313

Comparison of Upright lnspiratory

and Expiratory Chest Radiographs
Downloaded from www.ajronline.org by 36.84.29.82 on 08/15/16 from IP address 36.84.29.82. Copyright ARRS. For personal use only; all rights reserved

for Detecting Pneumothoraces

- . -‘ .

Albert Seow1 OBJECTIVE. Expiratory films are regarded as being superior to inspiratory films for
Ella A. Kazerooni pneumothorax detection, yet this has not been proved. In the current study, we com-
Philip N. Cascade pared inspiratory versus expiratory chest radiographs for pneumothorax detection.
MATERIALS AND METHODS. Eighty-five paired inspiratory and expiratory radio-
Perry G. Pernicano
graphs with pneumothoraces and 93 pairs without pneumothoraces were randomly
Maureen Neary
arranged and reviewed independently by three radiologists. A score of 1-5 was
assigned for each hemithorax (5 = definite pneumothorax, 1 = definitely no pneu-
mothorax). Results were compared for inspiration and expiration using receiver oper-
ating characteristic (ROC) analysis.
RESULTS. The average area under the ROC curves for all readers was .973 for
inspiration and .972 for expiration (nonsignificant). McNemar’s test and an alpha level
of .05 also yielded no significant difference in sensitivity and specificity. Four of the
85 cases were scored as definite pneumothorax on inspiration and as definitely not
on expiration by all readers, and three of the 85 cases were scored as definite pneu-
mothorax on expiration and as definitely not on inspiration.
CONCLUSION. lnspiratory and expiratory upright films are equally sensitive for
pneumothorax detection. Given the limitations of expiratory films, Inspiratory films
are recommended as the initial examination of choice for pneumothorax detection.

AJR 1996;166:313-316

The early and accurate diagnosis of pneumothorax is essential in preventing

respiratory compromise and potential death. There has been continued debate in
the radiologic literature as to which radiographic technique is most accurate for
detecting pneumothoraces [1 2]. Many
, experts recommend the expiratory,
upright chest radiograph as the technique of choice [2]; however, no scientific val-
idation exists in the literature. Other experts advocate a decubitus film with the
suspected side nondependent as a more sensitive technique [1]. Either of these
radiographic methods may suffice when the only information necessary from the
examination is a diagnosis of pneumothorax (e.g., after placement of a central
venous catheter).
However, evaluating for other conditions with symptoms that may mimic pneu-
mothorax-such as pneumonia, pulmonary embolism, neoplasm, musculoskele-
tal injuries, pleurisy, and pericarditis-requires optimal radiographic technique.
Received July 1 , 1 995; accepted after revision
Abnormalities on decubitus or expiratory chest radiographs may be disregarded
August23, 1995.
as hypoventilatory changes. Conversely, the opacities of atelectasis on expiratory
1AII authors: Department of Radiology, Universi-
ty of Michigan Hospital, 1500 E. Medical Center radiographs may simulate pulmonary disease, such as pneumonia, and result in
Dr., Ann Arbor, Ml 48109-0326. Address correspon- false-positive interpretations. At our institution, it is common for physicians to
dence to E. A. Kazerooni.
order paired inspiratory and expiratory chest radiographs when a pneumothorax
Presented at the annual meeting of the Amen- is suspected. This practice increases the cost and length of the examination and
can Roentgen Ray Society, New Orleans, April
1994. doubles the radiation dose to a patient.
0361-803X/96/1662-313
The purpose of the current study was to compare paired upright inspiratory and
© American Roentgen Ray Society expiratory chest radiographs for detecting pneumothoraces.
 314 SEOW ET AL. AJR:166, February 1996

Materials and Methods tory radiographs were the same, this score was considered to be the
truth. If readers disagreed by a numerical value of only 1 , the score
To compare the sensitivity of inspiratory and expiratory chest
given by the majority of the readers was considered to be the truth.
radiographs, we retrospectively reviewed reports of all three view
For the 39 radiographs for which readers’ scores disagreed by 2 or
(i.e. , inspinatony posteroantenior, expiratory posteroanterior, and
more, the truth was reached by consensus among the three radiolo-
inspiratory lateral) chest radiographs obtained at our institution
gists. If a pneumothorax was scored positive on either the inspira-
between June 1992 and December 1992, resulting in a total of
tory or expiratory radiograph of a film pair, a pneumothorax was
approximately 1500 examinations. From these reports, the 116
considered to be present on both films for purposes of analysis.
Downloaded from www.ajronline.org by 36.84.29.82 on 08/15/16 from IP address 36.84.29.82. Copyright ARRS. For personal use only; all rights reserved

paired inspinatory and expiratory radiographs interpreted as positive

The sensitivity and specificity of each reader for detecting a pneu-
for pneumothonax (and that met certain criteria) were examined.
mothorax was calculated for the inspiratory radiographs and the
From this culled group of examinations, a final cohort of 86 pairs of
expiratory radiographs for each lung. ROC curves [4] were drawn
radiognaphs was compiled on the basis of the following criteria: (1)
and areas under the curves were measured to analyze each
The pneumothorax had to be small (i.e. , an average intenpleunal dis-
reader’s performance on the two sets of radiographs, using ROCFIT
tance of less than 2.25 cm, which was estimated to be less than a
and CORROC2 [5], by fitting bin#{243}rmalcurves to each reader’s confi-
25% pneumothorax as defined by Rhea et al. [3]). (2) The inspina-
dence ratings by maximum-likelihood estimation. To evaluate
tory lung volume had to be greater than the expinatory volume. The
whether the two techniques (inspiratory and expiratory) would result
latter criterion was based on a decrease in the vertical distance from
in similar detection rates, we compared the ROC areas using a
the apex of the lung to the dome of the diaphragm of at least 1 .5 cm
univariate z score
[6]. In addition, the number of cases in which a
from inspiration to expiration. Although 1 .5 cm is arbitrary, we chose
pneumothorax was seen only on the inspiratory film or only on the
it to ensure that a change in the lung volume occurred between the
expiratory film by all three readers was noted.
two radiographs. There also were three exclusionary criteria that
applied to all experimental groups: (1) hydnopneumothorax; (2) prior
lung resection or partial resection; and (3) presence of a chest tube. Results
A control group of 93 pairs of inspiratory and expiratory chest radio-
graphs interpreted as negative for pneumothorax also was obtained Comparisons of the two techniques
for sensitivity, specific-
from the same series of approximately 1500 cases. A total of 179 ity, and positive predictive value in pneumothorax detection
pairs, or 358 films, were reviewed. The lateral-view radiograph was are shown in Table 1 The three readers’ sensitivity
. for pneu-
not considered. Patients were 18-80 years old (mean, 49.6 years); mothorax ranged from 80.2% to 85.7% for inspiratory radio-
61% of the patients were male and 39% were female. graphs and from 83.5% to 84.6% for expiratory radiographs.
All films were arranged by computer-generated, random-number Specificity was 99.6% for inspiratory films and 99.6-100%
assignment. In addition, all labels on each film indicating the type of
for expiratory films. These values were obtained for a 95%
study (inspiratory or expiratory) were concealed to reduce inter-
confidence level. The differences between inspiratory and
preter bias. All radiographs were taken in the upright position using
120 kVp, 2-4 mAs, and a 72-inch (1 .83-m) source-to-film distance.
expiratory techniques for each reader were 4.4%, 0%, and
The inspiratory and expiratory films were typically obtained within 5 2.2% for sensitivity and 0%, 0%, and 0.4% for specificity.
mm of each other. Using McNemar’s test for paired proportions and an alpha
The radiographs were reviewed independently by three board- level of .05, we found no detectable difference in the propor-
certified radiologists with recognized expertise in thoracic radiology. tion of positive diagnoses of a pneumothorax between
Groups of 50 radiographs were reviewed at each sitting, each of inspiratory and expiratory techniques.
which lasted 30-60 mm. For each radiograph, the readers indicated The areas under each ROC curve are given in Table 1.
the presence or absence of a pneumothorax on a 5-point confi- The areas under the curves for each reader ranged from
dence rating scale for each hemithorax, right and left (5 = definitely
.947 to .979 for inspiratory films and from .960 to .981 for
pneumothorax, 4 = probable pneumothorax, 3 = indeterminate, 2 =
expiratory films. (For reference, an area under the curve of
probably no pneumothorax, and 1 = definitely no pneumothorax).
1 .0 represents a perfect test.) The difference in the areas
For purposes of analysis, a score of 4 or 5 was considered to be
positive for pneumothorax, and scores of 1 , 2, and 3 were consid- under the ROC curves, calculated as the inspiratory (I) area
ered to be negative. Given the lack of a gold standard for a pneu- minus the expiratory (E) area divided by expiratory area [(I -

mothorax, the truth was determined as follows. When the readers’ E)/E], ranged from -2.1% to 1.9% (mean, -0.1%). Using a
scores for an individual hemithorax on both inspiratory and expira- univariate z test, we found that the ROC curve areas of the

TABLE 1 : Comparison of Inspiratory and Expiratory Films in Pneumothorax Detection: Sensitivity, Specificity, Positive Predictive
Value, and ROC Curve Areas

Sensiti vity (%) Specifi city (%) PPV ROC Curve Areas
R eaderNo.
I E I E I E I E #{176}‘
Difference

1 80.2 84.6 99.6 99.6 98.7 98.7 .976 .981 -0.4

2 83.5 83.5 99.6 99.6 98.7 98.7 .974 .966 0.9
3 85.7 83.5 99.6 100 98.7 100 .968 .971 -0.3
Mean 83.1 83.9 99.6 99.7 98.7 99.1 .973 .972 0.1

Note-I = inspiration, E = expiration, PPV = positive predictive value, ROC = receiver operating characteristic, % difference = percentage difference in area
between I and E [(I - E)/E].
 AJR:166, February 1996 UPRIGHT INSPIRATORY VERSUS EXPIRATORY X-RAYS 315

inspiratory and expiratory techniques did not demonstrate a overwhelming majority of patients by using this technique.
statistically significant detectable difference (p = .226). He also stated that the inspiratory film is more useful in eval-
In seven cases, the three readers detected a pneumotho- uating the rest of the chest and for comparing subsequent
rax only on either the inspiratory or the expiratory radiograph and prior films and that even when a pneumothorax is
of a film pair. In four cases, the pneumothorax was detected missed, rarely will there be an adverse consequence.
only on the inspiratory radiograph. In three cases, the pneu- Our study confirms Hall’s statements. We found no statisti-
mothorax was detected only on the expiratory radiograph. cally significant difference in the sensitivity of inspiratory versus
Downloaded from www.ajronline.org by 36.84.29.82 on 08/15/16 from IP address 36.84.29.82. Copyright ARRS. For personal use only; all rights reserved

None of these cases required treatment. expiratory radiographs for detecting pneumothoraces. Bradley
et al. [16] demonstrated similar results by examining 79 paired
inspiratory and expiratory chest films of patients with pneu-
Discussion
mothorax. In their study, all pneumothoraces were clearly visi-
A pneumothorax is defined as the presence of gas or air ble on both paired chest radiographs. However, their readers
from any source in the pleural cavity. If intrapleural pressure were aware of experimental conditions and the study lacked
exceeds pressure in the lung, as in a bronchopleural check- an essential part of any experimental design: a control group.
valve mechanism, a tension pneumothorax exists; this may Also, in their study, inspiratory and expiratory chest radio-
compromise systemic venous return and reduce cardiac out- graphs were viewed together. In our study, we included a set of
put, resulting in a medical emergency [7]. For this reason, it is control radiographs without pneumothoraces, thus eliminating
important that detection be as accurate as possible. A small, the bias generated by knowing that all films contained a pneu-
undetected pneumothorax can rapidly progress to a tension mothorax. All radiographs in our study were randomized and
pneumothorax, particularly in intubated patients, patients on analyzed independently and separately from their paired coun-
positive-pressure ventilation, and patients requiring high peak terparts; the readers were therefore not influenced by interpret-
inspiratory pressures to maintain oxygenation. Although there ing the other film from the same examination pair. Finally, by
are many causes of pneumothorax, including underlying lung keeping the readers from knowing which films were inspiratory
disease, trauma, and ruptured apical blebs, surgery and iatro- and which were expiratory, we attempted to prevent predeter-
genic etiologies are particularly important in a hospital setting. mined bias for or against a specific technique [17]. We
These etiologies include central venous access procedures, acknowledge, however, that in some cases expiratory films
right heart catheterization, percutaneous lung biopsy, bron- could be recognized by the reader because of hypoventilatory
choscopy, and thoracentesis [7, 8]. Chest radiographs are changes seen on the radiograph.
obtained after such procedures, but outpatients should be In our series of cases, a pneumothorax was identified on
given proper instruction in the event of a delayed pneumotho- either the inspiratory or expiratory radiograph in seven
rax before being allowed to return home. cases. A retrospective review of these cases revealed no
The radiographic diagnosis of a pneumothorax on an definitive reason why this occurred. In four cases, the pneu-
upright chest radiograph relies on identification of a visceral mothorax was missed on the expiratory radiograph and was
pleura line separated from the parietal pleura by a radiolu- detected only on the inspiratory radiograph. In addition,
cent airspace [7, 8]. Pulmonary vessels are followed to the three pneumothoraces were missed on the inspiratory radio-
visceral pleural line, but not beyond. In the supine patient, graph and were detected only on the expiratory radiograph.
gas collects anteriorly within the lower hemithorax, appear- All of these pneumothoraces spontaneously resolved on fol-
ing as a relatively lucent area, especially along the juxtacar- low-up chest radiographs without the need for chest tube
diac regions, the diaphragm, and extending into the lateral intervention. Combined with the fact that there was no statis-
costophrenic sulcus, the so-called “deep sulcus” sign [9]. tically significant difference between the inspiratory and
However, even a large pneumothorax may be difficult to expiratory radiographs in detecting pneumothoraces, we
detect on supine chest radiographs [10, ii]. conclude that obtaining an expiratory radiograph alone has
The results of recent studies have created controversy little benefit as a replacement for the inspiratory film and that
about the most accurate radiographic technique for diagnos- obtaining both radiographs at the same time is unnecessary.
ing a pneumothorax [12, 13]. Carr et al. [1] demonstrated that Also, there are technical advantages to using the inspiratory
the lateral decubitus view is superior to the upright expiratory radiograph. A properly done inspiratory film allows better
film in detecting an experimentally induced pneumothorax in evaluation of other conditions-such as pneumonia and dis-
cadavers. Beres and Goodman [2] refuted this, demonstrating ease of the heart, pericardium, and pleura-all of which can
that upright expiratory films are more sensitive than decubitus mimic symptoms of a pneumothorax. By contrast, hypoventi-
films in detecting a pneumothorax in the clinical setting. 5ev- lation of the lungs on expiratory film can mask these disease
eral authors have suggested that an upright expiratory chest entities and may create false-positive opacities that mimic
radiograph is more sensitive than an upright inspiratory chest pneumonia or pulmonary edema.
radiograph in detecting a small pneumothorax [7, 8, 14]. How- According to Fraser et al. [7],at fullexpiration “the volume
ever, to our knowledge, there is no well-controlled study that of air in the pleural space is relatively greater in relation to the
substantiates this claim. lung volume, providing better separation of the pleural sur-
In a letter to the editor, Hall [1 5] recommended the upright faces’ thereby making a pneumothorax easier to see. How-
inspiratory chest radiograph as the initial imaging examina- ever, other physical changes in the thorax due to expiration
tion in the evaluation of a suspected pneumothorax, stating may actually hinder the detection of a pneumothorax. On expi-
that the diagnosis of a pneumothorax can be made in the ratory film, the ribs are closer together and the chest volume
 316 SEOW ET AL. AJR:166, February 1996

is smaller, increasing the likelihood that the visceral pleural delayed pneumothorax. If a small pneumothorax is detected,
line may overlap a rib margin, which may reduce the ability to the results of this study as well as clinical practice suggest
detect a pneumothorax. Also, respiration changes the relation that treatment is usually unnecessary. We recommend obtain-
of the pleural line to the overlying ribs. Whether this change ing an inspiratory upright chest radiograph as the initial test of
improves or obscures visualization of a pneumothorax on choice for detecting a pneumothorax.
expiratory film depends on chance. The visceral pleural line
may lie within a rib interspace on inspiratory film and lie under REFERENCES
Downloaded from www.ajronline.org by 36.84.29.82 on 08/15/16 from IP address 36.84.29.82. Copyright ARRS. For personal use only; all rights reserved

a rib on expiratory film, thus making the pneumothorax more 1. Cam JJ, Reed JC, Choplin RH, Pope TL Jr, Case LD. Plain and computed
visible on one film and not the other. radiography for detecting experimentally induced pneumothorax in cadav-
One limitation of our study was the method of determining ens: implications for detection in patients. Radiology 1992:183:193-1 99
“the truth” for whether a pneumothorax was actually present. 2. Beres RA, Goodman LR. Pneumothorax: detection with upright versus
decubitus radiography. Radiology 1993:186:19-26
(This is an inherent problem in any human in vivo study of
3. Rhea JT, DeLuca SA, Greene RE. Determining the size of pneumothonax
pneumothoraces.) The gold standard used for determining the in the upright patient. Radiology 1982;144:733-736
truth in our study was the readers’ consensus readings of the 4. Vining DJ, Gladish GW. Receiver operating characteristic curves: a basic
inspiratory and expiratory radiographs. If a pneumothorax was understanding. RadioGraphics 1992:12:1147-1154
5. Metz CE, Shen JH, Wang PL, Kronman HB. ROCFITand CORROC2 pro-
present on either the inspiratory or expiratory film of a film
grams (IBM-PC version). Chicago: University of Chicago, Department of
pair, the patient was considered to have a pneumothorax at Radiology, 1989
the time both films were obtained. In an ideal study, an inde- 6. Fleiss JL. Statistical methods for rates and proportions. New York: Wiley,
pendent, highly sensitive method such as CT scanning would 1981:1 12-1 37
be used as the standard for determining the true disease sta- 7. Fraser RG, Pare JA, Pare PD, Fraser RS, Generoux GP, eds. Pneu-
mothorax. In: Diagnosis of disease of the chest (3rd ed). Philadelphia:
tus. However, this was not feasible because of additional
Saunders, 1991:2741-2750
costs and radiation exposure. The lack of an absolute gold 8. Felson B. Pneumothorax. In: Chest roentgenology. Philadelphia: Saun-
standard raises another issue: It is possible that the cases for dens, 1973:366-371
which we determined that a pneumothorax was missed might 9. Gordon R.The deep sulcus sign. Radio/ogy 1980:136:25-27
10. Tocino IM. Pneumothorax in the supine patient: radiographic anatomy.
not have been true-negatives but false-positives.
RadioGraphics 1985:5:557-585
In summary, independent upright inspiratory and indepen- 1 1 . Tocino IM, Miller MH, Fairfax WR. Distribution of pneumothonax in the
dent expiratory chest radiographs are each highly sensitive for supine and semirecumbent critically ill adult. AiR 1985;144:901-905
detecting pneumothoraces. Obtaining both an inspiratory and 12. Carr JJ, Reed JC, Choplin RH, Case LD. Pneumothorax detection: a
expiratory chest radiograph increases the time and cost of the problem in experimental design. Radiology 1993;186:23-25
13. Beres RA, Goodman LR. Pneumothorax detection: clarifications and
examination, as well as doubles the radiation exposure. On
additional thoughts. Radiology 1993;186:25-26
the other hand, the use of expiratory radiographs alone hin- 14. Greene R, McLoud TC, Stark P. Pneumothorax. Semin Roentgenol 1977;
ders interpretation of the remainder of the chest, creating 12:31 3-324
false-positive opacities. Patients who have recently under- 15. Hall FM. Radiographic diagnosis of pneumothonax (letter). Radiology
gone a procedure associated with a risk of pneumothorax 1993;188:583
1 6. Bradley M, Williams C, walshaw MJ. The value of routine expiratory chest
often are discharged with a small pneumothorax. Even in the films in the diagnosis of pneumothorax. Arch Emerg Med 1991 :8:11-116
absence of a pneumothorax, these patients should always be 17. Begg CB, McNeil BJ. Assessment of radiologic tests: control of bias and
given appropriate instructions about the possibility of a other design considerations. Radiology 1988:167:565-569