Cost-Effectiveness Analysis of Diagnostic Strategies for Suspected Pulmonary Embolism Including Helical Computed Tomography
Medical Clinic 1; Division of Medical and Surgical Emergencies; Division of Angiology and Hemostasis; Department of Internal Medicine, Division of Radiodiagnosis; and Department of Radiology, Geneva University Hospital, Geneva, Switzerland^4?&[a\, 百拇医药
ABSTRACT^4?&[a\, 百拇医药
TOP^4?&[a\, 百拇医药
ABSTRACT^4?&[a\, 百拇医药
INTRODUCTION^4?&[a\, 百拇医药
METHODS^4?&[a\, 百拇医药
RESULTS^4?&[a\, 百拇医药
DISCUSSION^4?&[a\, 百拇医药
REFERENCES^4?&[a\, 百拇医药
We performed a formal decision analysis to evaluate the cost-effectiveness of various strategies for pulmonary embolism, including helical computed tomography (CT), and determined the most cost-effective schemes for each clinical probability of pulmonary embolism. Other tests included D-dimer (DD), lower limb venous ultrasound (US), ventilation–perfusion (/) scan, and angiography. Outcome measures were 3-month survival and costs per patient managed. Baseline sensitivity of CT was 70%, corresponding to the performance of single-detector CT, and that figure was raised in sensitivity analysis to account for the expected higher sensitivity of newer multidetector CT scanners. All strategies were compared with a reference strategy, namely the / scan in all patients followed when nondiagnostic by an angiogram. For low clinical probability patients, the most cost-effective strategy was DD, US, and / scan, patients with a nondiagnostic / scan being left untreated. Replacing / scan by CT was also cost-effective. For intermediate and high clinical probability patients, a fourth test must be added, either CT or angiography in patients with nondiagnostic / scan, or angiography in patients with a negative helical CT. When using sensitivity figures above 85% (in the multidetector range), DD, US, and CT became the most cost-effective strategy for all clinical probability categories. Helical CT as a single test was not cost-effective. In summary, including helical CT in diagnostic strategies for pulmonary embolism is cost-effective provided that it is combined with DD and US. In contrast, helical CT as a single test is not cost-effective.
Key Words: pulmonary embolism • fibrin fibrinogen degradation products • ultrasonography, doppler, duplex • tomography scanners, x-ray computed • cost-effectiveness analysis7x+3&, http://www.100md.com
INTRODUCTION7x+3&, http://www.100md.com
TOP7x+3&, http://www.100md.com
ABSTRACT7x+3&, http://www.100md.com
INTRODUCTION7x+3&, http://www.100md.com
METHODS7x+3&, http://www.100md.com
RESULTS7x+3&, http://www.100md.com
DISCUSSION7x+3&, http://www.100md.com
REFERENCES7x+3&, http://www.100md.com
Pulmonary embolism remains a diagnostic challenge. Until recently, a ventilation–perfusion (/) lung scan was the cornerstone of the diagnostic workup, a normal or near-normal scintigram excluding pulmonary embolism (1, 2), and a high-probability lung scan considered sufficient to establish the disease (3). However, lung scan is diagnostic only in 30 to 40% of patients, and pulmonary angiography, although still the definitive diagnostic standard for pulmonary embolism, is costly and invasive (4). Hence, numerous strategies for pulmonary embolism, including noninvasive instruments, have been proposed (5–7). Clinical assessment allows stratifying patients into three categories corresponding to an increasing prevalence of pulmonary embolism and individualizing the diagnostic workup according to the clinical probability category (1, 3, 5). Lower limb venous compression ultrasonography is noninvasive and highly specific for deep-vein thrombosis (8) and reveals a deep-vein thrombosis in 30 to 50% of patients with pulmonary embolism (9, 10). Plasma D-dimer (DD) measurement is increasingly used to exclude pulmonary embolism, and enzyme-linked immunosorbent assays (11, 12) and some automated turbidimetric assays (13, 14) have a sensitivity of 97 to 99% for acute pulmonary embolism. Finally, helical computed tomography (CT) is already widely used to diagnose pulmonary embolism (15, 16).
Cost-effectiveness analysis is ideally suited to compare schemes combining those diagnostic tools, and several have been published previously (17–20). Although they evaluated numerous strategies, they did not incorporate the patient's clinical probability of pulmonary embolism and adopted optimistic figures for the performance of helical CT (17, 18). Indeed, baseline sensitivity of CT was 89% (18) and 95% (17), whereas it was only 70% in two recent and more methodologically robust series from the literature (21, 22). Therefore, we studied the cost-effectiveness of incorporating single-detector helical CT in the diagnostic approach of pulmonary embolism, either as a single test, or as a substitute for / lung scan or pulmonary angiography in a sequential diagnostic strategy, including DD and ultrasound (US). Moreover, we evaluated the most cost-effective strategy for the three levels of clinical probability of pulmonary embolism. Finally, we assessed whether using multidetector CT, which should have a higher sensitivity, might modify the ranking of the strategies in sensitivity analysis.
METHODS^@\.c, 百拇医药
TOP^@\.c, 百拇医药
ABSTRACT^@\.c, 百拇医药
INTRODUCTION^@\.c, 百拇医药
METHODS^@\.c, 百拇医药
RESULTS^@\.c, 百拇医药
DISCUSSION^@\.c, 百拇医药
REFERENCES^@\.c, 百拇医药
Decision Model^@\.c, 百拇医药
We used a decision analysis software (Decision Analysis by TreeAge, version 3.0.5; TreeAge Software, Inc., Williamston, MA) to create a model representing alternative strategies for diagnosing pulmonary embolism. The decision model is represented in .^@\.c, 百拇医药
fig.ommitted^@\.c, 百拇医药
Figure 1. Structure of the decision tree. Angio = pulmonary angiography; DVT = deep-vein thrombosis; neg = negative; NonDx = nondiagnostic; normal-near N = normal or near normal; PE = pulmonary embolism; pos = positive; Rx = treatment.^@\.c, 百拇医药
Assumptions.^@\.c, 百拇医药
The following assumptions were made:^@\.c, 百拇医药
Patients with a normal DD level (below 500 µg/L as determined by an enzyme-linked immunosorbent assay) were left untreated, whereas those with an abnormal DD level ( 500 µg/L) proceeded to other tests.
Patients in whom lower limb venous US showed a deep-vein thrombosis were treated without further testing, whereas those with a normal US proceeded to further tests.0.s-7v, 百拇医药
In strategies including / lung scan, patients with a high-probability lung scan were treated, and those with a normal or near-normal lung scan were left untreated.0.s-7v, 百拇医药
All tests were conditionally independent.0.s-7v, 百拇医药
To facilitate the results' interpretation, we selected / scan ± angiography (discussed later here) as the reference strategy. Arbitrarily, we considered any strategy with a less than 0.5% difference in 3-month survival compared with the reference strategy of acceptable effectiveness.0.s-7v, 百拇医药
All effective strategies that were cost-saving compared with the reference strategy (/ scan ± angiography) were considered cost-effective.0.s-7v, 百拇医药
Strategies.0.s-7v, 百拇医药
The following strategies were evaluated :
/ scan ± angiography: All patients undergo a lung scan. Patients with a nondiagnostic lung scan undergo pulmonary angiography.ammkq, http://www.100md.com
CT: All patients undergo a helical CT as a single test and are treated according to the CT scan result.ammkq, http://www.100md.com
US ± CT: Lower limb venous US is the initial test. Patients with a normal US undergo a helical CT and are treated according to the CT scan result.ammkq, http://www.100md.com
DD ± US ± CT: DD is the initial test in all patients, followed by US in patients with an abnormal DD level. Those with a normal US undergo a helical CT and are treated according to the CT scan result.ammkq, http://www.100md.com
DD ± US ± CT ± angiography: This strategy is similar to DD ± US ± CT except that patients with a negative CT are submitted to a pulmonary angiogram and are treated according to the angiogram result.ammkq, http://www.100md.com
DD ± US ±/ scan: DD is the initial test in all patients, followed by US in patients with an abnormal DD level. Those with a normal US undergo a / lung scan. Patients with a nondiagnostic / lung scan are left untreated.
DD ± US ± / scan ± angiography: This strategy is similar to the DD ± US ± / scan, but patients with a nondiagnostic /scan undergo a pulmonary angiogram and are treated according to the angiogram result.([, 百拇医药
DD ± US ± / scan ± CT: This strategy is similar to DD ± US ± / scan ± angiography in which helical CT replaces pulmonary angiography.([, 百拇医药
Because patients are managed differently according to the clinical probability of pulmonary embolism in most recent algorithms (1, 5, 23), all analyses were performed for three levels of prevalence of pulmonary embolism corresponding to a low, intermediate, and high clinical probability of pulmonary embolism.([, 百拇医药
Summary of Data Used for Analysis([, 百拇医药
and summarize the baseline probabilities, utilities and costs used in the analysis, and the range of values tested in sensitivity analysis. Those data were extracted from a literature search of the Medline database completed by a manual search of the references cited in the retrieved articles. The data are presented and discussed in more detail in the online data supplement.
fig.ommittedof*, http://www.100md.com
TABLE 1. Test characteristics: values (baseline and range used in sensitivity analyses) used in the decision modelof*, http://www.100md.com
fig.ommittedof*, http://www.100md.com
TABLE 2. Outcomes: values (baseline and range used in sensitivity analyses) used in the decision modelof*, http://www.100md.com
Outcome Measuresof*, http://www.100md.com
We selected the 3-month quality-adjusted expected survival as the main outcome measure for effectiveness. For example, untreated patients free of pulmonary embolism had a 100% 3-month expected survival, whereas those with treated pulmonary embolism had an expected survival of 100% minus the mortality associated with treated pulmonary embolism (8%) minus that of a 3-month course of anticoagulants (0.2%), that is, 91.8%. A quality-adjustment factor used in the previously published literature was used to take into account the consequences of permanent sequelae after a major bleed (24) (). The strategy / scan ± angiography, in which all patients with a nondiagnostic lung scan undergo an angiogram (3), was selected as the reference strategy in terms of effectiveness. The proportion of angiograms required in each strategy was also calculated. Sensitivity analyses were performed to test the stability of the results over a wide range of clinically relevant values ( and ).
Costsdg$d, http://www.100md.com
We considered the direct costs, that is, those involving the health care system. Indirect costs such as the loss of earnings were not considered. Cost data were actual costs, not charges, extracted from the database of our hospital, a 1,300-bed primary to tertiary-care urban facility, during the fiscal year 1996. All of these costs are expressed in United States dollars. Mean costs of treatment per patient with pulmonary embolism in our institution are $5,982 and include costs of tests, hospital stay, inpatient and outpatient treatment and monitoring, and costs of major bleeding complications. The costs of individual tests vary widely between countries, and costs are fixed values in each location. Therefore, rather than evaluating random cost variations in sensitivity analysis, we calculated the expenses associated with each strategy for three sets of costs: those from our center and recent cost estimates from Canada (18) and the United States (20) () .
fig.ommitted9pa.q;, 百拇医药
TABLE 3. Costs of tests and treatment in various locations (in usd)9pa.q;, 百拇医药
RESULTS9pa.q;, 百拇医药
TOP9pa.q;, 百拇医药
ABSTRACT9pa.q;, 百拇医药
INTRODUCTION9pa.q;, 百拇医药
METHODS9pa.q;, 百拇医药
RESULTS9pa.q;, 百拇医药
DISCUSSION9pa.q;, 百拇医药
REFERENCES9pa.q;, 百拇医药
Baseline Analyses9pa.q;, 百拇医药
Results are presented independently for each clinical probability of pulmonary embolism (low, intermediate, and high) () .9pa.q;, 百拇医药
fig.ommitted9pa.q;, 百拇医药
Figure 2. Three-month survival and average costs per patient yielded by diagnostic strategies for patients with low clinical probability (A), intermediate clinical probability (B), and high clinical probability (C) of pulmonary embolism. The solid line corresponds to the effectiveness of the reference strategy (/ scan ± angiography), and the dotted line represents an arbitrary 0.5% difference in 3-month survival with the reference strategy within which a strategy is considered as acceptable. The data used to plot the graphs are available on the online data supplement.
Low clinical probability of pulmonary embolism.m, http://www.100md.com
In patients with a low clinical probability (10%), shows that all strategies were reasonably effective, considering that the 3-month survival remained within a range of 0.5% mortality (see ASSUMPTIONS) compared with the reference strategy (/ scan ± angiography, $1,728). The cheapest strategy was DD ± US ± / scan ($845), followed by DD ± US ± CT ($1,230). All other strategies required the performance of a fourth test in the sequence (CT or angiography) without yielding a significant survival benefit. CT as a single test was the least effective strategy (0.34% difference in 3-month survival with the reference strategy) and was less cost saving ($1,503). In summary, for patients with a low clinical probability of pulmonary embolism, the most cost-effective strategy was DD ± US ± / scan, followed by DD ± US ± CT.m, http://www.100md.com
Intermediate clinical probability of pulmonary embolism.m, http://www.100md.com
When clinical probability of pulmonary embolism was intermediate (), the most cost-effective strategy was DD ± US ± / scan ± CT, in which a nondiagnostic / scan is followed by helical CT ($2,674). That strategy allowed us to forego pulmonary angiography in all patients. DD ± US ± / scan ± angiography was also cost saving compared with the reference strategy ($2,832 versus $3,164, respectively) and required an angiogram in 29% of patients. Of note, DD ± US ± CT ± angiography ($3,214) was not cost saving compared with the reference strategy and required an angiogram in 35% of patients. All other strategies were associated with a higher 3-month mortality in the intermediate clinical probability subgroup. CT as a single test was clearly associated with the highest mortality. In summary, the most cost-effective strategies for intermediate clinical probability patients were DD ± US ± / scan ± CT, followed closely by DD ± US ± / scan ± angiography. DD ± US ± CT ± angiography was not cost saving.
High clinical probability of pulmonary embolism.anh@y, http://www.100md.com
In patients with a high clinical probability of pulmonary embolism (), DD ± US ± / scan ± CT remained the cheapest strategy ($4,308) but was of marginal effectiveness (difference in 3-month survival with the reference strategy, 0.47%). DD ± US ± / scan ± angiography, requiring an angiogram in 25% of patients, was also cost-effective ($4,598 versus $4,866 for the reference strategy). DD ± US ± CT ± angiography was effective, but it was as expensive ($4,867) as the reference strategy while requiring an angiogram in 25% of patients.anh@y, http://www.100md.com
Sensitivity Analysesanh@y, http://www.100md.com
We performed sensitivity analyses on all variables of the model for each clinical probability of pulmonary embolism (low, intermediate, and high). Overall, the ranking of strategies was not affected by any variation within the range of values described in . However, some of the variables had an impact on the strategies that were marginally effective in the baseline analysis.
Diagnostic performance of helical CT.(%?eu'^, 百拇医药
Lowering the sensitivity of CT to 62%, the lower value of the 95% confidence interval found in our previous study using single-detector CT (21), did not affect the ranking of strategies in the low clinical probability of pulmonary embolism category. Nevertheless, to account for the expected higher sensitivity of multidetector CT, we also performed the analysis using higher sensitivity values for helical CT. Angiography was no longer necessary after CT and US (DD ± US ± CT strategy) for a CT sensitivity above 76% in intermediate clinical probability patients and above 85% in the high clinical probability subgroup. The corresponding figures for the CT alone strategy were 87% and 92%. shows these values in intermediate clinical probability patients. A lower specificity of CT had little impact on effectiveness but entailed higher costs due to a higher proportion of patients treated unnecessarily, whatever the sensitivity of helical CT.
fig.ommitted+/, 百拇医药
Figure 3. Influence of the sensitivity of helical CT on the 3-month mortality in patients with an intermediate clinical probability of pulmonary embolism. Results are expressed as the difference in mortality with the reference strategy (/ scan ± angiography). The dotted line represents an arbitrary 0.5% difference in 3-month survival with the reference strategy (/ scan ± angiography) within which a strategy is considered acceptable.+/, 百拇医药
Prevalence of pulmonary embolism in clinical probability categories.+/, 百拇医药
Although the CT alone strategy seemed acceptable when the prevalence of pulmonary embolism in the low clinical probability subgroup was 10% (effectiveness 98.7%), its effectiveness dropped to 98.0% (0.57% less than the reference strategy) when that prevalence reached 15%. In patients with an intermediate clinical probability of pulmonary embolism, US ± CT and DD ± US ± CT, which were significantly less effective in the baseline analysis, became marginally acceptable (difference in effectiveness 0.40% with reference strategy) when the prevalence of pulmonary embolism in that category dropped to the lower limit of 25%. Finally, in the high clinical probability category, the DD ± US ± / scan ± CT strategy, which was only marginally effective in the baseline analysis, became associated with a more than 0.5% 3-month mortality difference with the reference strategy when the prevalence of pulmonary embolism became 70% or higher in that subgroup.+/, 百拇医药
Diagnostic performance of scan.+/, 百拇医药
The analysis was performed using the Canadian (1) and the Prospective Investigation On Pulmonary Embolism Diagnosis (PIOPED) (3) values of the performance of(Arnaud Perrier, Mathieu R. Nendaz, François P. Sarasin, Nigel Howarth and Henri Bounameaux)
ABSTRACT^4?&[a\, 百拇医药
TOP^4?&[a\, 百拇医药
ABSTRACT^4?&[a\, 百拇医药
INTRODUCTION^4?&[a\, 百拇医药
METHODS^4?&[a\, 百拇医药
RESULTS^4?&[a\, 百拇医药
DISCUSSION^4?&[a\, 百拇医药
REFERENCES^4?&[a\, 百拇医药
We performed a formal decision analysis to evaluate the cost-effectiveness of various strategies for pulmonary embolism, including helical computed tomography (CT), and determined the most cost-effective schemes for each clinical probability of pulmonary embolism. Other tests included D-dimer (DD), lower limb venous ultrasound (US), ventilation–perfusion (/) scan, and angiography. Outcome measures were 3-month survival and costs per patient managed. Baseline sensitivity of CT was 70%, corresponding to the performance of single-detector CT, and that figure was raised in sensitivity analysis to account for the expected higher sensitivity of newer multidetector CT scanners. All strategies were compared with a reference strategy, namely the / scan in all patients followed when nondiagnostic by an angiogram. For low clinical probability patients, the most cost-effective strategy was DD, US, and / scan, patients with a nondiagnostic / scan being left untreated. Replacing / scan by CT was also cost-effective. For intermediate and high clinical probability patients, a fourth test must be added, either CT or angiography in patients with nondiagnostic / scan, or angiography in patients with a negative helical CT. When using sensitivity figures above 85% (in the multidetector range), DD, US, and CT became the most cost-effective strategy for all clinical probability categories. Helical CT as a single test was not cost-effective. In summary, including helical CT in diagnostic strategies for pulmonary embolism is cost-effective provided that it is combined with DD and US. In contrast, helical CT as a single test is not cost-effective.
Key Words: pulmonary embolism • fibrin fibrinogen degradation products • ultrasonography, doppler, duplex • tomography scanners, x-ray computed • cost-effectiveness analysis7x+3&, http://www.100md.com
INTRODUCTION7x+3&, http://www.100md.com
TOP7x+3&, http://www.100md.com
ABSTRACT7x+3&, http://www.100md.com
INTRODUCTION7x+3&, http://www.100md.com
METHODS7x+3&, http://www.100md.com
RESULTS7x+3&, http://www.100md.com
DISCUSSION7x+3&, http://www.100md.com
REFERENCES7x+3&, http://www.100md.com
Pulmonary embolism remains a diagnostic challenge. Until recently, a ventilation–perfusion (/) lung scan was the cornerstone of the diagnostic workup, a normal or near-normal scintigram excluding pulmonary embolism (1, 2), and a high-probability lung scan considered sufficient to establish the disease (3). However, lung scan is diagnostic only in 30 to 40% of patients, and pulmonary angiography, although still the definitive diagnostic standard for pulmonary embolism, is costly and invasive (4). Hence, numerous strategies for pulmonary embolism, including noninvasive instruments, have been proposed (5–7). Clinical assessment allows stratifying patients into three categories corresponding to an increasing prevalence of pulmonary embolism and individualizing the diagnostic workup according to the clinical probability category (1, 3, 5). Lower limb venous compression ultrasonography is noninvasive and highly specific for deep-vein thrombosis (8) and reveals a deep-vein thrombosis in 30 to 50% of patients with pulmonary embolism (9, 10). Plasma D-dimer (DD) measurement is increasingly used to exclude pulmonary embolism, and enzyme-linked immunosorbent assays (11, 12) and some automated turbidimetric assays (13, 14) have a sensitivity of 97 to 99% for acute pulmonary embolism. Finally, helical computed tomography (CT) is already widely used to diagnose pulmonary embolism (15, 16).
Cost-effectiveness analysis is ideally suited to compare schemes combining those diagnostic tools, and several have been published previously (17–20). Although they evaluated numerous strategies, they did not incorporate the patient's clinical probability of pulmonary embolism and adopted optimistic figures for the performance of helical CT (17, 18). Indeed, baseline sensitivity of CT was 89% (18) and 95% (17), whereas it was only 70% in two recent and more methodologically robust series from the literature (21, 22). Therefore, we studied the cost-effectiveness of incorporating single-detector helical CT in the diagnostic approach of pulmonary embolism, either as a single test, or as a substitute for / lung scan or pulmonary angiography in a sequential diagnostic strategy, including DD and ultrasound (US). Moreover, we evaluated the most cost-effective strategy for the three levels of clinical probability of pulmonary embolism. Finally, we assessed whether using multidetector CT, which should have a higher sensitivity, might modify the ranking of the strategies in sensitivity analysis.
METHODS^@\.c, 百拇医药
TOP^@\.c, 百拇医药
ABSTRACT^@\.c, 百拇医药
INTRODUCTION^@\.c, 百拇医药
METHODS^@\.c, 百拇医药
RESULTS^@\.c, 百拇医药
DISCUSSION^@\.c, 百拇医药
REFERENCES^@\.c, 百拇医药
Decision Model^@\.c, 百拇医药
We used a decision analysis software (Decision Analysis by TreeAge, version 3.0.5; TreeAge Software, Inc., Williamston, MA) to create a model representing alternative strategies for diagnosing pulmonary embolism. The decision model is represented in .^@\.c, 百拇医药
fig.ommitted^@\.c, 百拇医药
Figure 1. Structure of the decision tree. Angio = pulmonary angiography; DVT = deep-vein thrombosis; neg = negative; NonDx = nondiagnostic; normal-near N = normal or near normal; PE = pulmonary embolism; pos = positive; Rx = treatment.^@\.c, 百拇医药
Assumptions.^@\.c, 百拇医药
The following assumptions were made:^@\.c, 百拇医药
Patients with a normal DD level (below 500 µg/L as determined by an enzyme-linked immunosorbent assay) were left untreated, whereas those with an abnormal DD level ( 500 µg/L) proceeded to other tests.
Patients in whom lower limb venous US showed a deep-vein thrombosis were treated without further testing, whereas those with a normal US proceeded to further tests.0.s-7v, 百拇医药
In strategies including / lung scan, patients with a high-probability lung scan were treated, and those with a normal or near-normal lung scan were left untreated.0.s-7v, 百拇医药
All tests were conditionally independent.0.s-7v, 百拇医药
To facilitate the results' interpretation, we selected / scan ± angiography (discussed later here) as the reference strategy. Arbitrarily, we considered any strategy with a less than 0.5% difference in 3-month survival compared with the reference strategy of acceptable effectiveness.0.s-7v, 百拇医药
All effective strategies that were cost-saving compared with the reference strategy (/ scan ± angiography) were considered cost-effective.0.s-7v, 百拇医药
Strategies.0.s-7v, 百拇医药
The following strategies were evaluated :
/ scan ± angiography: All patients undergo a lung scan. Patients with a nondiagnostic lung scan undergo pulmonary angiography.ammkq, http://www.100md.com
CT: All patients undergo a helical CT as a single test and are treated according to the CT scan result.ammkq, http://www.100md.com
US ± CT: Lower limb venous US is the initial test. Patients with a normal US undergo a helical CT and are treated according to the CT scan result.ammkq, http://www.100md.com
DD ± US ± CT: DD is the initial test in all patients, followed by US in patients with an abnormal DD level. Those with a normal US undergo a helical CT and are treated according to the CT scan result.ammkq, http://www.100md.com
DD ± US ± CT ± angiography: This strategy is similar to DD ± US ± CT except that patients with a negative CT are submitted to a pulmonary angiogram and are treated according to the angiogram result.ammkq, http://www.100md.com
DD ± US ±/ scan: DD is the initial test in all patients, followed by US in patients with an abnormal DD level. Those with a normal US undergo a / lung scan. Patients with a nondiagnostic / lung scan are left untreated.
DD ± US ± / scan ± angiography: This strategy is similar to the DD ± US ± / scan, but patients with a nondiagnostic /scan undergo a pulmonary angiogram and are treated according to the angiogram result.([, 百拇医药
DD ± US ± / scan ± CT: This strategy is similar to DD ± US ± / scan ± angiography in which helical CT replaces pulmonary angiography.([, 百拇医药
Because patients are managed differently according to the clinical probability of pulmonary embolism in most recent algorithms (1, 5, 23), all analyses were performed for three levels of prevalence of pulmonary embolism corresponding to a low, intermediate, and high clinical probability of pulmonary embolism.([, 百拇医药
Summary of Data Used for Analysis([, 百拇医药
and summarize the baseline probabilities, utilities and costs used in the analysis, and the range of values tested in sensitivity analysis. Those data were extracted from a literature search of the Medline database completed by a manual search of the references cited in the retrieved articles. The data are presented and discussed in more detail in the online data supplement.
fig.ommittedof*, http://www.100md.com
TABLE 1. Test characteristics: values (baseline and range used in sensitivity analyses) used in the decision modelof*, http://www.100md.com
fig.ommittedof*, http://www.100md.com
TABLE 2. Outcomes: values (baseline and range used in sensitivity analyses) used in the decision modelof*, http://www.100md.com
Outcome Measuresof*, http://www.100md.com
We selected the 3-month quality-adjusted expected survival as the main outcome measure for effectiveness. For example, untreated patients free of pulmonary embolism had a 100% 3-month expected survival, whereas those with treated pulmonary embolism had an expected survival of 100% minus the mortality associated with treated pulmonary embolism (8%) minus that of a 3-month course of anticoagulants (0.2%), that is, 91.8%. A quality-adjustment factor used in the previously published literature was used to take into account the consequences of permanent sequelae after a major bleed (24) (). The strategy / scan ± angiography, in which all patients with a nondiagnostic lung scan undergo an angiogram (3), was selected as the reference strategy in terms of effectiveness. The proportion of angiograms required in each strategy was also calculated. Sensitivity analyses were performed to test the stability of the results over a wide range of clinically relevant values ( and ).
Costsdg$d, http://www.100md.com
We considered the direct costs, that is, those involving the health care system. Indirect costs such as the loss of earnings were not considered. Cost data were actual costs, not charges, extracted from the database of our hospital, a 1,300-bed primary to tertiary-care urban facility, during the fiscal year 1996. All of these costs are expressed in United States dollars. Mean costs of treatment per patient with pulmonary embolism in our institution are $5,982 and include costs of tests, hospital stay, inpatient and outpatient treatment and monitoring, and costs of major bleeding complications. The costs of individual tests vary widely between countries, and costs are fixed values in each location. Therefore, rather than evaluating random cost variations in sensitivity analysis, we calculated the expenses associated with each strategy for three sets of costs: those from our center and recent cost estimates from Canada (18) and the United States (20) () .
fig.ommitted9pa.q;, 百拇医药
TABLE 3. Costs of tests and treatment in various locations (in usd)9pa.q;, 百拇医药
RESULTS9pa.q;, 百拇医药
TOP9pa.q;, 百拇医药
ABSTRACT9pa.q;, 百拇医药
INTRODUCTION9pa.q;, 百拇医药
METHODS9pa.q;, 百拇医药
RESULTS9pa.q;, 百拇医药
DISCUSSION9pa.q;, 百拇医药
REFERENCES9pa.q;, 百拇医药
Baseline Analyses9pa.q;, 百拇医药
Results are presented independently for each clinical probability of pulmonary embolism (low, intermediate, and high) () .9pa.q;, 百拇医药
fig.ommitted9pa.q;, 百拇医药
Figure 2. Three-month survival and average costs per patient yielded by diagnostic strategies for patients with low clinical probability (A), intermediate clinical probability (B), and high clinical probability (C) of pulmonary embolism. The solid line corresponds to the effectiveness of the reference strategy (/ scan ± angiography), and the dotted line represents an arbitrary 0.5% difference in 3-month survival with the reference strategy within which a strategy is considered as acceptable. The data used to plot the graphs are available on the online data supplement.
Low clinical probability of pulmonary embolism.m, http://www.100md.com
In patients with a low clinical probability (10%), shows that all strategies were reasonably effective, considering that the 3-month survival remained within a range of 0.5% mortality (see ASSUMPTIONS) compared with the reference strategy (/ scan ± angiography, $1,728). The cheapest strategy was DD ± US ± / scan ($845), followed by DD ± US ± CT ($1,230). All other strategies required the performance of a fourth test in the sequence (CT or angiography) without yielding a significant survival benefit. CT as a single test was the least effective strategy (0.34% difference in 3-month survival with the reference strategy) and was less cost saving ($1,503). In summary, for patients with a low clinical probability of pulmonary embolism, the most cost-effective strategy was DD ± US ± / scan, followed by DD ± US ± CT.m, http://www.100md.com
Intermediate clinical probability of pulmonary embolism.m, http://www.100md.com
When clinical probability of pulmonary embolism was intermediate (), the most cost-effective strategy was DD ± US ± / scan ± CT, in which a nondiagnostic / scan is followed by helical CT ($2,674). That strategy allowed us to forego pulmonary angiography in all patients. DD ± US ± / scan ± angiography was also cost saving compared with the reference strategy ($2,832 versus $3,164, respectively) and required an angiogram in 29% of patients. Of note, DD ± US ± CT ± angiography ($3,214) was not cost saving compared with the reference strategy and required an angiogram in 35% of patients. All other strategies were associated with a higher 3-month mortality in the intermediate clinical probability subgroup. CT as a single test was clearly associated with the highest mortality. In summary, the most cost-effective strategies for intermediate clinical probability patients were DD ± US ± / scan ± CT, followed closely by DD ± US ± / scan ± angiography. DD ± US ± CT ± angiography was not cost saving.
High clinical probability of pulmonary embolism.anh@y, http://www.100md.com
In patients with a high clinical probability of pulmonary embolism (), DD ± US ± / scan ± CT remained the cheapest strategy ($4,308) but was of marginal effectiveness (difference in 3-month survival with the reference strategy, 0.47%). DD ± US ± / scan ± angiography, requiring an angiogram in 25% of patients, was also cost-effective ($4,598 versus $4,866 for the reference strategy). DD ± US ± CT ± angiography was effective, but it was as expensive ($4,867) as the reference strategy while requiring an angiogram in 25% of patients.anh@y, http://www.100md.com
Sensitivity Analysesanh@y, http://www.100md.com
We performed sensitivity analyses on all variables of the model for each clinical probability of pulmonary embolism (low, intermediate, and high). Overall, the ranking of strategies was not affected by any variation within the range of values described in . However, some of the variables had an impact on the strategies that were marginally effective in the baseline analysis.
Diagnostic performance of helical CT.(%?eu'^, 百拇医药
Lowering the sensitivity of CT to 62%, the lower value of the 95% confidence interval found in our previous study using single-detector CT (21), did not affect the ranking of strategies in the low clinical probability of pulmonary embolism category. Nevertheless, to account for the expected higher sensitivity of multidetector CT, we also performed the analysis using higher sensitivity values for helical CT. Angiography was no longer necessary after CT and US (DD ± US ± CT strategy) for a CT sensitivity above 76% in intermediate clinical probability patients and above 85% in the high clinical probability subgroup. The corresponding figures for the CT alone strategy were 87% and 92%. shows these values in intermediate clinical probability patients. A lower specificity of CT had little impact on effectiveness but entailed higher costs due to a higher proportion of patients treated unnecessarily, whatever the sensitivity of helical CT.
fig.ommitted+/, 百拇医药
Figure 3. Influence of the sensitivity of helical CT on the 3-month mortality in patients with an intermediate clinical probability of pulmonary embolism. Results are expressed as the difference in mortality with the reference strategy (/ scan ± angiography). The dotted line represents an arbitrary 0.5% difference in 3-month survival with the reference strategy (/ scan ± angiography) within which a strategy is considered acceptable.+/, 百拇医药
Prevalence of pulmonary embolism in clinical probability categories.+/, 百拇医药
Although the CT alone strategy seemed acceptable when the prevalence of pulmonary embolism in the low clinical probability subgroup was 10% (effectiveness 98.7%), its effectiveness dropped to 98.0% (0.57% less than the reference strategy) when that prevalence reached 15%. In patients with an intermediate clinical probability of pulmonary embolism, US ± CT and DD ± US ± CT, which were significantly less effective in the baseline analysis, became marginally acceptable (difference in effectiveness 0.40% with reference strategy) when the prevalence of pulmonary embolism in that category dropped to the lower limit of 25%. Finally, in the high clinical probability category, the DD ± US ± / scan ± CT strategy, which was only marginally effective in the baseline analysis, became associated with a more than 0.5% 3-month mortality difference with the reference strategy when the prevalence of pulmonary embolism became 70% or higher in that subgroup.+/, 百拇医药
Diagnostic performance of scan.+/, 百拇医药
The analysis was performed using the Canadian (1) and the Prospective Investigation On Pulmonary Embolism Diagnosis (PIOPED) (3) values of the performance of(Arnaud Perrier, Mathieu R. Nendaz, François P. Sarasin, Nigel Howarth and Henri Bounameaux)