Contemporary management of pulmonary embolism: the answers to ten questions


  • H. Bounameaux

    1. From the Division of Angiology and Hemostasis, Department of Internal Medicine, University Hospitals of Geneva and Faculty of Medicine, Geneva, Switzerland
    Search for more papers by this author

Prof Henri Bounameaux, Division of Angiology and Hemostasis, University Hospitals of Geneva, CH-1211 Geneva 14, Switzerland.
(fax: +41 22 3729299; e-mail:


Abstract.  Bounameaux H (Division of Angiology and Hemostasis, Department of Internal Medicine, University Hospitals of Geneva and Faculty of Medicine, Geneva, Switzerland). Contemporary management of pulmonary embolism: the answers to ten questions (Review). J Intern Med 2010; 268: 218–231.

Pulmonary embolism (PE) cannot be diagnosed solely on a clinical basis, because of the lack of sensitivity and specificity of clinical signs and symptoms. Pulmonary angiography is invasive and resource demanding. Because the prevalence of PE is relatively low (20% or less) amongst individuals who are clinically suspected of having the disease, submitting all of them to imaging (multi-detector CT angiography or ventilation/perfusion lung scintigraphy) would not be cost-effective. Therefore, diagnostic algorithms have been developed that include clinical probability assessment and D-dimer measurement to select the patients who require noninvasive imaging. Once the diagnosis is suspected or confirmed, therapy must be started to avoid potentially fatal recurrence. Treatment starts for an initial 3-month period with a 5-day course of parenteral unfractionated or low-molecular-weight heparin or fondaparinux overlapping with and followed by oral vitamin K antagonists monitored to maintain an international normalized ratio of 2–3. This initial period of 3 months may then be followed by a long-term secondary prevention period in patients who experience an idiopathic thromboembolic event and are at low risk of bleeding. New oral anticoagulants that do require patient monitoring and might exhibit a more favourable benefit–risk balance are currently under extensive clinical testing and might change the situation in the near future. A critical appraisal of the contemporary management of suspected PE is given in this overview with the discussion of 10 practical questions.


Pulmonary embolism (PE), the source of which is predominantly thrombosis of the deep veins (DVT) of the legs, is the third most common cause of mortality because of cardiovascular disease, after coronary artery disease and stroke. In addition, late sequelae of DVT may produce disabling leg symptoms in a substantial proportion of patients, including venous ulcers in a minority, resulting in a considerable economic burden. Furthermore, chronic thromboembolic pulmonary hypertension may develop as a late complication of PE in a small subset of patients.

During the past two decades, the approach to diagnosing DVT and PE has changed considerably. First, it has been accepted that these two conditions are different manifestations of a single entity that was known as venous thromboembolism (VTE). Second, novel noninvasive diagnostic tools such as venous compression ultrasonography (CUS), fibrin D-dimer (DD) measurement and multi-detector computed tomography angiography (MDCTA) have become available. These techniques have drastically reduced the need for invasive tools such as phlebography and pulmonary angiography. Third, the index of suspicion has progressively increased and resulted in at least 80% of patients with suspected VTE being referred to a diagnostic centre but found not to have the disease. Fourth, efficient mainly noninvasive strategies have been validated in large-scale outcome studies, and their costs have become an issue [1]. In parallel, the therapeutic management has also been modified; risk stratification has been recommended and new drugs as well as new regimens of old drugs have emerged that might change our practice in the near future.

The aim of the present overview is to discuss these diagnostic and therapeutic changes by answering 10 questions regarding nonmassive PE, which represents the vast majority (about 95%) of all PE events.

How should we manage a patient with clinically suspected PE?

Once the possibility of PE has been raised, the initial step of any diagnostic work-up consists of assessing the clinical probability of the patient having the disease. The aim of this step is to (i) identify the patients (probably those with a high and intermediate clinical probability) who require anticoagulant treatment whilst awaiting the results of diagnostic tests and (ii) select the patients (those with a nonhigh clinical probability) in whom the diagnosis of PE can be excluded based on a negative DD test result (Fig. 1).

Figure 1.

Multi-detector computed tomography angiography in a 32-year-old pregnant woman showing multiple bilateral pulmonary emboli, including (arrows) in the right pulmonary artery and in the left inferior lobar artery.

Pulmonary embolism cannot be diagnosed on a clinical basis alone, because of the lack of sensitivity and specificity of isolated clinical signs and symptoms [2]. Indeed, clinical symptoms, signs and abnormalities of blood gases, chest radiograph and electrocardiogram have a low predictive value for suspected PE when considered individually. Nevertheless, clinicians can combine these findings effectively either implicitly (or empirically) or by prediction rules (or scores) to classify patients according to their probability of having the disease; this is usually referred to as ‘clinical probability’.

Two scores have been proposed [3, 4] and externally validated [5] that are as, but not more, accurate than implicit judgment [6, 7]. Nevertheless, in a study of 110 patients who received duplicate assessments, Rodger et al. [8] showed that interobserver reliability was higher when using an explicit clinical model (kappa coefficient of agreement = 0.62 vs. 0.33 for implicit evaluation). In addition, explicit clinical models or scores are useful educational tools. All these means of assessing clinical likelihood of PE allow a fairly accurate classification of patients into three categories corresponding to a prevalence of the disease of 5–10% (low clinical probability), 20–30% (intermediate clinical probability) and 60–80% (high clinical probability). Most patients with suspected PE have a low or intermediate clinical probability of having the disease. Those with a low or intermediate probability of PE can usually be investigated by entirely noninvasive algorithms. However, these rules also have limitations. Computation of the original Geneva score [4] requires arterial blood gas values whilst breathing room air, a variable that was not available in 15% of the patients in the external validation sample [5]. The Canadian score (Table 1) includes the clinician’s judgment of whether an alternative diagnosis is more likely than that of PE [3]. This criterion has major weight in this score and can clearly not be standardized, thereby hampering its use by nonclinical health professionals.

Table 1. Description of the most widely used scores for clinical probability assessment in patients with clinically suspected pulmonary embolism (PE)
Original Geneva score [4]Wells’ score [3]Revised Geneva score [9]
Elements of the scorePointsElements of the scorePointsElements of the scorePoints
Previous PE or DVT+2Previous PE or DVT+1.5Age >65 years+1
Heart rate >100 min−1+1Heart rate >100 min−1+1.5Previous DVT or PE+3
Recent surgery+3Recent surgery or immobilization+1.5Surgery (under general anaesthesia) or fracture (of the lower limbs) within 1 month+2
Age 60–79 years+1Clinical signs of DVT+3Active malignancy (solid or haematological malignancy, currently active or considered as cured within <1 year)+2
Age ≥80 years+2Alternative diagnosis less likely than PE+3Unilateral lower limb pain+3
PaCO2 < 4.8 kPa (36 mmHg)+2Haemoptysis+1Haemoptysis+2
PaCO2 4.8–5.19 kPa (36–38.9 mmHg)+1Cancer+1Heart rate 75–94 bpm+3
PaO2 < 6.5 kPa (49 mmHg)+4  Heart rate ≥ 95 bpm+5
PaO2 6.5–7.99 kPa (49–60 mmHg)+3  Pain on lower-limb deep vein palpation and unilateral oedema+4
PaO2 8.0–9.49 kPa (61–71 mmHg)+2    
PaO2 9.5–10.99 kPa (71–82 mmHg)+1    
Elevated hemi-diaphragm+1    
Clinical probability
  1. DVT, deep vein thrombosis.

Initial rule
Dichotomized rule

To obviate the logistical and standardization problems of the two aforementioned scores, the Geneva group derived a new prediction rule from a large multicentre cohort of patients admitted to the emergency ward for clinically suspected PE [9]. The rule is entirely based on readily available clinical variables and is independent of physicians’ implicit judgment (Table 1). In an external validation set of the new score, the prevalence of PE was 8% in the low-probability category (0–3 points), 28% in the intermediate group (4–10 points) and 74% in the high-probability group (11 points or more) [9]. Although the clinical data in the validation sample were collected prospectively, this new score, the so-called revised Geneva score, was calculated retrospectively. To be considered fully validated, it needed be used prospectively in a formal outcome study with patient follow-up. This has been realized in two independent cohorts. First, Klok et al. [10] reported the same predictive accuracy in a series of 300 consecutive patients as in the derivation and initial retrospective validation samples. Moreover, the diagnostic accuracy was similar to that of the Wells’ score, which was also assessed in this series. A second prospective validation of this score was obtained in a cohort of 1819 patients with suspected PE [11], which again showed that the accuracy was similar to that obtained with the Wells’ rule.

In two recent outcome studies, the Wells’ score for PE has been used with a single cut-off of four points instead of two cut-offs [12, 13]. This resulted in a classification of patients into two clinical probability categories (PE unlikely or likely) instead of the traditional three levels of probability (low, intermediate or high). This would theoretically allow the use of a less-sensitive DD assay in a higher proportion of patients than when using the three-level probability scheme but is further from implicit clinical judgment. Nevertheless, the dichotomized score is also valid for discriminating between patients at low and higher clinical probability of PE. Further simplifications of both the Wells’ rule and the revised Geneva score, in which all items carry the same weight, have also been proposed and have undergone external validation [14–16].

Recently, the Geneva group performed a systematic review and meta-analysis of the performance of all available clinical prediction rules for suspected PE [17]. It was concluded that these rules have a similar accuracy. However, they are not identical, and the choice between the various prediction rules and classification schemes must be guided by the local prevalence of PE, the type of patients considered (outpatients or inpatients) and the type of DD assay used. When using a highly sensitive DD assay, a three-level classification scheme is preferred because it mimics more closely clinicians’ empirical reasoning and potentially rules out VTE in a higher proportion of patients. Less-sensitive DD assays should be used in combination with the Wells’ score with two levels of probability (VTE likely or unlikely) as these assays are not able to exclude VTE in the 40% of patients in the intermediate clinical probability category, and the VTE unlikely category includes a higher proportion of patients than the low-probability class. The performance of the Wells’ score for PE is sensitive to the overall prevalence of PE in the population of patients with suspected disease: in a series in which the prevalence of PE was higher than 20%, the prevalence of PE in the low and intermediate probability categories was significantly higher (16%, pooled average) than in series with a lower prevalence of the disease (3.4%). This is not a problem when using a highly sensitive DD assay but might be with less-sensitive assays. We therefore advise using the revised Geneva score in populations with a prevalence of PE above 20% as that rule has been derived and validated in such settings. Finally, the Wells’ score for PE (two- or three-level) is the only validated score in inpatients.

What is the place of DD measurement in the diagnostic work-up?

D-dimer is a degradation product of cross-linked fibrin, and its blood concentration is increased in patients with acute VTE. When assayed using a quantitative enzyme-linked immunosorbent assay (ELISA) or an automated turbidimetric assay, the DD level has been shown to be highly sensitive (more than 95%) in acute DVT or PE, usually with a cut-off value of 500 μg L−1. Hence, a DD level below this value reasonably excludes acute VTE, at least in patients with a low or intermediate clinical probability, as recently reviewed [18]. Data from two recent large-scale multicentre management studies have definitively established the safety of two DD assays. The Vidas D-dimer Exclusion® test (bioMérieux, Marcy-l'Etoile, France), an ELISA, has now been used in more than 5000 patients with suspected PE in clinical trials [11, 13, 19–22]. The disease was ruled out by a negative result in 34% patients who had a nonhigh clinical probability or were classified as ‘PE unlikely’ in a Dutch series [13]. The 3-month thromboembolic risk was only 0.1% (95% CI, 0–0.4), and this low rate was fully confirmed in a subsequent meta-analysis [23]. The Tinaquant® test (Roche Diagnostics, Mannheim, Germany), an immunoturbimetric assay, has also been validated in three outcome studies that included more than 2000 patients [13, 24, 25]. The result was negative ruling out PE in 41% of patients who had either a low clinical probability of PE or were classified as ‘PE unlikely’ according to the dichotomized Wells’ rule [15]. The 3-month thromboembolic risk was 0.6% (95% CI, 0.2–1.4). Finally, the SimpliRed® assay (Agen Biomedical Inc., Brisbane, Australia) has also been well validated [26, 27] but interobserver variability may be a problem [27], and exclusion of the disease is only possible in patients with low clinical probability.

Conversely, patients with a high clinical probability should not be tested for DD because a concentration below the diagnostic cut-off is rare in this category of patients, and above all because the upper limit of the 95% confidence interval for the 3-month occurrence of PE remains above 3% despite a negative result.

Although DD is very specific for fibrin, the specificity of fibrin for VTE is poor. Indeed, fibrin is produced in a wide variety of conditions such as cancer, inflammation, infection and necrosis. Hence, a DD concentration above 500 μg L−1 (the diagnostic cut-off for many assays) has a poor positive predictive value for VTE and cannot reliably rule in the disease. Nevertheless, based on the results of a retrospective analysis of a large patient sample, Bosson et al. [28] suggested that a DD level above 2000 μg L−1 was predictive of the presence of PE, independently of the clinical score with an odds ratio of 6.9. They noted, however, that this result requires clinical validation in a prospective study, particularly because the PE prevalence in their series was high: 27%, 58% and 85% in the low, intermediate and high clinical probability categories, respectively. We have previously reported that patients with an intermediate or high clinical suspicion of PE and a DD level above 4000 μg L−1 had a positive likelihood ratio of 5 of having the disease, but few clinicians would accept the diagnosis just on this basis [29].

What is the contemporary reference imaging modality for PE?

During the past few years, MDCTA has largely replaced ventilation/perfusion (V/Q) lung scintigraphy as the main imaging modality in suspected PE. The rapid acquisition of high-contrast images by MDCTA allows an adequate visualization of the pulmonary arteries up to at least the segmental level [30] (Fig. 1). It was immediately recognized that the sensitivity of MDCTA was higher in central pulmonary emboli than in segmental and subsegmental arteries, but the sensitivity of the single-detector technique was only 70%– a figure reported [31] and confirmed [32] several years ago – precluding its use as a single diagnostic test. Its implementation in a sequential approach (to replace V/Q lung scanning) has been studied in two large management studies [6, 21]. In these studies, the requirement for pulmonary angiography was <2%, and the 3-month thromboembolic risk was below 2%, but the work-up included venous CUS of the lower limbs, to minimize the false-negative rate because of the low sensitivity of single-detector computed tomography (CT). Indeed, in the ESSEP study [6], 55 (16%) of the 349 patients finally diagnosed with PE had a DVT despite a negative helical CT angiogram. The safety of the combination of single-detector CT angiography and CUS to rule out PE was further substantiated by a large-scale Canadian study that resulted in a 0.5% 3-month thromboembolic risk amongst the subgroup of 409 patients who had a negative CT angiography and no proximal DVT on CUS [33].

Multi-detector computed tomography angiography allows both a thinner collimation (1–2 mm slice thickness) and a better definition without increasing image acquisition time, and is therefore potentially more sensitive than single-detector techniques. This development has raised the possibility that PE might be safely ruled out without the use of lower-limb venous CUS, at least in patients without a high probability of PE – a change in strategy that could save both money and other resources. If this was the case, the diagnosis of DVT in a patient with clinically suspected PE and negative findings on MDCTA should be uncommon, and the 3-month risk of VTE in patients with a negative MDCTA scan should be low. This possibility was confirmed in a prospective, multicentre Swiss–French study [22] of patients admitted to the emergency department for clinically suspected PE. The proportion of patients with proximal DVT despite negative findings on MDCTA was very low (0.9%; 95% CI, 0.3–2.7). Therefore, the improvement in the overall detection rate of PE by CUS was marginal in this series, and the 3-month thromboembolic risk in patients left untreated if PE had been excluded on the sole basis of a negative MDCTA would have been 1.5% (95% CI, 0.9–2.7), i.e. similar to that using pulmonary angiography [32] and other recent outcome studies [6, 21, 26, 34]. Strong support for the safety of using MDCTA as a single imaging modality comes from the CHRISTOPHER study [13]. In that series, patients with an elevated DD level or in whom PE was considered likely according to the dichotomized Wells’ rule (n = 1436) underwent helical CT (MDCTA, 88% of the patients). The 3-month thromboembolic risk in patients not treated based on a negative CT angiogram was 1.3% (95% CI, 0.7–2.0). Moreover, in the study by Ghanima et al. [35], of 432 consecutive outpatients clinically suspected of PE who were sequentially submitted to clinical assessment, DD measurement and MDCTA, a definitive diagnosis was reached in 96.5%, with a 3-month thromboembolic risk of 0.6% (95% CI, 0–2.2). Together, these results support the safety of ruling out PE by DD measurement and MDCTA in nonhigh clinical probability patients without resorting to lower-limb CUS. The final proof was provided by a multicentre, randomized, noninferiority study that compared the strategies with and without CUS (see Is there any place for CUS of the lower-limb veins for diagnosing PE?) [11]. A Canadian randomized trial compared V/Q scan-based and MDCTA-based strategies for investigating suspected PE (mostly MDCTA) [12]. Very few patients developed VTE during follow-up, and there was no outcome difference between the two strategies. Whilst CT technology will be further refined in the near future with reduced irradiation and increased slice number, magnetic resonance angiography (MRA) was recently assessed for diagnosing PE in the North American PIOPED III study [36]. Of 371 patients included in the study, 104 (28%) were diagnosed with PE. MRA was technically inadequate in 25% of patients, and the sensitivity of technically adequate imaging was 78% with a specificity of 99%. Clearly, this technique is unlikely to become a standard test for diagnosing PE and should be reserved for centres that routinely perform it well and only for patients for whom standard tests are contraindicated.

Is there any place for CUS of the lower-limb veins for diagnosing PE?

Ultrasonography is not without costs, is labour-intensive and not all centres have access to the technique, therefore excluding ultrasonography may reduce costs. Based on the results of the most recent study by the Swiss–French consortium [22] and supported by the results of the CHRISTOPHER study [13], a randomized trial addressed this question by comparing two strategies based on clinical assessment, DD measurement and MDCTA, with or without CUS [11]. The incidence of PE was 21% in both study arms. In the per-protocol analysis, the 3-month thromboembolic risk was 0.3% (95% CI, 0.1–1.1) in the with CUS arm and 0.3% (95% CI, 0.1–1.2) in the without CUS arm, a nonsignificant difference (0.0; 95% CI, −0.9–0.8). Results were similar in the intention-to-diagnose analysis. In the with CUS arm, the presence of DVT was observed with CUS, hence MDCTA, was unnecessary in 9% of patients. The mean cost per patient was significantly higher in the with CUS strategy [11]. In addition, Elias et al. [37] suggested that early complete (not just proximal) ultrasonography examination of the lower-limb veins was safe and resulted in a substantial reduction in the number of CT scans: 101 of 274 outpatients (37%) clinically suspected of PE had a DVT demonstrated by ultrasonography (65 proximal, 36 distal). However, the specificity of complete CUS for PE (84%) was lower than that of proximal CUS, raising concern of a high number of false positives and that patients would be given anticoagulant treatment unnecessarily. Indeed, it is uncertain whether a patient with a negative MDCTA and a distal DVT should be treated with anticoagulants.

In addition, a secondary analysis of the ESSEP study [6] showed that the 3-month risk of thromboembolic recurrence or death was 6.5% in patients with PE and a CUS-detectable DVT, compared to 2.7% in those patients with PE but without DVT [38]. Although the difference was not statistically significant (P = 0.15), a beta error cannot be excluded. On the other hand, this supports the hypothesis that underlying DVT might negatively influence the prognosis of patients with established PE, as shown in the prognostic score developed [39] and externally validated [40] during the last decade.

How does age influence the diagnostic modalities for PE?

Pulmonary embolism diagnosis in elderly patients is a major issue, because not only the prevalence of the disease increases with age but also diagnosis may be more difficult because of cardiopulmonary comorbidities that are associated with ageing and can mimic clinical presentation of PE. Sohne et al. [41] retrospectively analysed data from two prospective studies of consecutive inpatients and outpatients clinically suspected of PE by categorizing them into three age groups (<65, 65–75, >75 years). They concluded that the proportion of patients >75 years of age with a nonhigh clinical probability and a normal DD level was only 14%, whereas the proportion was 22% and 41% in patients aged 65–75 and <65, respectively. In a secondary analysis of the ESSEP study [6], Couturaud et al. [42] analysed the effect of age on the performance of a diagnostic strategy based on clinical probability, spiral CT and venous CUS. They found that the distribution of inconclusive spiral CT or CUS examinations was not different amongst the three age groups studied (<54, 54–73, >73). Finally, more elderly patients had a positive result with both spiral CT and CUS and could have been diagnosed by CUS alone. In a comprehensive review, Righini et al. [43] studied the influence of age on all commonly used diagnostic tests and strategies for suspected PE. Briefly, they concluded that age reduces the clinical usefulness of DD measurement and V/Q lung scintigraphy. DD enables PE to be excluded in only 5% of patients aged 80 or above, compared with 60% in those younger than 40. Similarly, the proportion of inconclusive V/Q lung scans is almost twice as high (58%) as in patients aged above 70 compared with those who are younger than 40 (32%). By contrast, age does not change the diagnostic accuracy of clinical probability assessment (either implicitly or explicitly), or appears to influence the diagnostic performance of venous CUS and MDCTA.

To increase the exclusion rate of DD in elderly patients, a European group recently devised an age-dependent DD exclusion cut-off for PE that was based on 5132 patients aged 50 or above from four previous prospective cohort studies [16]. The individualized DD cut-off value was defined as (patient’s age × 10) μg L−1 in patients aged above 50. In 1331 patients in the derivation set with an ‘unlikely’ score from clinical probability assessment, PE could be excluded in 42% with the new cut-off value versus 36% with the old cut-off value (<500 μg L−1). In the two validation sets, the absolute increase in the proportion of patients with a DD below the new cut-off value compared with the old value was 5% and 6%, respectively. This increase was largest amongst patients aged above 70 years, ranging from 13% to 16% in the three datasets. The failure rates (all ages) were 0.2% (95% CI, 0–1.0) in the derivation set and 0.6% (95% CI, 0.3–1.3) and 0.3% (95% CI, 0.1–1.1) in the two validation sets, respectively. Thus, it appears that an age-adjusted DD cut-off point, combined with clinical probability, greatly increases the proportion of older patients in whom PE could be safely excluded; this is presently being externally further validated in a prospective clinical outcome study. It is clear that this rule will be valid only for the two tests that were used in the original studies, the VIDAS and the Tinaquant DD tests.

Is there a validated diagnostic algorithm for suspected PE?

In an elegant systematic review, Roy et al. [44] assessed the likelihood ratios of diagnostic tests for PE and determined their clinical application according to the pretest clinical probability. Positive likelihood ratios for diagnostic tests were high-probability V/Q lung scan 18.3 (95% CI, 10.3–32.5), spiral CT 24.1 (95% CI, 12.4–46.7) and CUS of leg veins 16.2 (95% CI, 5.6–46.7). In patients with a moderate or high pretest probability, these findings are associated with a >85% post-test probability of PE. Negative likelihood ratios were normal or near normal appearance on lung scan, 0.05 (95% CI, 0.03–0.10); a negative result on single-detector spiral CT along with a negative result on ultrasonography, 0.04 (95% CI, 0.03–0.06) and a DD concentration <500 μg L−1 measured by quantitative ELISA, 0.08 (95% CI, 0.04–0.18). In patients with a low or moderate pretest probability, these findings were associated with a post-test probability of PE of <5%. Single-detector helical CT alone, a low-probability V/Q lung scan, MRA, and quantitative latex and haemagglutination DD tests had less low negative likelihood ratios and could therefore only exclude PE in patients with a low pretest probability. Roy et al. [44] concluded that the accuracy of tests for suspected PE varies greatly, but that it is possible to estimate the range of pretest probabilities over which each test or strategy can confirm or exclude PE. Based on all these considerations, a simple diagnostic algorithm can be proposed for suspected PE (Fig. 2).

Figure 2.

A simple diagnostic algorithm for suspected deep vein thrombosis (DVT) or pulmonary embolism (PE). 1CUS (lower-limb venous compression ultrasonography) in case of suspected DVT. 2Multi-detector CT angiogram (MDCTA) in case of suspected PE. 3In case of negative CUS or MDCTA and high prior clinical probability, consider additional imaging, e.g. venography (suspected DVT) or lung V/Q scintigraphy or pulmonary magnetic resonance angiography (suspected PE). D-dimer (DD) refers to highly sensitive DD assays. If less-sensitive assays are used, a negative DD result rules out PE or DVT only in the presence of a low (or unlikely) clinical probability. (Modified from Ref. [18]). CUS, venous compression ultrasonography.

Why should PE patients be stratified according to prognosis?

Pulmonary embolism can be stratified according to prognosis. According to European guidelines [45], high-risk PE (formerly ‘massive’ PE) is a life-threatening emergency that requires a specific diagnostic and therapeutic strategy (short-term mortality of about 15%). High-risk or massive PE implies presence of shock or haemodynamic instability. Nonhigh-risk PE can be further stratified according to the presence of markers of right ventricular dysfunction (RVD) and/or myocardial injury into intermediate- and low-risk PE. Intermediate-risk PE (formerly ‘submassive’ PE, a subpopulation of the ‘nonmassive’ form) is diagnosed if at least one RVD or one myocardial injury marker is positive. Low-risk PE (formerly ‘truly nonmassive’ PE) is diagnosed when all measured RVD and myocardial injury markers are negative (short-term PE-related mortality of 1%). The implications of this prognostic risk stratification are (i) high-risk patients (who represent about 5% of all symptomatic PE patients) should be treated aggressively with thrombolytic agents or surgical embolectomy; (ii) low-risk PE patients (the majority of PE patients) might benefit from early discharge or even outpatient treatment if the results of the ongoing outpatient treatment of pulmonary embolism (OTPE) study validate this type of treatment; (iii) intermediate-risk PE patients (who represent about 30% of all symptomatic PE patients) should probably be hospitalized, and the potential benefit of thrombolytic treatment in this subset of patients still awaits confirmation from ongoing clinical trials.

How should we treat patients with PE?

Except for the small minority of high-risk PE patients who should be treated aggressively (thrombolysis or surgical embolectomy), unfractionated heparin (UFH), low-molecular-weight heparin (LMWH) and fondaparinux are the cornerstones of treatment for patients with PE. The efficacy of UFH for treating established PE was first demonstrated as early as 1960 [46]. Heparins act by binding to the natural anticoagulant antithrombin, thereby dramatically accelerating the inactivation by antithrombin of thrombin and of several other activated coagulation factors (including activated factor X, FXa). Even though UFH can be administered subcutaneously [47], it has mostly been given as continuous intravenous infusion. Because of a large individual difference in binding to plasma proteins, including platelet factor 4, the dosage must be adapted to the result of blood tests such as the activated partial thromboplastin time or, more recently, the anti-FXa activity. The relation between these test results and efficacy (thrombosis recurrence) or safety (bleeding) has, however, never been convincingly demonstrated, at least for the individual patient. Nevertheless, the test results enable the avoidance gross overdosage or underdosage.

During the 1980s, UFH was progressively replaced by its low-molecular-weight fractions, which have the main advantages of being administered subcutaneously in weight-adjusted doses without requiring monitoring in most cases [48]. The mechanism of action of LMWH is similar to that of UFH with a more pronounced effect on FXa, when compared to thrombin. The clinical equivalence of LMWH and UFH for treating DVT has been suggested in several studies and shown conclusively in a meta-analysis [49]. One study confirmed this conclusion in the setting of PE [50]. Dosage schemes are summarized in Table 2.

Table 2. Dosage schemes of classical anticoagulants for treatment of venous thromboembolism
LMWHTrade nameDaily dosage (sc)Number of daily injections
  1. LMWH, low-molecular-weight heparin; sc, subcutaneously. a5.0 mg if body weight <50 kg; 10 mg if weight >100 kg.

Enoxaparin (Sanofi Aventis, Paris, France)Lovenox®,
2 mg kg−1
1.5 mg kg−1
Nadroparin (GSK, London, UK)Fraxodi®, Fraxiforte®171 anti-Xa IU kg−11
Tinzaparin (Leo Pharma, Copenhagen, Denmark)Innohep®175 anti-Xa IU kg−11
Dalteparin (Pfizer, New York, NY, USA)Fragmin®200 anti-Xa IU kg−11
Fondaparinux (GSK)Arixtra®7.5 mga1

Fondaparinux is a synthetic pentasaccharide that is almost identical to the smallest natural component of heparin that can still bind to antithrombin to specifically inhibit FXa. In the MATISSE PE study, fondaparinux was shown to be noninferior at the dose of 7.5 mg day−1 to UFH (continuous infusion) for treatment of established PE [51]. Compared to LMWHs that are derived from the porcine intestinal tract, fondaparinux has the advantage of being synthetic. In addition, it seems that heparin-induced thrombocytopenia is extremely rarely even though one well-documented case has recently been reported [52].

Because of their renal elimination, LMWHs and fondaparinux should be administered with caution in patients with impaired kidney function, especially when the calculated creatinine clearance (CrCl) is below 30 mL min−1. In such patients, alternative options include FXa activity monitoring or use of UFH [48], which is cleared via the liver. CrCl can be approximated by means of the Cockcroft or the Modification of Diet in Renal Disease (MDDR) formula. In patients with a CrCl below 30 mL min−1, fondaparinux is contraindicated although it could be administered at a reduced dosage for patients with a CrCl between 30 and 80 mL min−1, but data are scarce.

In most cases, administration of heparins or fondaparinux should overlap with that of vitamin K antagonists (VKAs) from the first day of treatment, and heparin can be stopped as soon as the anticoagulant level induced by VKA has reached an international normalized ratio (INR) of 2.0 on two consecutive days. Heparin/fondaparinux treatment should, however, last for at least 5 days. It has also been recommended that cancer patients should be treated for several months with LMWH rather than VKAs [53].

Vitamin K antagonists block a late step in the synthesis of four plasma coagulation factors (prothrombin or factor II, and factors VII, IX and X) by the liver. Because of the relatively long and different half-lives of circulating factors, a stable level of anticoagulation cannot be reached before 4–7 days. VKAs include substances with a short (acenocoumarol, Sintrom®), intermediate (warfarin, Coumadin®; fluindione, Previscan®) or long (phenprocoumone, Marcoumar®) half-life. This feature, along with a genetically induced metabolic variability [54, 55], the influence of environmental variables such as the vitamin K content of food, and a narrow therapeutic index, means that VKA treatment requires close monitoring. Monitoring has been standardized, and the therapeutic level corresponds to an INR between 2 and 3 (target 2.5). At an INR <2.0, the risk of thromboembolic recurrence increases, and >3.0, the bleeding risk becomes much higher.

The safety of VKA treatment can be improved by supporting patients’ compliance, avoiding concurrent drugs with potential interactions, or excessive ingestion of alcohol, and, in some patients, by using self-assessment or even self-treatment after careful teaching. In addition, large loading doses should be avoided to overcome an initial paradoxical prothrombotic state because of the depletion of protein C, a vitamin K-dependent coagulation inhibitor with a very short half-life. Instead, VKA treatment should be initiated with doses likely to be close to the maintenance dose, and the INR should be checked after two administrations, at least for VKAs with a short half-life (e.g. acenocoumarol). Whether analysis of genetic determinants of the response to various warfarin doses will be clinically useful in the future remains to be established [55].

A common adverse effect of all anticoagulant drugs is bleeding that occurs more frequently at the initiation of treatment (‘demasking’ of lesions) and can have devastating consequences (intracerebral or retroperitoneal bleeds). During that initial period, heparin is associated with a major bleeding risk of 0.8% per day [56]. Major bleeding associated with VKAs occurs at an age-dependent [57] monthly rate of about 0.4% [58]. Clinical scores, such as the Hemorr2hages score [56] (Table 3), have been prospectively validated and may guide estimation of the haemorrhagic risk under VKA treatment. However, they are derived from cohorts of patients given VKAs because of atrial fibrillation, not VTE. More recently, the RIETE group proposed a new score entirely derived from a very large cohort of patients with VTE [59] (Table 3).

Table 3. Bleeding scores
Hemorr2hages bleeding risk scoreaRIETE bleeding risk scoreb
  1. PE, pulmonary embolism. aRate of major bleeding varies from 1.9 per 100 patient-years with a score of 0–2.5 (score of 1), 5.3 (2), 8.4 (3), 10.4 (4) and 12.3 (≥5). bRate of major bleeding varies from 0.3 per 100 patient-trimesters (score of 0) to 2.6 (1–4) and 7.3 (>4).

Prior bleed2 pointsRecent major bleed2 points
Hepatic or real disease1 pointCreatininaemia > 1.2 mg dL−1 (110 μmol L−1)1.5 point
Alcohol abuse1 pointHaemoglobin < 13 (men) or 12 (women) g dL−11.5 point
Malignancy1 pointMalignancy1 point
Age >75 years1 pointClinically overt PE1 point
Uncontrolled hypertension1 pointAge >75 years1 point
Anaemia1 point  
Excessive risk of fall1 point  
Stroke1 point  
Reduced platelet count or function1 point  
Prior bleed2 points  

How long should we treat patients with PE?

The duration of anticoagulant treatment following DVT and PE remains controversial. Nevertheless, several facts have been highlighted in the past two decades that should help to establish guidelines based on evidence rather than on changing opinions of leaders in the field. It is clear that the duration of anticoagulation therapy should be dictated by the balance between two risks: the risk of recurrent VTE with and without treatment, and the risk of treatment-induced haemorrhage.

In fact, recurrent VTE is rare during anticoagulant treatment and has a fatality rate of only 0.4% (95% CI, 0.2–0.6) [60]. However, anticoagulant treatment is associated with a definite bleeding risk: heparin induces major bleeding at a rate of 0.8% per day (with a daily fatality rate of 0.05%) [61] and oral anticoagulants at a rate of 0.4% per month [58]. In addition, in an authoritative review, Kearon et al. [53] noted that a 3-month course of anticoagulant treatment was as efficacious as a 6- to 12-month course and that VTE related to transient (reversible) risk factors (e.g. surgery, trauma) is associated with a lower risk of recurrence. Based on all these considerations, the 8th American College of Chest Physicians (ACCP) Expert consensus [53] recommended the duration of anticoagulant treatment under various conditions, which is summarized in Table 4.

Table 4. Recommended duration of anticoagulant treatment for venous thromboembolism events according to the 8th American College of Chest Physicians (ACCP) evidence-based clinical practice guidelines
Type of eventRecommended treatment durationGrade of recommendation
  1. PE, pulmonary embolism; DVT, deep vein thrombosis.

First DVT or PE secondary to a transient (reversible) risk factor (‘provoked’ event)3 months1A
First idiopathic (‘unprovoked’) DVT or PEAt least 3 months1A
At the end of the 3-month periodEvaluate for long-term treatment1C
In the absence of contra-indicationLong-term treatment1A
During long-term treatmentEvaluate risk–benefit balance periodically1C
Recurrent DVT or PE or strong thrombophiliaLong-term treatment1A

Another interesting approach to better tailor individually the duration of anticoagulant treatment would be to recognize which patients are at a particularly low or high risk of a recurrent event. For example, the aim of the ongoing REVERSE II study is to validate a clinical prediction rule to identify patients who might be safely left untreated after the initial 3- to 6-month course of anticoagulants. Some authors have tried to use DD levels to predict the risk of recurrent events. Palareti et al. [62] showed in the PROLONG study that following idiopathic DVT, a normal DD value (<500 ng mL−1) 1 month after discontinuation of oral anticoagulant treatment, had a high negative predictive value for VTE recurrence, whereas the risk of a recurrent event in patients with a DD level above the cut-off was 15% after 18 months, which may not be considered sufficient to indicate life-long anticoagulant treatment. In a similar attempt to better evaluate the risk of recurrence after a first DVT, Prandoni et al. [63] showed that the presence of residual venous thrombosis is a risk factor for recurrent thromboembolism. However, the role of these interesting new options in clinical practice remains to be established [64].

At present, all patients with PE should be treated for 3 months. If the event was secondary to a transient or removable risk factor (Fig. 3), especially if strong, anticoagulant treatment may then be stopped. In patients without a triggering risk factor (the so-called idiopathic or unprovoked events), anticoagulant therapy should be continued as long as the benefit–risk balance is judged to be favourable, whilst patients with PE and cancer should also receive anticoagulant therapy until the cancer is considered to be under control. In patients with a weak triggering factor, continuation of anticoagulant treatment should also be prolonged for as long as possible, but the bleeding risk should be taken into consideration.

Figure 3.

Types of venous thromboembolism events and risk of recurrence.

In some patients in whom the treatment should ideally be prolonged but for some reasons cannot be, targeting a lower INR value (1.5–2.0) might be an acceptable alternative strategy [65], as endorsed by the ACCP guidelines [53].

Are there alternatives to the classical anticoagulant treatment with heparin/LMWH followed by coumarin derivatives?

In the acute phase, fondaparinux, a pure, subcutaneously administered, indirect (acting via antithrombin) FXa inhibitor is now recommended by the ACCP guidelines (Grade 1A) as well as UFH or LMWHs [53]. In addition, several new oral anticoagulants are presently under clinical development (Table 5). These direct (i.e. antithrombin-independent) inhibitors of FXa (e.g. rivaroxaban, apixaban) or thrombin (e.g. dabigatran) avoid most of the drawbacks of heparin and have the potential to replace both heparins and VKAs in the future in a substantial proportion of patients. Recently, the RE-COVER study was published [66]. In this randomized, double-blind, noninferiority trial involving patients with acute VTE who were initially given parenteral anticoagulation therapy for a median of 9 days (interquartile range, 8–11), oral nonmonitored dabigatran, administered at a dose of 150 mg twice daily, was compared with warfarin, which was dose-adjusted to achieve an INR of 2.0–3.0. A total of 30 of the 1274 patients randomly assigned to receive dabigatran (2.4%), when compared with 27 of the 1265 patients randomly assigned to warfarin (2.1%), had recurrent VTE; the difference in risk was 0.4 percentage points (95% CI, −0.8 to 1.5; P < 0.001 for the prespecified noninferiority margin). The hazard ratio with dabigatran was 1.10 (95% CI, 0.65–1.84). Major bleeding episodes occurred in 20 patients assigned to dabigatran (1.6%) and in 24 patients assigned to warfarin (1.9%) (hazard ratio for dabigatran, 0.82; 95% CI, 0.45–1.48), and episodes of any bleeding were observed in 205 patients assigned to dabigatran (16.1%) and 277 patients assigned to warfarin (21.9%; hazard ratio for dabigatran, 0.71; 95% CI, 0.59–0.85). The numbers of deaths, acute coronary syndromes, and abnormal liver function tests were similar in the two groups. Adverse events leading to discontinuation of the study drug occurred in 9.0% of patients assigned to dabigatran and in 6.8% of patients assigned to warfarin (P = 0.05). A fixed dose of dabigatran was thus found to be as effective as warfarin, with a similar safety profile, with no need for monitoring [67]. In the multicentre, randomized, double-blind, placebo-controlled, event-driven EINSTEIN-EXT study [68], 1197 patients who had already received anticoagulant treatment for 6 or 12 months were treated for an additional 6- or 12-month period either with oral unmonitored rivaroxaban (20 mg once daily) or placebo. The primary efficacy end-point of recurrent symptomatic objectively confirmed VTE was reached in eight patients (1.3%) in the rivaroxaban arm and 42 patients (7.1%) in the placebo arm, resulting in a highly significant relative risk reduction of 82%, and the number needed to treat to prevent one primary efficacy outcome of only 15. Major bleeding was not different amongst the two groups (P = 0.11), but clinically relevant nonmajor bleeds occurred in 32 patients (5.4%) treated with rivaroxaban compared with seven (1.2%) with placebo (P < 0.01).

Table 5. Comparison of four new oral anticoagulants in development (phase III studies completed or ongoing) for treating deep vein thrombosis and pulmonary embolism
Drug (class)Company (acronym of ongoing studies)Half-life (h)Bioavailability (%)Renal elimination (%)Oral dosage
  1. DTI, direct thrombin inhibitor; anti-Xa, direct anti-activated factor X; b.i.d., twice daily; o.d., once daily.

Dabigatran (DTI)Boehringer-Ingelheim
(Ingelheim, Germany)
Rivaroxaban (anti-Xa)Bayer (Bayer Schering Pharma,
Leverkusen, Germany)/
(Johnson & Johnson,
New Brunswick, NJ, USA)
5–13>8033% (unchanged) 33% as inactive metaboliteso.d.
Apixaban (anti-Xa)BMS (New York, NY, USA)/Pfizer
Edoxaban (anti-Xa)Daïchi-Sankyo (Tokyo, Japan)

Conclusions and perspectives

Diagnosing VTE relies on several, mainly noninvasive diagnostic tools that must be used sequentially. According to Roy et al. [69], using validated diagnostic algorithms is associated with a lower risk of complications and should therefore be implemented in all institutions if possible, depending on local availability and expertise and taking into account the issue of costs. With the development of potentially more sensitive diagnostic tests such as MDCTA, and possibly MRA (despite the disappointing results of the PIOPED III study [36]), or calf vein CUS, clinicians will face the risk of overdiagnosis, and hence overtreatment with its associated iatrogenic risk.

Development of new anticoagulant drugs might modify this situation in the near future, should the benefit–risk balance of their use be improved. Therefore, the important issue in the future may no longer be the detection of clots alone, but also the identification of patients who need to be treated with anticoagulant drugs for a short or indefinite duration, which may not be easy.

Conflict of interest statement

No conflict of interest was declared.