Pulmonary embolism (PE) is a common cause of death in patients with cancer.1 In recent years, the extensive use of scheduled CT for tumor assessment has led to increasing diagnoses of incidental PE in this population. Thus, the current clinical spectrum of PE may vary from sudden life-threatening events to unsuspected radiologic findings on scheduled CT scans.2,3 In several series of patients with acute symptomatic PE, cancer has been identified as a predictor for adverse outcomes.4,5 Several factors, such as the greater risk of recurrence and bleeding, may influence this poor prognosis compared with similar patients without cancer.6 Similar to the introduction of outpatient therapy for deep venous thrombosis (DVT), the latest evidence-based guidelines from the American College of Chest Physicians7 suggest that early discharge may be appropriate for selected patients with low-risk PE (evidence-based grade 2B). However, this recommendation is based on observational and retrospective studies8-10 and some randomized trials including only symptomatic events11,12 in which patients with cancer were underrepresented.
Several prognostic scales, such as the Pulmonary Embolism Severity Index (PESI),13 the Geneva Prognostic Score (GPS),14 a simplified PESI version,15 and others,16 have been validated to predict short-term PE-related mortality. However, these scales classify almost all patients with cancer as high-risk, limiting their discriminatory power in these patients. This fact has led to the search for specific prognostic scales, such as the POMPE-C17 and the model derived from the Registro Informatizado de la Enfermedad Tromboembólica Venosa (RIETE registry).18 However, these scales have not been validated for incidental PE, and their clinical utility in selecting patients suitable for home treatment has not been prospectively evaluated.
Therefore, it remains of great interest to identify patients with cancer and low-risk PE who could be candidates to receive ambulatory treatment, and to integrate this practice within the continuum of care in the cancer population. The goal of the present study was to describe the feasibility of ambulatory treatment of symptomatic or incidental PE in a prospective cohort of consecutive patients with cancer selected for outpatient management based on a pragmatic set of exclusion criteria. The authors also retrospectively compared the prognostic performance of the PESI, GPS, POMPE-C, and RIETE scales to predict mortality and identify patients who could be safely treated at home.
Patients and Methods
Patients and Setting
From May 2006 to December 2009, the authors conducted a prospective study in the Medical Oncology Department at the Hospital Clinic in Barcelona, a tertiary care hospital with a reference population of more than 500,000 inhabitants. The authors consecutively enrolled patients with cancer older than 18 years diagnosed with PE who were selected for outpatient treatment according to predefined exclusion criteria.
All PEs were confirmed using objective radiologic methods. Patients clinically suspected of having PE routinely underwent a CT scan with pulmonary angiography. Incidental or symptomatic emboli were defined as an intraluminal filling defect in lung arteries on at least 2 consecutive transverse images, confirmed by 2 senior radiologists. Three radiologic features were used to classify PEs: the number (single or multiple) and type of vessels involved (main and lobar arteries as central arteries vs segmental or subsegmental branches as peripheral arteries) and the extension (unilateral or bilateral lung involvement). In patients with renal failure or known hypersensitivity to contrast medium, a ventilation/perfusion pulmonary scintigraphy with technetium 99m was performed. Studies were interpreted according to the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED) study criteria, and only scans categorized as having a high probability were considered diagnostic of PE.19
Patients were eligible for inclusion if they (1) developed either a symptomatic or an incidental PE and (2) had active cancer or were receiving adjuvant chemotherapy. Patients were recruited both in the emergency department and in the cancer outpatient clinic. Incidental PE was detected on scheduled CT scans performed for cancer evaluation as part of the usual clinical practice. The diagnosis of incidental PE was reported to an on-call oncologist, who contacted the patient for a thorough clinical evaluation according to an established protocol.
Patients younger than 18 years, pregnant women, and patients with PEs detected during previous hospitalization were not eligible for inclusion. The study was approved by the local ethics committee and all patients provided written informed consent.
All patients underwent a detailed medical interview, assessment of vital signs (temperature, blood pressure, and heart and respiratory rates), an electrocardiogram, a chest radiograph, and pulse oximetry assessment as part of routine care. The authors designed standardized data collection sheets to collect the clinical information.
Anticoagulant therapy was prescribed by treating physicians according to international recommendations at the time of the study.20,21 Low-molecular-weight heparin (LMWH) for a minimum of 3 months was the standard treatment. All of the patients selected for home treatment were discharged in less than 12 hours after PE diagnosis. Patients were trained for the self-injection of LMWH by a specialist nurse and were given a number for 24-hour telephone support.
The set of exclusion criteria for home treatment included systolic blood pressure less than 100 mm Hg, arterial oxygen pressure less than 60 mm Hg or pulse-oximetry less than 90%, active bleeding, platelet count of 50,000/mm3 or less, renal failure, lack of social support, poor treatment compliance, or the presence of other admission criteria according to treating physicians.
Outcomes and Follow-up
All of the patients were followed up in the outpatient clinic. The database was updated during scheduled medical visits, hospital admissions, electronic health record input, and/or telephone calls if necessary.
The primary outcomes were all-cause and PE-specific mortality at 30 and 90 days after PE diagnosis. The mortality attributable to PE was classified into 2 categories: (1) PE-caused death, which involved a clear causal relationship, and (2) PE-related death, which was considered when other complications (eg, cancer progression or infection) may have contributed to the outcome.
Secondary outcomes were the development of thromboembolic recurrence and major bleeding events at the same time points. Thromboembolic recurrence was defined either as a second venous thrombotic event or a clinically relevant progression of the previous embolism during anticoagulation. A bleeding event was classified as major if it was associated with death, occurred at a critical site (intracranial, intraspinal, intraocular, retroperitoneal, or pericardial), required a blood transfusion, or resulted in a reduction of hemoglobin of at least 20 g/L.22 An independent adjudication committee consisting of 2 physicians evaluated all possible end points (ie, recurrent venous thromboembolism, major bleeding, or death). Any dispute was resolved by a third opinion.
Assessment of Prognostic Scales for Acute PE
The performance of the PESI, GPS, POMPE-C, and RIETE models was compared to predict the 30- and 90-day risk of mortality. The PESI score uses 11 factors to classify patients with PE into 5 classes.13 Patients in risk classes I-II (score ≤85) have a low risk of mortality. The GPS is based on 6 variables to predict mortality, rethrombosis, and major bleedings.14 Patients with a GPS score less than 3 are classified as low risk. POMPE-C and RIETE were specifically developed for patients with active cancer and PE. The POMPE-C scale included 8 risk factors to determine the risk of mortality in the next 30 days. A cutoff lower than 10% is used to categorize patients as low-risk.17 The RIETE scale comprises 6 clinical variables. A cutoff score of less than 5 was used to identify patients with a low or intermediate 30-day mortality.18 The performance of the exclusion criteria for outpatient treatment of PE in patients with cancer was also evaluated.
Standard statistics were used to describe the sample. For the univariate analysis, both the χ2 and the Fisher exact tests were used to identify the potential risk factors for death. The Mantel-Haenszel test for linear association was applied with ordered categorical variables. The Student t test was used for continuous variables. All tests were 2-sided and P values less than 0.05 were considered significant.
To predict 30-day mortality, the predictive values of the PESI, GPS, POMPE-C, RIETE, and the authors’ set of exclusion criteria were calculated, namely sensitivity; specificity; positive and negative predictive values (PPV and NPV, respectively); and positive and negative likelihood ratios (pLR and nLR, respectively). The receiver operating characteristic (ROC) curves were calculated, and the method proposed by DeLong et al23 was used to test for statistically significant differences in area under the curves (AUCs). Univariate and stepwise Cox regression analyses were performed to evaluate the effects of the baseline traits, radiologic findings, and prognostic scales. Candidate covariates were the clinical context of PE detection (incidental vs symptomatic), the exclusion criteria, the presence of PE-related symptoms, pulse rate greater than 110 beats per minute, systolic arterial hypotension, hypoxemia, enlargement of the right ventricle, and RIETE and POMPE-C scores. Relative hazard ratios and corresponding 95% confidence intervals were reported. Statistical analyses were performed using the SPSS 17 software (SPSS Inc., Chicago, IL, USA).
Patients and PE Characteristics at Recruitment
The authors recruited 138 consecutive patients with cancer and PE (81 men and 57 women) with a mean age of 63 ± 11 years. All of the eligible patients were included in this study. A total of 62 patients (45%) were managed as outpatients, whereas 76 (55%) were admitted to the hospital. Both groups were balanced with respect to age, chronic comorbidities, ECOG performance status, type of tumor, and disease stage. Table 1 shows the baseline characteristics of the patients according to the site of PE treatment (home vs hospital). Patients with good performance status (ECOG 0-2) and advanced tumors, and who were being treated with palliative chemotherapy constituted most of the sample.
The reasons for hospital admission were hypoxemia (74%), hemodynamic instability (9%), other cancer-related complications (11%), insufficient social support (8%), workup for the initial diagnosis of cancer (5%), and low platelet count (3%). The mean time of hospitalization was 10 ± 8 days (range, 1-42 days). Admission to the intensive care unit was required for 4 patients (5%), 2 of whom (3%) received initial thrombolytic therapy. The flowchart of patients is represented in Figure 1.
Baseline Characteristics of Patients With Cancer and Pulmonary Embolism According to Treatment Location
Table 2 summarizes the main clinical and radiologic findings associated with the PE and the scores obtained from the prognostic scales at recruitment. Most of the patients treated at home had an incidentally detected PE (89%), with no PE symptoms at all (77%). Notably, 10 of 66 patients (15%) in whom the PE was diagnosed during scheduled CT scans revealed mild PE-specific symptoms during a detailed assessment. Thus, incidental and asymptomatic PE did not coincide in all cases. With regard to the radiologic burden of the PEs, patients admitted to the hospital more frequently had (1) bilateral lung involvement (65% vs 44%; P=.014); (2) associated DVT (33% vs 18%; P=.044); and (3) right ventricular dilatation on CT scans (9% vs 0%; P=.014).
Patients admitted to the hospital had higher scores on all the prognostic scales (PESI, GPS, POMPE-C, and RIETE) than those treated as outpatients. However, a large proportion of patients treated at home showed high-risk scores, especially on the POMPE-C and PESI scales (83.9% and 83.0%, respectively), whereas this proportion was lower on the GPS (22.6%) and RIETE (29.0%) scales.
Mortality rates were higher for patients admitted to the hospital compared with outpatients at 30 days (18% vs 3%; P=.006) and 90 days (34% vs 10%; P=.001). Detailed outcome variables are summarized in Table 3. Notably, the 2 patients who died at home were being treated by palliative care teams in the context of cancer progression. None of the patients selected for home treatment required further admission because of PE complications. No differences were seen either in the development of recurrent venous thrombosis or in major bleeding events, according to the place of treatment.
The exclusion criteria yielded a higher PPV (17.8%) and pLR (1.68) than the other prognostic models (Table 4). More interestingly, the results showed a significant decrease in the likelihood of death (nLR, 0.25 and NPV, 96.8%), which is remarkable considering that only 45% of the patients were classified as low risk. Moreover, the Cox proportional hazards model showed that only the exclusion criteria (hazard ratio [HR], 3.7; P=.04) and arterial hypotension (HR, 4.8; P=.006) were independently associated with the risk of 90-day overall mortality.
Table 4 summarizes the comparison of the accuracy and discriminatory power of the 4 prognostic scales to predict all-cause death at 30 days of follow-up. Most patients were classified as high risk using the PESI and POMPE-C scales (92% and 89%, respectively), with lower rates obtained with the GPS and RIETE scales (45% and 42%, respectively).
Notably, when these scales were dichotomized according to conventional cutoffs (low vs high risk), none was significantly associated with the 30-day mortality. However, both the scenario of detection (incidental vs symptomatic) and the set of exclusion criteria showed a clear prognostic value. Improved survival outcomes were observed in incidentally detected PEs compared with symptomatic events (mortality rates, 3.2% vs 18.4%, respectively; P=.006). POMPE-C apparently showed the best results for identifying a low-risk group: high sensitivity, NPV of 100%, and nLR of 0, but this scale classified 123 of 138 patients (89.1%) as high risk and the accuracy was low (21.0%).
The AUCs for the exclusion criteria were 0.679 (95% CI, 0.595-0.756) for POMPE-C; 0.621 (95% CI, 0.534-0.702) for RIETE; 0.650 (95% CI, 0.565-0.730) for PESI; 0.641 (95% CI, 0.555-0.721) for GPS; and 0.683 (95% CI, 0.599-0.756) for the exclusion criteria. Pairwise comparison of ROC curves was performed but no differences were significant (Figure 2).
This study showed that home treatment of patients with cancer and low-risk PE seems to be safe and feasible, with candidates being considered using simple criteria easily obtained at the bedside. The exclusion criteria for outpatient treatment were similar to those of previous cancer-specific series by Siragusa et al24 and Ageno et al,25 including DVT and PE, and recent studies specifically involving PE.8-12,26 Although many prognostic scales have been validated to predict short-term mortality in patients with symptomatic PE, the authors decided to rely on a pragmatic set of exclusion criteria because of the heterogeneity and complexity of the management of patients with cancer, in which several clinical and practical constraints must be taken into account for individualized decision-making.
Clinical, Radiologic, and Prognostic Characteristics of Pulmonary Embolisms According to Treatment Location
The study population was mainly composed of patients with a good performance status undergoing active anticancer treatments for advanced malignancies who were eligible to be managed as outpatients. The distribution of tumors and baseline traits in this series is similar to that reported by other investigators,27,28 but notably, incidental PE accounted for almost half of all events. Although this proportion is representative of the authors’ clinical practice and is in agreement with other recent studies,29-32 it should be taken into account when comparing reports exclusively including patients with symptomatic PE.4,5,8-18 In contrast to previous reports,33,34 incidental PE was associated with a lower mortality rate than nonincidental events in the present series. As a result, the 3-month mortality rate was 23% in this study compared with the expected 15% in noncancer studies,2 but it was lower than the mortality rate in cancer-specific series involving symptomatic events.24,25
Considering the authors’ sample, it is noteworthy that home and hospital groups did not differ in distribution of tumors, disease stage, therapy, and baseline comorbidities. In addition, the burden of PE was roughly similar, although patients with bilateral involvement and right ventricular enlargement were more likely to be hospitalized. All of these facts suggest that clinical features are overwhelmingly more important than radiologic characteristics in decisions regarding the optimal place of treatment.
The short-term PE-specific mortality among outpatients was zero in this cohort, highlighting the safety of the exclusion criteria. Moreover, 2 patients died within 1 month, although this outcome occurred in the setting of progressive cancer and was not directly attributable to PE. Furthermore, other medium-term complications were manageable with the support of the outpatient clinics and home palliative care facilities.
Although outpatient management of low-risk PE is increasingly implemented, evidence in cancer populations remains scarce. Notably, most of the deaths in this cohort (20/32) occurred concurrently with other factors (mainly cancer progression) that potentially influenced the outcome, whereas pure PE-caused deaths were scarce. However, because of the absence of autopsies, it is difficult to be certain of the exact causes of death. Data on outpatient management from non-disease-specific studies show that most PE-related deaths occur in patients with cancer,9,10,26 but whether the excess of mortality is attributable to the increased severity of the thrombotic event itself or to other cancer-related complications remains unclear. Unfortunately, the randomized trials available that could have addressed this question included few patients with cancer or directly excluded these patients.11,12
Outcomes According to Treatment Location
If ambulatory care is proven feasible, identifying low-risk groups that could be suitable for this strategy remains a priority. None of the available prognostic models developed for symptomatic patients with PE (PESI, GPS, POMPE-C, and RIETE) have been tested in mixed populations of patients with cancer. Predictive scales have recently been generated for patients with cancer (POMPE-C and RIETE), but their usefulness has not been validated externally. The results show that POMPE-C classifies more efficiently than other scales, although the fact that roughly 90% of patients were classified as high risk raises questions about its practical utility to select potential outpatients. Moreover, the exclusion criteria only ruled out half of the cohort for home treatment, but the accuracy was not significantly inferior. No significant differences were found among the prognostic scales as quantitative decision tools for stratifying patients with PE at risk of death.
Prognostic Values of PESI, POMPE-C, Modified GPS, and RIETE Scales and Exclusion Criteria for 30-Day Mortality
One must also consider that the probability of death might be uncoupled with the suitability of ambulatory management, which depends on many other clinical, social, or logistic constraints. Therefore, future research should focus on the development of more pragmatic tools that have a genuine ability to modify clinical decisions.
In addition to the limitation of being a single-center experience, the criteria for home treatment were subject to the experience of the clinicians. Therefore, the results might not be as reproducible as those obtained by objective scales. However, similar sets of exclusion criteria have already proven successful in selecting low-risk patients with PE for outpatient treatment.9,10,26 In addition, the relatively small sample size does not allow the assessment of specific prognostic factors (ie, subsets of incidental PE).
This study provides further evidence supporting the feasibility of home management of PE in selected low-risk patients with cancer. Prospective studies designed to validate cancer-specific scales would avoid unnecessary hospital admissions and contribute to the optimal management of PE in this setting.
The authors have disclosed that they have no financial interests, arrangements, affiliations, or commercial interests with the manufacturers of any products discussed in this article or their competitors. These findings were partially reported at the 6th International Conference on Thrombosis and Hemostasis: Issues in Cancer (ICTHIC) in Bergamo, Italy in April 2012 and at the 2012 International Symposium of the Multinational Association of Supportive Care in Cancer (MASCC/ISOO) in New York City in June 2012 (abstract 55).
HuttenBPrinsMGentM. Incidence of recurrent thromboembolic and bleeding complications among patients with venous thromboembolism in relation to both malignancy and achieved international normalized ratio: a retrospective analysis. J Clin Oncol2000;18:3078–3083.