Casey O’Connell, Jane Anne Nohl Division of Hematology, Norris Comprehensive Cancer Center, 1441 Eastlake Avenue, Los Angeles, CA 90033, USA. Tel.: +1 323 865 3950; fax: +1 323 865 0060. E-mail: firstname.lastname@example.org
Background: While symptomatic venous thromboembolism adversely impacts survival among cancer patients, the outcome of cancer patients with unsuspected pulmonary embolism (UPE) found on routine cancer staging multi-row detector computed tomography (MDCT) scans is unknown. Objective: To determine whether UPE detected on routine staging MDCT scans impacts overall survival among cancer patients. Patients and methods: We performed a matched cohort study of cancer patients diagnosed with UPE on routine staging scans between May 2003 and August 2006. Two controls (n = 137) were individually matched by age (± 5 years), cancer type and stage for each UPE patient (n = 70). We used Cox’s proportional hazard models to compare the mortality between UPE patients and their matched controls. Results: The hazard ratio (HR) for death among UPE patients was 1.51 (95% CI 1.01–2.27, P = 0.048). Compared with their matched controls, patients with UPE more proximal than the subsegmental arterial branches had a HR for death at 6 months of 2.28 (95% CI 1.20–4.33, P = 0.011) and an overall HR of 1.70 (95% CI 1.06–2.74, P = 0.027). Survival among UPE patients with isolated subsegmental PE (ISSPE) was not significantly different than that of matched controls (HR 1.04 95% CI 0.44–2.39, P = 0.92). Conclusions: UPE identified more proximal than the subsegmental arterial branches has a significant negative impact on survival among cancer patients.
Multi-row detector computed tomography (MDCT) scanning has significantly enhanced radiographic visualization of the peripheral pulmonary vasculature . This has increased the detection of unsuspected pulmonary embolism (UPE) on lung MDCT performed for other diagnostic purposes [2–4]. Cancer patients frequently undergo serial CT scanning to assess for treatment response and for recurrent disease, and MDCT has replaced single-row detector CT scans in most institutions. Although one study suggests that the prevalence of UPE may not be higher in cancer patients than in non-cancer patients undergoing MDCT , the absolute numbers are higher, accounting for the finding that about two-thirds of UPE occur in patients with cancer [4,5]. Whereas the significance of and treatment algorithm for symptomatic venous thromboembolism (VTE) are well established in cancer patients [6,7], the clinical relevance and long-term outcome of UPE are not, so therapeutic decision-making remains a vexing problem for clinicians .
We previously reported the results of a comprehensive chart review of cancer patients with UPE in which 52% of affected patients had major thrombi in the main or lobar pulmonary arteries . In order to further evaluate the impact of UPE on outcomes among cancer patients, we expanded on our previous study to include a larger number of patients with additional data on treatment, follow-up MDCT scans and survival.
We designed a retrospective matched cohort study to assess the impact of UPE on survival among cancer patients undergoing routine MDCT scan staging. This study was approved by the Institutional Review Board of the University of Southern California.
Ninety-six consecutive cases of UPE were prospectively identified among all patients who underwent MDCT scanning for routine staging between May 2003 and August 2006 by the Department of Radiology at the Norris Comprehensive Cancer Center, a tertiary referral center in Los Angeles, CA, USA. Of these, 26 were excluded from further analysis for the following reasons: VTE diagnosis within 1 year before the index MDCT scan (n = 15), already on anticoagulation for atrial fibrillation (n = 1), incomplete medical records (n = 3), UPE picked up on abdominal CT without complete chest CT (n = 1) or no available control patient with the same type and stage of tumor (n = 6).
The files of the Department of Radiology were used to identify 465 cancer patients who had staging MDCT scans without findings of UPE between May 2003 and August 2006. For each UPE patient, two controls were individually matched by age (± 5 years), cancer type and stage at the time of the MDCT scan. Matched controls were selected from patients whose staging MDCT scan occurred within 1 year of the case patients’ UPE. Potential control patients were excluded if they had a VTE diagnosis within 1 year of the index MDCT scan, if they were on anticoagulation or if their medical records were incomplete. In all, 137 control patients were identified for the 70 patients with unsuspected PE. For three UPE patients (one with stage IV bladder cancer, one with stage IV pancreatic and one with stage IV melanoma), only one matched control patient could be identified. For 11 UPE patients, the controls exceeded the 5-year age range; in eight of these sets, one control was more than 5 years older (four 5–10 years older and four > 10 years older) and in three sets, the controls were more than 5 years younger than the cases (three 5–10 years younger and three > 10 years younger). We were not able to identify a stage-matched control for five of the cases, so controls with the same tumor but a more advanced stage were selected for those cases (one stage II breast cancer matched to two stage IV breast cancer controls; one stage I bladder cancer matched to one stage I and one stage III control; one stage II mesothelioma matched to two stage III mesothelioma controls; one stage III liver cancer matched to two stage IV controls; and one stage III colorectal cancer matched to one stage III and one stage IV control). Excluding these sets from the analysis did not change our results.
Medical records were reviewed and data were extracted by seven of the authors (M.G., C.O., S.V., S.B., A.C., L.M., K.G.) using structured data collection forms. Patient demographics, tumor characteristics, sites of metastases, treatment information, and signs and symptoms were extracted from nursing and physician notes and from medication administration records. Stage was defined as the TNM staging that best described the extent of disease at the time of the index MDCT scan. VTE symptoms including fatigue, cough, dyspnea on exertion, shortness of breath, chest pain, limb pain or limb swelling were obtained from clinician notes recorded within 2 weeks before the index MDCT scan. Signs of PE included tachycardia, defined as heart rate more than 100 beats per minute on two consecutive visits within 4 weeks before the index MDCT scan, limb tenderness and limb swelling. A complete blood count was recorded from laboratories drawn within 2 weeks before the day of the staging MDCT scan. Recorded VTE risk factors include: any deep vein thrombosis (DVT) or PE occurring more than 1 year before the index MDCT scan, major surgery within 2 months of the index MDCT scan, immobilization defined as confinement to bed or chair for more than 75% of the time for at least 2 weeks before the index MDCT, or an indwelling central line noted on the index MDCT scan. Chemotherapy was recorded as administered within 30 days, between 30 and 90 days, or more than 90 days from the index MDCT scan. The use of erythropoietin and hormonal agents was recorded if given within 1 month of the index MDCT. We recorded data regarding treatment of the UPE, including the use of inferior vena caval filters and low-molecular weight heparins (LMWH), heparin and warfarin. For UPE patients treated with anticoagulation, the type, dose (prophylactic or treatment) and duration was recorded. Survival information was obtained from the medical records or from public vital statistics databases and included date last seen, date of death and cause of death when available. Survival length was measured in months from the index MDCT scan to death or most recent active follow-up appointment or physician contact. Follow-up continued until death or an arbitrary cutoff date of 1 September 2008.
A 16-slice MDCT (Philips Brilliance) scanner was used throughout the study period. Our staging protocol utilizes 120–150 cc of Isovue 370 contrast administered intravenously at an injection rate of 2.0–2.5 cc s−1. Chest scanning is performed in 2 mm slices over 40–50 s. PE were defined as intraluminal filling defects located in the right or left main pulmonary artery, lobar arteries, segmental arteries and/or subsegmental arteries not associated with parenchymal abnormalities which could represent tumor-related obstruction. Chronic PE characterized by intravascular web-like structures or by partial luminal patency without widening of the affected arterial segment was not included. The first follow-up staging MDCT scan and the last recorded staging MDCT were assessed for resolution of the original PE and/or development of new PE.
For the analysis, proximal PE included all clots identified in the main, lobar or segmental arteries. Distal clots involving one or more subsegmental branches of the pulmonary arterial tree were categorized as isolated subsegmental PE (ISSPE). All UPE cases were independently confirmed by independent but non-blinded review by one of the authors (W.B.). A second independent blinded review was performed on the cases with ISSPE and their controls by a second author (V.D.).
The main purpose of the analysis was to evaluate the effect of being diagnosed with UPE on the survival of the cancer patients. We used conditional Cox proportional hazard models stratified by the matching variables to estimate the hazard ratios (HR) and 95% confidence intervals (CI) in order to compare the all-cause mortality between patients diagnosed with UPE and their matched pairs. We used this method and accounted for individual matching in our analysis because the matching factors were strongly associated with the survival and the censoring pattern of the subjects. We further adjusted the models for gender and metastasis location at the time of the MDCT scan (brain, liver and lung). Further adjustment for other possible confounders including history of recent surgery, history of VTE, time between diagnosis and MDCT scan, and levels of hemoglobin, platelet, and leukocytes at the time of MDCT scan did not change our findings and thus the results are not presented in the present study. We tested the assumption of proportional hazards for the main effect and each of the adjusting variables using a likelihood ratio test of interaction with the time-scale. We observed no departures from proportional hazards assumption. We also performed a sensitivity analysis using the same models further adjusted for time of index MDCT to compare survival between cases and controls from the time of their malignancy diagnosis. To graphically present the adjusted survival curves and to generate survival probability estimates for the stratified Cox proportional hazard models, we used the methods described by Zhang et al. .
We used conditional logistic regression models to examine the association between the baseline characteristics of patients, signs and symptoms, potential VTE risk factors and risk of being diagnosed with UPE. For all analyzes, non-UPE patients were considered as the referent group. The α error was 0.05 and the reported P-values are two-sided. Tests for trends and interactions were performed using likelihood ratio tests. SAS 9.2 (SAS Institute, Cary, NC, USA) was used to conduct all analyzes.
The demographic information of the patients included in this analysis is shown in Table 1. Colorectal and pancreatic cancers accounted for 49% of the cancers in the UPE cohort. Genitourinary malignancies including bladder, renal, prostate and testicular made up 17% of the cancers. Cancers affecting fewer than three UPE patients are included in the ‘other’ category. The time between the malignancy diagnosis and the date of the index MDCT scan was slightly shorter among the UPE cases compared with their matched controls (median 7 vs. 13 months). However, this difference was not statistically significant (P = 0.11).
Table 1. Baseline characteristics of UPE patients and matched controls without UPE
UPE (n = 70)
No UPE (n = 137)
UPE, unsuspected pulmonary embolism; HGB, hemoglobin; WBC, white blood cell count; PLT, platelet count. *P-value based on conditional logistic regression models matched on age (± 5 year), cancer type and cancer stage. †Matching factor. ‡Months between diagnosis of malignancy and the index MDCT scan.
Age, median (25th, 75th percentile)
64.5 (56, 72)
65 (56, 71)
Hematologic data, median (25th, 75th percentile)
HGB (g dL−1)
11.7 (10.8, 13.2)
12.3 (10.9, 13.3)
WBC (× 109 L−1)
7.3 (4.6, 9.7)
6.0 (4.6, 8.3)
PLT (× 109 L−1)
240 (181, 348)
240 (165, 285)
Cancer diagnosis year
Index CT year
Diagnosis – MDCT scan interval‡, Median (25th, 75th percentile)
7 (4, 20)
13 (5, 25)
PE characteristics and treatment
Among 70 patients with UPE, 53 had proximal PE (7 main pulmonary artery emboli, 26 lobar emboli and 20 segmental emboli) and 17 had ISSPE. Fifty-nine (84%) patients, 46 with proximal PE and 13 with ISSPE, received some form of anticoagulation: 35 with LMWH alone (3 received only prophylactic dosing and 32 received full treatment doses) and 24 with LMWH or unfractionated heparin followed by warfarin. Eighteen of the warfarin-treated patients had interruptions in warfarin therapy and therefore received additional LMWH at some time during the course of anticoagulation. Nine patients with UPE had inferior vena cava (IVC) filters placed; of those, only two did not receive some form of anticoagulation. Of the 17 patients with ISSPE, 13 received some form of anticoagulation whereas four were untreated. Data regarding the duration of anticoagulation and reasons for interruption or discontinuation could not be reliably obtained for the majority of patients.
Risk factors, signs and symptoms
The distribution of VTE risk factors between UPE patients and controls is shown in Table 2. UPE patients did not differ significantly from matched controls in terms of the presence of lung, liver or brain metastases. The most common symptom among cases and controls was fatigue (44% vs. 21%, respectively). Among patients who disclosed signs or symptoms, shortness of breath, cough and fatigue were significantly more prevalent among the UPE cases than controls. We did not observe any statistically significant difference between cases and controls with respect to more specific signs and symptoms of PE, such as tachycardia or chest pain. Recent surgery and a past history of VTE were most closely associated with UPE (P-values < 0.001 and 0.007, respectively).
Table 2. Signs, symptoms and risk factors present at the time of index MDCT scan among 70 UPE patients and 137 matched controls
UPE, unsuspected pulmonary embolism; MDCT, multi-row detector computed tomography; VTE, venous thromboembolism. *P-values are based on conditional logistic regression models with subjects without the sign/symptom at the time of staging MDCT as the referent group. Each variable was fitted separately. †Major surgery within two months of the index MDCT scan. ‡Chemotherapy within 3 months of the index MDCT scan. §Includes aspirin or clopidogrel. ¶Confinement to bed or chair more than 75% of the time for at least 2 weeks before the index MDCT scan. **Use within one month of the index MDCT scan.
Signs and symptoms
Shortness of breath
Dyspnea on exertion
Of the 17 patients with unsuspected ISSPE, four had recent major surgery and two additional patients had a past history of VTE. Eleven patients were receiving chemotherapy when the ISSPE was detected, three were receiving erythropoietin and eight had central lines. Among the ISSPE patients, seven (41%) were asymptomatic whereas four had complained of fatigue without other signs or symptoms, four had tachycardia and two had respiratory symptoms (dyspnea on exertion and cough in one, dyspnea on exertion and hypoxia along with fatigue in one). Three of the patients with ISSPE had simultaneous DVT; but as with the cohort as a whole, the majority of the ISSPE group (11/17) did not have dedicated studies to detect DVT.
The HR for death among UPE patients as compared with matched controls was 1.51 (95% CI 1.01–2.27; median overall survival 8 vs. 12 months; Table 3, Fig. 1). The impact of UPE on overall survival was more pronounced among the 53 patients with proximal clots (HR 1.70, 95% CI 1.06–2.74; median overall survival 7 vs. 12 months; Table 3, Fig. 2A). In fact, the more proximal the UPE, the more significant was the adverse impact on survival (PTrend = 0.019; Fig. 2B). As a sensitivity analysis, we also compared the survival between cases and controls from the time of diagnosis of malignancy. The HR was 1.63 (95% CI 1.08–2.48, P = 0.019). We performed an additional analysis excluding those patients who received no anticoagulation for their UPE and found a HR of 1.72 (95% CI 1.03–2.87, P = 0.038) for the treated patients with proximal UPE and a HR of 1.09 (95% CI 0.43–2.75, P = 0.86) for the treated patients with ISSPE. Patients with proximal PE were more than twice as likely to have died by 6 months (HR 2.28, 95% CI 1.20–4.33; Table 3). Although the HR at 2 months was 2.03, the difference was not statistically significant (95% CI 0.87–4.70, P = 0.10). There was no statistically significant difference in survival between patients with ISSPE and matched control patients (HR 1.04, 95% CI 0.44–2.39, P = 0.92; Table 3, Fig. 2A). Four (23.5%) of these patients were not treated with anticoagulation and in all four, the first follow-up CT scan revealed resolution of the ISSPE. In one of these untreated patients, a new UPE developed on a subsequent scan.
Table 3. Survival analysis for 70 cancer patients with UPE and 137 matched controls without UPE
6 months survival
UPE, unsuspected pulmonary embolism; SE, standard error; HR, hazard ratio; 95% CI, 95% confidence interval. *Based on Cox’s proportional hazard models adjusted for metastases location (brain, lung, and liver) and gender; stratified by matching variables including: age (± 5 years), cancer type and cancer stage.
In this retrospective matched cohort study, identification of UPE was associated with shortened survival among cancer patients undergoing routine staging MDCT scans. The lack of impact on survival at 2 months with a significant and sustained impact starting at 6 months suggests that most mortality was not directly related to the acute thromboembolic event. There was, however, a significant trend towards the more proximal UPE having a greater impact on survival.
Moreover, cancer patients with unsuspected ISSPE did not have poorer outcomes than matched controls. In fact, the four patients with ISSPE who did not receive anticoagulation seemed to resolve these clots by the next staging MDCT scan. There is controversy as to whether ISSPE merit anticoagulation in the population at large. In an editorial review of the subject, Goodman  suggests that because of the intrinsic fibrinolytic activity present in the pulmonary vasculature, as opposed to the lower extremity deep venous system, small asymptomatic PE may not require anticoagulation in certain cases. In cancer patients, however, it is not known whether ISSPE represent an early manifestation of the hypercoagulable state that has been associated with greater metastatic  and prothrombotic potential . Furthermore, half (3/6) of the ISSPE patients who had a screening lower extremity ultrasound had DVT, raising the possibility that ISSPE are a manifestation of clot burden elsewhere. Although the present study corroborates the suggestion that ISSPE are not life-threatening, the majority of these ISSPE patients did receive anticoagulation so the effect of treatment cannot be discounted.
Previous studies suggest that the development of VTE is a poor prognostic marker among cancer patients  and biomarkers of thrombin generation, even in the absence of documented VTE, seem to correlate with more aggressive cancer biology and poorer outcomes [14–16]. The negative impact of UPE on the overall outcomes of affected cancer patients supports the hypothesis that the development of VTE may be a surrogate marker of aggressive cancer biology . Based on the present findings, UPE should be included as relevant adverse outcomes in studies which document the incidence of VTE in patients with malignancy.
The major limitation of the present study is the retrospective design. By choosing the control patients from the same diagnosis era, who underwent MDCT staging during the same period, with the same type and stage of cancer, and of approximately the same age and by including the matching in our analysis, we tried to mitigate the impact of these factors on the survival data. We did not include performance status or cancer treatment among the covariates we accounted for in our analysis of survival, which could have affected the results. We assumed, given the patients were treated by oncologists in the same institution, an National Cancer Institute-designated comprehensive cancer center, during a relatively short study period, that patients were treated on standard chemotherapy protocols.
Another potential source of bias in the present study was the use of the date of the index MDCT scan, rather than the date of the cancer diagnosis, to calculate survival time for cases and controls. While there was no statistically significant difference between cases and controls in terms of the time from diagnosis of their malignancy to the date of the index MDCT scan, we did perform an additional sensitivity analysis using survival from date of cancer diagnosis as our outcome. The results of this analysis confirmed our main finding that being diagnosed with UPE was associated with significantly poorer survival.
Our study was not designed to determine if treating UPE will favorably impact survival, and prospective trials are needed to determine if this is true. We were unable to systematically evaluate the impact anticoagulation may have had on survival and this is an important source of potential bias. However, restricting our survival analysis to those patients who received some form of anticoagulation did not change our results. It was not our intent to describe the incidence of UPE among cancer patients, as this was previously attempted by other authors [3–5]. Neither can we presume to use these findings or those from any other single-institution study to identify which cancers are most likely to be associated with UPE, as tertiary referral cancer centers typically have skewed referral patterns based on their unique areas of expertise or the particular needs of the referral community.
Ours is the first study to demonstrate that UPE has major clinical relevance and an impact on survival. Our results strongly suggest that proximal UPE are associated with poor outcome in affected cancer patients, and thus support the inclusion of properly defined UPE as relevant adverse outcomes in studies which report or address cancer-related VTE. As 50% of the patients with unsuspected ISSPE who had lower extremity ultrasonography did actually have DVT, we feel such testing should be performed before deciding whether anticoagulation is indicated in this group.
C. O’Connell, H. Liebman and W. Boswell were responsible for the concept and design of the study. C. O’Connell, M. Ghalichi, S. Boyle, S. Vasan, L. Mark A. Caton and K. Grabow, were responsible for data collection. V. Duddalwar and W. Boswell were responsible for radiology review of all scans. P. Razavi was responsible for the statistical analysis, preparation of tables and creation of figures. C. O’Connell, P. Razavi and H. Liebman wrote the manuscript. All authors reviewed and made editorial revisions on the manuscript.
We would like to thank S. Groshen, for her helpful comments on the statistical analysis and D. Feinstein, for his editorial comments on the manuscript.
Disclosure of Conflict of Interests
H. Liebman has received research support from Celgene, Sanofi-Aventis, Glaxo Smith Kline, Esai, Pfizer, Bristol-Myers-Squibb, and the Max Gondon Family Foundation. H. Liebman has also received honoraria from Glaxo Smith Kline, Sanofi Aventis, Esai and Pfizer. All other authors have nothing to disclose.