Characterization of Crash-Induced Thoracic Loading ... · a case review at which time the crash mechanics, biomechan-ics, and clinical aspects of injury are assessed. CIREN is one

Characterization of Crash-Induced Thoracic LoadingResulting in Pulmonary ContusionF. Scott Gayzik, PhD, R. Shayn Martin, MD, H. Clay Gabler, PhD, J. Jason Hoth, MD, Stefan M. Duma, PhD,J. Wayne Meredith, MD, and Joel D. Stitzel, PhD

Background: Pulmonary contusion(PC) is commonly sustained in motor ve-hicle crash. This study utilizes the CrashInjury Research and Engineering Net-work (CIREN) database and vehicle crashtests to characterize the occupants andloading characteristics associated with PC.A technique to match CIREN cases to vehi-cle crash tests is applied to quantify the tho-racic loading associated with this injury.

Methods: The CIREN database andcrash test data from the National HighwayTraffic Safety Administration were usedin this study. An analysis of CIREN datawere conducted between three study co-horts: patients that sustained PC and anyother chest injury (PC� and chest�), pa-tients with chest injury and an absence ofPC (PC� and chest�), and a controlgroup without chest injury and an absenceof PC (PC� and chest�). Forty-one lat-

eral impact crash tests were analyzed andthoracic loading data from onboard crashtests dummies were collected.

Results: The incidence of PC in CIRENdata were 21.7%. Crashes resulting in PCdemonstrated significantly greater mortal-ity (23.9%) and Injury Severity Score(33.1 � 15.7) than the control group. Theportion of lateral impacts increased from27% to 48% between the control groupand PC� and chest� cohort, promptingthe use of lateral impact crash tests for thecase-matching portion of the study. Crashtests were analyzed in two configurations;vehicle-to-vehicle tests and vehicle-to-poletests. The average maximum chest com-pression and deflection velocity from thedummy occupants were found to be25.3% � 2.6% and 4.6 m/s � 0.42 m/s forthe vehicle-to-pole tests and 23.0% �4.8% and 3.9 m/s � 1.1 m/s for the vehicle-

to-vehicle tests. Chest deflection versustime followed a roughly symmetric andsinusoidal profile. Sixteen CIREN caseswere identified that matched the vehiclecrash tests. Of the 16 matched cases, 12(75%) sustained chest injuries, with halfof these patients presenting with PC.

Conclusions: Quantified loading atthe chest wall indicative of PC and chestinjury in motor vehicle crash is valuableboundary condition data for bench-topstudies or computer simulations focusedon this injury. In addition, because PCoften exhibits a delayed onset, knowingthe population and crash modes highlyassociated with this injury may promoteearlier detection and improved manage-ment of this injury.

Key Words: Pulmonary contusion,Thorax, Motor vehicle crash, Biomechan-ics, CIREN, Compression, Experiment.

J Trauma. 2009;66:840–849.

Motor vehicle crash (MVC) and associated injuriesremain a major public health problem world wide. Inthe United States, in 2005, there were 6 million

police reported crashes resulting in nearly 2 million injuriesand 40,000 fatalities.1 Analysis of occupant injury after MVCreveals that thoracic trauma is second only to head trauma interms of frequency.2,3 For patients sustaining blunt chesttrauma, pulmonary contusion (PC) is the most commonlyidentified injury, affecting 10% to 17% of all traumaadmissions.4 The mortality associated with PC is difficult todetermine but is estimated to be 5% to 25% for adults and ashigh as 43% for children.5–7 PC has been demonstrated as an

independent predictor of acute respiratory distress syndromeand pneumonia.8

It is well established that MVC is the most commoncause for PC in the civilian population, yet there are currentlyno data in the literature characterizing the occupants, crashmodes, and thoracic loading most commonly associated withthis injury. This article uses the Crash Injury Research andEngineering Network (CIREN) database and matched vehiclecrash tests to better characterize PC and associated thoracicinjuries.

Although PC has a mechanical cause, the response to theinsult demonstrates a strong inflammatory component.9,10 Asurvey of recent trauma and critical care research indicatesthat there is strong interest in mechanistic studies of PC.11–16

Submitted for publication July 26, 2007.Accepted for publication July 9, 2008.Copyright © 2009 by Lippincott Williams & WilkinsFrom the Departments of Biomedical Engineering (F.S.G., J.D.S.) and

General Surgery (R.S.M., J.J.H., J.W.M.) Wake Forest University School ofMedicine, Winston-Salem, North Carolina; Department of BiomedicalEngineering, (F.S.G., J.D.S.), Virginia Tech – Wake Forest University Cen-ter for Injury Biomechanics, and Department of Mechanical Engineering,Virginia Tech (H.C.G., S.M.D.), Blacksburg, Virginia.

Funding has been provided by Toyota Motor North America Inc. underCooperative Agreement Number DTNH22-05-H-61001. This publicationwas also supported by Grant Number 1 R49 CE000664-01 from the Centersfor Disease Control and Prevention (CDC).

Work was performed for the Crash Injury Research and EngineeringNetwork (CIREN) Project at Wake Forest University School of Medicine incooperation with the United States Department of Transportation/NationalHighway Traffic Safety Administration (USDOT/NHTSA).

Views expressed are those of the authors and do not represent the viewsof any of the sponsors, NHTSA, or the CDC. Its contents are solely theresponsibility of the authors and do not necessarily represent the officialviews of the National Center for Injury Prevention and Control.

Address for reprints: F. Scott Gayzik, MS, Medical Center Blvd., MRI2, Winston-Salem, NC 27157; email: [email protected].

DOI: 10.1097/TA.0b013e318186251e

The Journal of TRAUMA� Injury, Infection, and Critical Care

840 March 2009

These studies introduced animal models for the study of PCor described components of the host inflammatory response.The studies acknowledge MVC as a common cause of PC,yet little discussion is devoted to similarities between theloading applied to the model and that experienced by avehicle occupant.

The purpose of this work is therefore to quantify theloading conditions resulting in PC and its sequelae afterMVC-induced blunt chest trauma. The study proceeds inthree parts. The first is to use the CIREN database to inves-tigate the occupants, vehicle types, and crash modes mostcommonly associated with PC. The strength of the CIRENdatabase is that it contains detailed information on: 1. thecrash that caused Abbreviated Injury Scale (AIS) score of 3or greater injury in the enrolled patient and 2. the individual’shospital course. CIREN data are collected through an in-depth case review process conducted by participating physi-cians, radiologists, engineers, and crash investigators.

After the CIREN database study, data from lateral impactvehicle crash tests were analyzed to quantify crash parame-ters of interest including the mean vehicle crush, mean crash-induced change in vehicle velocity (Delta-V), and occupantloading data from the chest of an anthropometric test devicehereafter referred to as a dummy. For reasons based on theoutcome of the CIREN analysis, this analysis was restrictedto passenger vehicles in lateral impact tests.

The final aspect of this study was to query the CIRENdatabase for cases whose crash characteristics matched theanalyzed vehicle crash tests. Matching was conductedthrough analysis of the Collision Deformation Classification(CDC) code from the CIREN case reconstruction and bymatching exterior crush and Delta-V within one SD of thevalues determined through analysis of the crash tests. TheCDC code is a comprehensive crash classification standardestablished by the Society of Automotive Engineers (standardSAE-J224), including the direction of force, general area ofvehicle damage, and specific horizontal location of damagealong the struck side of the vehicle.

The hypothesis behind this unifying aspect of the study isthat injury data from matching CIREN cases coupled withthoracic loading data from the dummy situated in the vehiclecrash test will provide two important results: realistic loadingdata for future mechanistic studies of PC and associatedthoracic injuries, and insight into injury mechanisms associ-ated with MVC-induced PC and thoracic injury.

MATERIALS AND METHODSDatabases

The CIREN network was established by the NationalHighway Traffic Safety Administration (NHTSA) in 1998.The CIREN database includes roughly 3,500 MVCs, gener-ally with maximum AIS of 3 or greater. CIREN’s mission isto bring together engineering and medical knowledge duringa case review at which time the crash mechanics, biomechan-ics, and clinical aspects of injury are assessed. CIREN is one

of the most detailed databases of injuries sustained in carcrash-induced trauma and contains operative notes, radio-graphic images, and photographs of crash victims.

NHTSA’s Vehicle Crash Test Database is a comprehen-sive source of engineering data generated through vehiclecrash testing. It includes acceleration, force, and deflectiontraces from the vehicle; location and type of crash test dum-mies in the test vehicle; and additional media describing thetests such as crash test video and reports. This data areavailable online in a searchable archive maintained byNHTSA.17

CIREN Data CollectionThe CIREN database is continually updated as additional

case studies are completed from the eight centers throughoutthe United States. The CIREN database was downloadedcontaining all case information from 1996 through 2006. Atthe time of acquisition, the database contained 3,567 totalcases. Case occupants less than 15 years of age were notconsidered in the analysis, eliminating 963 cases.

The CIREN data were imported into SAS statisticalanalysis software, (version 9.1e, Cary, NC). The cases weresorted into one of three groups: occupants who had sustaineda PC and any other chest injury (PC� and chest�), occupantswho did not present with PC but experienced some form ofchest trauma (PC� and chest�), and a control group who didnot present with chest injury whatsoever (PC� and chest�).The first group is by definition positive for chest injurybecause PC is a chest region injury. Individuals in this groupmay have sustained additional chest injuries but wereplaced in this category based on the presence of PC alone.Cases were categorized after analysis of the patient’s AISfor each injury. Injury codes are derived directly from thediagnosis provided by the patient’s attending physiciansand radiologists.

Patient demographics (age, gender, height, weight), in-jury characteristics (Maximum Value on the AbbreviatedInjury Scale [MAIS], Injury Severity Score [ISS], mortality),and vehicle demographics were compared between cohorts.Based on the wheelbase length, vehicles were classified intothree groups: with small or compact vehicles having a wheel-base less than 265 cm, intermediate or full size vehicleshaving a wheelbase between 265 cm and 291 cm, and largevehicles having wheelbases exceeding 291 cm. Crash char-acteristics including the Principal Direction of Force of thecrash, Delta-V, and the complete CDC code allowed forthorough description of the crash characteristics of all casesin CIREN.

Statistical comparisons between groups were conductedvia analysis of variance, with Tukey’s post hoc test forpairwise comparisons between groups. �2 tests were used forcomparisons between categorical data. A significance level of0.05 was used in all statistical tests.

Thoracic Loading for Pulmonary Contusion

Volume 66 • Number 3 841

Crash Test Data CollectionVehicle Crash test data were downloaded from NHTSA’s

research and development website.17 This database contains asummary of nearly 6,000 tests from model year 1965 to thepresent containing data on the crash configuration, vehiclemodel, and type of dummy used in the test.

All data used in case comparisons originated from lateralimpact crash tests. Tests from four distinct lateral impactcrash test configurations were analyzed to broaden the rangeof crashes investigated. Two of these, the Lateral Impact NewCar Assessment Program (LINCAP) tests and the FederalMotor Vehicle Safety Standard 214 (FMVSS) tests, replicatevehicle-to-vehicle lateral impacts. The remaining two test con-figurations, FMVSS 201 pole and FMVSS 214 pole, replicatevehicle-to-pole type lateral impacts. The crash test configu-rations, parameters for each test, and unique test numbersused in this study are provided in Table 1. A completedescription of these lateral impact tests can be found in theliterature.18 The tests used in this study were commissionedby NHTSA for research purposes rather than specifically forregulatory purposes.

The analysis was limited to crash tests of passengervehicles (excluding light trucks and vans). In addition, onlytests using the EuroSID-2 (European Side Impact Dummy,2nd generation) dummy, or an updated version of the samedummy, the EuroSID-2re, were used. Potentiometers embed-ded in the thorax of this model record the lateral thoracicdisplacement during the crash. Data from the selected testswere obtained through NHTSA’s Signal Browser softwareand the vehicle crash test report.17 Crash test data collectedfor this study included the vehicle Delta-V (calculated fromonboard accelerometers),19 crush profile, and dummy ribdeflection. Peak rib deflection was normalized by half thewidth of the dummy thorax module and is expressed in termsof a percent compression.20 The maximum deflection veloc-ity was determined by differentiation of the deflection trace.Chest module loading data are presented for a subset of crashtests that were not equipped with side impact airbags, or inwhich the dummy did not contact the airbag.

Video and a written summary of each crash test were alsodownloaded through NHTSA’s research and developmentweb site. The data were used to verify the crash mode,cross-check test data, and ensure that no malfunction of thedummy or instrumentation was noted in the test.

Matching CIREN Cases to Crash Test CasesThe pool of potential CIREN cases was restricted to

passenger cars. This was performed to reduce variabilitybecause of the height of the occupant from the road and tomore closely match the crash tests analyzed. Matched casesmet three criteria. The case occupant was the front seatoccupant on the near side lateral impact position. The CDCcode, including direction of force, general area of damage,and other parameters describing the crash configuration, was Ta

ble

1Su

mm

ary

ofL

ater

alIm

pact

Tes

tsU

sed

inM

atch

ing

Pro

toco

l

Test

Bar

rier

Imp

act

Ang

leIm

pac

tS

pee

dTe

stD

umm

yM

ean

Whe

elb

ase

(cm

)

Mea

nW

eigh

t(k

g)N

HTS

ATe

stnu

mb

ers

(n)

FMV

SS

214

Mov

ing

def

orm

able

bar

rier

(1,3

67kg

)27

deg

rees

CW

from

late

ral,

driv

er-s

ide

54km

/h(3

3.5

mp

h)E

uroS

ID-2

/E

uroS

ID-2

re26

7�

121,

433

�14

45,

483,

5,48

2,5,

461,

5,48

1,5,

485,

4,54

7,4,

551,

4,86

2,(8

)

LIN

CA

PM

ovin

gd

efor

mab

leb

arrie

r(1

,367

kg)

27d

egre

esC

Wfr

omla

tera

l,d

river

-sid

e

62km

/h(3

8.5

mp

h)E

uroS

ID-2

/E

uroS

ID-2

re27

0�

121,

361

�22

93,

819,

3,79

9,4,

456,

4,09

7,4,

455,

4,38

0,3,

803,

(7)

FMV

SS

201

(late

ralp

ole)

Rig

idp

ole

(254

mm

dia

.)La

tera

lim

pac

t,d

river

-sid

e29

km/h

(18

mp

h)E

uroS

ID-2

/E

uroS

ID-2

re27

3�

3.7

1,50

7�

103

3,70

8,3,

703,

3,70

7,3,

740,

3,80

2,3,

818,

3,82

0,(7

)FM

VS

S21

4(o

bliq

uep

ole)

Rig

idP

ole

(254

mm

dia

.)15

deg

rees

CW

from

late

ral,

driv

er-s

ide

32km

/h(2

0m

ph)

Eur

oSID

-2/

Eur

oSID

-2re

270

�9.

01,

484

�13

65,

300,

5,43

9,5,

458,

5,40

7,5,

408,

5,47

2,4,

860,

4,38

9,4,

498,

4,37

8,4,

497,

4,42

3,4,

284,

4,28

5,4,

365,

4,24

6,4,

313,

4,85

9,5,

317,

(19)

Sid

eim

pac

tco

mp

aris

ons

are

rest

ricte

dto

pas

seng

erca

rs.

Test

ssh

own

inb

old

wer

ein

clud

edin

the

ribm

odul

elo

adin

gst

udy.


842 March 2009

consistent with one of the four crash tests analyzed. TheDelta-V and weighted average crush of the CIREN casevehicle were within one SD of the mean of these values fromthe analyzed crash tests.

The specific coding used to match each test is providedin Table 2. To be consistent with the vehicle crash test dataand to avoid a potentially confounding factor, matchingCIREN cases in which the occupant was protected by a sideairbag were excluded.

Vehicle crush data in CIREN was conditioned beforecase matching. The average crush was weighted differentlydepending on the type of crash analyzed. To quantify thevehicle’s crush profile, six equally spaced measurements ofthe deformation are taken along the extent of the vehicledamage. This standardized technique is applied to both theCIREN cases during vehicle inspection and after all crashtests. Rather than using the average of these crush measuresto compare vehicle crush between CIREN cases and the crashtests, weighted averages were used for both the vehicle-to-vehicle and vehicle-to-pole crash tests. The formulas forcalculating the weighted average crush are found in Table 2.The purpose of the weighted average was to accurately reflect

the specific type of crush. Depending on the CDC codeclassification (i.e. vehicle-to-vehicle or vehicle-to-pole typecrash), the appropriate weighted crush was applied to thematching algorithm.

RESULTSCIREN Data

Group populations, patient demographics, injury charac-teristics, and vehicle characteristics from the CIREN databaseare shown in Table 3. The incidence of PC in the total CIRENpopulation was 21.7% (566 of 2,604). Study cohorts wereprimarily examined relative to the control group (PC� andchest�). The PC� and chest� group were significantlyolder, and demonstrated a lower male-to-female ratio than thecontrol group. In terms of the injury data, the MAIS, ISS, andmortality were significantly elevated relative to the controlgroup.

MAIS, ISS, and mortality were highest in the PC� andchest� group. This group exhibited significantly greater in-jury data than the control group and PC� and chest� group.The male-to-female ratio was also majority male in the PC�

Table 2 Collision Deformation Classification (CDC) Code and Crush Calculations Used for Matching Crash InjuryResearch and Engineering Network (CIREN) Cases to Selected Lateral Vehicle Crash Tests

Vehicle to Vehicle Type Vehicle to Pole Type

FMVSS 214 LINCAP FMVSS 201 Pole FMVSS 214 Pole

Principal direction of force Lateral � 20 degrees Lateral � 20 degreesGeneral area of deformation Left or right (L or R)* Left or right (L or R)*Specific horizontal location Side center (P, Y, or Z) Side center (P, Y, or Z)Specific vertical location All, everything below beltline, or top of

frame to top of vehicle (A, H, E)All, everything below beltline, or top of

frame to top of vehicle (A, H, E)Type of damage Wide NarrowWeighted average crush calculation (cm) 10% � C1 � 20% � C2 � 20% � C3 �

20% � C4 � 20% � C5 � 10% � C6

10% � C1 � 10% � C2 � 30% � C3 �30% � C4 � 10% � C5 � 10% � C6

Weighted crush range (cm) 12.2–20.0 10.4–26.9Delta-V range (km/h) 24.1–27.7 26.8–32.0 30.1–35.2 32.3–38.3

* Depending on case occupant’s seating position.

Table 3 Crash Injury Research and Engineering Network (CIREN) Pulmonary Contusion Study Results

PC� and Chest Injury�(n � 951)

PC� and Chest Injury�(n � 1,087)

PC� and Chest Injury�(n � 566) p

Age (yr) 37.2 � 17.4 46.6 � 19.1 39.0 � 20.1 �0.0001*Male:female 1.09:1 0.88:1 1.17:1 0.02†

Height (cm) 171.0 � 10.2 169.6 � 11.6 170.8 � 12.7 0.01*Weight (kg) 79.7 � 23.1 81.1 � 23.1 80.6 � 37.2 0.51*MAIS 3.1 � 0.8 3.5 � 1.1 4.1 � 0.9 �0.0001*ISS 14.3 � 9.1 24.0 � 17.7 33.1 � 15.7 �0.0001*Mortality (%) 3.2 16.8 23.9 �0.0001�

Model year 1,998 � 3 1,997 � 4 1,998 � 4 �0.0001*Wheelbase (cm) 273 � 28 273 � 26 269 � 21 0.03*Weight (kg) 1,426 � 341 1,404 � 497 1,394 � 281 0.34*

Mean patient demographics, injury characteristics, and vehicle demographics. Significant differences from the control group (PC� andchest injury�) are in bold, � � 0.05.

* ANOVA with Tukey’s post hoc test.† �2.



and chest� group. No differences in height and weight werenoted among population groups.

The incidence of PC was also investigated with respectto vehicle parameters. Significant differences were noted inthe mean vehicle model year for the PC� and chest� group,but the difference was 1 year. The mean wheelbase of allvehicles was in the intermediate size range, but a significantreduction in wheelbase size was noted for the PC� group.This is reflected in Figure 1, which shows the incidence of PCby vehicle class. The incidence of PC remains steady acrossthe small and intermediate class and drops substantially inlarger wheelbase vehicles.

Crash characteristics from the CIREN database are foundin Figure 2, which shows the crash classification based on the

general area of damage (a part of the CDC code) for allcohorts. The most common crash mode for all groups wasfrontal, but the proportion of lateral crashes steadily increasesacross study cohorts. In the PC� and chest� cohort, nearlyhalf (48%) of crashes were lateral impacts.

Analyzing only those patients with PC in side impactreveals a strong influence between the impacted side of thevehicle and the location of the contused lung. Figure 3 showsthe breakdown of contusion classification based on struckside of the vehicle. Contused lung is strongly dependent onstruck side for both right and left impacts. Bilateral lungcontusions constitute substantial portions of both right andleft side impact.

The results of the CIREN portion of the study were usedto focus the search criteria for the vehicle crash test database.The results of this study demonstrated a sharp increase in theproportion of lateral impacts associated with PC. Althoughthe crash modes for the PC� and chest� cohort were essen-tially equally split between frontal and side impact, it isimportant to consider relative frequency of these crashmodes. It is well documented that frontal impacts occur withmore than twice the frequency of side impact crashes,1 yetside impact crashes represented nearly half of all PC cases inCIREN (Fig. 2). This disproportionate increase justified thestudy of thoracic loading in side impact.

Because the incidence of PC was highest in compact andintermediate vehicle types (Fig. 1), and as these vehicles aretypically passenger cars, limiting the crash test analysis topassenger cars was deemed a justifiable approach. The meanheight and weight of the three cohorts did not differ signifi-cantly and approximated the 50th percentile male values of170 lbs (77.3 kg) and 69 inch (175.3 cm).21

Vehicle Crash Test DataForty-one lateral impact crash tests were examined for

this study. Each test met all inclusion criteria for the study(passenger car, crash test configuration, and use of the spec-ified dummy). Impact speed and impact angle were tightly

Fig. 1. Incidence of pulmonary contusion in the Crash Injury Re-search and Engineering Network (CIREN) Database by vehicle class.

Fig. 2. Crash mode distribution by study cohort. (A) Without PC orchest injury (PC� and chest�). (B) Without PC but with chest injury(PC� and chest�). (C). With PC and chest injury (PC� and chest�).Labels indicate frontal (F), right (R), left (L), and rear (B) impacts.

Fig. 3. Distribution of pulmonary contusion versus side of vehiclestruck for lateral impact cases in the PC� and chest injury� cohort.


844 March 2009

controlled per the test stipulations. All vehicles selected foranalysis were common models. Figure 4 shows the averagecrush (nonweighted) and average Delta-V for lateral crashesin CIREN and for the analyzed crash tests. The mean non-weighted crush values from the crash tests were significantlybelow the same values from the CIREN database. The meanDelta-V aligned more closely, although it should be notedthat the Delta-V for the crash tests was measured via onboardsensors whereas the Delta-V for the CIREN data were esti-mated from the vehicle crush profile.

All chest compression data were extracted from the leftfront seat occupant and therefore represent the near-side crashscenario. Figure 5 shows exemplar traces of the occupantloading at the chest wall for each crash test type analyzed. Alltraces show a roughly sinusoidal loading pattern. The meanand SD of the maximum medial-lateral compression andmaximum deflection velocity are provided in Table 4. Crashtest cases used for analysis of the rib module loading areshown in bold in Table 1. Statistical tests (t test, � � 0.05) formaximum compression and for maximum deflection velocity

were conducted within each test type (i.e. within the vehicle-to-vehicle and vehicle-to-pole type tests). These tests indi-cated that maximum rib velocity was significantly differentwithin tests but the compression was not.

Matched CIREN and Vehicle Crash TestsMatching was performed independently for each crash test

type (Table 2). The inclusive ranges for matched Delta-V andweighted average crush for all test types are shown in Table 2.The average weighted crush between the two vehicle-to-vehicletests and the two vehicle-to-pole tests were not significantlydifferent, so a common range was used for this parameter.

A total of 16 matching CIREN cases were identified.Thirteen cases matched the vehicle-to-vehicle type crashtests. The breakdown of specific matching cases is shown inFigure 6. No cases were found to match all vehicle-to-poletest criteria because many pole-type cases in CIREN lacked acalculated Delta-V. When only the Delta-V criterion wasdropped, three cases were found to match the pole tests. Nodifferentiation among the two vehicle-to-pole tests weremade because the weighted crush was not found to be sig-nificantly different (Table 2).

In terms of injury data, 4 of the 16 cases did notsustain chest injury, but all of these patients sustainedpelvic fractures. All of the remaining 12 cases sustained achest injury, and half of these cases presented with PC. Figure6, B shows the mean age by cohort. The mean age for thePC� and chest injury� group, 33.4 � 13.0 years, was foundto be significantly less than the PC� and chest� and PC�and chest� cohorts (53.3 � 23.9 years and 60.6 � 32.3years, respectively) through t test between cohorts (� �0.05). The control group was significantly shorter than theremaining cohorts (p � 0.006, analysis of variance, � �0.05). No other demographic, injury, or vehicle data differedsignificantly by cohort. There was one fatality in the matchedcase group, which matched the Lateral Impact New Car

Fig. 4. Average crush (A) and Delta-V (B) for lateral impacts fromCIREN database and analyzed crash tests with one SD error barsshown. The legend is equivalent for (A) and (B).

Fig. 5. Exemplar medial-lateral compression versus time from thedummy chest module in analyzed crash tests. Traces begin at theinstant of compressive loading. Test numbers are 4,547, 3,799,3,708, and 4,313 respectively, based on legend order.



Assessment Program-type test loading belonging to the PC�and chest� cohort. The case occupant was an 89-year-oldmale, right front seat passenger who was restrained. Most ofthe patients in the matched cases were restrained by a 3-pointbelt (14 of 16). There was no indication from the case vehiclereports that unrestrained occupants were out of position. AllPC� patients were restrained.

DISCUSSIONThe aims of this study were twofold: to characterize the

occupants, injuries, and crash modes associated with PC and

apply a case-matching algorithm to the CIREN database toquantify the associated thoracic loading conditions via com-parisons with similar laboratory-conducted crash tests. Theresults of this study provide several valuable contributions tothe trauma research community regarding these aims. First,because PC can be an insidious injury exhibiting a delayedonset, knowing the population and crash modes associatedwith this injury could lead to earlier detection and improvedmanagement of the injury. Second, well-quantified loadingconditions at the chest wall indicative of traumatic PC (andmore generally, chest injury) will be a useful tool for design-ing future bench-top mechanistic studies of PC as well as forcomputer simulations focused on studying thoracic trauma.Additionally, this data will be useful to the automotive safetycommunity with respect to the goal of preventing or mitigat-ing thoracic injuries.

Characterization of the CIREN population was achievedthrough analysis of mean demographic, injury, and vehicledata for three distinct cohorts (Table 3). These were a controlgroup without chest injury, a group with some form of chestinjury in the absence of PC (typically rib fractures), and afinal group with PC. These three cohorts revealed severalinteresting trends. Although there was no significant differ-ence in age between the control group and the PC� andchest� group, the PC� and chest� group was found to besignificantly older than the controls (37.2 � 17.4 years vs.46.6 � 19.1 years). Because Delta-V was fairly uniformacross all crash tests, these results suggest that age plays arole in the likelihood of developing PC after blunt chesttrauma. In addition, significant gender differences were notedbetween cohorts in the characterization study with a greaternumber of males presenting with PC. As a whole, CIRENpatient data were nearly evenly split between sexes (male-to-female ratio: 0.99:1). Males represented a larger proportion ofdrivers (54.7%) and cases with male drivers were associatedwith a significantly greater Delta-V (43.3 � 19.7 km/h) thanfemale drivers (40.2 � 17.2 km/h) (t test, � � 0.05). Basedon the results of this study, it appears that the occupants’physical size, the vehicle wheelbase, or the vehicle weight arenot predictive factors for sustaining PC in a vehicle crash.

There were significant differences between cohort injurycharacteristics. Patients in the PC� and chest� group exhib-ited significantly greater MAIS, ISS, and mortality. Our re-sults suggest that this difference is predominantly because ofthe crash mode and crash severity. Figure 4 shows that lateral

Fig. 6. Matched case summary. (A) Breakdown of cohort popula-tion for cases in the Crash Injury Research and Engineering Net-work (CIREN) database that matched vehicle crash tests. Matchingtest configuration for each cohort is also shown. (B) Mean and SDof patient age by study cohorts for matched case subpopulation.

Table 4 Mean and Standard Deviation of Rib Module Compression and Velocity for Analyzed Cases

Test TypeTest (n)

Vehicle-to-Vehicle Type Vehicle-to-Pole Type

FMVSS 214 (4) LINCAP (4) FMVSS 201Pole (3)

FMVSS 214Pole (6)

Maximum M-L compression (%) 19.6 � 3.2 26.4 � 3.4 27.2 � 1.0 24.7 � 2.7Maximum deflection velocity (m/s) 2.9 � 0.3* 4.9 � 0.4* 4.3 � 0.3* 4.7 � 0.4*

Sample size for each test is given. Tests included in this analysis are shown in bold in Table 1.* Statistically significant differences within test types (t test, � � 0.05).


846 March 2009

Delta-V and average crush increase from the control groupand are greatest for the PC� and chest� group. Lateralimpact crash proportions are greatest for the PC� and chest�cohort. In lateral impact scenarios, the occupant is situated closeto the point of momentum transfer of the impacting object.19

This leads to severe loading of the near side of the occupant(typically the chest, pelvis, or head). As a result of the detailedinjury analysis in the CIREN database, correlation of internalinjuries to external loading was possible. The influence of theoccupant’s position within the vehicle on the resulting contusionis clear in Figure 3, through the high correlation between theaspect of the contusion and the crash mode.

The chest loads associated with the crashes were calcu-lated through analysis of the vehicle crash tests. The loadingof the dummy chest module was quantified using two vari-ables; maximum medial-lateral compression and the maxi-mum deflection velocity. Table 4 provides the averages ofthese values by test type. The data from the dummy chestmodule during the crash test in conjunction with the outcomedata from matched CIREN cases provide a good estimate ofthe loading that caused these patients’ chest injuries.

The approach taken here provides valuable informationwith regard to loading boundary conditions for blunt chesttrauma models. The percent compression, deflection velocity,and general shape of these loading curves are ideal inputparameters for future experiments. In addition, as computersimulations of the human body for predicting injury becomemore prevalent, this type of data will be useful for recreatingthe appropriate injurious event.22

The average peak chest compression and rib deflectionvelocity (25.3% � 2.6% and 4.6 m/s � 0.42 m/s) was slightlygreater in the pole tests than in vehicle-to-vehicle type tests(23% � 4.8% and 3.9 m/s � 1.1 m/s). The difference indeflection velocity between pole and vehicle configurationswere significant (p � 0.05, t test, � � 0.05), and the differ-ence in compression approached significance (p � 0.11).Whether the data were examined from a pole-type or vehicle-type test, the loading data were roughly sinusoidal in nature,with a symmetric loading and unloading phase. Peak velocitywas achieved before peak compression. Given the lung’srelative compliance and coupling to the chest wall, theseexternally measured loading parameters should be interpretedas loading conditions at the surface of the lung. Previousresearch on thoracic loading in side impact has shown that thedoor velocity-time profile heavily influences the severity ofthe loading.23 Although it is door deformation and movementthat causes chest loading, measuring deflection directly fromdisplacement sensors embedded in the chest module of thedummy captures this induced loading in the most directmanner possible.

It is useful to compare these findings with recent mech-anistic studies of PC using animal models.11–16 In four ofthese six studies, blast-rate loading was applied to the studysubject, an insult characterized by extremely high deflectionvelocity (50 m/s) and low compression (�5%).24 Our results

show that this type of loading is markedly different from thatof MVC, which would be characterized by loading at areduced deflection velocity with increased chest compres-sion. Although blast rate loading is appropriate for ballisticstudies, our results indicate that the insult does not replicateloading in MVC. Two of the aforementioned studies appliedchest loads that approximate what was found in thiswork.11,12 The former used an electronically driven piston todirectly strike the lung of the subject at 2.8 m/s to a depth of6.3 mm (correlates to roughly 30% compression). The latterused a drop device to achieve loading speeds between 3.4 m/sand 4.2 m/s, but did not control chest compression. Methodsof blast-type loading included the use of a captive-bolt hand-gun, nail driver, and blast wave generator. In the studies usingblast-type loading, neither deflection velocity nor chest com-pression was reported.

Twelve of the 16 matching CIREN cases (75%) exhib-ited chest trauma and half of these cases presented with PC.Clearly, factors beyond chest wall loading alone influence thespecific types of thoracic injuries that will be present afterblunt chest trauma, but the results strongly indicate that thedummy chest loading data are indicative of PC. When con-sidering all matched cases, regardless of which type of testeach matched, statistical tests comparing the mean values ofthe subpopulations (categories in Table 3) showed that onlytwo parameters, age and height, were significantly differentbetween groups. Although hindered by the limited samplesize of matched cases, these findings again suggest that ageplays a role in determining thoracic injury outcome given asimilar loading profile. This finding is consistent with studiesfocusing on the aging thorax, which shows that aging isassociated with structural and morphologic changes that canincrease susceptibility to disease and decrease the body’sability to withstand traumatic insults.25,26

In an effort to normalize the study results, only passengercars were considered in the matched case analysis. This wasa reasonable restriction because the incidence of PC washigher in smaller vehicles (Fig. 1), and these vehicles tend tobe passenger cars. Furthermore, in the majority of CIRENcases (76%) the case vehicle was a passenger car. The studyfindings are also consistent with the observation that there arecrash compatibility deficiencies between many passengercars and light trucks and vans. In 56% (9 of 16) of matchedcases, the striking vehicle was a light truck, van, or SUV. Inall PC� matched cases, the striking vehicle fell into thiscategory or was a vehicle-to-pole type crash. These resultssuggest that loading was either uniform along the verticalextent of the vehicle or presented at the height of the occu-pant’s chest. The dummy data also support these findingsbecause the pole type exhibited a slightly greater peak com-pression and deflection velocity than the vehicle-to-vehicletests. This is not meant to dismiss the severity of the vehicle-to-vehicle test data presented in this study. In fact, the lonepatient fatality among the matched CIREN case subset wasinvolved in a vehicle-to-vehicle lateral impact.



Although CIREN data do not represent a broad, population-based sample, the database does have unique strengths(namely extensive injury and crash reconstruction data).These attributes make CIREN data especially well suited forthis type of study where the injury in question and vehiclecrash data are closely linked and equally important to theconclusions.

The presence of multiple injuries could be viewed as apotentially confounding factor between cohorts. As evi-denced by the elevated ISS of the PC� group, it is possiblethat shock induced release of proinflammatory mediatorscontributed to elevated diagnosis of PC in that cohort.27

CIREN cases are frequently severe crashes and in many casespatients have suffered multisystem trauma. Thus, althoughthere is the possibility that a systemic inflammatory responsecould be misconstrued as PC, injury data are only entered inthe database after detailed case review by the patient’s at-tending physicians and radiologists, ensuring that the diag-noses in CIREN are as accurate as possible.

All the thoracic loading data presented in this work arederived from a 50th percentile male dummy. Crash testsmeeting the inclusion criteria that utilized female dummieswere unavailable. We do not believe that this result had alarge influence on the results, although 10 of the 16 matchedCIREN case occupants were female. The mean height andweight of the matched case individuals were 168 cm � 9.4cm and 72.6 kg � 15 kg, respectively, well within one SD ofthe standard height and weight of the 50th percentile male(175.3 cm and 77.3 kg)21 In addition, the dummy used inthis work has demonstrated improved biofidelity over pre-vious models28 and NHTSA has adopted its latest iteration(EuroSID2-re) for use in federally mandated vehicle sideimpact testing.

CONCLUSIONSA study using the CIREN database was conducted to

characterize the demographics, injury, and vehicle character-istics of patients presenting with PC. A procedure was devel-oped to identify cases in the CIREN database that closelymatched laboratory-conducted lateral vehicle crash tests.Compression and deflection velocity of the chest wall asso-ciated with these crashes were collected from the dummyoccupants seated in analyzed crash tests. This data are in-tended for use as boundary condition data for future mecha-nistic studies of PC and for computer simulations focused onthe study of PC and blunt thoracic trauma in MVCs.

The results showed that patients presenting with PCexhibit significantly greater MAIS, ISS, and mortality than allother cohorts. A separate cohort including only those indi-viduals with a chest injury in the absence of PC were alsofound to exhibit increased MAIS and ISS values over controls(no chest injury). The mean age of individuals in this group(PC� and chest�) was significantly greater than the controlgroup. CIREN cases presenting with PC were disproportion-ately involved in lateral impacts. A total of 16 CIREN cases

were found to match the analyzed crash tests. Seventy-fivepercent of these cases exhibited an AIS score of 3 or greaterchest injury. Half of the matched cases with chest injuriesalso presented with PC.

The average maximum chest compression and rib deflec-tion velocity from the dummies was slightly greater in vehicle-to-pole type tests (25.3% � 2.6% and 4.6 m/s � 0.42 m/s)than in vehicle-to-vehicle type tests (23% � 4.8% and 3.9m/s � 1.1 m/s). Loading was roughly sinusoidal and sym-metric, with peak velocity achieved before peak compression.Pole type impacts exhibited significantly greater deflectionvelocity. Like the total CIREN population, age was found tobe a significant factor between cohorts within the matchedcase subpopulation. The PC� and chest� cohort was signif-icantly younger (33.4 � 13.0 years) than the remaining co-horts (PC� and chest� at 53.3 � 23.9 years and PC� andchest� at 60.6 � 32.3 years).

Through correlation with CIREN injury data, the resultsprovide well-quantified chest loading data from MVC result-ing in PC. This loading is markedly different from blast-typeloading, which has been used in some mechanistic studies ofthis injury. We recommend using equipment that providesloading similar to that provided in this study for future bench-top studies of crash-induced PC. Computational studies ofthis injury can also use these findings to provide realisticboundary conditions for human body models.

ACKNOWLEDGMENTSThe authors gratefully acknowledge the reviewers of this manuscript

for their helpful comments during the revision process. The authors thankMatt Craig of NHTSA for his helpful suggestions and assistance in identi-fying appropriate crash tests for use in this study. Mao Yu of WFU aided inthe reviewing process. Randy Kelly of Denton ATD aided in data collectionof EuroSID-2 and EuroSID-2re ATDs.

REFERENCES1. United States Department of Transportation, National Highway

Traffic Safety Administration. Traffic Safety Facts 2006: ACompilation of Motor Vehicle Crash Data from the FatalitiesAnalysis Reporting System and General Estimates System.Washington, DC: United States Department of Transportation,National Highway Traffic Safety Administration; 2006. DOT HS810 818.

2. Blansfield J, Danis D, Gervasini A, Sheehy S. Manual of TraumaCare: The First Hour. St. Louis, MO: Mosby; 1999.

3. Cavanaugh KJ. Thoracic biomechanics. In: Nahum A, Melvin J, eds.Accidental Injury Biomechanics and Prevention. New York, NewYork: Springer-Verlag; 1993.

4. Cohn SM. Pulmonary contusion: a review of the clinical entity.J Trauma. 1997;42:973–979.

5. Stellin G. Survival in trauma victims with pulmonary contusion. AmSurg. 1991;57:780–784.

6. Balci AE, Balci TA, Eren S, Ulku R, Cakir O, Eren N. Unilateralpost-traumatic pulmonary contusion: findings of a review. SurgToday. 2005;35:205–210.

7. Allen GS, Coates NE. Pulmonary contusion: a collective review. AmSurg. 1996;62:895–900.

8. Miller PR, Croce MA, Kilgo PD, Scott J, Fabian TC. Acuterespiratory distress syndrome in blunt trauma: identification of


848 March 2009

independent risk factors. Am Surg. 2002;68:845–850; discussion850–841.

9. Stitzel J, Gayzik F, Hoth J, et al. Development of a finite elementbased injury metric for pulmonary contusion. I. Model developmentand validation. Stapp Car Crash J. 2005;49:271–289.

10. Gayzik FS, Hoth JJ, Daly M, Meredith JW, Stitzel JD. A finiteelement based injury metric for pulmonary contusion: investigationof candidate injury metrics through correlation with computedtomography. Stapp Car Crash J. 2007;51:189–209.

11. Raghavendran K, Davidson BA, Helinski JD, et al. A rat model forisolated bilateral lung contusion from blunt chest trauma. AnesthAnalg. Nov 2005;101:1482–1489.

12. Hoth JJ, Stitzel JD, Gayzik FS, et al. The pathogenesis of pulmonarycontusion: an open chest model in the rat. J Trauma. 2006;61:32–44; discussion 44–35.

13. Knoferl MW, Liener UC, Seitz DH, et al. Cardiopulmonary,histological, and inflammatory alterations after lung contusion in anovel mouse model of blunt chest trauma. Shock. 2003;19:519–525.

14. Wang ND, Stevens MH, Doty DB, Hammond EH. Blunt chesttrauma: an experimental model for heart and lung contusion.J Trauma. 2003;54:744–748.

15. Davis K, Fabian T, Croce M, Proctor KG. Prostanoids: earlymediators in the secondary injury that develops after unilateralpulmonary contusion. J Trauma. 1999;46:824–832.

16. Cohn SM. Experimental pulmonary contusion: review of theliterature and description of a new porcine model. J Trauma. 1996;41:565–571.

17. National Highway Traffic Safety Administration [Vehicle Crash TestDatabase] Available at: http://www.nrd.nhtsa.dot.gov. AccessedSeptember 1, 2007.

18. Chou C, Aekbote K, Le J. A review of side impact component testmethodologies. Int J Vehicle Saf. 2007;2:141–184.

19. Gabler H, Hackney J. Safety performance of production vehicles inside impact. Paper presented at: Tenth International Conference onEnhanced Safety of Vehicles; July 1985; Oxford, England.

20. Schuster P, Franz U, Stahlschmidt S, Pleschberger M, Eichberger A.Comparison of ES-2re with ES-2 and USSID dummy, considerationsfor ES-2re model in FMVSS tests. Paper presented at: 3rd AnnualEuropean Ls-Dyna User’s Conference; 14–15 October 2004;Bamberg, Germany.

21. Schneider L, Robbins D, Pflug M, Snyder R. Development ofanthropometrically based design specifications for an advanced adultanthropomorphic dummy family. Vol 1. Washington, DC:Department of Transportation; 1983. DOT/HS 806 715.

22. Roberts JC, O’Connor JV, Ward EE. Modeling the effect of non-penetrating ballistic impact as a means of detecting behind armorblunt trauma. J Trauma. 2005;58:1241–1251.

23. Kent R, Crandall J, Butcher J, Morris R. Sled system requirementsfor the analysis of side impact thoracic injury criteria and occupantprotection. Paper presented at: SAE World Congress (2001-01-0721);2001; Detroit, MI.

24. Viano D, Lau I. A viscous tolerance criteria for soft tissue injuryassessment. J Biomech. 1988;21:387–399.

25. Kemper A, McNally C, Kennedy EA, et al. Material property ofhuman rib cortical bone from dynamic tension coupon testing. StappCar Crash J. 2005;49:199–230.

26. Slice DE, Stitzel JD. Modeling age-related changes in human ribcage geometry. Am J Phys Anthrop. 2005;126:192.

27. Hoth JJ, Hudson WP, Brownlee NA, et al. Toll-like receptor 2participates in the response to lung injury in a murine model ofpulmonary contusion. Shock. 2007;28:447–452.

28. Byrnes K, Abramczyk J, Berliner J, et al. ES-2 dummy biomechanicalresponses. Stapp Car Crash J. 2002;46:353–396.



Documents

Characterization of Crash-Induced Thoracic Loading ... · a case review at which time the crash mechanics, biomechan-ics, and clinical aspects of injury are assessed. CIREN is one