Professional Documents
Culture Documents
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.
P
ulmonary embolism (PE) is a anticoagulation therapy on the basis of a posi-
common clinical diagnosis in pa- tive result, regardless of pretest probability
tients presenting to the emergen- [6], even in isolated subsegmental PE [7].
cy department and in hospitalized The risk of hemorrhage related to antico-
patients. It is normally treated with long-term agulation therapy is potentially significant.
Keywords: artifact, false-positive, misdiagnosis,
anticoagulation therapy to reduce the risks of A large meta-analysis in 2003 [8] found a
pulmonary CT angiography, pulmonary embolism death and the morbidity associated with 7% annual risk of major bleeding and a 0.4%
chronic pulmonary venous thromboembo- incidence of bleeding-related fatality in pa-
DOI:10.2214/AJR.14.13938 lism. Because the clinical presentation is of- tients treated with oral anticoagulation ther-
ten nonspecific and can be mimicked by a apy for venous thromboembolism for longer
Received October 10, 2014; accepted after revision
January 23, 2015. range of other conditions, in routine practice, than 3 months. The practical implications of
pulmonary CT angiography (CTA) is often long-term anticoagulation therapy for the pa-
1
Department of Radiology, University Hospital Galway, used as the imaging method of choice for fur- tient are also potentially significant, requir-
Newcastle Rd, Galway, Ireland. Address correspondence ther investigation [1]. Pulmonary CTA has ing frequent attendance to their medical
to B. D. Hutchinson (barryhutchinson82@gmail.com).
been shown to be highly sensitive and specific practitioners for blood tests, consequent time
2
Department of Chest Radiology, M. D. Anderson Cancer when pretest clinical diagnostic tools are used off from work, potential adverse drug inter-
Center, Houston, TX. [2] but surprisingly inaccurate in patients with actions with other medications, adjustments
low pretest probability, with false-positive to travel and lifestyle, implications for future
AJR 2015; 205:271–277 rates as high as 42% [3]. Unfortunately, ad- dental and medical procedures, and possible
0361–803X/15/2052–271
herence to referral guidelines for pulmonary negative effects on life insurance status.
CTA has repeatedly been shown to be low [4, With these considerations in mind, it is
© American Roentgen Ray Society 5]. Nonetheless, many clinicians will initiate important to minimize the misdiagnosis
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) 64 (17–99) 60 (23–91) radiologists were of the consensus opinion
Single-vessel PE 67 (38.5) 31 (46.2) that 45 (25.9%) of the original 174 index cases
Multiple-vessel PEs 107 (61.5) 14 (13.1)
were negative for the presence of PE. Interob-
server agreement between the panel members
Mean quality of contrast enhancement (HU) 327.0 291.3a
was almost perfect (weighted κ = 0.835).
Pulmonary trunk PE 6 (3.4) 0 (0) These discordant cases comprised 25 wom-
Main pulmonary artery PE 28 (16.1) 1 (4.0) en (none of whom were pregnant) and 20 men
Lobar pulmonary artery PE 37 (21.3) 6 (16.2)
(mean age, 60 years; range, 23–91 years).
Overall image quality was considered to be
Segmental pulmonary artery PE 71 (40.8) 19 (26.8) satisfactory for diagnosis in 170 examinations
Subsegmental pulmonary artery PE 32 (18.4) 19 (59.4) (98%) and inadequate for diagnosis in four
Note—Except where noted otherwise, data are number (%) of patients. examinations (2%). The average quality of
ap = 0.002.
contrast enhancement was 291.3 ± 60.9 HU,
versus a mean of 338.5 ± 91.8 HU in the group
TABLE 2: Comparison of Solitary Pulmonary Embolism (PE) in All and with a concordant diagnosis (p = 0.002).
Discordant Pulmonary CT Angiography Examinations There was discordance between the chest
Location of PE All Cases Discordant Cases radiologists and the original radiologist in 31
of 67 (46.2%) cases of reported solitary PE,
Total 67 (38.5) 31 (46.2)
whereas discordance occurred in only 14 of
Pulmonary trunk PE 1 (0.6) 0 (0.0) 107 (13.1%) cases where multiple PEs were
Main pulmonary artery PE 5 (2.9) 1 (20.0) originally reported. Discordance was highest
Lobar pulmonary artery PE 6 (3.4) 4 (66.7) for cases of reported peripheral PEs (38/103
[36.9%] cases of segmental or subsegmental
Segmental pulmonary artery PE 31 (17.8) 10 (32.3)
PEs), with the highest rate for reported sol-
Subsegmental pulmonary artery PE 24 (13.8) 16 (66.7) itary subsegmental PEs (16/24 [66.7%] of
Note—Except where noted otherwise, data are number (%) of patients. such cases). Discordance occurred most
commonly in the lower lobes, with the most
tifact from breathing or cardiac pulsation; poor PE by the original radiologist (Table 1), commonly involved vessels being the lateral
contrast opacification of the pulmonary arteries and comprised 84 male and 90 female pa- basal and posterior basal segmental arteries
due to Valsalva maneuver, cardiac insufficien- tients with a mean age of 64 years (range, of the left lower lobe; these vessels accounted
cy, or other cause of mixing of opacified and un- 17–99 years). PEs were reported as solitary for 35.5% of all discordant cases diagnosed at
opacified blood; beam-hardening attenuation ar- in 67 cases (38.5%) and multiple in 107 cas- the segmental level. Interestingly, no discor-
tifact caused by adjacent high-density structures es (61.5%) (Table 2). PEs were more fre- dant diagnoses were seen in the right middle
such as opacified veins, contrast material pooling quently reported in the peripheral segmen- lobe (where diagnostic difficulty might have
in the inferior vena cava or right ventricle, or bony tal and subsegmental arteries (103 cases; been expected because of the more horizon-
structures; and the presence of airspace disease
obscuring the underlying pulmonary vasculature.
7
Descriptive numeric values were used for pa-
tient and PE demographics (actual values, per- 6
centages, mean [± SD], and ranges). Comparisons
No. of Discordant Cases
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- A B
nosis of PE on pulmonary CTA varies from Fig. 4—90-year-old woman with pulmonary embolism in left lower lobe pulmonary artery.
study to study (e.g., 15.4% [14], 16.4% [13], A, Coronal pulmonary CT angiography image (mediastinal window) shows apparent filling defect (arrow) within
left lower lobe pulmonary artery.
and 17.8% [6]) but usually ranges between B, Same image on lung window shows stair-step artifact (arrowhead) related to respiratory motion in right
14% [20] and 22% [7]. Differing levels of lower lobe artery and right major fissure (arrow) with rapid position movement.
nary CTA findings with clinical outcomes tector computed tomography for acute pulmonary
(e.g., recurrent thromboembolism or death) embolism. N Engl J Med 2006; 354:2317–2327
is a crude measure of accuracy and would 4. Adams DM, Stevens SM, Woller SC, et al. Adher-
also be difficult to achieve in routine clini- ence to PIOPED II investigators’ recommenda-
cal practice. The difficulty in performing tions for computed tomography pulmonary angi-
regular audits of this very common imaging ography. Am J Med 2013; 126:36–42
test highlights the risk of unrecognized di- 5. Corwin MT, Donohoo JH, Partridge R, Egglin
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
agnostic drift, where an established diagnos- TK, Mayo-Smith WW. Do emergency physicians
tic test performs less well over time because use serum d-dimer effectively to determine the
of changes in practice and personnel and be- need for CT when evaluating patients for pulmo-
cause of an absence of feedback or correla- nary embolism? Review of 5,344 consecutive pa-
tive reference standard test. tients. AJR 2009; 192:1319–1323
Practical measures to reduce the risk of PE 6. Ranji SR, Shojania KG, Trowbridge RL, Auerbach
misdiagnosis could and should include any AD. Impact of reliance on CT pulmonary angiog-
of the following: systematic use of pretest raphy on diagnosis of pulmonary embolism: a
probability assessment (which would require Bayesian analysis. J Hosp Med 2006; 1:81–87
Fig. 7—59-year-old man with chest pain, hemoptysis, buy-in from clinicians and incorporation into 7. Eyer BA, Goodman LR, Washington L. Clini-
and elevated d-dimer assay. Axial image from imaging protocols); radiology technologists cians “response to radiologists” reports of isolated
pulmonary CT angiography shows cardiac pulsation
artifact causing inhomogeneity in contrast material in
being educated to optimize image quality, subsegmental pulmonary embolism or inconclu-
left upper lobe pulmonary artery (arrow), which was focusing on proper patient breathing tech- sive interpretation of pulmonary embolism using
confused with pulmonary embolus. Cardiac motion nique and repeating examinations where MDCT. AJR 2005; 184:623–628
also causes movement of walls of ascending aorta appropriate; increased familiarization by 8. Linkins LA, Choi PT, Douketis JD. Clinical im-
and of main pulmonary outflow tract (arrowheads).
radiologists with the range of potential diag- pact of bleeding in patients taking oral anticoagu-
might suggest that discrepancies in the diag- nostic pitfalls; encouragement of the use of lant therapy for venous thromboembolism: a me-
nosis of PE arose not because of unfamiliari- second opinions by interpreting radiologists, ta-analysis. Ann Intern Med 2003; 139:893–900
ty with anatomy or to a lack of attention to de- particularly for solitary subsegmental PEs; 9. Jones SE, Wittram C. The indeterminate CT pulmo-
tail when reading the scan but rather because and regular review of positive pulmonary nary angiogram: imaging characteristics and patient
of perceptual errors resulting from an under- CTA cases (e.g., at monthly discrepancy or clinical outcome. Radiology 2005; 237:329–337
recognition of the other causes of false-pos- audit meetings). Some of these measures are 10. Wittram C, Maher MM, Yoo AJ, Kalra MK,
itive examinations, as summarized already. easier to implement than others, but their im- Shepard JAO, McLoud TC. CT angiography of
Our study also highlights the difficulty of portance is underscored by the implications pulmonary embolism: diagnostic criteria and
performing audits of the accuracy of pulmo- of a false-positive diagnosis of PE. causes of misdiagnosis. RadioGraphics 2004;
nary CTA interpretation. The original deci- 24:1219–1238
sion by the treating physician to initiate an- Conclusion 11. Bruzzi JF, Rémy-Jardin M, Kirsch J, et al. Sixteen-
ticoagulation therapy would be taken within When compared with the consensus opin- slice multidetector computed tomography pulmo-
the clinical context of the patient’s presenta- ion of a panel of three chest radiologists, we nary angiography: evaluation of cardiogenic motion
tion, history of thromboembolic disease, car- found a high rate of overdiagnosis of PE by artifacts and influence of rotation time on image
diac workup, d-dimer levels, and so forth, and pulmonary CTA in a department of mixed quality. J Comput Assist Tomogr 2005; 29:805–814
is not based purely on the result of the pulmo- specialty radiologists, which appeared to be 12. Ghanima W, Nielssen BE, Holmen LO, Witwit A,
nary CTA scan. In our study, this clinical in- due to a lack of recognition of the imaging Al-Ashtari A, Sandset PM. Multidetector com-
formation was not included. In this regard, as pitfalls that can be encountered in the inter- puted tomography (MDCT) in the diagnosis of
a specific outcome from our study, it was con- pretation of pulmonary CTA examinations. pulmonary embolism: interobserver agreement
sidered to be neither clinically appropriate nor Increased education among radiography among radiologists with varied levels of experi-
ethical to revisit the original clinical diagnosis technologists, radiologists, and clinicians re- ence. Acta Radiol 2007; 48:165–170
several years later on the basis of the results of garding these pitfalls should be encouraged. 13. Pena E, Kimpton M, Dennie C, Peterson R, Le
an academic study that had not been designed Gal G, Carrier M. Difference in interpretation of
to reexamine all of the clinical information References computed tomography pulmonary angiography
that was originally available. Rather, the pur- 1. Rémy-Jardin M, Pistolesi M, Goodman LR, et al. diagnosis of subsegmental thrombosis in patients
pose of our study was to examine the diagnos- Management of suspected acute pulmonary embo- with suspected pulmonary embolism. J Thromb
tic difficulties in the use of pulmonary CTA as lism in the era of CT angiography: a statement from Haemost 2012; 10:496–498
a diagnostic study in isolation. the Fleischner Society. Radiology 2007; 245:315–329 14. Costa AF, Basseri H, Sheikh A, Stiell I, Dennie C.
In the absence of a practical true reference 2. van Belle A, Büller HR, Huisman MV, et al.; The yield of CT pulmonary angiograms to ex-
standard, we opted to rely on the consen- Christopher Study Investigators. Effectiveness of clude acute pulmonary embolism. Emerg Radiol
sus opinion of three experienced chest radi- managing suspected pulmonary embolism using 2014; 21:133–141
ologists, which would be difficult to repro- an algorithm combining clinical probability, d- 15. Yavas US, Calisir C, Ozkan IR. The interobserver
duce on a routine basis because of the time dimer testing, and computed tomography. JAMA agreement between residents and experienced radi-
involved in collating and reviewing the nec- 2006; 295:172–179 ologists for detecting pulmonary embolism and DVT
essary examinations. Correlation of pulmo- 3. Stein PD, Fowler SE, Goodman LR, et al. Multide- with using CT pulmonary angiography and indirect
CT venography. Korean J Radiol 2008; 9:498–502 quency and causes of false-positive CTPA exams in creasing the models utility with the SimpliRED
16. Shaham D, Heffez R, Bogot NR, Libson E, Brezis community hospitals. Chest 2009; 136(4_ d -dimer. Thromb Haemost 2000; 83:416–420
M. CT pulmonary angiography for the detection MeetingAbstracts):14S 22. Le Gal G, Righini M, Parent F, van Strijen M,
of pulmonary embolism: interobserver agreement 19. Boyden EA. Segmental anatomy of the Couturaud F. Diagnosis and management of sub-
between on-call radiology residents and special- lungs. New York, NY: McGraw-Hill, 1955 segmental pulmonary embolism. J Thromb Hae-
ists (CTPA interobserver agreement). Clin Imag- 20. Donato AA, Khoche S, Santora J, Wagner B. most 2006; 4:724–731
ing 2006; 30:266–270 Clinical outcomes in patients with isolated sub- 23. Lefebvre P, Laliberté F, Nutescu EA, et al. All-
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 segmental pulmonary emboli diagnosed by multi- cause and potentially disease-related health care
of interpretations of CT angiograms in the evalu- detector CT pulmonary angiography. Thromb Res costs associated with venous thromboembolism
ation of suspected pulmonary embolism by on- 2010; 126:e266–e270 in commercial, Medicare, and Medicaid benefi-
call radiology fellows and subsequently by radiol- 21. Wells PS, Anderson DR, Rodger M, et al. Deriva- ciaries. J Manag Care Pharm 2012; 18:363–374
ogy faculty. AJR 2004; 182:61–66 tion of a simple clinical model to categorize pa- 24. Wittram C. How I do it: CT pulmonary angiogra-
18.
Miller WT Jr, Marinari LA, Mahne A. Fre- tients probability of pulmonary embolism: in- phy. AJR 2007; 188:1255–1261