Femur Loading in Feet-First Fall Experiments using an Anthropomorphic Test Device

Abstract

Background

Femur fractures are a common orthopedic injury in young children. Falls account for a large portion of accidental femur fractures in young children, but there is also a high prevalence of femur fractures in child abuse, with falls often provided as false histories. Objective information regarding fracture potential in short distance fall scenarios may aid in assessing whether a child’s injuries are the result of abuse or an accidental fall. Knowledge of femur loading is the first step towards understanding likelihood of fracture in a fall.

Objective

Characterize femur loading during feet-first free falls using a surrogate representing a 12-month-old child.

Methods

The femur and hip joint of a surrogate representing a 12-month-old were modified to improve biofidelity and measure femur loading; 6-axis load cells were integrated into the proximal and distal femur. Femur modification was based upon CT imaging of cadaveric femurs in children 10–14 months of age. Using the modified 12-month-old surrogate, feet-first free falls from 69 cm and 119 cm heights onto padded carpet and linoleum were conducted to assess fall dynamics and determine femur loading. Femur compression, bending moment, shear and torsional moment were measured for each fall.

Results

Fall dynamics differed across fall heights, but did not substantially differ by impact surface type. Significant differences were found in all loading conditions across fall heights, while only compression and bending loads differed between carpet and linoleum surfaces. Maximum compression, bending, torsion and shear occurred in 119 cm falls and were 572 N, 23 N-m, 11 N-m and 281 N, respectively.

Conclusions

Fall dynamics play an important role in the biomechanical assessment of falls. Fall height was found to influence both fall dynamics and femur loading, while impact surface affected only compression and bending in feet-first falls; fall dynamics did not differ across carpet and linoleum. Improved pediatric thresholds are necessary to predict likelihood of fracture, but morphologically accurate representation of the lower extremity, along with accurate characterization of loading in falls are a crucial first step.

INTRODUCTION

Femur fractures are the most common hospitalized traumatic orthopedic injury in children. In a review of 84,000 pediatric patients admitted for orthopedic trauma, 22% had a femur fracture (incidence of hospitalization 27.2 per 10,000 children in the United States).¹ Incidence rates of femur fracture are highest for children aged 0–3 years. In a study of femur fractures in children < 6 years, Brown and Fisher² reported the highest incidence rates in children < 1 year and children 2 years of age. Heideken, Svensson³ reviewed femoral shaft fractures in children aged 0–14 years, and also reported the highest incidence in the 1–3 year age range.

Falls are the most common cause of femur fractures in young children. In children < 2 years of age, falls account for nearly 50% of all femur fractures⁴, but child abuse is the second highest cause of femur fractures in this age group, and may account for up to 35% of fractures.⁵ Falls are often a false history provided by caregivers to conceal abuse.^5–8 Knowing the risk of fracture for a given fall history is key to distinguishing between an accidental vs. abusive femur fracture. But knowledge of the type and extent of loading the femur is exposed to in falls is needed to understand likelihood of fracture.

Anthropomorphic test devices (ATDs) have been used as child surrogates in experimental simulations of falls to describe injury potential.^9–16 Measured ATD loads and accelerations are typically compared with published injury thresholds to assess the potential for various injuries including head injuries and long bone fractures. Though a few studies have investigated femur loading in short-distance fall events^{9, 11}, the biofidelity of pediatric ATD extremities is currently limited and thus reported loads (and associated assessment of fracture potential) have been questionable. For example, the Hybrid III 3-year-old and CRABI 12-month-old ATDs were used in previous studies to investigate femur fracture.^{9, 11} These ATDs were not designed to measure femur loading; strain gages were added to the plastic or aluminum rods representing the femur. Additionally, the limb motion of current pediatric ATDs is limited to the sagittal plane (i.e. no hip abduction or rotation). Several studies have described improvements to the biofidelity of existing ATDs to improve the accuracy of the event dynamics and loads measured.^17–19 Changes to anthropometry (shape), joint range of motion, and material properties (stiffness) to more closely represent humans have been shown to improve biofidelity.

The purpose of this study was to characterize femur loading during feet-first free falls using a surrogate representing a 12-month-old child. The lower extremities of a 12-month-old ATD were modified to (1) measure femur loading and (2) to more accurately represent lower limb anatomy and range of motion. The improved ATD was then used in feet-first fall simulations to demonstrate ATD measurement capabilities and assess femur fracture potential. Feet-first falls were chosen for this study as a first-step in understanding femur loading in falls as it was expected that femur compression would be the dominant loading condition and worse case compression loading would occur in free falls with a vertical body orientation.

METHODS

To characterize femur loading in pediatric short-distance falls, a Child Restraint Air Bag Interaction (CRABI, First Technology Safety Systems, Plymouth, MI) 12-month old anthropomorphic test device (ATD) was modified and used in experimental simulations of falls. The CRABI ATD represents a 50^th percentile 12-month old child in height (74 cm), mass (10 kg), and geometric and inertial properties of individual body segments. The CRABI lower extremity was modified to improve biofidelity and allow for measurement of femur loads. We implemented four key design modifications: 1)six-axis load cells were added to the femur to directly measure loading, 2) the newly designed femur shaft geometry was based on a computed tomography (CT) scan of an 11-month-old child (9.2 kg overall weight and 76 cm height), 3) hip joints were modified to represent a more realistic range of motion (compared to the original CRABI design which limits hip and knee motion to the sagittal plane), and 4) soft tissue material was replaced to more accurately simulate the properties of human soft tissue.

Feet-first free fall experiments were then conducted using the modified CRABI from two heights onto two different impact surfaces to assess potential fall scenarios.

Design and Development of the Hip-Femur-Knee Complex

Femur morphology

Femoral shaft morphology was based on CT imaging of an 11-month-old child. The de-identified CT scan (resolution 0.5 × 0.5 × 1.0 mm³) was acquired from the Radiology-Pathology Center for Forensic Imaging at the University of New Mexico (Albuquerque, NM). The overall height and weight for this child (9.2 kg and 76 cm) were closest to target parameters for a 50^th percentile 12-month-old child of the available pediatric CTs. To assess whether the femur morphology of the selected 11-month-old child was representative of a 50^th percentile 12-month-old child, femur diaphysis length (between growth plates) and mid-diaphysis diameter were measured and compared to those from CT (n=2), plain radiographs (n=15) and magnetic resonance images (MRI, n=3) available for twenty children aged 10–14 months (Figures 1 and 2). Femoral measures for the selected 11-month-old child were within one standard deviation of the mean femur diaphysis length and mid-diaphysis diameter.

Figure 1.

Femoral diaphysis length (distance between growth plates) vs. body weight for 20 children aged 10–14 months. The 11-month-old subject selected for this study indicated by an X. Horizontal line represents the mean length (136 +/− 17 mm SD) for this sample.

Figure 2.

Femur mid-diaphysis diameter vs. body weight for 20 children aged 10–14 months. The 11-month-old subject selected for this study indicated by an X. Horizontal line represents the mean diameter (10.5 +/− 1.1 mm SD) for this sample.

The CT images were reconstructed using Mimics software (Materialise, Plymouth, MI, USA) and a 3D model of the left femur diaphysis (between growth plates) was created (Figure 3a). This model was imported into Solidworks 2014 (Dassault Systèmes, Waltham, MA, USA) as a 3D surface mesh and used as a template for the new ATD femur design. A solid 3D model was generated from the mesh. To enable characterization of proximal and distal femur loading, load cells were incorporated into the proximal and distal metaphyseal locations; in doing so, only the central 60mm of the shaft was retained from the 3D femur model, and flanges were added to each end of this central portion to allow for positioning of load cells (Figure 3b and 3c).

Figure 3.

(a) 3D femur model derived from CT scan of 11-month-old child, (b) assembly drawing of new hip-femur-knee complex design, (c) exploded assembly drawing of new hip-femur-knee complex. Scale and orientation of 3D model and assembly (a and b) are equivalent for comparison.

Instrumentation

Six-axis load cells (Sunrise Instruments M3722C, Novi, MI, USA) were integrated into the femur design at distal and proximal metaphyseal locations (where buckle fractures commonly occur). The load cells were oriented so that the z-axis (with the greatest load capacity, 3000 N) was aligned with the long axis of the femur.

Knee

No changes were made to the lower leg of the CRABI ATD. Thus, the knee joint of the new femur was designed to interface with the existing lower leg assembly and the distal load cell. The range of motion of the knee joint was 0–85 degrees flexion; no extension, adduction, abduction, or rotation was permitted.

Hip

Targeted degrees of freedom and range of motion were based on measurements of passive hip range of motion in children.^20–23 Universal joints were utilized to allow for abduction/adduction motion and flexion/extension. A hip block representing the acetabulum was designed to interface with the ATD pelvis and receive one end of a universal joint. A bracket was designed to join the other end of the universal joint to the proximal load cell. The resulting ranges of motion of the modified hip were 90, 40, and 90 degrees in flexion, extension, and abduction, respectively. Unlike the human, the universal joints utilized in the ATD limb did not allow internal or external rotation of the hip.

Fabrication and Assembly

Brackets for attaching load cells to the femur diaphysis were created using CNC machining of aluminum 6061-T6 and all other parts were fabricated using Direct Metal Laser Sintering (DMLS) of aluminum alloy AlSi10Mg. Aluminum was selected for durability and repeatability in testing. The assembled hip-femur-knee complex design with load cells is shown in Figure 3b and 3c. The overall hip to knee length and weight of the lower extremity (including soft tissue) were maintained within 15 mm (10%) and 14 grams (2.5%), respectively.

Soft tissue

Two different compositions of synthetic ballistics gelatin (10% and 20% gelatin, Clear Ballistics, Fort Smith, AR, USA) were considered for use as soft tissue. Indentation testing of these materials was performed at 3 mm/s using a 6 mm diameter flat-tipped indenter and a materials testing machine (ADMET model #2608; Admet, Inc.; Norwood, MA, USA) to obtain force-displacement curves for comparison with human tissue (Figure 4). Force-deformation response of the 20% ballistics gelatin aligned more closely with the values for in vivo adult human tissue (overlying gastrocnemius) found by Silver-Thorn²⁴, so this material was chosen for the modified lower extremities.

Figure 4.

Force-displacement curves from indentation testing of ballistics gelatin samples at 3 mm/s loading rate. Shaded region represents range of values found in testing of human tissue ²⁴.

After selection of surrogate soft tissue material, molds of the ATD thigh were created. A solid model of soft tissue was created to approximate the shape and size of existing CRABI thigh soft tissue while enveloping the new femur and load cells. The model was symmetric such that it could be formed in two halves from a single mold. A two-piece mold was modeled (Figure 5), and this mold was 3D printed (CubePro printer, 3D Systems, Rock hill, SC, USA) using ABS. Polyurethane was applied to the interior surfaces of the mold to aid in release of material from the mold. The soft tissue was then formed by melting the ballistics gelatin (as per manufacturer instructions), pouring the molten gelatin into the mold, and allowing to fully cool and set. Once molds were removed, the two halves were joined together around the femur.

Figure 5.

Solid model of new lower extremity soft tissue (left) and mold (right)

As the ballistics gelatin was susceptible to tearing on impact, a flexible nylon “skin” enveloping the soft tissue was applied to protect the soft tissue during experiments, without significantly altering the load transfer to the soft tissue and femur. Following assembly of the lower extremity, a small youth nylon compression arm sleeve (CompressionZ, www.compressionz.com) was slipped over the soft tissue to hold the two halves in position and protect the soft tissue from damage. The assembled lower extremities were then attached to the ATD pelvis.

Fall Experiments with Modified ATD

The modified CRABI ATD was suspended in a vertical posture using a rope loop under the chin of the ATD and released to simulate a feet-first free fall. The rope was secured by a mechanism that released via an external trigger, providing for repeatable falls. The height was adjustable to allow for falls from heights of 69 cm and 119 cm measured from the ground to the ATD center of mass. These heights were selected for comparison with data from previous ATD feet-first fall experiments.^{10, 11} The 69 cm height is representative of a child standing on a short (~23 cm) stool and the 119 cm height representative of a child being carried by an adult. The ATD was in the same initial position for all falls with upper extremities and lower extremities in a vertical orientation achieved under gravity. 10–13 drops were performed for each height and surface combination for a total of 45 falls.

Two impact surfaces were tested: linoleum over wood and padded carpet. Both surfaces were placed over a 183 cm × 183 cm wooden platform. The platform consisted of 1.9 cm plywood covering 5.1 cm × 10.2 cm joists spaced 40.6 cm apart. Linoleum tile was 1mm thick self-adhesive no wax vinyl, and carpet was 1.3 cm thick open loop over 0.3175 cm thick padding. Further details of impact surfaces including coefficients of friction and coefficients of restitution were published previously.¹⁰

Data Analysis

Load cell data was sampled at 10,000Hz using a custom-developed LabView (National Instruments, Austin, TX, USA) program and filtered using a 600Hz 4th order low pass Butterworth filter according to SAE J211 standards.²⁵ Falls were recorded from two angles (a direct view of the sagittal plane 30cm above ground level, and a view from 152cm above ground level positioned anterior to the ATD at a 45 degree angle to the sagittal plane) using digital video cameras (GoPro HERO4 Silver, GoPro, Inc.) recording at 240 frames per second.

For each fall trial, peak compression, bending, shear and torsion loads at the proximal and distal femur were determined. The effects of surface type and fall height on the resulting femur loads were assessed using separate two-way ANOVAs for each dependent variable. Normality of data was assessed using the Shapiro-Wilk test. For dependent variables violating the assumption of normality, data was transformed using the Aligned Rank Transform, a technique developed to allow the use of factorial ANOVAs with non-parametric data. Due to the use of multiple statistical tests, a Bonferroni correction was applied resulting in a statistical significance level of p ≤ 0.0125. All statistical analysis was performed using IBM SPSS v22.0.0.2.

RESULTS

Fall Dynamics

Image sequences depicting fall dynamics for two representative falls (69 cm and 119cm) are shown in Figures 6 and 7. In all falls, the ATD first fell to a crouching position with hips and knees flexed (Figure 6B). Dynamics differed across the two tested fall heights, but did not substantially differ by impact surface type. Fall dynamics were grouped into 3 broad categories (Table 1); the majority of falls exhibited dynamics described by category A or C, depending upon fall height. Generally, in shorter-distance falls (69 cm), the ATD’s feet rebounded off the floor surface as the hips flexed and legs extended forward causing the ATD to land in a seated position (Figure 6C) before falling to a supine position or to one side (Table 1, Fall Category A). In a small number of falls (Table 1, Fall Category B), only one foot swung out while the other foot remained planted to the floor surface; after pelvis impact, the ATD torso rotated toward the floor at an angle (not directly rearward) and landed supine. Unlike the shorter-distance falls, in most falls from the 119 cm height, the feet stayed in contact with the floor surface (Table 1, Fall Category C). The heels came off the floor as the hips flexed and knees extended, but the foot rolled anteriorly onto the dorsal surface as the ankle plantar flexed (Figure 7B). The ATD then rebounded as the legs extended, launching the ATD rearward to land in a supine position. Two of the 119cm falls had similar dynamics to the 69 cm falls with the feet rebounding off the floor due to slight abduction of the legs as the hips and knees flexed (Table 1, Fall Category A).

Figure 6.

Fall dynamics sequence (A through E) and corresponding femur compression and bending moment time histories (A through E) from representative 69 cm fall onto carpet. (This sequence is representative of Fall Dynamic A in Table 1). A video of this fall is available in the electronic version of the manuscript.

Figure 7.

Fall dynamics sequence and corresponding compression and bending moment time histories from representative 119 cm fall onto linoleum. (This sequence is representative is representative of Fall Dynamic C in Table 1). A video of this fall is available in the electronic version of the manuscript.

Table 1.

Descriptions and frequencies of observed fall dynamics by fall type.

		Frequency
Fall Dynamics Category:		A	B	C
69 cm	Carpet	9	3	0
69 cm	Linoleum	11	2	0
119 cm	Carpet	0	0	10
119 cm	Linoleum	2	0	8

^AATD fell to crouching position with hips and knees flexed, knees then extended while feet rotated forward from beneath torso as ATD pelvis continued to move downward. ATD landed in a seated position with knees fully extended before rotating rearward into a supine position or to one side.

^BATD fell to crouching position with hips and knees flexed, left knee then extended while foot rotated forward from beneath torso, but right toes remained planted on floor surface as ATD pelvis continued to move downward. ATD landed in a seated position (left knee extended, right knee flexed), torso then rotated rearward into a supine position

^CATD fell to crouching position with hips and knees flexed, heels then lifted off floor, while toes remained planted resulting in plantar flexion of ankles and rolling onto the dorsal surface of the foot as ATD pelvis continued to move downward. Hips and knees extended after pelvis impact launching ATD rearward to land in supine position.

Lower Extremity Loading

Peak compression loads typically occurred upon initial foot contact with the floor, and were followed by peak bending moments as the ATD hips and knees flexed during descent (Figures 6 and 7). Peak compression loads ranged from 175–385 N in 69 cm falls and 173–572 N in 119 cm falls (Figure 8). In 69 cm falls, the maximum bending moment was 13 N-m and in 119 cm falls bending moment peaked at 23 N-m (Figure 9). Peak shear loading commonly occurred when ATD hips and knees were flexed during descent, and coincided with bending moment peaks. Maximum torsional loading typically was measured during maximum knee flexion and concurrent pelvis impact with the floor in the 119 cm falls. Torsional moments ranged from 1–7 N-m in 69 cm falls and from 3–11 N-m in 119 cm falls, while maximum shear loads ranged between 42–176 N for 69 cm falls and 115–281 N for 119 cm falls (Figures 8 and 9).

Figure 8.

Mean peak compression and shear loads measured in the four fall scenarios. Error bars represent 95% confidence intervals.

Figure 9.

Mean peak bending and torsion moments measured in the four fall scenarios. Error bars represent 95% confidence intervals.

Fall height had a significant influence on all femur loads (p<0.001 for compression, bending, shear and torsion) with greater loads resulting from greater heights. Compression and bending loads were significantly greater in falls onto linoleum than carpet (p=0.001 for compression and p=0.002 for bending), but no significant differences were found in shear and torsion loads across the two surfaces. Additionally, there was a significant interaction effect of fall height and impact surface type on femur compression loads (p=0.004). The highest levels of compression and bending moment occurred during 119 cm falls onto linoleum. Compression and bending loads were significantly reduced for the same falls onto padded carpet, as well as for 69 cm falls.

Discussion

To our knowledge, this is the first study to examine femur loading in short-distance falls using a child surrogate with lower extremities designed specifically for improved measurement of femur loading. Given the prevalence of femur fractures in accidents and abuse, knowledge of factors that influence femur loading during falls can be helpful when assessing whether a femur fracture is compatible with a stated fall history. Our findings indicate a difference in femur loading across the two studied fall heights; femur compression, bending moment, torsional moment and shear loading were significantly greater in the 119 cm falls as compared to the 69 cm falls. Femur loads were 1.5–1.7 times greater in the 119 cm falls than the 69 cm falls. These higher loads are consistent with the principle of conversion of potential energy (a function of height) just before the start of a fall to kinetic energy during a fall, which is then absorbed by the body upon impact. One would therefore expect higher height falls to lead to increased energy being transferred to the body on impact. Given the direct transference of energy to the feet and through the lower extremities in feet-first falls, it is expected that lower extremity loading would be directly related to fall height. Thus, in feet-first falls, higher femur loads would generally be associated with falls from greater heights as was found in our study. While not possible to definitely state whether fracture would have occurred from 119 cm falls, likelihood of femur fracture would certainly be higher relative to 69 cm falls.

Compression loads were found to peak on initial foot contact, while bending, shear and torsion subsequently peaked following initial foot contact in most falls. As the knees freely extended following initial foot contact in the lower height falls, impact energy was transferred from the feet/legs to the pelvis. Conversely, since the feet did not leave the floor in higher height falls, fall energy was transferred through the lower extremities throughout the impact duration with the legs remaining beneath the torso as the pelvis descended to the floor. In part, this accounts for the higher bending moments and shear loads measured in the 119 cm falls. However, because the CRABI ankle joint is constructed of rigid foam and lacks structure representing the tarsus, biofidelity of foot-ankle kinematics observed in 119 cm falls may not be representative of children.

Given the higher femur loads in the 119 cm falls, it is not surprising that dynamics during impact with the floor differed from the 69 cm falls. Higher falls, having greater energy and downward directed velocity, tended to cause the legs and feet to remain fixed to the floor beneath the weight of the ATD, while lower falls having less transferred energy on impact led to the feet slipping off the floor surface as the knees extended after initial foot contact. Crumpling of the legs and feet beneath the body would tend to cause higher loading on the lower extremities as was seen in the 119 cm falls. Conversely, unconstrained free extension of the knees, as was seen in the 69 cm falls, would tend to generate lower femur loading. Thus downward velocity, which is proportional to the square root of fall height, played an important role in determining whether the feet were able to rotate free from beneath and away from the descending torso. The higher downward velocity associated with 119 cm falls would be approximately 1.3 times that of the velocity in 69 cm falls, and tended to prevent the ATD feet from being able to rotate from beneath the ATD as the pelvis continued to descend. It is unclear whether this same phenomenon is true when comparing children free falling from higher heights. Foot-ankle impact kinematics also suggest that high levels of loading may also occur at the tibia and ankle during higher falls. However, lower leg loading was not measured in this study. Interestingly, flooring surface did not influence impact kinematics for either fall height. Though friction and energy absorbing properties of the flooring surface did not have a noticeable effect on kinematics, small differences in femur loading were evident at the greater fall height (119 cm) with the linoleum surface producing higher compression and bending loads.

Femur loads measured in this study differ from those measured in similar studies. In one study similar feet-first fall simulations were conducted with a 3-year-old ATD.¹¹ Falls from the lower height (69 cm) evaluated in our study onto similar surfaces (padded carpet and linoleum) with the 3-year-old ATD were associated with lower compression forces, lower bending moments, and higher torsional loads than those in this study (Table 2). Due to the larger mass of the 3-year-old ATD compared to the 12-month-old ATD, larger loads would be expected but only torsional loads were higher. Conversely, a separate study of feet-first falls using the 3-year-old ATD from a greater (119 cm) height onto linoleum found similar compression loads and lower torsional loads than those measured in this study.¹³ In addition to our lower ATD mass, differences may be due to the new femur design and load measurement capability of the CRABI ATD in this study. Unmodified ATDs typically use metal stock having a rectangular cross-section to represent the femur whereas our femur was consistent with the morphology of a child’s femur; this difference in femur geometry may cause differences in the magnitude and distribution of loading. Differences in hip joint range of motion across the ATDs would also be expected to cause differences in femur loading during falls. Additionally, in the previously mentioned studies^{11, 13}, loads were measured via strain gages placed at the mid-shaft of the femur, whereas the load cells in the redesigned CRABI were placed in the proximal and distal metaphyseal locations; this difference in load measurement location could lead to differing femur loading. In a previous study of horizontal bed falls from a 61 cm height using the unmodified CRABI 12-month-old ATD, mean peak femur compression loads onto carpet and linoleum were about one-third the values measured in the current study.⁹. Bending and torsion moments in the bed fall study were approximately one-half those measured in the feet-first falls. These results are expected as feet-first impact would be expected to generate greater lower extremity loads.²⁶ Computer simulation studies of pediatric falls have shown influence of fall height, surface properties, and impact position on lower extremity loads.^{26, 27} Additionally, the increase in lower extremity loads with increasing height seen in this study is consistent with prior ATD studies.^{11, 13}

Table 2.

Comparison of mean peak femur loads measured in this study vs. those reported in previous ATD fall studies

ATD age	Fall type	Fall Height (cm)	Impact Surface	Compression (N)	Bending (Nm)	Torsion (Nm)	Reference
12-month	Feet-first	69,119	Carpet, Linoleum	244–429	6.1–14.5	2.0–8.4	Present study
3-year	Feet-first	69	Carpet, Linoleum	125–200	0.9–2.5	4.7–9.5	Bertocci11
3-year	Feet-first	119	Linoleum	310–490	--	3.8–3.9	Deemer13
12-month	Horizontal	61	Carpet, Linoleum	75–100	3.0–3.8	1.5–2.0	Thompson9

To our knowledge, there are no published femur fracture thresholds for infants. Levine²⁸ published fracture thresholds of the adult femur in compression (7720 and 7110 N for males and females, respectively), bending (310 and 180 Nm for males and females, respectively), and torsion (175 and 136 Nm for males and females, respectively). Additionally, femur fracture compression thresholds for use with adult ATDs range from 6800 N for the 5th percentile female to 10,000 N for the 50th percentile adult male ATD.²⁹ Loads measured in this study were far below any of these adult fracture thresholds, but it is reasonable to assume that infant fracture thresholds would be lower than those of an adult.

A few studies have investigated the strength of the pediatric femur but did not report fracture loads.^30–32 Martin and Atkinson³³ estimated bending strength using properties measured on small femoral shaft specimens, and found in one 2.5-year-old subject, a maximum bending load of 53 Nm. Using data from quasi-static bending and compression tests of pediatric femoral shaft specimens, Sturtz³⁴ estimated the dynamic loads necessary to produce a fracture. This calculation was based on the assumption that dynamic load limits are 20% higher than quasi-static load limits. The dynamic bending fracture criteria for a 7-year-old and 3.6-year-old child were 116–131 Nm and 62–73 Nm, respectively. Dynamic axial (compression) fracture criteria were 1800 and 1000 N for a 6-year-old and 3-year-old, respectively. Another study measured the load necessary for fracture in pediatric cadaveric thighs (with soft tissue intact) in both quasi-static and dynamic bending tests.³⁵ In quasi-static tests, the femora from 18 subjects ranging from newborn to 6 years old were loaded in 3 point bending to fracture. Fracture moments tended to increase with age ranging from 14.1 Nm (in a 6-day old child) to 219 Nm (in the 6-year-old child). Dynamic tests were performed on 10 subjects aged 2–27 months. In these tests, the subjects (whole cadavers) were dropped from a height of 70–90 cm onto an impactor at the lateral mid-thigh. Impact forces ranged from 250 to 2370 N, and impact speeds ranged from 13.3 to 16.8 km/hr. A fracture occurred in only 2 cases. Bending and compression loads measured in our study were far below thresholds reported by Martin and Atkinson and Sturtz. In our study, several 119 cm falls onto linoleum produced bending moments greater than 14.1 Nm (the lowest fracture moment reported by Miltner and Kallieris, associated with a 1-month-old child); however, for children >6 months of age, failure moments reported by Miltner were all greater than 80 Nm, far above any measured bending moment in this study. These results suggest that the likelihood of fracture due to compression and bending loads is low in this study. However, these thresholds should be interpreted with caution as they are based on small sample sizes and, in the case of Martin and Atkinson and Sturtz, small samples of femoral cortical bone (rather than testing of whole femurs) from older children than the 12-month-old used in this study. The potential for fractures resulting from shear or torsional loading is unclear as there are no known pediatric fracture thresholds for these loading types. Future research is needed to improve understanding of pediatric bone mechanical properties and fracture potential under various loading conditions.

Clinical evidence suggests that although the incidence of femur fractures in short-distance falls is low, femur fractures sometimes occur in falls from heights or when femur impingement occurs. Hennrikus, Shaw³⁶ found 3 children with femur fractures of 115 patients <12 years of age with orthopedic injuries from short falls (<1.2 m). In a study by Tarantino, Dowd³⁷, 3 femur fractures were found in short falls involving 167 infants 0–10 months of age. Stair falls are a frequent cause of femur fractures in children; Pierce, Bertocci³⁸ reported a series of 29 femur fractures in children aged 3–31 months resulting from a stair fall (of these, abuse was suspected in 4 cases). Of the 25 plausible stair falls, 16 were falls with a caregiver and 9 were falls involving only the child. Buckle fractures were the most common fracture type, followed by transverse and spiral fractures of the diaphysis.³⁸ A separate study of stair falls involving a child held by a caregiver found 50% (8 of 16) involved a buckle or spiral femur fracture.³⁹ Other studies reporting the mechanism of femur fracture in short distance falls include one buckle fracture (of 85 children treated for falls from beds) in a 2-year-old child who bounced off a bed while playing⁴⁰ and one buckle fracture in a review of 79 household falls⁴¹ involving a 5-month-old who fell was unrestrained in an infant seat and fell from a table with the infant seat landing on top of her. The buckle fractures frequently seen in these cases would result from compressive loading directed through the long axis of the femur.⁴² This may occur when the child lands on their feet (as simulated in our ATD experiments) or impacts their knee. In many cases, particularly those involving infants, these occurred in falls from heights (e.g. a fall from a caregiver’s arms) or involved another person or object (e.g. falling with the caregiver or infant seat on top of the child) that would generate higher levels of loading. Spiral fractures result from torsion loads that may occur when the legs twist or bend underneath the child.⁴² Transverse and short oblique fractures may result from bending or shear loading such as when the leg is directly struck with an object or in stair falls with a caregiver, for example, when the child’s leg is impacted between the caregiver and step.⁴² At 12 months of age, children are typically cruising, but usually at low speeds which make fracture unlikely in ground-based falls.

Limitations

This study attempted to improve the biofidelity of the ATD lower extremity; however, some concerns regarding the realistic dynamics and measured loads (compared to human children) remain. We were not able to fully replicate hip range of motion in our modified design. In particular, there was no internal or external rotation of the lower extremity; this could lead to overestimating torsional moments. The joint stiffness in the hip was not controlled and joint laxity could result in underestimated loads. Additionally, no changes were made to the lower leg (including foot and ankle) of the CRABI which can influence transference of impact energy to the femur as well as fall dynamics. Impact dynamics associated with 119 cm falls led to rolling of the foot onto the dorsal surface in combination with extreme ankle plantar flexion. This dynamic may be an artifact given the CRABI ATD shank design. The unmodified shank consists of a tibial rod that extends to just proximal to the ankle, without connecting to the ankle or foot. No physical “ankle joint” exists in the CRABI ATD – it is simply a rigid foam structure. The absence of an ankle joint and an incomplete tibial structure contributed to the observed foot-shank kinematics on impact. Were a more biofidelic shank and ankle representation incorporated, differing kinematics may have been observed and femur loading may have been altered. Additionally, compliance of the rigid foam ankle is likely higher than that of a child, and probably accounts for the rebound following impact. Finally, as is a limitation with most human surrogates, the modified CRABI in this study is unable to simulate active muscle response. However, the active response of a 12-month-old child in a short distance free fall such as this is anticipated to be minimal.

Loads were measured at two locations along the femur, the proximal and distal metaphyses. It is possible that peak loads (particularly moments) occurred at other points along the femur shaft. Future studies using finite element analysis or applying additional instrumentation (e.g. strain gages) along the ATD femur shaft are recommended to assess how stresses and strains vary along the femur shaft to further evaluate fracture potential.

The simulated fall was a simple, vertical free fall. The position and orientation of the ATD was held constant for all falls. Variations in initial position could lead to different impact orientations and thus different loads transmitted to the femur. Compression loads were maximized in this study due to the vertical orientation. Changes in impact orientation from vertical (for a given height/surface combination) would likely lower compression loads but could increase other types of femur loads (bending, shear, torsion). Additional factors such as footwear or differing surface properties (e.g. friction, stiffness) may also affect both the impact dynamics and measured femur loads.

Conclusions

This study evaluated femur loading in feet-first falls using a 12-month-old CRABI ATD with a modified lower extremity under various height and impact surface conditions. Consistent with the notion that fall height affects injury severity, our study has shown that fall height has a direct effect on femur loading in feet-first falls, with greater fall heights typically leading to higher magnitudes of loading. In addition, fall height was found to influence fall dynamics, which in turn affected the timing and magnitude of loading (i.e. compression, bending moment, shear and torsion). Comparatively, surface only had an effect on compression and bending moments, but did not influence shear, torsional moments or dynamics at the fall heights evaluated in this study. While greater magnitudes of femur loading lead to increased likelihood of fracture, improved pediatric fracture thresholds are needed to be able to state whether femur fracture would occur in the evaluated falls. However, morphologically accurate representation of the femur and human-likehip joint range of motion, along with accurate characterization of loading during falls, as intended in this study, are also crucial to predicting fractures.

HIGHLIGHTS.

• The lower extremities of a child surrogate were modified to improve biofidelity.

• Femur loading was characterized in feet-first free falls for a 12-month-old child.

• Femur compression, bending, shear and torsion loads were measured during falls.

• Fall height and surface influenced magnitude of femur loading.

• Fall height influenced fall dynamics.

Appendix

Details of modified ATD lower extremity design

Figure A1.

Photos of modified ATD lower extremity without soft tissue, including modified hip joint, femur and load cells positioned proximally and distally (a), with soft tissue overlying femur (b), and with nylon sleeve overlying soft tissue (c).

Figure A2.

Detailed design of new hip-femur-knee complex with parts listing.