BRIEF REPORT foreign body, detection; radiography, foreign body
Radiographic Screening for Glass Foreign Bodies -- What Does a "Negative" Foreign Body Series Really Mean? This study was designed to provide percentages of radiographic detection of small glass foreign bodies and to compare the detection rates (sensitivity) of two-view posteroanterior and lateral with those of four-view posteroanterior, lateral, and right and left obliques screening radiographs. Three sizes (0.5, 1.0, and 2.0 ram, m a x i m u m dimension) of beer bottle glass fragments were inserted into chicken legs and radiographed. Three staff radiologists reviewed the films independently. With two-view radiography, the 2.0-, 1.0-, and 0.5-ram glass fragments exhibited 99%, 83%, and 61% average detection rates, respectively An analysis of variance and a Tukey's range test demonstrated that the variation in the average detection rate with respect to size was statistically significant (P < .01). X2 Analysis failed to demonstrate a statistically significant (P > .05) increase in average detection rates with four-view radiography. The results of this study indicate that the size of the glass foreign body is often the limiting factor for radiographic detection and that 0.5- to 2.O-ram fragments represent a "limited detection" size range. Furthermore, two-view radiography is appropriate and more cost-effective than four-view radiography for screening purposes. [Courter BJ: Radiographic screening for glass foreign bodies - What does a "'negative" foreign body series really mean? Ann Emerg Med September 1990;19:997-1000.]
Brad J Courter, ME) Indianapolis, Indiana From the Department of Emergency Medicine, Methodist Hospital, Indianapolis, Indiana. Received for publication November 16, 1989. Revision received February 13, 1990. Accepted for publication April 18, 1990. Address for reprints: Brad J Courter, MD, 121 Raintree Drive, Taylors, South Carolina 29687.
INTRODUCTION Numerous studies demonstrate that most glass foreign bodies are radiopaque.~ ~ The size of the glass foreign body, not the type of glass, is usually the limiting factor for radiographic detection. Tandberg ~ demonstrated that glass fragments as small as 0.5 m m could be visualized without difficulty. However, he also noted that overlying bone may completely obscure small fragments in the 0.5- to 2-ram range. ~ His study implied that the 0.5to 2-ram range was a limited detection range, but he did not provide percentages of radiographic detection of the various sizes. Furthermore, no study has compared various radiographic screening views to determine sensitivity or cost-effectiveness. The purpose of this study was to provide percentages of radiographic detection of minute glass foreign bodies in the limited detection range and to compare the detection rates (sensitivity) of two-view with four-view screening radiographs to detect glass foreign bodies. MATERIALS A N D M E T H O D S Beer bottle glass was chosen because it is likely to be an embedded foreign body. Furthermore, the density of beer bottle glass (2.43 g/mL3) is representative of various common glass densities that range from approximately 2.22 to 2.87 g/mL< 2 Miller ® beer bottle glass was chosen because it is clear and nonpigmented. The glass was broken up with a hammer, measured with a ruler, and sorted into various size groups (0.5, 1.0, and 2.0 m m maximum dimension). The glass fragments were inserted with tweezers into a puncture wound (created by iris scissors) in the leg muscle of cadaveric chicken legs. This experimental model, designed by Tandberg, was chosen because of the general similarity in size and anatomic structure between the chicken "drumstick" and the human thumb. 1
19:9 September 1990
Annals of Emergency Medicine
997/85
FOREIGN BODIES Courter
TABLE 1. Foreign body detection rates Foreign Body Size (mm) 2
1
0.5
Radiologist 1
Two Views*
Four Viewst
%
No.
%
No.
P
100
50/50
100
50/50
NS
2
96
48/50
98
49/50
NS
3
100
50/50
100
50/50
NS NS
1
92
46/50
92
46/50
2
72
36/50
70
35/50
NS
3
84
42/50
88
44/50
NS
1
80
40/50
72
36/50
NS
2
50
25/50
50
25/50
NS
3
54
27/50
62
31/50
NS
*Posteroanterior and lateral views. tPosteroanterior, lateral, and right and left oblique views.
FIGURE 1. Radiograph of 2-ram glass fragments in chicken legs I3-24. From zero to three glass fragments of a similar size were inserted into the chicken legs. Fifty 2-ram glass fragments were placed into chicken legs 1 through 24. Likewise, 50 1-mm fragments were placed into chicken legs 25 through 48, and 50 0.5-ram fragments were placed into chicken legs 49 through 72. The c h i c k e n legs were radiographed (12 legs per film) in four views, posteroanterior, lateral, and r i g h t and left o b l i q u e ( P A / L a t / Obliques) with our radiology department's standard foreign body techn i q u e (Kodak OG-1 film, Kodak Lanex regular screen, 300 mA, 0.003 seconds, 40 in. from table, 50 kV, and a 25 ° angle for oblique films). The three staff radiologists chosen devote most of their time to plain film radiology. These radiologists were blinded to the range of possible foreign bodies in each chicken leg (zero through three) or the total number of glass foreign bodies in each size category (50). Each radiologist reviewed the films separately. The PA/Lat films were first presented to the radiologists. Once a radiologist committed himself by writing down the number of definite foreign bodies he saw, he was provided with the oblique films in addition to t h e o r i g i n a l t w o to d e t e r m i n e whether any additional or any fewer definite glass foreign bodies could be identified with four views. Again, he 86/998
wrote down how many definite foreign bodies he saw. The second number could be higher or lower than the original assessment. The X2 method was used for statistical analysis, with P < .05 defined as statistical significance. D e t e c t i o n rates (%) or sensitivity {%) were defined as 100 × true positive/(true positive + false negative). An analysis of variance and a Tukey's range Annals of Emergency Medicine
test were used to analyze the variation in the average detection rate with respect to size of the glass foreign body.
RESULTS The beer bottle glass and chicken leg model used in this study radiographed well (Figure 1). The overall results are given (Table 1). A X2 analysis of each radiologist for 19:9 September 1990
Possible Glass Foreign Body
,/ Acute "Noncomplicated" Wound
/
Nonhealing, Infected, or Persistently Painful "Complicated" Wound
Two-View Radiography ~r
Two-View Radiography
If Negative
i
If Negative
Follow Clinically
Four-View Radiography ,
If Negative
Other Modalities, Including CT, MRI, Ultrasound, or Xeroradiography
FIGURE 2. Protocol to detect a possible glass foreign body.
TABLE 2. Average detection rates Foreign Body Size (mm)
Two Views*
Four Viewst
%
(%)
2 98.7 1 82.7 0.5 61.3 *Posteroanterior and lateral views. tPosteroanterior, lateral, and right and left oblique views. individual foreign body sizes failed to d e m o n s t r a t e a statistically significant increase in the individual detection rate with four-view radiography (Table 1). The average detection rates of the three radiologists for each foreign body size were calculated. The 2-ram fragments were easily identified by all three radiologists and had an average detection rate of 98.7% (PA/Lat 19:9 September 1990
99.3 83.3 61.3
P
NS NS NS
films). T h e l - r a m f r a g m e n t s were more difficult to identify and exhibited an average d e t e c t i o n rate of 82.7% (PA/Lat films). The 0.5 m m fragments were the most difficult to identify accurately and had an average detection rate of 61.3% (PA/Lat films). These rates of detection are given (Table 2). An analysis of variance and a Tukey's range test' both d e m o n s t r a t e d t h a t the variation in Annals of Emergency Medicine
the average detection rate with respect to size was significant (P < .01). A X2 analysis of the average detection rate for each foreign body size was calculated. Because the average detection rate combines across raters, it is a m u c h more reliable and therefore more valid assessment than indi~ vidual radiologists' detection rates. A X~ analysis of the average detection rate for each foreign body size also failed to demonstrate a statistically significant increase in the average detection rate with four-view radiography (Table 2). Specificity was not reported in this study because this experiment was not designed to analyze specificity. Only one of 24 chicken legs in each glass foreign body size category had no glass fragments for obvious cost and logistic reasons. Therefore, the 999/87
FOREIGN BODIES Courter
number of true-negatives (chicken legs w i t h o u t any glass fragments) was too small for significant statistical evaluation. An interesting p h e n o m e n o n displayed by this study is the interobserver variation among the three radiologists. The variation was not significant in the easily identified 2-mm fragments; however, it became more significant with the increasing difficulty of detection of the smaller sizes (Table 3). Because specificity was not reported, the number of false-positives (over-reads) of each radiologist was determined. With the exception of radiologist 1 viewing the 1-mm foreign body size category, the radiologists' thresholds for calling a film positive (positivity criteria) were rather uniform. DISCUSSION The size of the glass foreign body is most often the limiting factor for radiographic detection. Tandberg implied that the 0.5- to 2-ram range was the transition zone of radiographic detection, l This study provided percentages of detection proving that the 0.5- to 2-mm size range is indeed a limited detection range. Roberts s demonstrated that even 0.1 m m glass fragments could be visualized, but no percentage of detection was documented. W h e t h e r a piece of radiopaque glass will visualize on plain films is dependent primarily on its size and secondarily on its orientation with respect to overlying bone and the radiographic beam. Obviously, less overlying bone increases the detection rate. Less obvious is the fact that small fragments are more apparent if the greatest diameter of the glass is positioned parallel to the central ray of the radiographic beam. 3 In certain cases, oblique films may detect a previously unseen foreign body because the oblique angle m a y "free" the glass foreign body of overlying bone or may change the orientation of the glass foreign body with respect to the
88/1000
TABLE 3. I n t e r o b s e r v e r variation a m o n g radiologists
Foreign Body Size (mm) 2
Detection Rate Variation (Two Views) (%) 4
P
Detection Rate Variation (Four Views) (%)
P
NS
2
NS
1
20
< .05
22
< .05
0.5
30
< .01
22
< .05
radiographic beam. 3 The second purpose of this study was to compare two-view with fourview screening radiography to detect foreign bodies. Four-view radiography offers no statistically significant increase in sensitivity and therefore is not cost-efficient for mass screening of noncomplicated wounds (Table 2). However, if a patient returns to the ED days to weeks later with complications of a retained glass foreign body (infection or pain), the additional oblique films m a y be warranted. Although the average detection rate with four-view radiography was not statistically significant compared w i t h two-view radiography, there were cases (radiologist 1, five cases; radiologist 2, six cases; radiologist 3, three cases) in which the oblique films visualized a previously nondetected foreign body. In addition, oblique films may increase the d e t e c t i o n rate in m u l t i p l e - b o n e m o d e l s (ie, hand injuries) w h e r e many bones more easily obscure a foreign body. This study used only a one-bone model (ie, chicken leg or finger). Because oblique films are inexpensive, are readily available, and occasionally visualize a previously nondetected foreign body, four-view radiography may be warranted in "complicated wounds" before using other more expensive radiographic modalities. If four-view radiography fails to demonstrate a retained foreign body in a complicated wound, then various radiographic modalities, including ultrasound, computed tomogra-
Annals of Emergency Medicine
phy scan, magnetic resonance imagi n g , or x e r o r a d i o g r a p h y , are indicated. Figure 9. shows a proposed flow chart to detect a possible glass foreign body. CONCLUSION The size of a glass foreign body is most often the limiting factor for radiographic detection. A negative foreign body radiographic series means that probably no foreign body of 2 m m or more is present; however, a foreign body of less than 2 m m may be present. Two-view radiography is appropriate and c o s t - e f f i c i e n t for screening noncomplicated wounds. Four-view radiography may be helpful for screening complicated wounds to detect a glass foreign body before using more sophisticated and expensive radiographic modalities.
The author thanks the Radiology Department of Methodist Hospital; Earl Miller; Robert Prosser, MD; William Cordell, MD; James Jones, MD; Betty Watts; and Sue Home. REFERENCES 1. Tandbcrg D: Glass in the hand and foot. [AMA 1982;248:1872-1874. 2. deLaceyG, Evans R, Sandin B: Penetrating injuries: How easy is it to see glass (and plastic) on radiographs? Br [ R~idio] t985~58:27-30. 3. Felman AH, Fisher MS: The radiographicdetection of glass in soft tissue. Radiology 1969; 92:1529-1531. 4. Fordham SD: The detection of glass foreign bodies. South Med ] 1976;69:1484-1485. 5. Roberts WC: Radiographic characteristics of glass. Arch Ind Health 1958;18:470-472.
19:9 September 1990
Radiographic Screening for Glass Foreign Bodies -- What Does a "Negative" Foreign Body Series Really Mean? This study was designed to provide percentages of radiographic detection of small glass foreign bodies and to compare the detection rates (sensitivity) of two-view posteroanterior and lateral with those of four-view posteroanterior, lateral, and right and left obliques screening radiographs. Three sizes (0.5, 1.0, and 2.0 ram, m a x i m u m dimension) of beer bottle glass fragments were inserted into chicken legs and radiographed. Three staff radiologists reviewed the films independently. With two-view radiography, the 2.0-, 1.0-, and 0.5-ram glass fragments exhibited 99%, 83%, and 61% average detection rates, respectively An analysis of variance and a Tukey's range test demonstrated that the variation in the average detection rate with respect to size was statistically significant (P < .01). X2 Analysis failed to demonstrate a statistically significant (P > .05) increase in average detection rates with four-view radiography. The results of this study indicate that the size of the glass foreign body is often the limiting factor for radiographic detection and that 0.5- to 2.O-ram fragments represent a "limited detection" size range. Furthermore, two-view radiography is appropriate and more cost-effective than four-view radiography for screening purposes. [Courter BJ: Radiographic screening for glass foreign bodies - What does a "'negative" foreign body series really mean? Ann Emerg Med September 1990;19:997-1000.]
Brad J Courter, ME) Indianapolis, Indiana From the Department of Emergency Medicine, Methodist Hospital, Indianapolis, Indiana. Received for publication November 16, 1989. Revision received February 13, 1990. Accepted for publication April 18, 1990. Address for reprints: Brad J Courter, MD, 121 Raintree Drive, Taylors, South Carolina 29687.
INTRODUCTION Numerous studies demonstrate that most glass foreign bodies are radiopaque.~ ~ The size of the glass foreign body, not the type of glass, is usually the limiting factor for radiographic detection. Tandberg ~ demonstrated that glass fragments as small as 0.5 m m could be visualized without difficulty. However, he also noted that overlying bone may completely obscure small fragments in the 0.5- to 2-ram range. ~ His study implied that the 0.5to 2-ram range was a limited detection range, but he did not provide percentages of radiographic detection of the various sizes. Furthermore, no study has compared various radiographic screening views to determine sensitivity or cost-effectiveness. The purpose of this study was to provide percentages of radiographic detection of minute glass foreign bodies in the limited detection range and to compare the detection rates (sensitivity) of two-view with four-view screening radiographs to detect glass foreign bodies. MATERIALS A N D M E T H O D S Beer bottle glass was chosen because it is likely to be an embedded foreign body. Furthermore, the density of beer bottle glass (2.43 g/mL3) is representative of various common glass densities that range from approximately 2.22 to 2.87 g/mL< 2 Miller ® beer bottle glass was chosen because it is clear and nonpigmented. The glass was broken up with a hammer, measured with a ruler, and sorted into various size groups (0.5, 1.0, and 2.0 m m maximum dimension). The glass fragments were inserted with tweezers into a puncture wound (created by iris scissors) in the leg muscle of cadaveric chicken legs. This experimental model, designed by Tandberg, was chosen because of the general similarity in size and anatomic structure between the chicken "drumstick" and the human thumb. 1
19:9 September 1990
Annals of Emergency Medicine
997/85
FOREIGN BODIES Courter
TABLE 1. Foreign body detection rates Foreign Body Size (mm) 2
1
0.5
Radiologist 1
Two Views*
Four Viewst
%
No.
%
No.
P
100
50/50
100
50/50
NS
2
96
48/50
98
49/50
NS
3
100
50/50
100
50/50
NS NS
1
92
46/50
92
46/50
2
72
36/50
70
35/50
NS
3
84
42/50
88
44/50
NS
1
80
40/50
72
36/50
NS
2
50
25/50
50
25/50
NS
3
54
27/50
62
31/50
NS
*Posteroanterior and lateral views. tPosteroanterior, lateral, and right and left oblique views.
FIGURE 1. Radiograph of 2-ram glass fragments in chicken legs I3-24. From zero to three glass fragments of a similar size were inserted into the chicken legs. Fifty 2-ram glass fragments were placed into chicken legs 1 through 24. Likewise, 50 1-mm fragments were placed into chicken legs 25 through 48, and 50 0.5-ram fragments were placed into chicken legs 49 through 72. The c h i c k e n legs were radiographed (12 legs per film) in four views, posteroanterior, lateral, and r i g h t and left o b l i q u e ( P A / L a t / Obliques) with our radiology department's standard foreign body techn i q u e (Kodak OG-1 film, Kodak Lanex regular screen, 300 mA, 0.003 seconds, 40 in. from table, 50 kV, and a 25 ° angle for oblique films). The three staff radiologists chosen devote most of their time to plain film radiology. These radiologists were blinded to the range of possible foreign bodies in each chicken leg (zero through three) or the total number of glass foreign bodies in each size category (50). Each radiologist reviewed the films separately. The PA/Lat films were first presented to the radiologists. Once a radiologist committed himself by writing down the number of definite foreign bodies he saw, he was provided with the oblique films in addition to t h e o r i g i n a l t w o to d e t e r m i n e whether any additional or any fewer definite glass foreign bodies could be identified with four views. Again, he 86/998
wrote down how many definite foreign bodies he saw. The second number could be higher or lower than the original assessment. The X2 method was used for statistical analysis, with P < .05 defined as statistical significance. D e t e c t i o n rates (%) or sensitivity {%) were defined as 100 × true positive/(true positive + false negative). An analysis of variance and a Tukey's range Annals of Emergency Medicine
test were used to analyze the variation in the average detection rate with respect to size of the glass foreign body.
RESULTS The beer bottle glass and chicken leg model used in this study radiographed well (Figure 1). The overall results are given (Table 1). A X2 analysis of each radiologist for 19:9 September 1990
Possible Glass Foreign Body
,/ Acute "Noncomplicated" Wound
/
Nonhealing, Infected, or Persistently Painful "Complicated" Wound
Two-View Radiography ~r
Two-View Radiography
If Negative
i
If Negative
Follow Clinically
Four-View Radiography ,
If Negative
Other Modalities, Including CT, MRI, Ultrasound, or Xeroradiography
FIGURE 2. Protocol to detect a possible glass foreign body.
TABLE 2. Average detection rates Foreign Body Size (mm)
Two Views*
Four Viewst
%
(%)
2 98.7 1 82.7 0.5 61.3 *Posteroanterior and lateral views. tPosteroanterior, lateral, and right and left oblique views. individual foreign body sizes failed to d e m o n s t r a t e a statistically significant increase in the individual detection rate with four-view radiography (Table 1). The average detection rates of the three radiologists for each foreign body size were calculated. The 2-ram fragments were easily identified by all three radiologists and had an average detection rate of 98.7% (PA/Lat 19:9 September 1990
99.3 83.3 61.3
P
NS NS NS
films). T h e l - r a m f r a g m e n t s were more difficult to identify and exhibited an average d e t e c t i o n rate of 82.7% (PA/Lat films). The 0.5 m m fragments were the most difficult to identify accurately and had an average detection rate of 61.3% (PA/Lat films). These rates of detection are given (Table 2). An analysis of variance and a Tukey's range test' both d e m o n s t r a t e d t h a t the variation in Annals of Emergency Medicine
the average detection rate with respect to size was significant (P < .01). A X2 analysis of the average detection rate for each foreign body size was calculated. Because the average detection rate combines across raters, it is a m u c h more reliable and therefore more valid assessment than indi~ vidual radiologists' detection rates. A X~ analysis of the average detection rate for each foreign body size also failed to demonstrate a statistically significant increase in the average detection rate with four-view radiography (Table 2). Specificity was not reported in this study because this experiment was not designed to analyze specificity. Only one of 24 chicken legs in each glass foreign body size category had no glass fragments for obvious cost and logistic reasons. Therefore, the 999/87
FOREIGN BODIES Courter
number of true-negatives (chicken legs w i t h o u t any glass fragments) was too small for significant statistical evaluation. An interesting p h e n o m e n o n displayed by this study is the interobserver variation among the three radiologists. The variation was not significant in the easily identified 2-mm fragments; however, it became more significant with the increasing difficulty of detection of the smaller sizes (Table 3). Because specificity was not reported, the number of false-positives (over-reads) of each radiologist was determined. With the exception of radiologist 1 viewing the 1-mm foreign body size category, the radiologists' thresholds for calling a film positive (positivity criteria) were rather uniform. DISCUSSION The size of the glass foreign body is most often the limiting factor for radiographic detection. Tandberg implied that the 0.5- to 2-ram range was the transition zone of radiographic detection, l This study provided percentages of detection proving that the 0.5- to 2-mm size range is indeed a limited detection range. Roberts s demonstrated that even 0.1 m m glass fragments could be visualized, but no percentage of detection was documented. W h e t h e r a piece of radiopaque glass will visualize on plain films is dependent primarily on its size and secondarily on its orientation with respect to overlying bone and the radiographic beam. Obviously, less overlying bone increases the detection rate. Less obvious is the fact that small fragments are more apparent if the greatest diameter of the glass is positioned parallel to the central ray of the radiographic beam. 3 In certain cases, oblique films may detect a previously unseen foreign body because the oblique angle m a y "free" the glass foreign body of overlying bone or may change the orientation of the glass foreign body with respect to the
88/1000
TABLE 3. I n t e r o b s e r v e r variation a m o n g radiologists
Foreign Body Size (mm) 2
Detection Rate Variation (Two Views) (%) 4
P
Detection Rate Variation (Four Views) (%)
P
NS
2
NS
1
20
< .05
22
< .05
0.5
30
< .01
22
< .05
radiographic beam. 3 The second purpose of this study was to compare two-view with fourview screening radiography to detect foreign bodies. Four-view radiography offers no statistically significant increase in sensitivity and therefore is not cost-efficient for mass screening of noncomplicated wounds (Table 2). However, if a patient returns to the ED days to weeks later with complications of a retained glass foreign body (infection or pain), the additional oblique films m a y be warranted. Although the average detection rate with four-view radiography was not statistically significant compared w i t h two-view radiography, there were cases (radiologist 1, five cases; radiologist 2, six cases; radiologist 3, three cases) in which the oblique films visualized a previously nondetected foreign body. In addition, oblique films may increase the d e t e c t i o n rate in m u l t i p l e - b o n e m o d e l s (ie, hand injuries) w h e r e many bones more easily obscure a foreign body. This study used only a one-bone model (ie, chicken leg or finger). Because oblique films are inexpensive, are readily available, and occasionally visualize a previously nondetected foreign body, four-view radiography may be warranted in "complicated wounds" before using other more expensive radiographic modalities. If four-view radiography fails to demonstrate a retained foreign body in a complicated wound, then various radiographic modalities, including ultrasound, computed tomogra-
Annals of Emergency Medicine
phy scan, magnetic resonance imagi n g , or x e r o r a d i o g r a p h y , are indicated. Figure 9. shows a proposed flow chart to detect a possible glass foreign body. CONCLUSION The size of a glass foreign body is most often the limiting factor for radiographic detection. A negative foreign body radiographic series means that probably no foreign body of 2 m m or more is present; however, a foreign body of less than 2 m m may be present. Two-view radiography is appropriate and c o s t - e f f i c i e n t for screening noncomplicated wounds. Four-view radiography may be helpful for screening complicated wounds to detect a glass foreign body before using more sophisticated and expensive radiographic modalities.
The author thanks the Radiology Department of Methodist Hospital; Earl Miller; Robert Prosser, MD; William Cordell, MD; James Jones, MD; Betty Watts; and Sue Home. REFERENCES 1. Tandbcrg D: Glass in the hand and foot. [AMA 1982;248:1872-1874. 2. deLaceyG, Evans R, Sandin B: Penetrating injuries: How easy is it to see glass (and plastic) on radiographs? Br [ R~idio] t985~58:27-30. 3. Felman AH, Fisher MS: The radiographicdetection of glass in soft tissue. Radiology 1969; 92:1529-1531. 4. Fordham SD: The detection of glass foreign bodies. South Med ] 1976;69:1484-1485. 5. Roberts WC: Radiographic characteristics of glass. Arch Ind Health 1958;18:470-472.
19:9 September 1990
