilage thickness and peak pressure. Future studies should
2008; Pike et al., 2009), more common than coronoid fractures. This overcome that inaccuracy with optimized CT parameters, as studies
finding might suggest that the radius possibly fractures first and only showed that the cartilage geometry does impact contact mechanics
with sufficient remaining impact energy, the coronoid fractures second. ((Anderson et al., 2008); Willing et al. 2013).
Yet, the simulated load share, contact area and peak pressure By allowing for self-alignment in the experiment, contact at lower
magnitude only correlated weakly to moderately with the experiments. loads will be improved and the poor practice of linearization, based on
In our experiments, the radius took up 57.2 ± 18.2% of the applied loads over 400 N, can be avoided.
1000 N, in our simulation 51.7 ± 15.7%. These values are similar to the The simulated peak pressure magnitude might differ from the ex­
literature: (Hackl et al., 2017) observed load shares on the radius be­ periments due to the soft tissue missing in the model but being intact in
tween 55.0 ± 15.1% and 66.5 ± 16.1% of the applied 50–400 N in 14 the experiment, as far as inserting the pressure mapping sensor allowed
cadaveric specimens in full pronation. (Halls and Travill, 1964) reported for. Future studies should address this discrepancy by representing all
that in a study with 7 cadaveric specimens, on average 57% of the 150 N relevant structures. Apart from that, even though the pressure mapping
applied to the hand were transferred over radio-capitular column to the sensor is pre-conditioned and calibrated, diminishing load output over
humerus. time and with repeated use cannot be excluded (Jansson et al., 2013).
Further, our average contact area measured at 1000 N was, with Subjecting the FEM to realistic loads is necessary in order to transfer
395.4 mm2, similar to the strongly varying literature data: Willing et al. the results to real-life situations (Viceconti et al., 2020), which are re­
reported of 342–375 mm2 at 80 N whereas Smithson et al measured ported as 1600 N in the ulno-humeral and 800 N in the humero-radial
around 275 mm2 at100 N (Smithson et al., 2020; Willing et al., 2013). joint (Chadwick and Nicol, 2000). Balancing the real-life loads and the
Our simulations underestimated the contact area, and peak pressure risk of off-axis fractures starting at 1448 N (Troy and Grabiner, 2007),

8
S.L. Kahmann et al. Journal of Biomechanics 158 (2023) 111766

we decided for a maximum load of 1000 N. When analyzing the 95 % Chadwick, E.K.J., Nicol, A.C., 2000. Elbow and wrist joint contact forces during
occupational pick and place activities. J. Biomech. 33, 591–600. https://doi.org/
confidence interval of the predicted contact force (Fig. 6), one has to
10.1016/S0021-9290(99)00184-0.
consider the limits of this statistical parameter in non-normal distribu­ Cohen, J., 1977. Statistical Power Analysis for the Behavioral Sciences, second ed.
tions, particularly with small sample sizes (Puth et al., 2015). Elsevier. doi: 10.1016/C2013-0-10517-X.
Even though most steps of the Reporting checklist for verification Fornalski, S., Gupta, R., Lee, T.Q., 2003. Anatomy and biomechanics of the elbow joint.
Sports Med. Arthrosc. 11, 1–9.
and validation of finite element analysis in orthopedic and trauma Hackl, M., Wegmann, K., Kahmann, S.L., Heinze, N., Staat, M., Neiss, W.F., Scaal, M.,
biomechanics (Oefner et al., 2021) were followed, the workflow can be Müller, L.P., 2017. Radial shortening osteotomy reduces radiocapitellar contact
improved with a four-eyes principle observing the critical parameters. pressures while preserving valgus stability of the elbow. Knee Surgery, Sport.
Traumatol. Arthrosc. 25 https://doi.org/10.1007/s00167-017-4468-z.
Apart from that, more outcome parameters could be evaluated, e.g., full- Halls, A.A., Travill, A., 1964. Transmission of pressures across the elbow joint. Anat. Rec.
field optical measurements at the cost of soft tissue coverage or strain 150, 243–247. https://doi.org/10.1002/ar.1091500305.
gauges serving as local validation points. Haut Donahue, T.L., Hull, M.L., Rashid, M.M., Jacobs, C.R., 2002. A Finite element model
of the human knee joint for the study of tibio-femoral contact. J. Biomech. Eng. 124,
The relevance of the uncertainty quantification grows with 273–280. https://doi.org/10.1115/1.1470171.
increasing precision and accuracy of the model predictions and is related Hotchkiss, R.N., 1996. Fractures and dislocations of the elbow. In: Rockwood, C.A.,
to the research question the model is designed for. Green, D.P.J., Bucholz, R.W., Heckman, J.D. (Eds.), Rockwood and Green’s Fractures
in Adults. Lippincott-Raven, Philadelphia, pp. 929–1024.
The published workflow with parameters derived from one specimen Hwang, J.-T., Kim, Y., Bachman, D.R., Shields, M.N., Berglund, L.J., Fitzsimmons, A.T.,
can predict the stiffness, pressure distribution and peak pressure loca­ Fitzsimmons, J.S., O’Driscoll, S.W., 2018. Axial load transmission through the elbow
tion properly for the 7 specimens of the test cohort. It may serve as during forearm rotation. J. Shoulder Elb. Surg. 27, 530–537. https://doi.org/
10.1016/j.jse.2017.10.003.
groundwork for further research, which should focus on improvement of
Jansson, K.S., Michalski, M.P., Smith, S.D., LaPrade, R.F., Wijdicks, C.A., 2013. Tekscan
the cartilage contact. We believe that this study presents a solid base for pressure sensor output changes in the presence of liquid exposure. J. Biomech. 46,
simulations aiding in prosthesis design as well as fracture prediction. 612–614. https://doi.org/10.1016/j.jbiomech.2012.09.033.
Kaas, L., Van Riet, R.P., Vroemen, J.P.A.M., Eygendaal, D., 2008. The incidence of
associated fractures of the upper limb in fractures of the radial head. Strateg. Trauma
CRediT authorship contribution statement Limb Reconstr. 3, 71–74. https://doi.org/10.1007/s11751-008-0038-8.
Keyak, J.H., Kaneko, T.S., Tehranzadeh, J., Skinner, H.B., 2005. Predicting proximal
Stephanie L. Kahmann: Conceptualization, Methodology, Valida­ femoral strength using structural engineering models. Clin. Orthop. Relat. Res.
219–228 https://doi.org/10.1097/01.blo.0000164400.37905.22.
tion, Formal analysis, Investigation, Data curation, Writing – original Kim, S., Miller, M.C., 2016. Validation of a finite element humeroradial joint model of
draft, Writing – review & editing, Visualization, Project administration. contact pressure using fuji pressure sensitive film. J. Biomech. Eng. 138, 2016–2019.
Amelie Sas: Software, Writing – review & editing. Nils Große Hokamp: https://doi.org/10.1115/1.4031976.
Langohr, G.D.G., Willing, R., Medley, J.B., King, G.J.W., Johnson, J.A., 2015. Contact
Investigation, Writing – review & editing. G. Harry van Lenthe: analysis of the native radiocapitellar joint compared with axisymmetric and
Methodology, Formal analysis, Supervision, Writing – review & editing. nonaxisymmetric radial head hemiarthroplasty. J. Shoulder Elb. Surg. 24, 787–795.
Lars-Peter Müller: Resources, Funding acquisition. Kilian Wegmann: https://doi.org/10.1016/j.jse.2014.12.011.
Li, M., Venäläinen, M.S., Chandra, S.S., Patel, R., Fripp, J., Engstrom, C., Korhonen, R.K.,
Resources, Supervision, Funding acquisition, Writing – review & editing. Töyräs, J., Crozier, S., 2021. Discrete element and finite element methods provide
similar estimations for hip joint contact mechanics during walking gait. J. Biomech.
115, 110163 https://doi.org/10.1016/j.jbiomech.2020.110163.
Declaration of Competing Interest
Lindenhovius, A.L.C., Jupiter, J.B., Ring, D., 2008. Comparison of Acute Versus Subacute
Treatment of Terrible Triad Injuries of the Elbow 920–926. doi: 10.1016/j.
The authors declare that they have no known competing financial jhsa.2008.02.007.
Mononen, M.E., Jurvelin, J.S., Korhonen, R.K., 2013. Effects of radial tears and partial
interests or personal relationships that could have appeared to influence
meniscectomy of lateral meniscus on the knee joint mechanics during the stance
the work reported in this paper. phase of the gait cycle - a 3D finite element study. J. Orthop. Res. 31, 1208–1217.
https://doi.org/10.1002/jor.22358.
Oefner, C., Herrmann, S., Kebbach, M., Lange, H.E., Kluess, D., Woiczinski, M., 2021.
Acknowledgement
Reporting checklist for verification and validation of finite element analysis in
orthopedic and trauma biomechanics. Med. Eng. Phys. 92, 25–32. https://doi.org/
We thank Prof. Manfred Staat for providing the pressure mapping 10.1016/j.medengphy.2021.03.011.
sensors and Jutta Knifka for supporting us in the preparation of the Osawa, T., Moriyama, S., Tanaka, M., 2014. Finite element analysis of hip joint cartilage
reproduced from real bone surface geometry based on 3D-CT image. J. Biomech. Sci.
specimens and in the performance of the experiments. Eng. 9, 1–8. https://doi.org/10.1299/jbse.13-00164.
Ethical approval: VT 21-1514_1. Papatheodorou, L.K., Rubright, J.H., Heim, K.A., Weiser, R.W., Sotereanos, D.G., 2014.
Terrible triad injuries of the elbow: Does the coronoid always need to be fixed? Clin.
Orthop. Relat. Res. 472, 2084–2091. https://doi.org/10.1007/s11999-014-3471-7.
Funding Pellikaan, P., Giarmatzis, G., Vander Sloten, J., Verschueren, S., Jonkers, I., 2018.
Ranking of osteogenic potential of physical exercises in postmenopausal women
based on femoral neck strains. PLoS One 13, e0195463. https://doi.org/10.1371/
None. journal.pone.0195463.
Peña, E., Calvo, B., Martínez, M.A., Doblaré, M., 2006. A three-dimensional finite
References element analysis of the combined behavior of ligaments and menisci in the healthy
human knee joint. J. Biomech. 39, 1686–1701. https://doi.org/10.1016/j.
jbiomech.2005.04.030.
ABAQUS, 2013. ABAQUS 6.11 User’s documentation IV, 1100.
Pike, J.M., Athwal, G.S., Faber, K.J., King, G.J.W., 2009. Radial head fractures-an update.
Abel, A., Kahmann, S.L., Mellon, S., Staat, M., Jung, A., 2020. An open-source tool for the
J. Hand Surg. Am. 34, 557–565. https://doi.org/10.1016/j.jhsa.2008.12.024.
validation of finite element models using three-dimensional full-field measurements.
Puth, M.-T., Neuhäuser, M., Ruxton, G.D., 2015. On the variety of methods for
Med. Eng. Phys. 77 https://doi.org/10.1016/j.medengphy.2019.10.015.
calculating confidence intervals by bootstrapping. J. Anim. Ecol. 84, 892–897.
An, K.N., Kwak, B.M., Chao, E.Y., Morrey, B.F., 1984. Determination of muscle and joint
https://doi.org/10.1111/1365-2656.12382.
forces: a new technique to solve the indeterminate problem. J. Biomech. Eng. 106,
Renani, M.S., Rahman, M., Cil, A., Stylianou, A.P., 2017. Ulna-humerus contact
364–367. https://doi.org/10.1115/1.3138507.
mechanics: finite element analysis and experimental measurements using a tactile
Anderson, A.E., Ellis, B.J., Maas, S.A., Peters, C.L., Weiss, J.A., 2008. Validation of finite
pressure sensor. Med. Eng. Phys. 50, 22–28. https://doi.org/10.1016/j.
element predictions of cartilage contact pressure in the human hip joint. J. Biomech.
medengphy.2017.08.010.
Eng. 130 https://doi.org/10.1115/1.2953472.
Safran, M., Ahmad, C.S., Elattrache, N.S., 2005. Ulnar collateral ligament of the elbow.
Bozon, O., Chrosciany, S., Loisel, M., Dellestable, A., Gubbiotti, L., Dumartinet-
Arthrosc. - J. Arthrosc. Relat. Surg. 21, 1381–1395. https://doi.org/10.1016/j.
Gibaud, R., Obrecht, E., Tibbo, M., Sos, C., Laumonerie, P., 2022. Terrible triad
arthro.2005.07.001.
injury of the elbow: a historical perspective. Int. Orthop. https://doi.org/10.1007/
Smithson, K., Smith, J., Hogue, W., Mannen, E., Ahmadi, S., 2020. Biomechanics of axial
s00264-022-05472-4.
load transmission across the native human elbow. Shoulder Elb. 13, 671–676.
Brimacombe, J.M., Wilson, D.R., Hodgson, A.J., Ho, K.C.T., Anglin, C., 2009. Effect of
https://doi.org/10.1177/1758573220961025.
calibration method on Tekscan sensor accuracy. J. Biomech. Eng. 131, 1–4. https://
Stylianou, A., 2021. Computational approaches to study elbow biomechanics. Ann. Jt. 6,
doi.org/10.1115/1.3005165.
11. https://doi.org/10.21037/aoj.2020.04.01.
Bryce, C.D., Armstrong, A.D., 2008. Anatomy and biomechanics of the elbow. Orthop.
Clin. North Am. 39, 141–154. https://doi.org/10.1016/j.ocl.2007.12.001.

9
S.L. Kahmann et al. Journal of Biomechanics 158 (2023) 111766

Troy, K.L., Grabiner, M.D., 2007. Off-axis loads cause failure of the distal radius at lower Willing, R.T., Lalone, E.A., Shannon, H., Johnson, J.A., King, G.J.W., 2013. Validation of
magnitudes than axial loads: a finite element analysis. J. Biomech. 40, 1670–1675. a finite element model of the human elbow for determining cartilage contact
https://doi.org/10.1016/j.jbiomech.2007.01.018. mechanics. J. Biomech. 46, 1767–1771. https://doi.org/10.1016/j.
Viceconti, M., Juarez, M.A., Curreli, C., Pennisi, M., Russo, G., Pappalardo, F., 2020. jbiomech.2013.04.001.
Credibility of in silico trial technologies—a theoretical framing. IEEE J. Biomed.
Heal. Informatics 24, 4–13. https://doi.org/10.1109/JBHI.2019.2949888.

10