Missed Pulmonary Embolism on Abdominal CT

Cardiopulmonary Imaging Original Research Lim et al. Pulmonary Emboli Missed on CT of the Abdomen Cardiopulmonary Imaging Original Research Kun Young Lim 1 Seth J. Kligerman 2 Cheng Ting Lin 2 Charles S. White 2 Lim KY, Kligerman SJ, Lin CT, White CS Keywords: abdominal CT, missed diagnosis, pulmonary embolism DOI:10.2214/AJR.13.11436 Received June 25, 2013; accepted after revision August 17, 2013. 1 Department of Radiology, National Cancer Center, Goyang-si Gyeonggi-do, Republic of Korea. 2 Department of Diagnostic Radiology and Nuclear Medicine, University of Maryland School of Medicine, 22 South Greene St, Baltimore, MD 21201. Address correspondence to C. S. White (cwhite@umm.edu). AJR 2014; 202:738 743 0361 803X/14/2024 738 American Roentgen Ray Society Missed Pulmonary Embolism on Abdominal CT OBJECTIVE. The purpose of this study is to evaluate a series of missed pulmonary emboli (PE) identified on abdominal CT and to describe their characteristics and the clinical scenario. MATERIALS AND METHODS. All reports of chest CT scans performed during a 12-month period were searched for keywords indicative of PE. Among patients with PE, patients who also underwent an enhanced abdominal CT within 3 months were assessed for missed PE. Three radiologists reviewed the abdominal CT to confirm the presence of a missed PE. Missed PEs were classified as unknown or known. Each study was assessed for characteristics of the missed PE and the image quality of the PE study. The electronic medical record was used to document the clinical context in which the PE occurred. RESULTS. Eighteen patients (12 men and six women; average age, 58.8 years) were identified as having missed PE on abdominal CT. In seven patients (38.9%), the PE had not been previously diagnosed. Most of the missed PEs were segmental, but three missed PEs occurred in lobar vessels. In a slight majority of the cases, the reviewing radiologists judged the contrast bolus as good. The abdominal CT on which PE was overlooked was obtained for a variety of reasons, most commonly because of abdominal pain or to follow up a preexisting condition. CONCLUSION. This study shows that missed PE can occur on abdominal CT. It is recommended that interpretation include a careful search of the lower pulmonary arterial vasculature on contrast-enhanced abdominal CT scans. P ulmonary embolism (PE) is the third most common cause of cardiovascular death after myocardial infarction and stroke, with an estimated annual incidence in the United States of 69 205 cases per 100,000 personyears [1, 2]. Although the mortality rate of PE has decreased in the last 2 decades, untreated PE continues to cause significant mortality, with rates ranging from 18% to 35%, depending on the underlying risk factors and the characteristics of the thrombus [2, 3]. During the past decade, CT technology has advanced considerably, and pulmonary CT angiography (CTA) has become the diagnostic method of choice for PE in most institutions [4]. Image quality is sufficient in many instances to permit detection of unsuspected PE on other contrast-enhanced CT examinations of the chest [5]. Moreover, contrast-enhanced abdominal CT scans obtained for a variety of indications typically include the lower portions of the chest. Consequently, there is a risk that PE might be missed on CT scans of the abdomen. Such overlooked PEs may be the cause of symptoms initially thought to be due to abdominal pathology, and the potential exists for significant morbidity and possibly a fatal outcome in this situation. To our knowledge, there are no prior reports of missed PE on abdominal CT scans. In this study, we report a series of missed PEs on abdominal CT, their characteristics, and the clinical scenario in which they occurred. Materials and Methods This project was approved by our institutional review board and is HIPAA compliant. The requirement for informed consent was waived. Patient Selection Initial selection was based on the identification of patients who underwent CT scans of the chest that were positive for PE at a single large university hospital. All reports of contrast-enhanced chest CT scans performed during a 12-month period between September 1, 2011, and August 31, 2012, were searched for the keywords embolism, emboli, embolus, thromboembolism, thrombi, thrombus, filling, and PE using an institutionally developed 738 AJR:202, April 2014

Pulmonary Emboli Missed on CT of the Abdomen electronic word search system (Radsearch). All reports were reviewed to ensure the positivity for PE. Among the patients with chest CT reports positive for PE, those who also underwent an abdominal CT with contrast enhancement or abdominal CTA within 3 months (before or after) of the PE-positive chest CT scan were documented. Both chest CT scans and abdominal CT scans were retrospectively examined on a PACS (Impax Agility, Agfa) by a board-certified thoracic radiologist with 8 years of postfellowship experience. Each abdominal CT scan was classified in a binary fashion according to whether PE was visible or not visible. Candidate cases were excluded for one of the following reasons: first, the patient had undergone chest CT and abdominal CT on the same day, and, thus, the PE had already been described on the chest CT; second, the region of the PE was not covered on abdominal CT; or third, the image quality of the abdominal CT scan was nondiagnostic because of suboptimal enhancement of relevant pulmonary arteries or because of substantial motion artifact. In the remaining cases, if PE was deemed to be present and there was no mention of the PE positivity in the abdominal CT scan report, the case was classified as a potentially missed PE on abdominal CT. All potentially missed PE studies seen on abdominal CT were reviewed by two additional board-certified thoracic radiologists with 5 and 22 years of experience, respectively. All three radiologists had to independently agree on the presence of PE on the abdominal CT. If any of the three was not certain of the diagnosis of PE on the abdominal CT, the study was excluded. In instances where a patient underwent multiple abdominal CT scans, the abdominal CT with the shortest time interval from the chest CT on which the PE diagnosis was made was selected. If the abdominal CT scan on which a PE occurred was performed before the diagnosis of PE on a contrast-enhanced chest CT, the cases were categorized as missed unknown PE. Conversely, if the abdominal CT scans were performed after the diagnosis of PE had previously been made on a chest CT, the cases were categorized as missed known PE. CT Technique The abdominal CT scans were performed on MDCT scanners ranging from 40- to 256-detector configurations. Scanners including Brilliance and ict (both from Philips Healthcare) were used for each of the CT scans. A standard abdominal CT protocol used 120 kv; a reference tube current exposure time product of 180 mas; 40, 64, or 128 0.625 collimation; 0.993 pitch; and 0.5-second rotation time. One hundred milliliters of contrast material (350 mg I/mL; iohexol; Omnipaque 350, GE Healthcare) was injected through an antecubital vein at a rate of 3 ml/s using an automatic injector (Stellant, Medrad). Imaging was started 70 seconds after injection at end-expiration beginning at the level of the diaphragm. Images were reconstructed with 2- to 4-mm slice thickness using a soft-tissue algorithm (B filter). Image Review CT images were assessed for characteristics of PE, including the number of PEs present (single vs multiple); lobar distribution; the most proximal location within the lobar, segmental, or subsegmental pulmonary arteries; chronicity (acute vs chronic); presence of vascular occlusion; and the attenuation of thrombi and pulmonary artery enhancement in Hounsfield units. PE was identified as an intraluminal filling defect that showed a clear interface with intravascular contrast media [1]. A PE with luminal occlusion accompanied by arterial enlargement, central location, or acute angle with the arterial wall was considered acute. PEs were classified as chronic if they occurred in a vessel smaller in caliber than typical for pulmonary arteries of the same order of branching, as either an eccentric filling defect with an obtuse angle to the arterial wall or as a web within an artery [6]. The attenuation of thrombus was measured from the most reliable and largest thrombus identified on the abdominal CT using a region of interest diameter half the diameter of the thrombus being measured. Attenuation of the pulmonary artery adjacent or just proximal to the thrombus was also measured with a region of interest diameter of half the diameter of the pulmonary artery. The three radiologists graded the conspicuity of the PE using Fig. 1 53-year-old man with chronic pancreatitis and recent hernia surgery. Missed known pulmonary embolism (PE) was found on contrast-enhanced CT of abdomen and pelvis. Scan was performed 26 days after diagnosis of PE had been made on chest CT. A, Axial 4-mm-thick image shows filling defect (arrow) in segmental branch of right lower lobe pulmonary artery that was not initially reported on abdomen and pelvis CT. B, Coronal image shows that filling defect (arrow) is seen only on first few slices of scan. A B AJR:202, April 2014 739

Lim et al. Fig. 2 92-year-old woman with history of recent antrectomy due to duodenal ulcer. Missed unknown pulmonary embolism (PE) was found on contrast-enhanced CT of abdomen and pelvis. Scan was obtained 2 days before diagnosis of PE was made on chest CT. A, Axial 4-mm-thick image shows filling defect (arrow) in segmental branch of right lower lobe pulmonary artery that was not initially reported. Bilateral pleural effusions and passive atelectasis are also noted. B, Coronal image shows filling defect (arrow) in branch of right lower lobe pulmonary artery. a numeric scale of 1 to 3, where 1 is very subtle (i.e., the filling defect is evident only in retrospect and would not be expected to be identified during primary interpretation), 2 is subtle (i.e., the filling defect is difficult to visualize but would likely be identified by careful inspection during primary interpretation), and 3 is evident (i.e., the filling defect is clearly visible in retrospect). Other findings visible in the lower thorax, including atelectasis, pleural effusion, consolidation, ground-glass opacity, and dependent opacities, were recorded. To evaluate image quality, the presence or absence of respiratory motion artifact was evaluated, and the quality of the contrast enhancement of the pulmonary arteries was assessed by the three radiologists. The latter was graded on a scale of 1 to 3 using the following criteria: 1 is poor (i.e., suboptimal enhancement of the pulmonary arteries distal to the lobar branches), 2 is limited (i.e., suboptimal enhancement of the pulmonary arteries distal to the segmental branches), and 3 is good (i.e., optimal enhancement of the pulmonary arteries distal to the subsegmental branches). As noted, cases with the suboptimal enhancement of the pulmonary arteries (grade 1) were excluded. Clinical Information The electronic medical records of the patients enrolled in the study group were reviewed to identify medical history, the indication for the abdominal CT, and the time interval between the initial chest CT at the time of diagnosis of PE and abdominal CT. When available, the value of d-dimer at the time of the chest CT or abdominal CT was noted. Statistical Analysis Descriptive statistics were used to assess each of the data categories. For outcome variables, chisquare analysis was used. Differences in patient populations were assessed with an unpaired Student t test. A p value less than 0.05 was considered to indicate a statistically significant difference. Fig. 3 83-year-old man with left common iliac aneurysm rupture. Missed unknown pulmonary embolism (PE) was found on abdominal and pelvic CT. Axial 4-mm-thick image shows small posterior subsegmental filling defect (arrow) in right lower lobe that was not initially reported on CT angiography of abdomen. CT of abdomen was performed 11 days before chest CT where diagnosis of PE was made. A Results Between September 1, 2011, and August 31, 2012, 358 reports of chest CT scans positive for PE were identified in 329 patients. Of these, 157 (47.7%) patients had also undergone 272 abdominal CT scans with contrast enhancement within 3 months. Of the 272 abdominal CT scans, 172 studies were performed on the same day of chest CT and were thus excluded. An additional 58 studies found no PE, and 14 studies were nondiagnostic because of poor enhancement of pulmonary arteries or respiration motion. Of the remaining 28 abdominal CT scans, four scans in four patients were reported as positive for PE. In 24 scans of 18 patients, PE was not described, and these studies were classi- B 740 AJR:202, April 2014

Pulmonary Emboli Missed on CT of the Abdomen fied as being potentially missed PE cases on abdominal CT. The presence of PE was confirmed by the three radiologists for all 24 scans of the 18 patients. Six of the 24 missed PE scans were excluded because they were repeat scans and only the abdominal CT scan closest to the time of diagnosis of PE on the chest CT was included for study. Eleven patients underwent the abdominal CT scan after the chest CT on which the diagnosis of PE was made (missed known PE) (Fig. 1), and seven of 18 patients underwent the abdominal CT scan before the index chest CT (missed unknown PE) (Figs. 2 and 3). Presentation of Patients With Missed PE Eighteen patients (12 men and six women; average age, 58.8 years; range, 28 92 years) were identified as having potentially missed PE on abdominal CT. The clinical indications for the abdominal CT scan were abdominal pain (n = 4), none of which was attributable to PE; follow-up of preexisting diseases, including sigmoid abscess, pneumatosis, and psoas abscess (n = 3); evaluation for bleeding foci (n = 3); evaluation for thromboembolic disease (n = 2); staging or restaging of known malignancy (n = 2); postoperative follow-up (n = 2); fever (n = 1); and further evaluation for the liver lesion on prior chest CT scan (n = 1). In addition, one patient underwent abdominal CT for lung cancer staging, and four other patients had a history of malignancy at the time of the abdominal CT scan. The four known malignancies were bladder cancer, breast cancer, appendiceal cancer, and melanoma. Two patients had a history of remote previous PE, and two other patients had a history of both PE and deep vein thrombosis. The d-dimer value was available for two patients within 3 weeks of their abdominal CT and was more than 3999 ng/ml (normal range, 0 390 ng/ml) in both patients. Clinical Features and Outcome of Missed PE The average interval between abdominal CT and chest CT was 13.4 days (range, 1 92 days; SD, 21.9 days). The average interval was significantly shorter for patients with missed unknown PEs (8.0 days; range, 1 24 days; SD, 7.8 days) compared with the missed known PE group (23.1 days; range, 1 92 days; SD, 31.0; p = 0.02). Among patients with missed known PE, seven of 11 patients (64%) were receiving treatment, including system anticoagulation, thrombolysis or thrombectomy, and inferior vena cava (IVC) filter placement, either alone or in combination. The remaining four patients did not undergo treatment; three were consid- TABLE 1: Patient Outcomes Outcome All Patients (n = 18) Pulmonary CTA First (n = 11) Abdominal CT First (n = 7) Untreated 4 (22) 4 (36) 0 (0) 0.07 Treatment initiated according to 14 (78) 7 (64) 7 (100) 0.07 pulmonary CTA Anticoagulation therapy 12 (67) 6 (55) 6 (86) 0.17 Thrombolysis or thrombectomy 3 (17) 2 (18) 1 (14) 0.83 Inferior vena cava filter 7 (39) 5 (45) 2 (29) 0.47 Outcome Discharged 15 (83) 8 (73) 7 (100) 0.13 Death resulting from pulmonary 0 (0) 0 (0) 0 (0) 1.00 embolism Death resulting from other causes 3 (17) 3 (27) 0 (0) 0.13 Note Data are no. (%) of patients. CTA = CT angiography. a Chi-square test. TABLE 2: Pulmonary Embolism (PE) Characteristics Characteristics No. of Patients No. of PEs present Single 13 Multiple 5 Lobar distribution Right middle lobe 3 Right lower lobe 7 Left lower lobe 3 Right middle and right lower lobes 1 Right lower and left lower lobes 4 Most proximal location Lobar 3 Segmental 11 Subsegmental 4 Onset Acute 17 Chronic 1 Occlusion Present 2 Absent 16 Motion Present 3 Absent 15 Pulmonary artery enhancement score 2 8 3 10 PE confidence score 1 0 2 4 3 14 p a AJR:202, April 2014 741

Lim et al. ered to be at high risk for bleeding (one with brain metastases) and one patient already had an IVC filter placed before the initial CT. Three (27%) patients died of causes unrelated to PE, and the remaining eight (73%) were discharged from the hospital. The initial abdominal CT report included a description of the lower lungs in 10 of these 11 patients. With respect to patients with missed unknown PE, all ultimately received treatment. Six of seven (86%) were treated with anticoagulation therapy, and the remaining patient underwent catheter-directed thrombectomy and IVC filter placement. There was an average delay of 5.2 days between the missed PE and initiation of therapy. In this small cohort, there was no evident cost of delay, in part because of the short follow-up time. All patients were ultimately discharged home or to a subacute facility. There were no deaths due to PE in either cohort. Table 1 provides more complete details. Imaging Findings Most of the missed emboli were single, right-sided, and nonocclusive and were deemed to be acute (Table 2). The most proximal PE was in a subsegmental, segmental, and lobar vessel in four (22.2%), 11 (61.1%), TABLE 3: Attenuation of Missed Pulmonary Embolism (PE) and Adjacent Pulmonary Artery Category Average Attenuation in HU (SD) Total Embolus 47.8 (28.4) Pulmonary artery 166.7 (79.2) Missed new PE Embolus 41.6 (31.5) Pulmonary artery 193.5 (104.3) Missed known PE Embolus 51.2 (27.9) Pulmonary artery 151.8 (63.7) TABLE 4: Other Abdominal CT Findings in the Chest Finding No. of Patients Atelectasis 14 Effusion 11 Consolidation 2 Ground-glass opacity 2 Dependent opacity 1 and three (16.7%) patients, respectively. The average attenuation for the pulmonary arteries in the CT studies was less than 200 HU for both acute and chronic PE (Table 3). Chest CT found that in seven (38.9%) of the patients, the PE was limited to the lower lung FOV of the abdominal CT, whereas in the remaining 11 patients, other areas outside the abdominal CT FOV were affected by PE. Other findings were common in the lower chest on abdominal CT, most frequently effusions and atelectasis (Table 4). Discussion Over a 1-year period, we found 18 examples of PEs that were not described on initial interpretation of abdominal CT but that were visible in retrospect. Most of the missed PEs were segmental, but three missed PEs occurred in lobar vessels. In nearly half the patients, the PE had not been diagnosed previously. In a slight majority of the cases, the reviewing radiologists judged the contrast bolus as good. To our knowledge, this is the first systematic report on missed PEs found on abdominal CT. In contrast to the absence of data for abdominal CT, missed PE on pulmonary CT angiography has been described in at least two studies [7, 8]. Wittenberg et al. [7] reviewed 292 patients who underwent pulmonary CTA, of which 68 (23.3%) examinations were initially interpreted as positive for PE. Using a prototype computer-aided detection algorithm, they found an additional seven patients whose PE had been overlooked at initial interpretation. The missed embolic disease was segmental in two cases and subsegmental in the remaining five cases. Kligerman et al. [8] retrospectively evaluated pulmonary CTA studies over several years and reported 53 overlooked PEs, of which 24 were segmental and 29 were subsegmental. Multiple PEs were missed in 23 of 53 (43.4%) instances. A PE computer-aided detection device was able to identify the overlooked PE in over 75% of cases. In the present study, lobar or segmental PE accounted for 13 cases and subsegmental PE accounted for five cases, for a lower proportion of subsegmental emboli. This difference is likely due to the better contrast enhancement obtained during pulmonary CTA, which permits more consistent visualization of subsegmental PE. Pitfalls during pulmonary CTA that may lead to overlooking PE have been well described [1]. Patient-related factors, such as respiratory motion artifact and image noise, and technical issues, such as improper window settings and streak artifact, are well-recognized obstacles. Anatomic and pathologic factors, such as partial-volume effects with lymph nodes and mucus plugging, are also important pitfalls. A satisfactory IV contrast bolus is also important. In a study by Jones and Wittram [9] that assessed reasons for indeterminate detection of PE, poor contrast enhancement was the second most common reason. The average attenuation in nondiagnostic studies was 245 HU, compared with 339 HU in a control group with adequate contrast. The average attenuation of the pulmonary artery adjacent to the PE on our abdominal CT scans was 167 HU. The clinical situation also plays a role in the likelihood of encountering PE and, therefore, the level of suspicion when interpreting a study. Certain patient groups, such as those with malignancy and certain genetic predispositions, are at higher risk. However, incidental PE is not rare on CT. In one comparative study, PE occurred in 11.8% of patients with high clinical suspicion, compared with 1.8% in those in whom it was not suspected [10]. The challenge in identifying PE is clearly greater on abdominal CT than on chest CT. In addition to the multiple pitfalls described already, the lungs are typically not a primary target of abdominal CT interpretation, and only a limited part of the pulmonary anatomy is included. In the present study, most patients had other indications and substantial other disease that may have created a satisfaction-of-search error and contributed to overlooked PE. Nevertheless, even though it is not a primary area of focus, important pathologic abnormalities can be identified in the lower chest on abdominal CT. Pulmonary nodules identified in the lower chest are a well-defined phenomenon and a source of considerable discussion in the literature [11, 12]. Our study suggests that the search for pulmonary embolism should be given similar weight during interpretation of enhanced abdominal CT. The clinical implications of identifying PE on abdominal CT also merit consideration. This can be considered in the two scenarios that pertain to our study. In patients with missed known PE on abdominal CT, the clinical implications are likely to be limited because most patients will already be undergoing treatment. Nevertheless, in patients with a smaller embolic load, particularly those with subsegmental PEs, there is not universal agreement on the need to treat [13]. Attention to a 742 AJR:202, April 2014

Pulmonary Emboli Missed on CT of the Abdomen subsequent enhanced abdominal CT is advisable because of the small risk that PE may not resolve and ultimately progress to chronic PE. In contrast, patients with missed unknown PE on abdominal CT are at greater risk of a negative outcome. Although such an adverse outcome was not seen in our small cohort, a recent larger study of missed PE on chest CT found progression of PE in 14 of 26 patients (53.8%) for whom clinical follow-up was available [8]. This study has several limitations. The study design is retrospective and only 1 year s worth of data were gathered. Thus, the study cohort is rather small. We cannot calculate the prevalence of missed PE on abdominal CT because we identified our patients according to the occurrence of positive pulmonary CTA studies. Other abdominal CT examinations may have had overlooked PE but those examinations would not have been detected using our method of selection; thus, our study presumably underestimates the frequency of such missed PEs. A second drawback is that we cannot determine the full clinical implications of missed PE on abdominal CT because all were either known before the index study or were detected soon after and treated. A final limitation is that we were not able to convincingly determine the reasons for failure to detect PE on abdominal CT because of the relative scarcity of PE detected on abdominal CT. In fact, in only four abdominal CT examinations was PE detected in the absence of a concurrent chest CT, meaning that PE was substantially more likely to be overlooked than diagnosed in our study group. In reviewing these four cases, the only notable characteristic is that they extended quite cranially to include portions of the central pulmonary arteries, where PE is more easily diagnosed. In summary, our study shows that overlooked PE can occur on abdominal CT, even in sizeable vessels. We view it as advisable that the pulmonary arteries on enhanced abdominal CT examinations be scrutinized for PE. References 1. Wittram C, Maher MM, Yoo AJ, et al. CT angiography of pulmonary embolism: diagnostic criteria and causes of misdiagnosis. RadioGraphics 2004; 24:1219 1238 2. Estrada-Y-Martin RM, Oldham SA. CTPA as the gold standard for the diagnosis of pulmonary embolism. Int J Comput Assist Radiol Surg 2011; 6:557 563 3. Calder KK, Herbert M, Henderson SO. The mortality of untreated pulmonary embolism in emergency department patients. Ann Emerg Med 2005; 45:302 310 4. Henzler T, Barraza JM Jr, Nance JW Jr, et al. CT imaging of acute pulmonary embolism. J Cardiovasc Comput Tomogr 2011; 5:3 11 5. Browne AM, Cronin CG, English C, NiMhuircheartaigh J, Murphy JM, Bruzzi JF. Unsuspected pulmonary emboli in oncology patients undergoing routine computed tomography imaging. J Thorac Oncol 2010; 5:798 803 6. Wittram C, Maher MM, Halpern EF, Shepard JA. Attenuation of acute and chronic pulmonary emboli. Radiology 2005; 235:1050 1054 7. Wittenberg R, Peters JF, Sonnemans JJ, Prokop M, Schaefer-Prokop CM. Computer-assisted detection of pulmonary embolism: evaluation of pulmonary CT angiograms performed in an oncall setting. Eur Radiol 2010; 20:801 806 8. Kligerman SJ, Lahiji K, Galvin JR, Stokum C, White CS. Missed pulmonary emboli on CT angiography: assessment with pulmonary embolism computer-aided detection. AJR 2014; 202:65 73 9. Jones SE, Wittram C. The indeterminate CT pulmonary angiogram: imaging characteristics and patient clinical outcome. Radiology 2005; 237:329 337 10. Hui GC, Legasto A, Wittram C. The prevalence of symptomatic and coincidental pulmonary embolism on computed tomography. J Comput Assist Tomogr 2008; 32:783 787 11. Alpert JB, Fantauzzi JP, Melamud K, Greenwood H, Naidich DP, Ko JP. Clinical significance of lung nodules reported on abdominal CT. AJR 2012; 198:793 799 12. Wu CC, Cronin CG, Chu JT, et al. Incidental pulmonary nodules detected on abdominal computed tomography. J Comput Assist Tomogr 2012; 36:641 645 13. Carrier M, Kimpton M, Le Gal G, et al. The management of a sub-segmental pulmonary embolism: a cross-sectional survey of Canadian thrombosis physicians. J Thromb Haemost 2011; 9:1412 1415 AJR:202, April 2014 743