| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From the Division of Cardiology (Dr. Paterson), McGill University Health Centre, and Respiratory Epidemiology Unit (Dr. Schwartzman), McGill University, Montreal, Quebec, Canada.
Correspondence to: Kevin Schwartzman, MD, MPH, Respiratory Epidemiology Unit, McGill University, 1110 Pine University, Montreal, Quebec, Canada H3A 1A3: e-mail: KEVINS{at}MEAKINS.LAN.McGILL.CA
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterials and MethodsResultsDiscussionConclusionReferences |
|---|
Design: Computer-based cost-effectiveness analysis.
Patients: Simulated cohort of 1,000 patients with suspected acute pulmonary embolism (PE), with a prevalence of 28.4%, as in the Prospective Investigation of Pulmonary Embolism Diagnosis study.
Interventions: Using a decision-analysis model, seven
diagnostic strategies were compared, which incorporated combinations of
ventilation-perfusion (
/
) scans, duplex ultrasound of the
legs, spiral CT, and conventional pulmonary angiography.
Measurements and results: Expected survival and cost (in
Canadian dollars) at 3 months were estimated. Four of the strategies
yielded poorer survival at higher cost. The three remaining strategies
were as follows: (1) / ± leg ultrasound ±
spiral CT, with an expected survival of 953.4 per 1,000 patients and a
cost of $1,391 per patient; (2) / ± leg
ultrasound ± pulmonary angiography (the "traditional"
algorithm), with an expected survival of 953.7 per 1,000 patients and a
cost of $1,416 per patient; and (3) spiral CT ± leg ultrasound,
with an expected survival of 958.2 per 1,000 patients and a cost of
$1,751 per patient. The traditional algorithm was then excluded by
extended dominance. The cost per additional life saved was $70,833 for
spiral CT ± leg ultrasound relative to / ±
leg ultrasound ± spiral CT.
Conclusions:
Spiral CT can replace pulmonary angiography in patients with
nondiagnostic / scan and negative leg ultrasound
findings. This approach is likely as effective as—and possibly less
expensive than—the current algorithm for diagnosis of acute PE. When
spiral CT is the initial diagnostic test, followed by leg ultrasound,
expected survival improves but costs are also considerably higher.
These findings were robust to variations in the assumed sensitivity and
specificity of spiral CT.
Key Words: cost-effectiveness • CT • pulmonary embolism
|
Introduction |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionReferences |
|---|
The ventilation-perfusion (/) lung scan has been
considered the first-line investigation for suspected PE. A
"high-probability" scan finding, especially in a suggestive
clinical setting, is strongly associated with angiographically proven
PE. A "normal" scan finding virtually excludes the diagnosis.
However, most / scans fall into neither category and are
therefore nondiagnostic; in the Prospective Investigation of Pulmonary
Embolism Diagnosis (PIOPED) study,3
57.1% of scans in
patients with PE and 78.3% of scans in patients without PE were
nondiagnostic. As a result, patients with / scan findings
judged "intermediate" or "low" probability for PE have been
further investigated via noninvasive lower-extremity imaging and in
some cases pulmonary angiography, depending on the degree of clinical
suspicion for PE.4
5
Limitations of this approach include:
(1) the relatively low detection rate for above-knee deep venous
thrombosis (DVT) in the context of acute PE, and (2) reluctance to
refer patients for angiography because of perceived risks, despite its
status as the "gold standard."6
In 1992, Remy-Jardin and colleagues7 published the first prospective randomized trial using spiral CT angiography to diagnose acute PE. Using conventional pulmonary angiography as the "gold standard," they reported a sensitivity of 100% and a specificity of 96% for centrally located PE (second- to fourth-order branches of the pulmonary artery). Later studies8 9 10 11 12 have included all emboli, including clots lodged in fifth-order branches and beyond, and have documented slightly lower diagnostic accuracies. Despite the fact that this imaging technique is rapidly gaining acceptance as a diagnostic modality for PE, it is unclear how it relates to the earlier approach described above. We used decision-analysis techniques to model several possible strategies for the use of spiral CT in the diagnosis of PE, so as to evaluate its impact on expected survival and anticipated costs. To address the uncertainty surrounding the exact sensitivity and specificity of spiral CT, we modeled the impact of changes in these parameters (sensitivity analysis).
|
Materials and Methods |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionReferences |
|---|
/
scan followed by duplex ultrasound of the legs and possibly pulmonary
angiography (Fig 1
). The other six strategies involved use of spiral CT at various points
in the workup. Each strategy also considered the physician’s clinical
judgment when deciding how far to proceed with testing.
|
A time frame of 3 months was used, corresponding to the usual duration of follow-up in prospective clinical studies of PE and its diagnosis. Treatment for PE involved 10 days of unfractionated heparin therapy in the hospital, followed by 3 months of outpatient oral warfarin treatment. The perspective of the analysis was that of the third-party payer for health services, which in Canada is the provincial government.
Strategies
Our traditional algorithm involved minor modifications to the
approach taken by Wells et al13
in their investigation of
1,239 patients with suspected PE (Fig 1)
. To date, this is the only
prospective cohort study to evaluate the safety and efficacy of
/ scan, leg ultrasound, and conventional pulmonary
angiography as guided by bedside assessment in the diagnosis of acute
PE. In our analysis, we categorized / scan results
according to the PIOPED study criteria (normal, low, intermediate, and
high probability) and not the criteria of Hull et
al14 (normal, nonhigh, and high probability). Most
physicians are familiar with the PIOPED study classification, and it
allows identification of an important subgroup of patients for which
the diagnosis of PE can be safely excluded without further testing:
low-probability scan with a low clinical suspicion. Indeed, in the
study by Wells et al,13
none of the patients in this
subgroup had PE. Thus, in our model, serial leg ultrasound was not
performed in these patients. Another modification was the elimination
of leg ultrasound for patients with normal / scan
findings. The authors conceded that most physicians accept that normal
or near-normal scan findings exclude PE. The final change was the
removal of leg venography from our algorithm. This test is invasive,
and the authors concluded that it should be eliminated from their model
since it failed to reduce the need for angiography. This traditional
algorithm as well as the six other strategies involving CT are detailed
in Figure 1
.
Assumptions
Disease Prevalence:
We assumed that the prevalence of PE was
28.4% among patients whom the diagnosis was initially suspected, as in
the PIOPED study.3
Diagnostic Accuracies: The sensitivity and specificity of infused spiral CT for PE were obtained by pooling the data from five studies (599 scans).8 9 10 11 12 In each study, blinded readers prospectively compared spiral CT scans with conventional pulmonary angiograms. The likelihood of normal-, low-, intermediate-, and high-probability scan findings in the presence or absence of angiographically proven PE was obtained from the PIOPED study.
The estimated diagnostic accuracies for noninvasive testing of the lower extremities are shown in Table 1 . The probability of diagnosing DVT in patients with acute PE was derived from four studies11 15 16 17 (322 patients in total), which used impedance plethysmography (IPG) or ultrasonography to assess for proximal leg thrombosis in patients with angiographically proven PE. The overall probability of detecting DVT by IPG or leg ultrasound was 38.5% in this context. These figures are for a single examination of the femoral and popliteal veins and relate to occult proximal DVT.
|
/ scans. They detected DVT in 16 of 28 patients who were
ultimately found to have recurrent thromboembolic disease during
follow-up (sensitivity 57%). This value was the lower limit of the
range used in our sensitivity analysis. Because it is the "gold standard," it is difficult to estimate the sensitivity and the specificity of selective pulmonary angiography for PE. The probabilities of false-negative and false-positive angiograms were based on the PIOPED study and reflected (1) interobserver differences in angiogram interpretation, and (2) cases where PE was identified during follow-up despite negative pulmonary angiographic findings.19
Clinical Outcomes: The primary outcome evaluated was survival at 3 months after initial evaluation for PE. There are limited data documenting survival in untreated acute PE; retrospective studies1 20 21 22 suggest the mortality of untreated PE to be as high as 31%. Three prospective studies14 23 24 suggest that with anticoagulant therapy, all-cause mortality at 3 months in patients with diagnosed PE is 6.5%. However, patients with suspected PE in whom the diagnosis is ultimately excluded have a higher mortality rate than age-matched "healthy" control subjects. We used the 3-month mortality rate of 3.0% observed by Hull and colleagues14 among patients with suspected PE found to have normal scan findings.
Death from angiography is uncommon and is usually associated with underlying cardiac or respiratory compromise. During the PIOPED study, it occurred in 5 of 1,111 patients (0.5%) who underwent selective pulmonary angiography.19 Death due to spiral CT is even rarer and is related to an anaphylactic-type reaction.25 Bleeding deaths from anticoagulant therapy are similarly infrequent (0.49%).23 24 26 27 28 Major, nonfatal hemorrhage (requiring hospital admission and/or transfusion) usually affects the GI tract (7.4%)26 27 28 29 30 31 32 and less commonly the brain (0.36%).26 27 29 31 32 33 34 The hemorrhagic complications listed in Table 1 assume 5 to 10 days of unfractionated heparin treatment in the hospital, followed by 3 months of outpatient oral warfarin therapy. Risks of contrast-induced acute renal failure were derived from six studies of patients receiving IV ionic or nonionic contrast for a variety of imaging procedures.19 35 36 37 38 39
Costs: Costs are in 1996 Canadian dollars ($1 [Canadian] = $0.68 [United States]) and were derived from the Royal Victoria Hospital financial information services and from Quebec physician fee schedules (Table 2 ). The costs for each diagnostic test were derived from the sum of technical, professional, and capital costs. Capital costs were depreciated according to the life span of the equipment and its total number of yearly procedures as recommended by Drummond et al.40
|
Contrast-induced renal impairment is usually heralded by an elevation in serum creatinine and rarely requires dialysis. In no published report did affected patients require long-term dialysis. Hence, contrast-induced acute renal failure not requiring dialysis was assumed to require hospitalization for 4 days of observation. The 25.4 days of hospital admission for renal failure requiring dialysis reflects the average inpatient stay at our institution for patients presenting with unspecified acute renal failure.
Survivors of hemorrhagic cerebrovascular accidents were considered to require long-term hospitalization; therefore, the cost of 3 months of institutionalization was incorporated in this context. Bleeding was assumed to occur at the beginning of the 3-month time frame since most bleeding occurs during heparin treatment. For convenience, all major nonfatal, nonintracranial bleeding was considered GI in origin.
All test properties, clinical events and costs were varied using
one-way sensitivity analysis. In addition, two-way sensitivity analysis
was used to assess the impact of simultaneous variation in sensitivity
and specificity of spiral CT on estimated survival, and the impact of
variations in the cost of spiral CT and / scans on the
average total cost.
US Medicare costs for 1998 were obtained from a Midwest university-affiliated health center (Table 3 ). The upper limit of costs used in the sensitivity analyses was established from the US costs, from costs in previous cost-effectiveness analyses,41 42 or from tripling the base cost. The lower limit of each range was derived by taking one half of the base cost.
|
|
Results |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionReferences |
|---|
|
Cost and Cost-effectiveness:
One strategy appeared to be
cheaper than the traditional algorithm: / scan ±
leg ultrasound ± spiral CT. The estimated saving relative to the
traditional algorithm was $25 per patient evaluated. Strategies that
combined spiral CT and ultrasonography were considerably more
expensive. Use of spiral CT and pulmonary angiography also entailed
costs related to procedural complications, eg, $49 per
patient in the spiral CT-alone strategy and $30 per patient in spiral
CT ± leg ultrasound.
Incremental cost-effectiveness analysis identified two dominant
strategies: / scan ± leg ultrasound ± spiral
CT and spiral CT ± leg ultrasound. Four of the other strategies
were associated with equivalent or slightly poorer estimated survival
at higher cost, whereas the traditional algorithm was eliminated by
extended dominance.43
In other words, the additional cost
per life saved for the spiral CT ± leg ultrasound strategy was
less than that for the traditional algorithm, relative to the cheaper
/ ± leg ultrasound ± spiral CT strategy. This
finding is shown numerically in Table 5
and graphically in Figure 2
.
|
|
|
/ scan ± leg ultrasound ±
spiral CT fell to 949.1 and 948.8 per 1,000 patients, respectively. When the assumed mortality related to pulmonary angiography exceeded 0.5%, then adding pulmonary angiography to spiral CT ± leg ultrasound actually worsened expected survival. Varying the risk of death from anticoagulant-associated bleeding did not change the results.
Cost and Cost-effectiveness:
Decreasing the sensitivity of
spiral CT for acute PE reduced the cost of all strategies using spiral
CT, because fewer cases of PE were diagnosed and treated. As the
assumed specificity of spiral CT was increased from 93.1 to 100%, the
costs of all strategies using spiral CT decreased. At a specificity of
96%, the least expensive strategy became the spiral CT-alone strategy
($1,387). Conversely, if the assumed specificity was only 80%, then
the cost of all strategies that included spiral CT as a first-line or
second-line test rose, despite unchanged expected survival.
/ scan ± leg ultrasound ± spiral CT and
/ scan ± leg ultrasound ± pulmonary
angiography then became relatively more cost-effective. With this
reduction in assumed specificity, the cost of the spiral CT-alone
strategy rose from $1,472 to $1,842 per patient evaluated while
survival fell slightly, to 952.3 per 1,000 patients.
Not surprisingly, the cost-effectiveness of /
scan ± leg ultrasound ± spiral CT and /
scan ± leg ultrasound ± pulmonary angiography improved when
the cost of spiral CT was increased relative to other costs.
Conversely, when the assumed cost of the / scan exceeded
90% of the cost of spiral CT, the spiral CT-only strategy became the
cheapest, and replaced / scan ± leg ultrasound
± spiral CT as a dominant strategy. The impact of varying these two
costs is illustrated in Figure 4
.
|
/
scan ± spiral CT and spiral CT alone. Spiral CT alone then became
one of the dominant strategies. The results of the analysis did not
change with variations in the cost of pulmonary angiography, the cost
of hospital admission for acute PE, or the cost of morbidity related to
anticoagulant therapy or to angiography. |
Discussion |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionReferences |
|---|
/ lung scans, (2) noninvasive lower-extremity
imaging, and (3) pulmonary angiography was believed to represent the
most cost-effective approach to investigate suspected
PE.41
Some analyses42
44
suggested that
noninvasive testing should play a more prominent role. The present
analysis suggests that diagnostic strategies that include spiral CT can
confer either a survival or cost advantage when compared with these
earlier approaches.
The potential value of spiral CT for the diagnosis of acute PE was also
addressed by an earlier cost-effectiveness analysis: van Erkel and
colleagues45
estimated survival and cost for 12 strategies
combining spiral CT, / lung scans, leg ultrasound,
conventional angiography, and/or D-dimer assay. They concluded that the
most cost-effective algorithm for PE diagnosis was D-dimer assay
followed by spiral CT if D-dimer assay findings were positive. D-dimer
assays are similar to "abnormal" / scan findings in
that they appear sensitive but not specific for PE. Unfortunately,
quantitative D-dimer assays (eg, enzyme-linked immunosorbent
assay)—which are the most sensitive—are also time-consuming and
costly. Cheaper agglutination assays, which provide qualitative or
semiquantitative results, may be performed rapidly.46
Initial studies of the SimpliRED assay (Agen Biomedical Ltd;
Brisbane, Australia) suggested sensitivities as high as 95 to
100%.46
47
48
49
More recent reports have yielded sensitivity
estimates ranging from 52 to 85%; the largest study reported a
sensitivity of 85% and a specificity of 68% for the diagnosis of
acute PE.50
51
If a D-dimer assay with these properties
were added as the initial step in the workup, followed by either of our
two dominant strategies, then expected survival would drop by
approximately 10 per 1,000 patients for each strategy.
Hence in a revised version of their earlier analysis, van Erkel and
colleagues52
recognized that the specificity of D-dimers
for PE, particularly in patients with comorbid disease, had previously
been overestimated. They concluded that the most cost-effective
approach to diagnosing acute PE was leg ultrasound followed by spiral
CT if the ultrasound finding was negative. This strategy yielded
superior survival at a lower cost relative to the traditional algorithm
of / scan, leg ultrasound, and pulmonary angiography.
However, the investigators did not incorporate clinical (pretest)
probabilities, or consider the costs of complications (ie,
renal failure and bleeding). The cost-effectiveness of strategies
containing spiral CT may also have been overestimated since very high
sensitivity and specificity estimates for spiral CT were used (95.5%
and 97.6%, respectively).
Because of patient selection, the prevalence of PE in the first studies
to report the sensitivity and specificity of CT was higher than that
observed in the PIOPED study (range, 46 to 57%). The cohort reported
by Remy-Jardin and colleagues9
had already undergone
investigations suggesting the diagnosis of pulmonary embolism prior to
their enrollment. Goodman et al8
limited recruitment to
patients with nondiagnostic / scans. In the cohort of van
Rossum et al,11
patients with normal / scan
findings had no further imaging, obviating the need for spiral CT in a
surprisingly high proportion of patients initially suspected to have
pulmonary embolism (69%).
If these previous evaluations of spiral CT were distorted by patient
selection based on previous test results (/ scans or leg
ultrasound), it could be argued that the purported sensitivity of
spiral CT was overestimated or underestimated. However, in the more
recent study by Mayo and colleagues,12
all patients with a
clinical suspicion of PE underwent spiral CT at the outset. The
reported prevalence of PE (33%) and sensitivity and specificity of
spiral CT (87% and 95%, respectively) were similar to the estimates
used in our analysis.
For CT-based diagnostic strategies, patient survival is primarily influenced by the sensitivity of the test. There has been concern about the limited sensitivity of spiral CT for emboli in subsegmental pulmonary branches.8 9 10 11 12 The pooled sensitivity of spiral CT used in this study (88.4%) reflects its ability to detect all pulmonary emboli, based on the use of conventional pulmonary angiography as the "gold standard." The reported proportion of patients with emboli limited to subsegmental arteries at pulmonary angiography has ranged anywhere from 5 to 36%.9 10 53 54 However, the higher values are likely an overestimate, since presumably many instances of central PE in the same patient populations had already been detected by other modalities, obviating the need for angiography.
It has also been suggested that isolated peripheral emboli are less
important clinically and hence a lower detection rate can be
tolerated.9
43
55
Feretti et al56
reported 3
months of follow-up data for a cohort of patients with suspected PE who
underwent / scans ± leg ultrasound ± spiral
CT. In untreated patients where the findings of all three
investigations were "negative," subsequent evidence of
thromboembolism was observed in only 6 of 112 patients (negative
predictive value, 94.6%). In our analysis, the estimated negative
predictive value for this strategy was very similar (96.3%). More
recently, Goodman and colleagues57
followed up 198
patients evaluated for PE who did not receive anticoagulation therapy
because of negative spiral CT scan findings. PE was subsequently
detected in only two patients (negative predictive value, 99.0%), with
no associated deaths.57
However, the negative predictive
value may have been overestimated since several additional patients
with negative CT scan findings were nonetheless receiving
anticoagulation treatment from their treating physicians.
As spiral CT technology and interpretation improve, better visualization of subsegmental clots may be anticipated. In any event, our sensitivity analyses suggest that for all reasonable estimates of the sensitivity of spiral CT for PE, the strategies achieving the highest patient survival are those that include spiral CT (Fig 2) .
In our analysis, all seven strategies evaluated yielded 3-month
survival estimates that clustered together (range, 95.3 to 95.8%), so
relative costs were of paramount importance. The three least expensive
strategies in our model were those that included / scans.
However, it could be argued that these costs were underestimated since
in our US costs and in previous cost-effectiveness analyses, the cost
of / scans was higher relative to spiral CT
scans.41
42
45
58
In both the analysis of van Erkel et
al45
and our US Medicare cost data (Table 3)
, the cost of
a / scan was approximately 135% the cost of a spiral CT
scan, whereas in our base case analysis, it was a mere 50% (Table 2)
.
Consequently, when the price of / scans was increased
proportionately, the spiral CT-alone strategy became the least
expensive and thus emerged as a dominant strategy.
Estimated costs for many procedures and services at our center were
lower than reported elsewhere; our cost estimates were based on
resource use rather than charges to third-party payers. The US Medicare
costs were in fact charges that may not have reflected the true costs
of resources consumed. In our cost-effectiveness analysis, one of the
key parameters proved to be the estimated cost of the /
scan relative to that of the spiral CT scan. The higher the relative
cost of the / scan, the stronger the case for replacing
the traditional algorithm with a spiral CT-based strategy.
Not all benefits derived from using spiral CT angiography were
quantified in this analysis. Unlike / scans and leg
imaging, which are used exclusively to diagnose thromboembolic disease,
the true cost-effectiveness of spiral CT is likely underestimated by
this type of analysis, because spiral CT may detect other conditions
that may be responsible for patients’ symptoms and signs. In one
study, spiral CT led to an alternative diagnosis (eg,
pneumonia) in 18 of the 164 patients investigated.56
By the same token, baseline and follow-up / scans and leg
ultrasounds may have additional uses in patients with thromboembolic
disease: they can document resolution and aid in subsequent
identification of recurrences. However, modeling these additional uses
was beyond the scope of this analysis.
|
Conclusion |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionReferences |
|---|
/ scans and
negative ultrasound findings of the lower extremities. This approach is
associated with equivalent survival and, in our health-care center,
reduced cost relative to the traditional approach. In other health-care
centers, where / scans may be relatively more expensive,
an approach using spiral CT as a single test could be considered. When
used as a first-line test, followed by leg ultrasound, spiral CT may be
associated with a modest survival increase, albeit at a substantial
incremental cost. This last observation is consistent with the only
other published cost-effectiveness analysis that evaluated spiral CT in
this context.46 It will never be feasible to compare all these strategies in a prospective trial because of prohibitive costs and sample size requirements. Ongoing refinement and further clinical evaluation of spiral CT will permit better estimates of its test properties, which can then be incorporated into subsequent cost-effectiveness analyses. However, the information already available suggests that spiral CT can safely replace pulmonary angiography in the workup of suspected PE.
|
Footnotes |
|---|
/ = ventilation/perfusion Dr. Schwartzman is supported by a Chercheur Boursier-Clinicien award from the Fonds de la Recherche en Santé du Québec.
Received for publication June 8, 2000. Accepted for publication December 5, 2000.
|
References |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  Y. Hayashino, M. Goto, Y. Noguchi, and T. Fukui Ventilation-Perfusion Scanning and Helical CT in Suspected Pulmonary Embolism: Meta-Analysis of Diagnostic Performance Radiology, March 1, 2005; 234(3): 740 - 748. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. E. Cox, S. S. Carson, and A. K. Biddle Cost-effectiveness of Ultrasound in Preventing Femoral Venous Catheter-associated Pulmonary Embolism Am. J. Respir. Crit. Care Med., December 15, 2003; 168(12): 1481 - 1487. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Kameyama, C.-l. Huang, D. Liu, T. Okamoto, E. Hayashi, Y. Yamamoto, and H. Yokomise Pulmonary embolism after lung resection: diagnosis and treatment Ann. Thorac. Surg., August 1, 2003; 76(2): 599 - 601. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Perrier, M. R. Nendaz, F. P. Sarasin, N. Howarth, and H. Bounameaux Cost-Effectiveness Analysis of Diagnostic Strategies for Suspected Pulmonary Embolism Including Helical Computed Tomography Am. J. Respir. Crit. Care Med., January 1, 2003; 167(1): 39 - 44. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 O. Dabbagh and C. Kaplan Spiral CT Is Not the Final Answer Chest, June 1, 2002; 121(6): 2083 - 2083. [Full Text] [PDF]
|
|
|
|
 |
|
 M. Pistolesi and M. Miniati Imaging techniques in treatment algorithms of pulmonary embolism Eur. Respir. J., February 1, 2002; 19(35_suppl): 28S - 39s. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Is Spiral CT a Cost-Effective Test for Diagnosing PE? Journal Watch Emergency Medicine, July 19, 2001; 2001(719): 3 - 3. [Full Text]
|
|
| ||||||||||||||||||||||||||
