Cardiopulmonar y Imaging • Original Research

Hutchinson et al.
Overdiagnosis of Pulmonary Embolism by CTA

Cardiopulmonary Imaging
Original Research
Downloaded from www.ajronline.org by 36.73.92.133 on 06/14/17 from IP address 36.73.92.133. Copyright ARRS. For personal use only; all rights reserved

Overdiagnosis of Pulmonary
Embolism by Pulmonary
CT Angiography
Barry Donald Hutchinson1 OBJECTIVE. The purpose of this study is to evaluate the rate of overdiagnosis of pulmo-
Patrick Navin1 nary embolism (PE) by pulmonary CT angiography (CTA) in a tertiary-care university hospital.
Edith M. Marom 2 MATERIALS AND METHODS. This study is a retrospective review of all pulmonary
Mylene T. Truong2 CTA examinations performed in a tertiary-care university hospital over a 12-month period. Stud-
John F. Bruzzi1 ies originally reported as positive for PE were retrospectively reinterpreted by three subspecialty
chest radiologists with more than 10 years’ experience. A pulmonary CTA was considered nega-
Hutchinson BD, Navin P, Marom EM, Truong MT, tive for PE when all three chest radiologists were in agreement that the pulmonary CTA study
Bruzzi JF was negative for PE. The location and potential causes for PE overdiagnosis were recorded.
RESULTS. A total of 937 pulmonary CTA studies were performed over the study period.
PE was diagnosed in the initial report in 174 of these cases (18.6%). There was discordance
between the chest radiologists and the original radiologist in 45 of 174 (25.9%) cases. Discor-
dance occurred more often where the original reported PE was solitary (46.2% of reported
solitary PEs were considered negative on retrospective review) and located in a segmental or
subsegmental pulmonary artery (26.8% of segmental and 59.4% of subsegmental PE diagno-
ses were considered negative on retrospective review). The most common cause of diagnostic
difficulty was breathing motion artifact, followed by beam-hardening artifact.
CONCLUSION. In routine clinical practice, PEs diagnosed by pulmonary CTA are fre-
quently overdiagnosed, when compared with the consensus opinion of a panel of expert chest
radiologists. Improvements in the quality of pulmonary CTA examination and increased fa-
miliarity with potential diagnostic pitfalls in pulmonary CTA are recommended to minimize
misdiagnosis of PE.

Keywords: artifact, false-positive, misdiagnosis,
pulmonary CT angiography, pulmonary embolism
DOI:10.2214/AJR.14.13938
Received October 10, 2014; accepted after revision
January 23, 2015.
1
Department of Radiology, University Hospital Galway,
Newcastle Rd, Galway, Ireland. Address correspondence
to B. D. Hutchinson (barryhutchinson82@gmail.com).
2
Department of Chest Radiology, M. D. Anderson Cancer
Center, Houston, TX.
AJR 2015; 205:271–277
0361–803X/15/2052–271
© American Roentgen Ray Society
P
ulmonary embolism (PE) is a
common clinical diagnosis in pa-
tients presenting to the emergen-
cy department and in hospitalized
patients. It is normally treated with long-term
anticoagulation therapy to reduce the risks of
death and the morbidity associated with
chronic pulmonary venous thromboembo-
lism. Because the clinical presentation is of-
ten nonspecific and can be mimicked by a
range of other conditions, in routine practice,
pulmonary CT angiography (CTA) is often
used as the imaging method of choice for fur-
ther investigation [1]. Pulmonary CTA has
been shown to be highly sensitive and specific
when pretest clinical diagnostic tools are used
[2] but surprisingly inaccurate in patients with
low pretest probability, with false-positive
rates as high as 42% [3]. Unfortunately, ad-
herence to referral guidelines for pulmonary
CTA has repeatedly been shown to be low [4,
5]. Nonetheless, many clinicians will initiate
anticoagulation therapy on the basis of a posi-
tive result, regardless of pretest probability
[6], even in isolated subsegmental PE [7].
The risk of hemorrhage related to antico-
agulation therapy is potentially significant.
A large meta-analysis in 2003 [8] found a
7% annual risk of major bleeding and a 0.4%
incidence of bleeding-related fatality in pa-
tients treated with oral anticoagulation ther-
apy for venous thromboembolism for longer
than 3 months. The practical implications of
long-term anticoagulation therapy for the pa-
tient are also potentially significant, requir-
ing frequent attendance to their medical
practitioners for blood tests, consequent time
off from work, potential adverse drug inter-
actions with other medications, adjustments
to travel and lifestyle, implications for future
dental and medical procedures, and possible
negative effects on life insurance status.
With these considerations in mind, it is
important to minimize the misdiagnosis

AJR:205, August 2015 271


of PE. Common artifacts that can lead to a
false-positive diagnosis of PE have been well
described in the published literature [9–11].
Despite this, however, reported interobserver
agreement varies widely, especially in the di-
agnosis of subsegmental PE [12]. Wide vari-
ations in concordance between general and
Downloaded from www.ajronline.org by 36.73.92.133 on 06/14/17 from IP address 36.73.92.133. Copyright ARRS. For personal use only; all rights reserved
subspecialist radiologists have been report-
ed (89–100%) [13, 14], as well as between
residents, fellows, and attending radiologists
(87–93%) [15–17]. Although pulmonary
CTA examinations are frequently inter-
preted by general radiologists in most cen-
ters, limited data exist on the interobserver
agreement between general and subspecialist
chest radiologists. A small number of stud-
ies [13, 18] have directly compared pulmo-
nary CTA interpretation by general radiolo-
gists with that of a single subspecialist chest
radiologist. There is very limited analysis of
these discrepant cases in terms of PE loca-
tion within the pulmonary arterial system
and potential causes of misdiagnosis. In ad-
dition, the absence of a practical reference
standard examination makes it difficult to
draw conclusions regarding the accuracy of
pulmonary CTA in routine clinical practice.
The purpose of this study was to evaluate
the rate of overdiagnosis of PE by pulmonary
CTA in a tertiary-care university hospital by
assessing the degree of discordance between
the original reporting radiologists and an ex-
pert panel of subspecialty chest radiologists
and to attempt to establish patterns of misdi-
agnosis to try to understand the causes un-
derlying pulmonary CTA misinterpretation.
Materials and Methods
This retrospective study was conducted at Uni-
versity College Hospital Galway, which is a spe-
cialist oncology center and a university-affiliated
tertiary-care medical center in Galway, Ireland.
Approximately 130,000 imaging studies are per-
formed annually in the University College Hospi-
tal Galway Radiology Department, which is staffed
by 15 attending and nine resident radiologists. The
Galway University institutional ethical review
board approved this retrospective study and waived
the requirement for written informed consent.
An electronic search was performed of the ap-
proved finalized reports of all consecutive pul-
monary CTA examinations performed over a
12-month period between August 1, 2012, and
July 31, 2013. Data were collected by both elec-
tronic query and manual review of the electronic
medical record.
All scans were acquired on a 64-MDCT scan-
ner (Somatom Sensation 64, Siemens Healthcare)
272 AJR:205, August 2015
Hutchinson et al.
Fig. 1—75-year-old
man with dyspnea.
Axial pulmonary CT
angiogram (mediastinal
window) shows
density measurement
in pulmonary trunk and
main pulmonary arteries
to calculate quality of
contrast enhancement.
Circles denote ROIs.
Avg = average, Dev =
deviation.
in the craniocaudal direction with a collimation of
0.6 mm and gantry rotation of 500 milliseconds.
Automated dose control software was used with
120 kVp and 200 mA maximum; 80–120 mL of
low-osmolar contrast medium (350 Omnipaque,
GE Healthcare) was injected through an 18-gauge
cannula sited in the antecubital fossa, at a rate
of 4–5 mL/s, followed by a 20-mL saline bolus
chaser injected at 4 mL/s. Optimal scan acquisi-
tion time was determined using a bolus-tracking
technique with an ROI placed over the pulmonary
trunk. Images were reconstructed with a 512 × 512
matrix and a smooth kernel, with 1-mm axial and
1.5-mm coronal slice thickness and 0.8-mm slice
overlap. Images were reviewed using IMPAX
(version 6.5, AGFA Healthcare).
All studies in which a definite diagnosis of PE was
reported were selected for further analysis. Studies
reported as nondiagnostic or negative for the pres-
ence of PE were excluded (because the purpose of
our study was to evaluate the potential rate of over-
diagnosed PEs, rather than the overall diagnostic ac-
curacy of pulmonary CTA). All studies were anony-
mized for independent interpretation on stand-alone
workstations by a panel of three subspecialist chest
radiologists, each with at least 10 years’ experience in
pulmonary CTA interpretation. One radiologist was
among the 15 on-site attending radiologists. The oth-
er two panel members were reviewers from another
tertiary referral center. The final consensus opinion
of these three chest radiologists was used as a surro-
gate reference standard for the diagnosis of PE.
Each examiner was blinded to the index report,
PE location, clinical history, and other diagnos-
tic test results. An initial interpretation was per-
formed by each of the three chest radiologists inde-
pendently, in which they recorded the presence or
absence of PE, the most proximal level of PE, the
lobar location of PE, and the overall quality of the
examination (satisfactory or unsatisfactory for di-
agnosis). After this initial interpretation, a second
analysis was then made of those studies in which
there was any disagreement among the three chest
radiologists (to minimize the risk of overlooking
PEs because of interpretation fatigue after read-
ing a large number of pulmonary CTA studies in
succession). Where there was any persistent dis-
cordance among the three chest radiologists af-
ter this second review, the original report was ac-
cepted as being correct (i.e., positive for PE). Next,
a third and final analysis was performed of those
studies for which there was a discrepancy between
the consensus opinion of the three chest radiolo-
gists and the original report, guided by a partial
unblinding of the original report to direct atten-
tion to the original reported PE. Where there was
unanimous agreement among the three chest radi-
ologists that a pulmonary CTA was negative, a fi-
nal outcome of negative for PE was recorded.
In addition, the following final data were re-
corded: patient demographics (age and sex); the
most proximal PE location according to the modi-
fied Boyden classification [19] (pulmonary trunk,
main pulmonary artery, lobar pulmonary artery,
segmental pulmonary artery, or subsegmental pul-
monary artery); number of PEs (solitary vs mul-
tiple); quality of contrast enhancement, assessed
by calculating the average of the CT number mea-
sured in the pulmonary trunk and the right and
left main pulmonary arteries with a circular ROI
equal to the diameter of the vessel (Fig. 1); and
interobserver agreement (modified kappa index)
among the three chest radiologists and between
the final consensus opinion of the three chest ra-
diologists and the original reporting radiologists.
Finally, the individual discordant cases (those
that were considered to be negative for PE) were
analyzed separately to attempt to establish a po-
tential underlying cause for the misdiagnosis of
PE on the original report, such as movement ar-
 Overdiagnosis of Pulmonary Embolism by CTA

TABLE 1: Comparison of Contrast Enhancement, Patient Age, and 59.2%) than in the more central and lobar
­Pulmonary Embolism (PE) Location in All and Discordant arteries (71 cases; 40.8%) (Table 1). Twen-
­Pulmonary CT Angiography Examinations Reported as Positive for ty-four patients (13.8%) had a reported soli-
PE by the Original Radiologist tary subsegmental PE. The average quality
Variable All Cases Discordant Cases of contrast enhancement was 327.0 ± 88 HU
(range, 127.1–625.2 HU).
Total 174 (100) 45 (25.9)
On final analysis, the panel of three chest
Downloaded from www.ajronline.org by 36.73.92.133 on 06/14/17 from IP address 36.73.92.133. Copyright ARRS. For personal use only; all rights reserved

Patient age (y), mean (range)
Single-vessel PE
Multiple-vessel PEs
Mean quality of contrast enhancement (HU)
Pulmonary trunk PE
Main pulmonary artery PE
Lobar pulmonary artery PE
Segmental pulmonary artery PE
Subsegmental pulmonary artery PE
Note—Except where noted otherwise, data are number (%) of patients.
ap = 0.002.
TABLE 2: Comparison of Solitary Pulmonary Embolism (PE) in All and
­Discordant Pulmonary CT Angiography Examinations
Location of PE
Total
Pulmonary trunk PE
Main pulmonary artery PE
Lobar pulmonary artery PE
Segmental pulmonary artery PE
Subsegmental pulmonary artery PE
Note—Except where noted otherwise, data are number (%) of patients.
tifact from breathing or cardiac pulsation; poor
contrast opacification of the pulmonary arteries
due to Valsalva maneuver, cardiac insufficien-
cy, or other cause of mixing of opacified and un-
opacified blood; beam-hardening attenuation ar-
tifact caused by adjacent high-density structures
such as opacified veins, contrast material pooling
in the inferior vena cava or right ventricle, or bony
structures; and the presence of airspace disease
obscuring the underlying pulmonary vasculature.
Descriptive numeric values were used for pa-
tient and PE demographics (actual values, per-
centages, mean [± SD], and ranges). Comparisons
64 (17–99) 60 (23–91) radiologists were of the consensus opinion
67 (38.5) 31 (46.2) that 45 (25.9%) of the original 174 index cases
107 (61.5) 14 (13.1)
were negative for the presence of PE. Interob-
server agreement between the panel members
327.0 291.3a
was almost perfect (weighted κ = 0.835).
6 (3.4) 0 (0) These discordant cases comprised 25 wom-
28 (16.1) 1 (4.0) en (none of whom were pregnant) and 20 men
37 (21.3) 6 (16.2)
(mean age, 60 years; range, 23–91 years).
Overall image quality was considered to be
71 (40.8) 19 (26.8) satisfactory for diagnosis in 170 examinations
32 (18.4) 19 (59.4) (98%) and inadequate for diagnosis in four
examinations (2%). The average quality of
contrast enhancement was 291.3 ± 60.9 HU,
versus a mean of 338.5 ± 91.8 HU in the group
with a concordant diagnosis (p = 0.002).
There was discordance between the chest
All Cases Discordant Cases radiologists and the original radiologist in 31
of 67 (46.2%) cases of reported solitary PE,
67 (38.5) 31 (46.2)
whereas discordance occurred in only 14 of
1 (0.6) 0 (0.0) 107 (13.1%) cases where multiple PEs were
5 (2.9) 1 (20.0) originally reported. Discordance was highest
6 (3.4) 4 (66.7) for cases of reported peripheral PEs (38/103
[36.9%] cases of segmental or subsegmental
31 (17.8) 10 (32.3)
PEs), with the highest rate for reported sol-
24 (13.8) 16 (66.7) itary subsegmental PEs (16/24 [66.7%] of
such cases). Discordance occurred most
commonly in the lower lobes, with the most
PE by the original radiologist (Table 1), commonly involved vessels being the lateral
and comprised 84 male and 90 female pa- basal and posterior basal segmental arteries
tients with a mean age of 64 years (range, of the left lower lobe; these vessels accounted
17–99 years). PEs were reported as solitary for 35.5% of all discordant cases diagnosed at
in 67 cases (38.5%) and multiple in 107 cas- the segmental level. Interestingly, no discor-
es (61.5%) (Table 2). PEs were more fre- dant diagnoses were seen in the right middle
quently reported in the peripheral segmen- lobe (where diagnostic difficulty might have
tal and subsegmental arteries (103 cases; been expected because of the more horizon-
7
6
No. of Discordant Cases

between groups were performed using the paired t 5

test and for ordinal categories using the chi-square
Fig. 2—Bar chart 4
test. A p value of 0.05 or less was considered sta- depicting number of
tistically significant. All statistics were performed discordant cases of 3
using SPSS (version 16, IBM). pulmonary embolism
diagnosis per attending 2
radiologist. Each letter
Results anonymously represents
1
There were 937 pulmonary CTA exami- each of 15 attending
nations performed over the course of the radiologists. Asterisk
denotes member of 0
12-month study period. Of these, 174 stud- panel of subspecialist
A B C D E F G H I J* K L M N O
ies (18.6%) were reported as positive for Attending Radiologist
chest radiologists.

AJR:205, August 2015 273


20
18
16
No. of Discordant Cases
14
12
10
Downloaded from www.ajronline.org by 36.73.92.133 on 06/14/17 from IP address 36.73.92.133. Copyright ARRS. For personal use only; all rights reserved
8 devastating or fatal) [8], interactions with oth-
6 er medications, inconvenience in terms of at-
4 tendance for repeated blood tests (which may
2 require time off work), and cost (to both the
0
Breathing Cardiac
(42.2%) (11.1%)
Artifact Simulating PE
tal course of the pulmonary arteries and their
greater susceptibility to partial volume aver-
aging effects). The distribution of instances
of discordance between the chest radiologists
and the original reporting radiologists was
relatively even, varying from zero to six cas-
es per radiologist (median, three cases), indi-
cating that this was a generalized rather than
an individual phenomenon (Fig. 2).
Causes for the 45 cases of discordance in-
cluded the following: 24 (53%) cases were
due to motion artifact from breathing (19
[42.2%]) or cardiac pulsation (5 [11.1%]),
eight (18%) cases were due to poor con-
trast opacification from Valsalva maneuver
(3 [6.7%]) or contrast material mixing (5
[11.1%]), 10 cases (22.2%) were due to atten-
uation artifact secondary to beam hardening
from adjacent high-density structures, and
three cases (6.7%) were due to effects from
adjacent airspace disease (Fig. 3).
Discussion
This study shows an unexpectedly high
rate of overdiagnosis of PE by pulmonary
CTA in a tertiary-care university hospital,
with an overall rate of 25.9% of all positive
pulmonary CTA examinations, increasing to
as high as 66.7% of cases where a solitary
subsegmental PE was originally reported.
Discordance was greatest for solitary PEs,
PEs located in segmental and subsegmental
pulmonary arteries, and in the lower zones
of the lungs. The positive predictive value of
pulmonary CTA for the diagnosis of PE was
only 74.1% in this study.
The published overall rate of positive diag-
nosis of PE on pulmonary CTA varies from
study to study (e.g., 15.4% [14], 16.4% [13],
and 17.8% [6]) but usually ranges between
14% [20] and 22% [7]. Differing levels of
274 AJR:205, August 2015
Hutchinson et al.
Fig. 3—Bar chart ment and either inconclusive or possibly er-
showing relative roneous pulmonary CTA results can cause
frequencies of artifact
thought to have difficulties in patient diagnosis and manage-
been responsible ment, often leading to repeat imaging and
for misdiagnosis in unnecessary anticoagulation therapy.
discordant cases. PE =
pulmonary embolism.
The risks and disadvantages of anticoagula-
tion therapy include hemorrhage (occasionally
patient and society) [23]. Furthermore, a diag-
Mixing Valsalva Beam Airspace nosis of PE carries with it implications for life
(11.1%) (6.7%) Hardening Disease insurance coverage, travel plans, and prepara-
(22.2%) (6.7%)
tion for other medical or surgical procedures.
The diagnosis of PE also places the patient in
a higher risk category for future events, which
adherence to referral guidelines, which has can influence investigations and management
been shown to significantly affect positivi- if the patient again seeks medical attention for
ty rates, may explain some of this variation; similar symptoms. The significance of a false-
for example, 30% of all pulmonary CTA ex- positive pulmonary CTA examination should
aminations were positive for PE in the mul- be considered in this context.
ticenter Christopher Study [2], which used Previously published studies have shown
strict adherence to a basic pretest risk strati- differences between chest- and non-chest-
fication tool. The rate of index positive cas- trained radiologists in the diagnostic accu-
es in our center was 18.6% for the 12-month racy of pulmonary CTA interpretation. In a
period studied. After review by the panel of 2011 study of 70 isolated subsegmental PEs
chest radiologists, this was revised down- by Pena et al. [13], a reviewing thoracic radi-
ward to 13.8%. In our institution, there is no ologist reinterpreted 11% of these examina-
systematic use of pretest probability scor- tions as negative. In a separate abstract pub-
ing (e.g., Well or Revised Geneva scores [21, lished by Miller et al. [18], a single thoracic
22]) and inconsistent use of d-dimer assays. radiologist found a false-positive or probable
Furthermore, many pulmonary CTA exami- false-positive rate of 11% at all pulmonary
nations in our institution are ordered by the artery levels in 508 cases. Compared with
emergency department before assessment by these previous studies, the current study is
the admitting medical team. The combina- larger and uses a panel of three subspecial-
tion of a lack of pretest probability assess- ty chest radiologists as a more robust surro-
A B
Fig. 4—90-year-old woman with pulmonary embolism in left lower lobe pulmonary artery.
A, Coronal pulmonary CT angiography image (mediastinal window) shows apparent filling defect (arrow) within
left lower lobe pulmonary artery.
B, Same image on lung window shows stair-step artifact (arrowhead) related to respiratory motion in right
lower lobe artery and right major fissure (arrow) with rapid position movement.
 Overdiagnosis of Pulmonary Embolism by CTA

The second-most-common confounding ar-

tifact was beam-hardening attenuation artifact
(Fig. 5) from high-density structures, includ-
ing pooled contrast agent in the SVC or oth-
er adjacent vessels, metallic structures such as
pacemakers, or the patient’s arms if they can-
not be elevated above the chest. The use of a
Downloaded from www.ajronline.org by 36.73.92.133 on 06/14/17 from IP address 36.73.92.133. Copyright ARRS. For personal use only; all rights reserved

saline chaser helps clear pooled contrast agent

from the SVC [10]. Apart from their prox-
imity to a high-density structure, regions of
low attenuation related to streak artifact have
much higher densities (> 78 HU) than real
thrombus and form indistinct borders with
contrast agent in the vessels [10]. We also ob-
served that beam-hardening attenuation arti-
A B fact could often be tracked in a radial pattern
Fig. 5—24-year-old woman presenting with pleuritic chest pain and deep calf tenderness. from the source of the artifact and could also
A, Axial image from pulmonary CT angiography shows apparent filling defect (arrow) in segmental pulmonary be identified in other nearby structures.
artery in right lower lobe, which is due to streak artifact secondary to beam hardening from high-density
contrast material in right ventricle.
Other artifacts responsible for misinter-
B, Example of beam attenuation artifact from high-density contrast material in superior vena cava, creating pretation included cardiac pulsatility (Figs. 6
pseudoembolus (arrow) in truncus anterior pulmonary artery. and 7), which is most often seen in regions
of the lung adjacent to the heart, such as the
gate reference standard, rather than relying cause identified for the misdiagnosis of PE lingula and the paracardiac segments of the
on one single radiologist’s opinion. Although was motion artifact due to breathing, which lower lobes; reduced mixing of contrast agent
this was a single-center study, our depart- accounted for 42.2% of cases. Breathing arti- with unopacified blood, which can be due to
ment does not differ in any significant way fact has previously been shown to be the most excessive inflow of unopacified blood from
from any other university hospital imaging common mimic of PE [18] as well as the most the inferior vena cava or other veins, exces-
center, with the same mix of inpatients, out- common cause of equivocal pulmonary CTA sive breath-holding resulting in a Valsalva
patients, emergency department patients, findings in up to 74% of cases [9]. Breathing maneuver, or poor cardiac function and poor
and pregnant patients as might be found in artifact can most easily be identified on a lung mixing of contrast agent; and obscuration of
any equivalent tertiary referral center, and window by the presence of the seagull arti- the pulmonary arteries by adjacent paren-
with a modern radiology department us- fact, the stair-step artifact, and rapid chang- chymal disease. The latter is attributed to in-
ing conventional MDCT technology and a es in position of vessels on contiguous image creased local vascular resistance, which leads
PACS for the performance and interpretation slices [10] (Fig. 4). Ways to reduce the level of to reduced flow and flow artifacts [10]. Inter-
of pulmonary CTA examinations, staffed breathing artifact include administering sup- estingly, in our study, there were no discrep-
by a general mix of experienced subspecial- plemental oxygen and scanning in the caudo- ancies due to confusion between PEs and pul-
ty fellowship-trained radiologists. The high cranial direction [24]. monary veins or mucus-filled bronchi. This
rates of discrepant pulmonary CTA inter-
pretations found in this study raise concerns
about the diagnostic accuracy of radiologists
in the wider community. However, the gener-
alizability of our results should be confirmed
with a larger multicenter study.
Causes of diagnostic difficulty in the in-
terpretation of pulmonary CTA examinations
are well recognized [9–11]. A full description
of such interpretative pitfalls is beyond the
scope of this discussion, but potential false-
positive findings are known to occur because
of partial volume averaging effects secondary
to motion (breathing and cardiogenic), poor
contrast opacification from mixing of opaci-
fied and unopacified blood, beam-hardening
attenuation artifact from high-density struc- A B
tures (e.g., contrast agent in the superior vena Fig. 6—61-year-old man with suspected pulmonary embolism.
cava [SVC] and right atrium), and confusion A, Axial image from pulmonary CT angiography at level of left lower lobe shows effects of cardiac pulsation
artifact, which results in partial volume averaging effect between high density in left lower pulmonary arteries
with venous structures and mucus-filled bron- and low density of adjacent lung (arrows), which was confused with pulmonary emboli.
chi [9, 10, 24]. In our study, the most common B, On lung window, this artifact can be recognized by blurring of walls of affected arteries (arrows).

AJR:205, August 2015 275


Downloaded from www.ajronline.org by 36.73.92.133 on 06/14/17 from IP address 36.73.92.133. Copyright ARRS. For personal use only; all rights reserved
Fig. 7—59-year-old man with chest pain, hemoptysis,
and elevated d-dimer assay. Axial image from
pulmonary CT angiography shows cardiac pulsation
artifact causing inhomogeneity in contrast material in
left upper lobe pulmonary artery (arrow), which was
confused with pulmonary embolus. Cardiac motion
also causes movement of walls of ascending aorta
and of main pulmonary outflow tract (arrowheads).
might suggest that discrepancies in the diag-
nosis of PE arose not because of unfamiliari-
ty with anatomy or to a lack of attention to de-
tail when reading the scan but rather because
of perceptual errors resulting from an under-
recognition of the other causes of false-pos-
itive examinations, as summarized already.
Our study also highlights the difficulty of
performing audits of the accuracy of pulmo-
nary CTA interpretation. The original deci-
sion by the treating physician to initiate an-
ticoagulation therapy would be taken within
the clinical context of the patient’s presenta-
tion, history of thromboembolic disease, car-
diac workup, d-dimer levels, and so forth, and
is not based purely on the result of the pulmo-
nary CTA scan. In our study, this clinical in-
formation was not included. In this regard, as
a specific outcome from our study, it was con-
sidered to be neither clinically appropriate nor
ethical to revisit the original clinical diagnosis
several years later on the basis of the results of
an academic study that had not been designed
to reexamine all of the clinical information
that was originally available. Rather, the pur-
pose of our study was to examine the diagnos-
tic difficulties in the use of pulmonary CTA as
a diagnostic study in isolation.
In the absence of a practical true reference
standard, we opted to rely on the consen-
sus opinion of three experienced chest radi-
ologists, which would be difficult to repro-
duce on a routine basis because of the time
involved in collating and reviewing the nec-
essary examinations. Correlation of pulmo-
276 AJR:205, August 2015
Hutchinson et al.
nary CTA findings with clinical outcomes
(e.g., recurrent thromboembolism or death)
is a crude measure of accuracy and would
also be difficult to achieve in routine clini-
cal practice. The difficulty in performing
regular audits of this very common imaging
test highlights the risk of unrecognized di-
agnostic drift, where an established diagnos-
tic test performs less well over time because
of changes in practice and personnel and be-
cause of an absence of feedback or correla-
tive reference standard test.
Practical measures to reduce the risk of PE
misdiagnosis could and should include any
of the following: systematic use of pretest
probability assessment (which would require
buy-in from clinicians and incorporation into
imaging protocols); radiology technologists
being educated to optimize image quality,
focusing on proper patient breathing tech-
nique and repeating examinations where
appropriate; increased familiarization by
radiologists with the range of potential diag-
nostic pitfalls; encouragement of the use of
second opinions by interpreting radiologists,
particularly for solitary subsegmental PEs;
and regular review of positive pulmonary
CTA cases (e.g., at monthly discrepancy or
audit meetings). Some of these measures are
easier to implement than others, but their im-
portance is underscored by the implications
of a false-positive diagnosis of PE.
Conclusion
When compared with the consensus opin-
ion of a panel of three chest radiologists, we
found a high rate of overdiagnosis of PE by
pulmonary CTA in a department of mixed
specialty radiologists, which appeared to be
due to a lack of recognition of the imaging
pitfalls that can be encountered in the inter-
pretation of pulmonary CTA examinations.
Increased education among radiography
technologists, radiologists, and clinicians re-
garding these pitfalls should be encouraged.
References
1. Rémy-Jardin M, Pistolesi M, Goodman LR, et al.
Management of suspected acute pulmonary embo-
lism in the era of CT angiography: a statement from
the Fleischner Society. Radiology 2007; 245:315–329
2. van Belle A, Büller HR, Huisman MV, et al.;
Christopher Study Investigators. Effectiveness of
managing suspected pulmonary embolism using
an algorithm combining clinical probability, d-
dimer testing, and computed tomography. JAMA
2006; 295:172–179
3. Stein PD, Fowler SE, Goodman LR, et al. Multide-
tector computed tomography for acute pulmonary
embolism. N Engl J Med 2006; 354:2317–2327
4. Adams DM, Stevens SM, Woller SC, et al. Adher-
ence to PIOPED II investigators’ recommenda-
tions for computed tomography pulmonary angi-
ography. Am J Med 2013; 126:36–42
5. Corwin MT, Donohoo JH, Partridge R, Egglin
TK, Mayo-Smith WW. Do emergency physicians
use serum d-dimer effectively to determine the
need for CT when evaluating patients for pulmo-
nary embolism? Review of 5,344 consecutive pa-
tients. AJR 2009; 192:1319–1323
6. Ranji SR, Shojania KG, Trowbridge RL, Auerbach
AD. Impact of reliance on CT pulmonary angiog-
raphy on diagnosis of pulmonary embolism: a
Bayesian analysis. J Hosp Med 2006; 1:81–87
7. Eyer BA, Goodman LR, Washington L. Clini-
cians “response to radiologists” reports of isolated
subsegmental pulmonary embolism or inconclu-
sive interpretation of pulmonary embolism using
MDCT. AJR 2005; 184:623–628
8. Linkins LA, Choi PT, Douketis JD. Clinical im-
pact of bleeding in patients taking oral anticoagu-
lant therapy for venous thromboembolism: a me-
ta-analysis. Ann Intern Med 2003; 139:893–900
9. Jones SE, Wittram C. The indeterminate CT pulmo-
nary angiogram: imaging characteristics and patient
clinical outcome. Radiology 2005; 237:329–337
10. Wittram C, Maher MM, Yoo AJ, Kalra MK,
Shepard JAO, McLoud TC. CT angiography of
pulmonary embolism: diagnostic criteria and
causes of misdiagnosis. RadioGraphics 2004;
24:1219–1238
11. Bruzzi JF, Rémy-Jardin M, Kirsch J, et al. Sixteen-
slice multidetector computed tomography pulmo-
nary angiography: evaluation of cardiogenic motion
artifacts and influence of rotation time on image
quality. J Comput Assist Tomogr 2005; 29:805–814
12. Ghanima W, Nielssen BE, Holmen LO, Witwit A,
Al-Ashtari A, Sandset PM. Multidetector com-
puted tomography (MDCT) in the diagnosis of
pulmonary embolism: interobserver agreement
among radiologists with varied levels of experi-
ence. Acta Radiol 2007; 48:165–170
13. Pena E, Kimpton M, Dennie C, Peterson R, Le
Gal G, Carrier M. Difference in interpretation of
computed tomography pulmonary angiography
diagnosis of subsegmental thrombosis in patients
with suspected pulmonary embolism. J Thromb
Haemost 2012; 10:496–498
14. Costa AF, Basseri H, Sheikh A, Stiell I, Dennie C.
The yield of CT pulmonary angiograms to ex-
clude acute pulmonary embolism. Emerg Radiol
2014; 21:133–141
15. Yavas US, Calisir C, Ozkan IR. The interobserver
agreement between residents and experienced radi-
ologists for detecting pulmonary embolism and DVT
with using CT pulmonary angiography and indirect
 Overdiagnosis of Pulmonary Embolism by CTA
CT venography. Korean J Radiol 2008; 9:498–502
16. Shaham D, Heffez R, Bogot NR, Libson E, Brezis
M. CT pulmonary angiography for the detection
of pulmonary embolism: interobserver agreement
between on-call radiology residents and special-
ists (CTPA interobserver agreement). Clin Imag-
ing 2006; 30:266–270
Downloaded from www.ajronline.org by 36.73.92.133 on 06/14/17 from IP address 36.73.92.133. Copyright ARRS. For personal use only; all rights reserved
17. Ginsberg MS, King V, Panicek DM. Comparison
of interpretations of CT angiograms in the evalu-
ation of suspected pulmonary embolism by on-
call radiology fellows and subsequently by radiol-
ogy faculty. AJR 2004; 182:61–66
18.
Miller WT Jr, Marinari LA, Mahne A. Fre-
AJR:205, August 2015 277
quency and causes of false-positive CTPA exams in
community hospitals. Chest 2009; 136(4_
MeetingAbstracts):14S
19. Boyden EA. Segmental anatomy of the
lungs. New York, NY: McGraw-Hill, 1955
20. Donato AA, Khoche S, Santora J, Wagner B.
Clinical outcomes in patients with isolated sub-
segmental pulmonary emboli diagnosed by multi-
detector CT pulmonary angiography. Thromb Res
2010; 126:e266–e270
21. Wells PS, Anderson DR, Rodger M, et al. Deriva-
tion of a simple clinical model to categorize pa-
tients probability of pulmonary embolism: in-
creasing the models utility with the SimpliRED
d -dimer. Thromb Haemost 2000; 83:416–420
22. Le Gal G, Righini M, Parent F, van Strijen M,
Couturaud F. Diagnosis and management of sub-
segmental pulmonary embolism. J Thromb Hae-
most 2006; 4:724–731
23. Lefebvre P, Laliberté F, Nutescu EA, et al. All-
cause and potentially disease-related health care
costs associated with venous thromboembolism
in commercial, Medicare, and Medicaid benefi-
ciaries. J Manag Care Pharm 2012; 18:363–374
24. Wittram C. How I do it: CT pulmonary angiogra-
phy. AJR 2007; 188:1255–1261