Radiographic Detection of Gravel in Soft Tissue

GENERAL CLINICAL INVESTIGATION/ORIGINAL CONTRIBUTION

Radiographic Detection of Gravel in Soft Tissue

From the Emergency Medlcrne and Trauma Center, * RadIOlogy Depm tment, 1 and Depm tment of Me,hwl RClcarch, q Methodist HospItal of Indiana, IndianapolIs, IN

Carey DChisholm. MD' Christopher 0 Wood' Gonzolo Chua. MD* William HCordell, MD' David RNelson, MS§

ReCflvcd j'lr pllblrc,rtlOn Mav 20, 1996 Revillml rcceived Novcmro 18, 1996 Accepted for PllbllcatlOn DeLembo 2, 1996 P, esmted at the SOCIety for Aeadfll!lc

Emergcncv Medrerne An/wal Meetrng, Denver, May 1996

Funded rn part bv a [vIetllOdlst Hospital of Indwna 1995 Summo Student RClfLlrch Gurnt Copvnght © bv the Amenean College of Emfl,~cncv Phvslcram

Study objective: We sought to quantify the detectable size of varying compositions of gravel uSing acadaveric chicken leg wound model and standard plain-film two-view radiographs Methods: We conducted a randomized, blinded, descriptive study With the assistance of faculty from the emergency medicine and radiology residency programs of a private urban teaching hospital, A standardized wound was created in each of 160 cadaver chicken legs. Zero, one, or two pieces of gravel of four differing composItions, ranging in size from .25 to 2,0 mm, were inserted into the wounds as determined with computer-generated randomization. The legs were then radiographically imaged (anteroposterior and lateral views). Three faculty physicians independently interpreted the radiographs to determine the number of foreign bodies and rated the ease of viSibility, We calculated sensitivity, specificity, and interobserver reliability, Results: The accuracy with which gravel was detected ranged from an average of 97.7% for 2-mm and 1-mm particles to less than 75% for ,5-mm and ,25-mm particles, Visibility ratings were also lower for particles in the smaller ranges. Sensitivity was greater for the emergency physicians than for the radiologists, but their specificity was lower Salt-and-pepper gravel was the most easily identified foreign body. Conclusion: In thiS wound model, gravel particles of less than 1 mm were not accurately identified,

[Chisholm CD, Wood CO, Chua G, Cordell WH, Nelson DR: Radiographic detection of gravel In soft tissue. Ann Emerg Med June 1997;29:725-730,]

JUNE 1997

296

ANNALS OF EMERGENCY MEOICINE

725

GRAVEL ChLsholm et al

INTRODUCTION A wound foreIgn body IS defmed as any substance introduced mto soft tlssue, elther accidentally or mtentlOnally. 1 ForeIgn bodIes can vary in compositlOn, and they mayor may not be visualized with standard imaging techmques. Some foreign bodies cause complicatlons, whereas others remain asymptomatlc and undetected. Injury resultmg from a foreIgn body IS one of the most common mJury types m the hand 2 and is a hlgh-nsk mjury in the foot, knee, and elbow. Detection and management of foreign materials m soft tlssues are common problems m the ED 2 .3 In 200 consecutlve cases reviewed by Anderson 1 mvolvmg foreIgn body-related hand mJunes, retamed foreign bodIes were mIssed in 75 patients. The foreIgn body was radiopaque in 58 of the 75 cases Failure to dIagnose a retained foreign body is a leading cause of malpractlce htigation against emergency physIcians. 3, 4 The authors of past studies have focused on the detectlon of nonmetal foreign bodies ranging from glass S- 11 to thoms,lO wood,s,8,9 cactus spines,lO plastlC,8,1l and other materials. Although several mvestlgators have addressed the radiopaque properties of glass and the detection of glass foreign bodies in wounds usmg plam radlOgraphy; only qualitative data exist about the propernes of gravel embedded in wounds.1,5,9,1l In thIS study we sought to gather quantitative data about the detection of gravel with plam radlOgraphy and to compare the accuracy of detection by radiologIsts and emergency physlClans. We attempted to determme the smallest gravel that can be reliably detected using standard two-VIew plam radlOgraphy. A chlCken leg model was used

because It simulates the human hand in terms of the proximity of bone, tendon, and muscle. 6 ,9

MATERIALS AND METHODS One hundred sixty refrigerated chicken legs were obtained for use m thIS study. Four gravel types were selected from four regions of the United States on the basis of varymg mmeral composition and regional geology. Volcanic crater rock was selected from Cahforma, Meramec rock (granite) from Mlssoun, pea gravel from Indiana, and salt-and-pepper rock from North Carolina. Usmg a hammer, chIsel, magmfying glass, and ruler, we broke these rocks into four sizes rangmg from 2.0 mm x 2.0 mm to .25 mm x .25 mm (Figure I). We chose these sizes on the basis of pdot studIes in which the ease of radiographic detection on was assessed on plam films 0- to 5-mm fragments). The chicken legs were placed in a standardized arrangement in groups of 10 on a cardboard tray, dependmg on assignment by a computer-generated random-numbers table. Usmg a no. II surglCal scalpel, we mClsed each leg m a predetermmed anatomlC location to create a umform amount of tIssue dIstortion regardless of whether a gravel fragment was to be inserted. Two incisions were made in legs recelVmg two gravel fragments. We made the incisions by puncturmg the leg to the depth of the scalpel blade and creating an mClsion the length of the scalpel blade (fmal wound size, approximately 15 x 15 mm). The hrst incision was always made through the largest muscle group, located along the posterolateral border of the leg. A second mClsion, when

Figure 1.

RelatIve gravel sizes Gcft to right): 2 mm, 1 mm, .5 mm, and 25 mm.

1 726

2

ANNALS OF EMERGENCY MEDICINE

296

JUNE 1997

GRAVEL ChIsholm et al

required, was made approximately 1 cm below the first inClsion or opposite the first incislOn We mserted gravel fragments according to the randomization generated by the computer program mentloned earlier. Group A, the control group, compnsed 40 legs glVen standard double inCiSlOns but no gravel. Groups B through G each comprised 20 legs divided mto four subsets of five. Each of these subsets contamed only one of the four types of gravel. Group B comprised legs containing 2.0 X 2.0-mm gravel fragments. The group C legs contamed 1.0 X 1.0-mm gravel fragments. The group D legs contained .5 X .5-mm gravel fragments. The group E legs contamed .25 x .25-mm gravel fragments. Two fOreign bodies (in different incisions) were lffiplanted in each leg in groups F and G. Group F legs were each implanted with one 2.0 x 2.0-mm fragment and one .5 x .5-mm fragment, whereas group G legs recelVed one 1.0 x l.O-mm fragment and one .25 x .25-mm fragment. We inserted gravel to the maXimum depth of the wound using tweezers. We then closed the wound by pressing the sides of the incision together and replacmg any overlying skin. The legs were subjected to radiography on the cardboard trays m randomized groups of 10. All legs

were imaged in the anteropostenor and lateral positions (Figure 2). Three board-certified faculty physicians from the radiology department interpreted the radiographs. Each was blinded to the other physiClans' mterpretatlons The physicians were told that each chlCken leg contamed an undisclosed number of gravel fragments, in some cases zero. Each physiClan recorded the presence or absence of foreign bodies and the number of foreign bodies present m each leg. In addition, each physiClan was mstructed to rate the ease of visibility of any detected foreign body on a scale of zero (definitely not visible) to 5 (defmitely visible). Three emergency medicine residency faculty then followed the same procedure m reviewmg the radlOgraphs. Magmfymg lenses were not used by the reviewers. We compared the accuracy rate (percentage of radiographs m which the presence or absence of any foreign body was correctly assessed) and the rate of correct estimation of the number of foreign bodies (percentage of radiographs in which the correct number of foreign bodies was correctly assessed) among physicians using lOgiStic regresSlOn. Contrasts between emergency physicians and radiologists were used to compare the two speClaltles. Similarly,

Figure 2.

Radiograph of ch!cken legs w!th !mplanted gravel fragments.

JUNE 1997

296

ANNALS OF EMERGENCY MEDICINE

727

GRAVEL Clmh"lm ct ol

10gistic-regresslOn analyses mvolvmg two subsets' radlOgraphs (only those WIth foreIgn bodIes and only those without foreIgn bodIes) were used to compare sensitivity and specIficity, respectIvely, of physlClans and speClalnes. We evaluated agreement among physlClans usmg average Kvalues. To assess the relanonshlps between the characteristics of the foreIgn bodIes and the abilIty to correctly detect theIr presence and number, we performed 10glstlC-regression analyses usmg only those radlOgraphs containmg partlcles. These analyses assessed physician, size of foreign bodIes, number of foreIgn bodIes, and gravel type. We also used loglsnc regression to compare the eight groups of foreIgn body SIze and number PairWISe tests with Bonferrom-adjusted P values were used for mtravanable compansons. VisibIlity ratmgs were compared WIth ANOVA and Tukey's mulnple-companson test results. All computations were performed WIth PC-SAS verSlOn 6 10 stansncal software (SAS, Incorporated).

RESULTS The SIX physIcIans vaned slgmfIcantly m theIr accuracy, sensItiVIty, speCificIty, and estimatlOn of the number of foreign bodies (P< 001, Table 1) However, the overall accuracy rate dId not differ sigmfIcantly between radlOloglsts (80.0%) and emergency physicians (81.0%) (P=.60, Table 1). We did note significant differences, however, m the senslnvity and speciflClty rates of the two speClalnes. Emergency physicians were sigmficantly hIgher m SenSltlVlty (90.3%) Figure 3.

Sensitivity and spcciftuty of dctcctmg prcscnce/absencc % Accurately Detected

Table 1.

DetcctlOn of gravel fragments among all phVS1CWIl5 and betwcen speCltlltlcS.

120 100 0% 100

992%

975%

Comparison

80 60 40 20

a

2 and 5 mm

10mm 1 and 25 mm 2 0 mm

Sensitivity

728

compared WIth radlOlogists (78.1 %) (P<.OOl), whereas radIOlOgIStS were slgmficantly higher in speClfiClty (85.8%) than emergency physiClans (53.3%) (P<.OOl). RadlOloglsts correctly assessed the correct number of foreign bodies signifIcantly more often (64.4%) than emergency physlClans (546%) (P = .003). The mterobserver agreement (K-value) among all partlClpants was .49 (95% confidence interval, .35 to .62); It dId not dIffer between specialties. The sensttivltles and speClficines of detection of the various partlcle sizes and numbers are Illustrated m Figure 3. Loglsnc regression indicated that SIze and number of foreign bodies was sigmflCantly related m the interpretatlOn accuracy of presence and number of foreIgn bodIes (Table 2). Loglsnc regresslOn also mdlCated that as partlcle SIze mcreased, the physlClan's abilIty to detect (P<.OOl) and count (P< 001) mcreased. As the number of particles mcreased, the abIlIty to detect the presence of any particles increased (P=. 01 ); however, the physlClan's ability to accurately count partIcles decreased (P=.02). We noted a slgmfIcant dlmmunon of mterpretanon accuracy for particles between 1 and 5 mm. ThIS dlmmution was manifested m three ways. In detection of the presence of foreIgn bodies, accuracy rates were sIgnificantly hIgher for radiographs of legs contaming at least one foreign body of 1 or 2 mm (overall accuracy, 95.6%) than for those of legs containing only .5-mm (71.7% accuracy) and .25-mm (58.3% accuracy) foreign bodies (Table 3). Second, accuracy in correctly esnmatmg the number of foreIgn bodies was sigmflCantly less If at least one .5-mm or .25-mm partlCle was present, whereas the hIghest counting accuracy rates were observed when only a I-mm or 2-mm partlCle was present (Table 3). In addition, of the radiographs of legs containing foreign bodIes, the physiClans' VISIbIlIty ratmgs

5mm

25mm

No Foreign Bodies Specificity

P*

Overall accuracy (all radiographs) PhysiCian Emergency phySICians (81 0%) versus radiologists (80 0%) Sensitivity (only radiographs of legs containing fragments) PhySICian Emergency phYSICians (90.3%1 versus radiologists (78 1%1 Specificity (only radiographs of legs containing no fragments) PhySICian Emergency phySICians (53 3%1 versus radiologists (85 8%1 Correct estimation of number of particles lall radiographs) PhySICian Emergency phYSICians (54 6%1 versus radiologists (64 4%) *Loglstlc regression

ANNALS OF EMERGENCY MEDICINE

296

< 001 60 <.001 < 001

< 001 < 001 < 001 003

JUNE 1997

GRAVEL

ChIsholm ,t al

were slgnificantly lower 1f only a .S-mm or .2S-mm particle was present. We also detected slgmhcant d1fferences among the four types of gravel. The highest rates of accurate identihcatlOn of the presence and number of fOrelgn bod1es were observed for salt-and-pepper gravel, whereas crater rock had the lowest accuracy rates (Table 4) These two groups d1ffered Slgmf1cantly Wlth regard to accuracy rates, whereas pea gravel and Meramec rock were intermed1ate V1S1b1hty ratings for salt-and-pepper rock were also h1ghest, whereas those for Meramec and crater rock were signihcantly lower (Table 4).

DISCUSSION Foreign bod1es are a high-nsk area of wound care. Although focused h1story questions and phys1cal examination may prov1de clues to the presence of a retamed fOrelgn body, its presence may not always be ObVlOUS 12 Many beheve that plam radlOgraphy 1S mdicated if a retamed fore1gn body is suspected, parucularly 1f the object under Susp1Clon 1S known to be radlOpaque. Anderson 1 reported that 80% of foreign bod1es m the hand m1ssed on initial physlCal exam1natlOn were detectable on plam radlOgraphy Because radiography 1S based on unequal absorption of rays as they pass through objects of d1ffenng densltles and effective atomlC numbers? a forogn body 1S detectable on the basis of 1tS denslty relative to that of the surroundmg tissues. Gravel parucles are cons1dered radiopaque because their density 1S so much greater than that of soft tissue. However, th1S does not mean that all gravel parucles w1ll be routmely detected by plam radlOgraphy The detection of a radlOpaque fore1gn body becomes a functlOn of its SlZe, location, and composltlon. Small fragments of gravel overlymg bone are Table 2. S!gn~icance offactors m the correct assessment of the presence of foreign bodIes and number offomgn bodies in mdLOgraph

P*

Variable Presence offragments PhySICian Size of fragment Gravel type No of fragments No. offragments PhySICian Size of fragment Gravel type No of fragments

< 001 < 001 < 001 01 001 < 001 02 < 001

296

Table 3.

DetectLOn and countmg offomgn bodies, by presCllcc/absence and number Foreign Body Group Presence/absence of foreign body Two fragments 12 and 5 mm) Two fragments 11 and 25 mm) One fragment (1 mm) One fragment (2 mml One fragment ( 5 mm) No fragments One fragment ( 25 mm) No. of foreign bodies One fragment (1 mml One fragment (2 mml No fragments One fragment I 5 mml Two fragments (2 and 5 mm) One fragment ( 25 mml Two fragments 11 and 25 mm) Average visibility rating Two fragments 12 and 5 mml One fragment (2 mm) One fragment (1 mm) Two fragments (1 and 25 mm) One fragment I 5 mml One fragment 125 mml No fragments

Finding

% Correct 1000 950 950 925

J

717 592 583 % Correct

J

817 775 592 575 508 400] 300 Mean:!:SEM

l

J

4 2± 1 4 O± 1 4 O± 1 3 5± 1 2 O± 2] 15±2 7± 1

Brackets denote groups that did not differ Significantly on pairwise comparison

*LOgIStIC regression

JUNE 1997

more diff1cult to detect than a large stone embedded in muscle and not obscured by bone. Because gravel fragments contam heterogeneous components, V1S1b1hty on plam hlms becomes dependent on the composltlon and variable denSlty of each specific fragment. 11 Our data suggest that the Slze of the smallest gravel fragment that can be rehably detected on plain radlOgraphs 1S I 0 mm. The detectlOn rates were comparable between the emergency physlCians and radlOlogists in our study We d1d note a small difference m detection rates based on the mineral composltlon of the gravel fragments, with salt-andpepper gravel the most accurately identif1ed. In a slm1lar study usmg glass foreign bod1es, Courter 6 found that the accuracy of detection decreased sign1hcantly for particles of less than 2 mm The diagnostic SenSltlV1ty of the emergency physicians' mterpretation was higher than that of the radlOlog1sts The emergency physlCians may have "overinterpreted" the radlOgraphs because they are sens1tized to the medlColegal consequences of fa1lmg to d1agnose retained fore1gn bodies dunng emergency wound care. RadlOlog1sts demonstrated better speClf1Clty than emergency physicians and were more

ANNALS OF EMERGENCY MEDICINE

729

GRAVEL

elml101m et al

accurate m determining the actual number of foreign bodies in each leg. Our data must be interpreted in light of severallimitations. Because we used a chicken leg model, the results may not be dIrectly applIcable to human bemgs. In addition, anatomlC dIfferences in various locations of the body may change the size of gravel that may be reliably excluded. In certam anatomic areas, such as the palm of the hand or the plantar surface of the foot, oblique films may mcrease the sensitivity of gravel detectlOn. 6 Although the physicians in our study were blinded to the location and number of foreign bodies in each leg, they probably tended to overinterpret the fIlms because they were participating in a study This overinterpretation would tend to exaggerate the falsepositive reading rate and decrease overall specificities. In addition, because wound location was standardized, study partlClpants may have experienced a "learning curve" that may have mcreased theIr overall abIlIty to detect the foreign bodIes. Our study IS the first to attempt to quantify the smallest gravel fragment that can be accurately detected Wlth the use of standard two-view plain radiography Our data showed that in this wound model, gravel of 1 mm or more can be accurately idennfIed by emergency physlClans and radiologists. Gravel fragments of less than 1 mm may be missed on two-view plain radlOgraphs. For thIS reason, when treatmg a wound under hIgh suspicion of harboring a foreIgn body, the physician should never tell the panent that no foreign body remams m the wound. Instead, it should be explamed to the patient that no foreign body found with the use of Table 4.

Dctectwn and counting offoreIgn bodIes, by type ofgravel Foreign Body Group Presence/absence of foreign body Salt-and-pepper rock Pea gravel Meramec rock Crater rock

No. offoreign bodies Salt-and-pepper rock Pea gravel Meramec rock Crater rock

Average visibility rating Salt-and-pepper rock Pea gravel Meramec rock Crater rock

Finding

% Correct

J 833 J 90 6

822 80 6

572 528

REFERENCES 1 Anderson MA, Newmeyer WL, Kilgore ES Jr DiagnOSIs and treatment of retamed foreign bodIes mthe hand Am J Surg 19B2, 144 63-67 2 Donaldson JS Radiographic Imagmg of foreign bodies mthe hand Hand Clm 1991,7 125-134 3 Flom LL, ElliS GL RadiologiC evaluation of foreign bodies Emerg Med Clm North Am 1992,10163-177 4 Schlager 0, Sanders AB, Wlggms 0, et al Ultrasound for the detection of foreign bodies Ann Emerg Med 1991 ,20 189-191 5 Charney DB, ManZI JA, Turllk M, et al Nonmetallic foreign bodies In the foot Radiography versus xeroradiography J Foot Surg 1986,25 44-49 6 Courter BJ Radiographic screenmg for glass foreign bodies What does a "negative" foreign body serres really mean? Ann Emerg Med 1990, 19997-1000 7 Felman AH, Fisher MS The radiographic detection of glass In soft tissue RadIOlogy 1969,921529-1531 8 Gmsburg MJ, ElliS GL, Flom LL Detection of soft-tissue foreign bodies by plam radiography, xerography, computed tomography, and ultrasonography Ann Emerg Med 1990, 19701-703 9 Manthey DE, Storrow A, Milbourn JM, et al Ultrasound versus radiography mthe detection of soft-tissue foreign bodies Ann Emerg Med1996,27 7-9 10 Pond GO, Lmdsey 0 Localization of cactus, glass, and other foreign bodies In soft tissues Anzona Medlcme 1977,34 700-702 11 Russell RC, Williamson DA, Sullivan JW, et al Detection of foreign bodies In the hand J Hand Surg [Am] 1991,16 2-11 12 Lammers RL Soft tissue foreign bodies Ann Emerg Med 1988, 17 1336-1347 13 Fitzgerald RH Jr, Cowan JD Puncture wounds of the foot Orthop Clm North Am 1975,6965972 14 Smoot EC, Robson MC Acute management of foreign body mJurres of the hand Ann Emerg Med 1983, 12 434-437 15 Stem F Foreign body mJurres of the hand Emerg Med Clm North Am 1985,3383-390 We thank Earl Toler and Mark Sawyer at the Stone Center of Indiana, Indianapolis, for their expertise and gravel donations, and Hayden Murray, PhD, and Karen Keith at the Geology Department of Indiana University, Bloomington, for their information on minerals and rock compOSition

Reprint no. 47/1/80643 Address for reprints: Carey D Chisholm, MD Emergency MediCine and Trauma Center Methodist Hospital of Indiana

% Correct 661

the most suitable detection technique but that a small foreign body could still be present. The clinical significance of undetected gravel fragments smaller than 1.0 mm is uncertain and necessitates further study

JJ

1701 North Senate Boulevard Indianapolis, IN 46202

489

Mean:!:SEM

3 6± 1J 3 3± 1 ] 3 o± 1]

2 9± 1

Brackets denote groups that did not differ Significantly on pairwise comparrson

730

ANNALS OF EMERGENCY MEDICINE

296

JUNE 1997