INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: diagnosis; pulmonary angiography; pulmonary embolism; radionuclide imaging; sensitivity and specificity; venous thrombosis

Pulmonary embolism (PE) is the most severe complication of deep venous thrombosis (DVT). The prevalence of PE in patients with proximal DVT has been estimated as 40-50%, and asymptomatic ("silent") PE is a recognized entity with a prevalence ranging from 25 to 88% in prospective series (1). However, because these studies selected asymptomatic patients and/or did not use pulmonary angiography in patients with nondiagnostic lung scans, the real prevalence of PE and the specificity of PE symptoms in patients with known proximal DVT remained unknown.

During anticoagulant therapy, the occurrence of new symptoms suggestive of PE may indicate either a new episode of PE, delayed symptoms of a pre-existing one (symptomatic or silent), or any other diagnosis consistent with these new symptoms. Occurrence of a new PE during adequate anticoagulation indicates failure of treatment and therefore a possible need for inferior vena cava (IVC) interruption (10). Consequently, routine assessment of "baseline" pulmonary perfusion in proximal DVT patients has been advocated to distinguish between a true new PE and an already established one, should pulmonary symptoms occur during anticoagulant treatment (4, 6). The clinical usefulness of this approach, however, remains uncertain (11), mainly because the incidence and the specificity of PE symptoms during anticoagulant therapy have not been carefully evaluated in large prospective clinical series (4, 6).

The "Prévention du Risque d'Embolie Pulmonaire par Interruption Cave" (PREPIC) trial (12) prospectively studied the efficacy of IVC filters in prevention of PE by evaluating, in 400 patients with phlebography-proven proximal DVT, the incidence of PE during anticoagulant therapy with and without an IVC filter. Evaluation of this end point required the use of strict criteria for the diagnosis of on-treatment PE. So a careful evaluation of baseline pulmonary perfusion, including the performance of pulmonary angiography in patients with nondiagnostic lung scans, was essential. Therefore, the PREPIC study provides unique data about the diagnosis of PE in patients with DVT. These data should allow determination of the specificity of PE symptoms and lung scan perfusion defects in patients with known proximal DVT, both at baseline and during anticoagulant therapy.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

(See additional information in the Online Data Supplement.)

The methods used for the PREPIC study have been described in a previous report (12). Briefly, over a 40-mo period, 400 patients with venography-proven DVT were entered into a multicenter prospective randomized trial with a factorial design comparing insertion of a permanent IVC filter with no filter, and subcutaneous enoxaparin with intravenous unfractionated heparin. Oral anticoagulants were started on Day 4 and continued for at least 3 mo in all patients (international normalized ratio [INR] between 2.0 and 3.0).

The primary outcome measure was the incidence of PE within the first 12 d of inclusion. Secondary efficacy end points included symptomatic recurrent venous thromboembolism up to 2 yr after inclusion in the study. In the present analysis, only symptomatic events indicative of PE during the initial 3 mo of anticoagulant therapy were considered.

Baseline Evaluation

Patients over 18 yr of age were eligible if they had an acute, proximal lower limb DVT confirmed by venography, with or without symptoms of PE. Patients with life-threatening PE were not included in the study. The presence or absence of PE symptoms on entrance into the study was noted prospectively for all patients.

All patients underwent ventilation-perfusion (/) lung scanning and/or pulmonary angiography within 48 h of enrollment. Pulmonary angiography was recommended for patients with nondiagnostic lung scans. In patients who had undergone pulmonary angiography as the initial evaluation of pulmonary perfusion, the performance of a baseline lung scan was also recommended.

The diagnosis of PE was based on the presence of at least one intralumenal filling defect or abrupt arterial cutoff on pulmonary angiograms, or on a high-probability lung scan according to PIOPED (Prospective Investigation of Pulmonary Embolism Diagnosis) criteria (13). Low- and intermediate-probability lung scans according to PIOPED criteria were merged into a single nondiagnostic category.

All baseline pulmonary imaging studies were evaluated by two independent central reading committees (one for the lung scans, one for the pulmonary angiograms). For patients who underwent both lung scanning and pulmonary angiography, angiographic data were considered over the scans.

On-treatment PE

Clinically suspected on-treatment PE was defined as any symptomatic thoracic event that led to reassessment of pulmonary perfusion by lung scanning and/or angiography, or any death within 3 mo of inclusion. After discharge, the patients and their physicians were instructed to seek help from the study centers in case of any suspected recurrent thromboembolic event. A follow-up visit was scheduled for Day 90 for all survivors.

The angiographic diagnosis of on-treatment PE was based on visualization of an intralumenal filling defect in an area not previously involved by the embolic process. The scintigraphic diagnosis of on-treatment PE required (arbitrarily) the visualization of new "high-probability" lung scan defects (13) without improvement in other areas (14, 15). For patients who died, the diagnosis of fatal on-treatment PE was based on strong clinical evidence and/or on autopsy findings.

Clinical, radiological, and lung scan data for all events indicative of PE were reviewed and validated by a central adjudication committee.

Statistical Methods

Kaplan-Meier curves were created to estimate risks of new pulmonary events.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Four hundred patients with venography-proven acute proximal DVT were included in the PREPIC study. Their main clinical features are listed in Table 1. No patients were lost to follow-up from Day 0 to 90. 

Baseline Assessment of Pulmonary Perfusion

The presence or absence of PE at baseline could be ascertained in 350 patients (87.5%), of whom 197 (56%) had PE (positive pulmonary angiography, n = 166; high-probability / lung scan, n = 31). Lung scan or pulmonary angiography findings were normal in 153 patients (normal / lung scan, n = 57; normal pulmonary angiography, n = 96) (Table 2). The presence or absence of PE at baseline could not be ascertained in only 50 patients: 33 patients with nondiagnostic lung scans did not undergo pulmonary angiography, and in 17 patients, angiograms (n = 13) or lung scans (n = 4) were the only baseline documents and could not be interpreted by the central reading committees (poor quality, only one side, or missing documents).

One hundred and ninety-three patients (48%) had no PE symptoms at inclusion. Table 2 shows the correlation of objective test results with the presence or absence of PE symptoms. Sensitivity and specificity of symptoms for the diagnosis of PE are 74 and 67%, respectively, and positive and negative predictive values are 70 and 53%, respectively (Table 2).

No significant correlation was found between the prevalence of symptomatic or asymptomatic PE and patient clinical characteristics listed in Table 1.

Among the 37 patients with PE symptoms and a nondiagnostic lung scan who underwent pulmonary angiography, only 8 (22%, 95% confidence interval [CI], 8-35%) had PE.

On-treatment PE

Between Day 0 and Day 90, 21 patients died of causes other than venous thromboembolism, and a total of 29 episodes of suspected on-treatment PE were recorded in 27 patients (2 patients had 2 episodes each). Most events (53%) occurred within 12 d of inclusion. Only eight (8 of 29, 28%; 95% CI, 11.3- 44.9%) proved to be "true" on-treatment PEs, and seven of them occurred within 12 d of beginning of treatment (Table 3 and Figure 1). Four patients died of PE, all within the first 10 d, and all with a positive baseline study, asymptomatic in two. The positive predictive values of PE symptoms occurring within 12 d and within 3 mo of beginning of anticoagulant therapy were 47 and 33%, respectively. The cumulated risk (Kaplan-Meier) of occurrence of PE symptoms during the first 3 mo of anticoagulant therapy was 6.8% (95% CI, 5.4-8.2%), whereas the cumulated risk for confirmed on-treatment symptomatic PE was only 2.0% (95% CI, 0.6-3.4%) (Figure 1).

View larger version (12K): [in this window] [in a new window]   Figure 1.   Risks of occurrence of new symptoms of pulmonary embolism and of confirmed pulmonary embolism during anticoagulant treatment (Kaplan-Meier) in 400 patients with acute proximal deep vein thrombosis. The numbers in parentheses are the numbers of patients still at risk at the corresponding follow-up dates (Day 10 and Day 90). Only 27 events are represented (in the two patients who each had two episodes of suspected pulmonary embolism, only the first episode is considered).

Scintigraphic or angiographic findings were abnormal at the time of suspected on-treatment PE in 20 of 25 nonfatal events (88%) (Table 4). Among the 19 abnormal / lung scans, 12 (63%; 95% CI, 41.5-84.8%) showed "high-probability" patterns, and were indeterminate in the other seven cases. Comparison with baseline documents confirmed PE in only four patients (Table 4): perfusion was unchanged or improved in 13 cases, and two patients showed "spurious" scintigraphic recurrences (worsened perfusion in some areas with contemporary improvement in other areas in patients with initial extensive perfusion defects [14, 15]).

The number of on-treatment PE suspicions was higher in the subgroup of 200 patients who were randomized to the "no filter" group (20, vs. 9 in the filter group, p < 0.05), but the rate of confirmed on-treatment PE was similar in both groups with and without a filter (5 of 20 [25%] vs. 3 of 9 [33%], respectively).

The occurrence of PE symptoms during therapy was significantly more frequent in patients with than without PE symptoms at the time of the baseline study (20 of 207 [10%] vs. 7 of 193 [4%], p < 0.02), but the rate of confirmed on-treatment PE was similar in the two groups. The incidence of suspected and confirmed PE were not significantly different between patients with and without PE at baseline.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Although silent PE is a recognized entity, the sensitivity and specificity of PE symptoms in patients with proximal DVT have not been carefully evaluated. One of the main reasons for this absence of data is that lung scanning alone was used to evaluate pulmonary perfusion in virtually all published series on this subject (Table 5), so a definitive diagnosis (presence or absence of PE) could not be reached for a significant proportion of these patients. By performing pulmonary angiography in most patients with nondiagnostic lung scan, we were able to confirm or refute PE at baseline in 350 of 400 patients (87.5%) with proximal DVT (Table 2). The finding that about one-fourth of proximal DVT patients with PE symptoms did not have detectable PE is therefore important new information. The 70% positive predictive value of PE symptoms found in these patients is only close to the 68% value found in patients with a high "clinical science" pretest probability of PE in the PIOPED study (13).

In patients with clinically suspected PE and nondiagnostic lung scan, the presence of a DVT has been considered sufficient to confirm the diagnosis of symptomatic PE (16, 17). Angiography data in such patients, however, are unavailable in the literature. In this study, angiography could confirm PE in only 8 of 37 patients (22%) with PE symptoms and nondiagnostic lung scan. Although 11 patients in the same situation did not undergo angiography (Table 2), this rate may appear surprisingly low. To our knowledge, the only comparable data were reported by Hull and colleagues: among 33 patients with suspected PE, nondiagnostic lung scan, and normal pulmonary angiography, 11 (33%) had DVT on systematic venography of the lower limbs (18). Whatever the interpretation of these findings (e.g., angiography may miss very small emboli and/or the prevalence of nondiagnostic lung scans and chest symptoms as well as the risk of DVT are increased in certain populations such as patients with cardiac or pulmonary disease [19, 20]), our data further suggest cautious interpretations of nondiagnostic lung scans, even in the presence of proximal DVT.

The clinical relevance of assessing pulmonary perfusion in all patients with DVT is a matter of debate. Because therapy depends little on the presence or absence of PE (10), most authors do not advocate further investigation once the presence of thrombi has been objectively confirmed either at the pulmonary or venous level (21). Others advocate routine baseline lung scanning in patients with proximal DVT (6, 15, 25) in order to separate true from spurious on-treatment PE, should pulmonary symptoms develop during anticoagulant therapy. This question is clinically important because PE during adequate anticoagulant therapy is a widely accepted indication for IVC interruption (6, 10, 26).

The incidence of confirmed PE (2% at 3 mo) as well as the timing of recurrent events (most within 2 wk of the start of anticoagulant therapy) are consistent with the findings of Douketis and colleagues in two studies (27, 28). But our findings that about three symptomatic events must be considered to identify one "true" on-treatment PE (Figure 1) has not been reported in the literature. Further, 88% of patients with suspected on-treatment PE who underwent lung scanning had abnormal perfusion studies, and 63% of them had "high-probability" lung scans at the time of the new symptomatic event (Table 4). These data are consistent with the fact that up to two-thirds of patients still have perfusion defects 3 mo after an acute PE (29). Consequently, only a comparison with baseline documents allowed a correct classification of the new events in these patients. In other words, even if other factors intervene in the final decision concerning the placement of an IVC filter, our data indicate that this procedure might be wrongly considered in most patients with suspected PE during anticoagulant therapy if baseline perfusion data are unavailable. Although these findings strongly support the statement that "patients . . . with DVT merit a lung scan as part of their initial evaluation" (6), the low incidence of events indicative of PE in this study (6.8% over 3 mo) raises the question of the cost- effectiveness of such an approach. This important issue should be addressed in specific studies, and our data constitute a unique basis for such a perspective.

The prognostic significance of PE may be seen as another reason for performing systematic evaluation of baseline pulmonary perfusion in patients with DVT. The prognostic value of PE was confirmed in a meta-analysis showing that patients with symptomatic PE had a 1.5% risk of fatal PE during anticoagulant therapy, compared with 0.4% (p < 0.01) for patients with proximal DVT but no PE symptoms (27). In the present study, the small number of confirmed embolic events (n = 8) did not allow a reliable analysis of the clinical risk factors for on-treatment PE. Notably, it is unknown whether the risk of recurrent PE varies between patients with and without PE symptoms at baseline. One can simply notice that two of four patients who died of recurrent PE had silent PE on baseline. These data and others (27, 28) emphasize the need for developing reliable criteria to identify the small subset of patients at "very high risk" for further potentially fatal PE despite adequate anticoagulant therapy.

Like any "post hoc" analysis, our study has inherent limitations. Notably, as only the distinction between symptomatic and asymptomatic PE was needed in the PREPIC trial (12), detailed information regarding what symptoms (chest pain and/or hemoptysis and/or dyspnea) were present or absent in the study patients at baseline was not collected. None of these symptoms, however, alone or in combination, were shown to have any specificity in the patients from the PIOPED study, even in those without underlying cardiac or pulmonary disease (30). So it is unlikely that more information about symptoms would change significantly the specificity values found in this study. Similarly, there were no predetermined criteria of clinical suspicion for PE during anticoagulant therapy. This likely resulted in varying "thresholds" among clinicians who ordered (or did not order) repeat lung scanning or angiography. Although our pragmatic approach, including such variability, reflects "real life" for patients with proximal DVT, the clinical interest of more objective clinical scores (31) could merit further study of patients with known DVT, both at baseline and during anticoagulant therapy.

Whatever its limitations, this study provides reliable data that may prove helpful both in everyday clinical practice and in the design and interpretation of clinical trials in venous thromboembolic disease. Specifically, our data indicate that even in patients with proximal DVT, the association of PE symptoms and a nondiagnostic lung scan is an insufficient basis to confirm the diagnosis of symptomatic PE. Similarly, during anticoagulant therapy for proximal DVT, the association of new PE symptoms and high-probability lung scan defects is unspecific, so a definitive diagnosis cannot be reached in most cases if baseline pulmonary perfusion has not been assessed. Finally, beyond the first 3 wk of anticoagulant therapy, the positive predictive value of new PE symptoms is so low (Figure 1) that the performance of any objective test may be questioned, especially if anticoagulant treatment is adequate.

Even in patients with known proximal DVT, the presence or occurrence of PE symptoms is unspecific and careful imaging studies are needed for diagnosis, both at baseline and during anticoagulant therapy. The availability of baseline lung perfusion data is helpful to confirm or refute suspected on-treatment PE, but the cost-effectiveness of such an approach needs further study. These findings may have important implications both in everyday clinical practice and in the design and interpretation of clinical trials in venous thromboembolic disease.