In view of the high rate of bleeding in patients with pulmonary embolism, much of which may be attributed to bleeding at the site of catheter insertion for the pulmonary angiogram, we assessed the relative risks for bleeding with thrombolytic therapy in patients who were managed using pulmonary angiograms compared with those managed using noninvasive tests, primarily the ventilation-perfusion lung scan.

Methods

Top Methods Results Discussion Author & Article Info References

The risks for pulmonary angiography combined with the risks for major bleeding from thrombolytic therapy in patients diagnosed invasively were assessed from the literature. The risk for major bleeding with thrombolytic therapy among patients diagnosed noninvasively was assessed from investigations of thrombolytic therapy in patients with acute myocardial infarction.

Definition of Major Bleeding

We defined major bleeding as a decrease in hemoglobin levels of 20 g/L (2 g/dL) or more, a blood transfusion requirement of 2 units or more, intracerebral bleeding, retroperitoneal bleeding, pericardial bleeding, bleeding that required a surgical intervention, bleeding into a major joint, or bleeding into the eye.

Risks of Thrombolytic Therapy in Patients Diagnosed by Pulmonary Angiography

Most investigations of thrombolytic therapy in patients with acute pulmonary embolism used tissue plasminogen activator (tPA). For consistency, we considered only patients treated with tPA. The frequency of major bleeding with tPA in patients with acute pulmonary embolism who had pulmonary angiography was based on data averaged from several studies [1-5]. The averaged data included data on patients who did and did not receive heparin. The frequency of major bleeding in both groups did not show a statistically significant difference. Data about the frequency of bleeding with tPA were included in the average, regardless of modest differences of the dose, site of injection, and rate of injection.

Trials of thrombolytic therapy in patients with pulmonary embolism in the 1980s and more recently showed lower rates of major bleeding than did earlier studies, regardless of the thrombolytic agent used [6]. Only more recent studies, therefore, were included in the analysis of the frequency of bleeding in patients with pulmonary embolism who received thrombolytic therapy. Studies related to bleeding after urokinase was administered over 12 hours or longer were not considered in this evaluation because the improvement of hemodynamic variables at 2 hours with tPA when compared with urokinase could be potentially important in patients with life-threatening massive pulmonary embolism [4, 5, 7]. It is possible that an accelerated regimen of urokinase given over 2 hours may be as effective as tPA in terms of lysis and hemodynamic improvement [5]. Data related to the bleeding rate with that regimen, however, are sparse and therefore were excluded from our evaluation. All studies included in our analysis of the literature indicated the amount of blood transfused. If the description of bleeding complications did not specify the amount of blood transfused, the study was eliminated from consideration.

All reported investigations excluded patients at high risk for bleeding, such as those with recent surgery, recent biopsy, peptic ulcer disease, blood dyscrasia, or severe hepatic or renal disease. Thus, the reported patients were at low risk for bleeding.

Risks of Thrombolytic Therapy in Patients Diagnosed Using Noninvasive Procedures

Few data are available on the use of thrombolytic agents for pulmonary embolism in patients who received treatment without an angiographically proven diagnosis [8, 9]. Among patients studied by Goldhaber and associates [8], 40 of 46 (87%) were treated on the basis of a diagnosis made by a ventilation-perfusion lung scan. Among the entire group, 3 of 46 (6.5%) had major bleeding after receiving 100 mg of tPA administered over 2 hours. Among patients treated by Levine and associates [9] with a bolus regimen of tPA, 11 of 33 (33%) were diagnosed on the basis of a ventilation-perfusion lung scan. No major bleeding was noted among the entire group of 33 patients. The mode of administration (bolus injection) differed from the mode of administration of tPA by other investigators.

Because of insufficient data on the use of tPA in patients with pulmonary embolism who were diagnosed noninvasively, the risk for bleeding after a noninvasive diagnosis was assessed from the Thrombolysis in Myocardial Infarction (TIMI) IIA study [10]. The relative risk for bleeding was assessed after an invasive compared with a noninvasive evaluation of acute myocardial infarction. We assumed that this risk ratio applied to the relative risks for invasive compared with noninvasive management of acute pulmonary embolism. In the TIMI-IIA study, patients were randomly assigned to receive coronary arteriography within 2 hours of the administration of tPA or to receive treatment with tPA in the absence of coronary arteriography (which was subsequently done after 18 to 48 hours). Major bleeding ( 2 units of blood transfused or intracranial hemorrhage) occurred in 23 of 195 (11.8%) of patients who had immediate coronary arteriography and in 7 of 194 (3.6%) who did not have this invasive procedure at the time of tPA administration. The risk ratio was 3.3 (11.8 ÷ 3.6 [CI, 1.5 to 9.8]) for major bleeding (invasive management of myocardial infarction compared with noninvasive management of myocardial infarction). Other comparative studies of tPA given with invasive and noninvasive management after myocardial infarction have been done, but the number of units of blood transfused was not given.

Major Complications of Pulmonary Angiography

The major complications of pulmonary angiography among 1111 patients were defined as death (0.45%), respiratory distress requiring cardiopulmonary resuscitation or intubation (0.36%), renal failure requiring dialysis (0.27%), and bleeding at the site of catheter insertion requiring blood transfusion of 2 units or more (0.18%) [11]. One or more major complications occurred in 1.3% of patients (CI, 0.6% to 1.9%). These major complications were independent of pulmonary artery pressure, volume of contrast material, and presence of pulmonary embolism.

Statistical Analysis

We did a decision analysis based on data from other studies. The 95% CI for relative risk was calculated using the method of Woolf [12], as described more recently by Schlesselman [13].

Results

Top Methods Results Discussion Author & Article Info References

The average frequency of major bleeding in patients with acute pulmonary embolism who had pulmonary angiography before the administration of tPA was 18 of 129 (14.0%; CI, 7.9% to 20.1%) (Table 1). Patients who received heparin with tPA did not have significantly more bleeding than patients in whom heparin was temporarily withheld (9 of 63 [14.3%] compared with 9 of 66 [13.6%] P = 0.9).

Based on a risk ratio of 3.3 (CI, 1.5 to 9.8) for major bleeding with tPA associated with invasive management of myocardial infarction compared with noninvasive management of myocardial infarction, the estimated risk was 4.2% (estimated CI, 1.4% to 9.3%) for major bleeding with tPA after a noninvasive diagnosis of pulmonary embolism.

View larger version (26K): [in this window] [in a new window]   Figure 1. Predicted major complications from pulmonary angiography plus major bleeding from thrombolytic therapy shown as a function of the prevalence of pulmonary embolism. Total major complications were major complications from angiography plus major bleeding from thrombolytic therapy. Predicted complications assume a rate of 1.3% for major complications from angiography and rates of 14.0% (observed value); 7.9%, and 20.1% (lower and upper ends of the 95% CI, respectively) for major bleeding among patients treated with thrombolytic therapy after an angiographic diagnosis. The total predicted complications describe a straight line with the slope = rate of major bleeding after angiography and the intercept = percentage of major complications of angiography. Major bleeding from thrombolytic therapy after a noninvasive diagnosis is shown by horizontal lines at assumed rates of 4.2% (observed value), 1.4%, and 9.3% (lower and upper ends of the 95% CI, respectively). PE = pulmonary embolism.

Assuming that the frequency of major bleeding with thrombolytic therapy after a noninvasive diagnosis is 9.3% based on the upper end of the estimated 95% CI and if the risk for major bleeding after an invasive diagnosis is 14%, Figure 1 shows that noninvasive therapy management is associated with fewer major complications than is invasive management if the estimated probability of pulmonary embolism is more than 57%. Assuming that the risk for major bleeding with thrombolytic therapy after a noninvasive diagnosis is 1.4% based on the lower limit of the estimated 95% CI, it would always be safer to treat on the basis of a noninvasive diagnosis (see Figure 1).

Sensitivity analysis was also done by assuming that the risk was 20.1% (the upper end of the 95% CI) for major bleeding with thrombolytic therapy after an angiographic diagnosis. Comparisons were made with a 4.2% assumed risk for major bleeding with thrombolytic therapy after a noninvasive diagnosis. Calculated total complications (major bleeding plus major complications from angiography) would be fewer if thrombolytic therapy were administered using a noninvasive diagnosis if the estimated probability of pulmonary embolism exceeded 14% (see Figure 1).

Sensitivity analysis was also done, assuming the risk was 7.9% (the lower end of the 95% CI) for major bleeding with thrombolytic therapy after an angiographic diagnosis compared with an assumed risk of 4.2% for major bleeding with thrombolytic therapy after a noninvasive diagnosis. Calculated total complications (major bleeding from thrombolytic therapy plus major complications of angiography) would be fewer if therapy was administered on the basis of a noninvasive diagnosis if the estimated probability of pulmonary embolism was more than 37% (see Figure 1).

If major bleeding with thrombolytic therapy after pulmonary angiography is only 1.0% more than major bleeding with a noninvasive diagnosis, that is, 5.2% compared with 4.2%, it would be safer to treat patients on a noninvasive basis if the probability of pulmonary embolism was more than 55% (Figure 1).

Discussion

Top Methods Results Discussion Author & Article Info References

Thrombolytic therapy for patients with massive acute pulmonary embolism is indicated for patients who are hypotensive, hypoxic when receiving high levels of oxygen, or clinically stable with echocardiographic evidence of right ventricular failure. Our calculations lead to a conclusion that differs from traditional practice about the necessity of an angiographically proven diagnosis. We conclude that in a patient who is unstable from what appears to be massive acute pulmonary embolism, at the appropriate level of probability of pulmonary embolism, it is safer to treat on the basis of a noninvasive diagnosis rather than an angiographically proven diagnosis. The relatively low rate of major bleeding of 6.5% after tPA reported by Goldhaber and associates [8] among patients with pulmonary embolism, 87% of whom were treated on the basis of a ventilation-perfusion lung scan, tends to support our conclusion. Unfortunately, if thrombolytic therapy is initiated on the basis of noninvasive diagnosis, some patients who do not have pulmonary embolism will receive thrombolytic drugs and will have bleeding complications. Overall, however, fewer patients will have major complications with the approach identified by these calculations.

The risk ratio is conservative for bleeding with thrombolytic therapy after invasive compared with noninvasive management of myocardial infarction, based on the TIMI-IIA study. Califf and associates [14], studying patients who received tPA followed in 90 minutes by coronary arteriography, reported transfusions of 2 units or more of blood or intracranial hemorrhage in 66 of 302 (21.8%) patients. For the noninvasive trials of tPA in myocardial infarction, the frequency was approximately 1% for blood transfusion of 2 units or more or for intracranial hemorrhage; for trials in which a transfusion of 1 unit or more was considered major, the bleeding rate was 1.5% [15-17]. This would result in a risk ratio of 14.5 for major bleeding with an invasive compared with a noninvasive diagnosis. Based on such a risk ratio, the estimated risk would be approximately 1% for major bleeding in patients with acute pulmonary embolism after a noninvasive diagnosis. With such a relative risk for major bleeding, it would always be safer to treat patients on the basis of a noninvasive diagnosis.

Although most studies did not indicate the site of bleeding among patients who received a blood transfusion, the groin was a frequent site [1, 5]. Our estimate of the risk ratio for invasive management of myocardial infarction compared with noninvasive management of myocardial infarction was based on a randomized trial [10]. The groups were similar; the only difference was an invasive procedure. The difference of bleeding with thrombolytic therapy resulted from this invasive procedure. We are confident, therefore, that the rate of bleeding with thrombolytic therapy after an invasive procedure is higher than with a noninvasive procedure. In patients with pulmonary embolism, the bleeding rate caused by recent surgery or underlying pathologic problems, regardless of whether a pulmonary angiogram was done, may be higher than among patients with myocardial infarction. Typically, patients who are at a higher risk for bleeding are readily identified by history and physical examination before therapy [18].

Most investigators who do pulmonary angiography insert the catheter percutaneously through a femoral vein. Perhaps when thrombolytic therapy is used, less bleeding would occur with catheterization of a superficial vein of the arm. Many useful suggestions related to the administration of thrombolytic agents and treatment of bleeding were made in a review of the hemorrhagic complications of thrombolytic therapy by Levine and associates [6].

The risk for bleeding with heparin or warfarin was not considered separately in this analysis. The risk was 4.9% (CI, 3.4% to 6.5%), based on averaged data, for major hemorrhagic complications from heparin among patients treated for thromboembolic disease, whether intermittent intravenous, continuous infusion, or subcutaneous heparin was used in doses ranging from 29 180 U to 40 320 U over 24 hours [19-22]. These risks have been refined according to patients at high and low risk for bleeding [18]. High risk was defined as surgery within the previous 14 days, history of peptic ulcer disease, gastrointestinal or genitourinary tract bleeding, disorders predisposing to bleeding, thrombotic studies within 14 days, or platelet counts less than 150 x 10⁹/L. The frequency of major bleeding from heparin in such high-risk patients was 10.8% (CI, 5.0% to 16.6%) [18]. Among patients at low risk for bleeding, the frequency of major bleeding with heparin was 1.1% (CI, 0% to 3.3%) [18]. The risk was 1.7% (CI, 0.3% to 3.0%), based on averaged data, for major hemorrhagic complications among patients with thromboembolic disease who were treated with "less intense" warfarin (international normalized ratio, 2 to 3) [18, 19, 23].

It would be ideal if, in patients who require thrombolytic therapy, a noninvasive diagnosis of pulmonary embolism could be based on a high-probability ventilation-perfusion lung scan combined with a high-likelihood clinical assessment. In the collaborative study from the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED) group, 96% of such patients had pulmonary embolism [24]. Unfortunately, however, only 11% of patients with pulmonary embolism had this concordance of a high-probability ventilation-perfusion lung scan and a high-likelihood clinical assessment [24]. Among all patients with a high-probability ventilation-perfusion lung scan, regardless of the clinical assessment, pulmonary embolism was present in 87%. Only 41% of patients with pulmonary embolism in the PIOPED study, however, had a high-probability ventilation-perfusion lung scan. Because most patients with pulmonary embolism do not have a high-probability ventilation-perfusion lung scan, the physician is often in the uncomfortable position of assessing the probability of pulmonary embolism based on less definitive data. It may be safer to administer thrombolytic therapy on the basis of a noninvasive diagnosis despite some uncertainty in the diagnosis.

The probability of pulmonary embolism can be assessed in several ways. In the PIOPED study [24], the probability of pulmonary embolism was shown on the basis of a clinical likelihood estimate and the ventilation-perfusion lung scan probability as follows: 1) If the ventilation-perfusion lung scan was read as intermediate probability [indeterminate probability] and the clinical assessment was a high likelihood of pulmonary embolism, the frequency of pulmonary embolism was 66%; 2) if the ventilation-perfusion lung scan was low probability with a high-likelihood clinical assessment or if the lung scan was intermediate probability with an intermediate-probability clinical assessment, the frequency of pulmonary embolism was 28% to 40%; 3) if the ventilation-perfusion lung scan was intermediate probability and the clinical assessment was low probability, or vice versa, the frequency of pulmonary embolism was 16%; 4) a ventilation-perfusion lung scan of low probability with a concordant low-likelihood clinical assessment or a nearly normal lung scan with an intermediate- or low-likelihood clinical assessment was associated with pulmonary embolism in 6% or fewer patients.

Based on these probabilities and assuming a 14.0% frequency of major bleeding with tPA after angiography, it would be safer to treat unstable patients on the basis of a noninvasive diagnosis unless they had a ventilation-perfusion lung scan of intermediate probability with a low probability clinical assessment or had a low-probability or nearly normal lung scan with an intermediate- or low-probability clinical assessment.

The probability of pulmonary embolism also can be assessed by the number of ventilation-perfusion scan defects among patients stratified according to previous cardiopulmonary disease and clinical risk assessment [25-27]. Neural network logic may assist in the clinical estimate of probability of pulmonary embolism [28]. In instances where the physician is uncomfortable with his or her assessment of the probability of pulmonary embolism, then he or she must either choose the traditional approach of angiography, with its attendant high risk for bleeding at the site of catheter insertion, or treat on the basis of a noninvasive diagnosis, recognizing that some patients who do not have pulmonary embolism will be treated inadvertently.

We recognize the uncertainty in assessment of the probability of acute pulmonary embolism. Even though physicians may be unable to accurately assess the likelihood of pulmonary embolism, it is useful to know the relative risks for thrombolytic therapy administered on the basis of invasive compared with noninvasive diagnoses. In the future, the calculations in this study may permit a broadening of the indications for thrombolytic therapy in acute pulmonary embolism. This will require prospective testing. Until this is accomplished, we recommend thrombolytic therapy for patients with massive acute pulmonary embolism who are hypotensive, hypoxic when receiving high levels of oxygen, or clinically stable with echocardiographic evidence of right ventricular failure.