WNWINNHIIHIHIIWUIJHN!IHIHIIHWIWWI

 

13
09
I HS.

Na)

 

   

W393
l

.‘ 7“
r/CVt-O

5?.3Wiil

This is to certify that the

thesis entitled

Meta-Analysis of ELISA 'D-dimer in the
Diagnosis of Acute Pulmonary EmboliSm

presented by

Michael D. Brown, M.D.

has been accepted towards fulfillment
of the requirements for

MS degree in Epidemiology

Major professor

 

 

 

Date October 31, 2002

 

0-7639 MSU is an Affirmative Action/Equal Opportunity Institution

 

 

 

 

LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINE return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/CIRC/Date0ue.p65—p.15

THE ACCURACY OF THE ELISA D-DIMER TEST IN THE DIAGNOSIS OF
PULMONARY EMBOLISM: A META-ANALYSIS
By

Michael D Brown, MD

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTERS OF SCIENCE
Department of Epidemiology

2002

ABSTRACT
THE ACCURACY OF THE ELISA D-DIMER TEST IN THE DIAGNOSIS OF
PULMONARY EMBOLISM: A META-ANALYSIS
By

Michael D Brown, MD

The objective was to determine the sensitivity and specificity of the enzyme-
linked immunosorbent assay (ELISA) D—dimer test in the diagnosis of pulmonary
embolism (PE) in the adult emergency department population. A search of MEDLTNE,
EMBASE and bibliographies of previous systematic reviews was conducted with no
language restriction. Two reviewers extracted data independently and assessed study
quality based on the patient spectrum and reference standard. The analysis was based on
a summary receiver operating characteristic (SROC) curve and pooled estimates for
sensitivity and specificity using a random-effects model. The search yielded 52
publications. Eleven studies met the inclusion criteria and provided a sample of 2,126
subjects. The SROC curve analysis found significant heterogeneity among the 11 studies.
Subgroup analysis of the 9 studies that used traditional ELISA D-dimer methods yielded
the most valid pooled estimates with a sensitivity of 0.94 (95% CI: 0.88, 0.97), and a
specificity of 0.45 (95% CI: 0.36, 0.55). Advanced age resulted in a lower specificity. A
prolonged duration of symptoms decreased both sensitivity and specificity. The ELISA
D-dimer test is highly sensitive but non-specific for the detection of PE in the acute care
setting. This test may help clinicians safely rule-out PE, especially in the face of low and

low-to-moderate pretest probabilities.

ACKNOWLEDGEMENTS

The author would like to thank Brian H Rowe, MD MSc, Division of Emergency
Medicine, University of Alberta, and Mathew J Reeves, PhD, Department of
Epidemiology, Michigan State University for assistance with the protocol development,
data collection and analysis, and preparation of the manuscript. I would also like to thank
J Michelle Benningham, MD, Grand Rapids MERC/ Michigan State University Program
in Emergency Medicine for assistance with translation, Samuel Z Goldhaber, MD,
Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, and
Harvard Medical School for the critical review of the protocol and manuscript, and
Joseph Lau, MD, New England Medical Center, for providing the Meta-Test statistical
software. In addition, I thank the following corresponding authors: Claire Barro MD,
Francois Bonnin MD, Henri Bounameaux MD, Philippe de Moerloose MD, Michele
Duet MD, Jeffrey Ginsberg MD, John Heit MD, Guy Meyer MD, Arnaud Perrier MD,
Deborah Quinn MD, Guido Rebcr MD, Paul Sijens PhD, Claudine Soria PhD, Bernard

Tardy MD.

iii

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................... v
LIST OF FIGURES ........................................................................................................... vi
INTRODUCTION ............................................................................................................... 1
METHODS .......................................................................................................................... 7
Research Question ................................................................................................... 7
Search Techniques ................................................................................................... 7
Study Selection ........................................................................................................ 7
Inclusion Criteria ..................................................................................................... 7
Final Inclusion ......................................................................................................... 8
Reference Standards ................................................................................................. 8
Quality Assessment .................................................................................................. 9
Statistical Analyses ................................................................................................ 10
RESULTS .......................................................................................................................... 13
Search ..................................................................................................................... 13
Inclusion ................................................................................................................. 14
Study Description ................................................................................................... 15
Quality Assessment ................................................................................................ 17
Analyses ................................................................................................................. 17
Sensitivity Analyses ............................................................................................... 17
DISCUSSION .................................................................................................................... 23
REFERENCES .................................................................................................................. 29

iv

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

LIST OF TABLES
Eighteen studies deemed ineligible after relevance screen (MEDLINE) ........ 13
Reasons for ineligibility following initial relevance search of MEDLINE ..... 13
Reasons for exclusion following full manuscript review ................................ 15
Articles excluded (12) after quality rating ....................................................... 15

Eleven studies of ELISA D-dimer in the diagnosis of pulmonary
embolism: study characteristics and diagnostic test performance ................... 16

Sensitivity analysis: random-effects pooled estimates of sensitivity and
specificity for various study subgroups including test for heterogeneity ........ 21

Figure 1

Figure 2

Figure 3

Figure 4

LIST OF FIGURES

Search, inclusion and exclusion flow diagram. The “grey
literature” was defined as those studies that were unpublished
or with limited distribution .............................................................................. 14

Sensitivity and Specificity plot with the 95% CI displayed as
horizontal lines. The circle at the bottom labeled REM is the
pooled sensitivity and specificity using a random-effects model .................... 18

Logit regression plot. (D = logit TPR — logit FPR) on the y-axis and

the sum (S = logit TPR + logit F PR) on the x—axis. The y-axis (D) is
equivalent to the log diagnostic odds ratio and the x-axis (S) is

a measure of how the test characteristics vary with the test threshold. ........... 19

Summary receiver-operating characteristic (SROC) curve analysis

of ELISA D-dimer in the diagnosis of PE. Plotted in each of the

SROC graphs are individual studies depicted as ellipses. The

x- and y-dimensions of the ellipses are proportional to the square

root of the number of patients available to study the sensitivity and

specificity, respectively, within the analysis. Also shown is the

unweighted SROC curve limited to the range where data are

available. The cross (x) represents the independent random-effects

pooling of sensitivity and specificity values of the studies with the

Shaded box marking the zone of the 95% CI. .................................................. 20

vi

INTRODUCTION

The goal of any diagnostic test is to allow the clinician to revise the patient’s
probability of having the disease to a level above the treatment threshold or below the
threshold that requires any further testing.1 Where the physician sets the treatment
threshold depends on the potential consequence of the disease (e.g., death vs. disability
vs. full recovery), expense of the diagnostic test, treatment or both, and the. associated
risk(s) of the therapy. Similarly, the diagnostic threshold depends on the morbidity and
mortality associated with a missed diagnosis. For conditions such as PE, the diagnostic
threshold is low since the “cost” of missing the correct diagnosis is high. For example,
the three-month mortality rate for untreated PE has been reported to be as high as 17.5%.2
In most tests in the emergency department (ED), the clinician must decide at what
diagnostic threshold the patient can be discharged home with appropriate follow-up, no
further testing and no treatment. Unfortunately, given the lack of continuity of care and
unreliable follow-up, the ED evaluation of patients who present with symptoms and signs
of suspected serious diseases, such as PE, is expensive and complex.

In considering the use of any diagnostic test, it is important to establish a pretest
probability of the disease being investigated.1 This is equivalent to the prevalence of the
disease among the ED population with similar presenting symptoms. The reported
prevalence of PE in the typical ambulatory patient population presenting to the ED with
signs and symptoms suspicious for PE is 15 to 30 %.3’5 This patient population
represents a select subset of patients presenting to the ED with dyspnea or chest pain that,
after initial evaluation, remains unexplained. Clinicians will often attempt to stratify these

patients into low, intermediate and high risk groups.6 Stratification is based on a

combination of risk factors, presenting symptoms, physical examination findings, and
initial screening tests. In an attempt to aid the physician in establishing a pretest
probability, clinical models have been developed for deep venous thrombosis (DVT) and
PE.7‘ 8 However, these risk stratification schemes are complex, difficult to remember, and
often not entirely driven by “evidence”. Not surprisingly, ED physicians find this a
confirsing clinical area for diagnostic testing.

Classic epidemiological risk factors associated with PE include a history of
thromboembolic disease, immobility, recent surgery or trauma, malignancy, age,
hypercoagulable state, oral contraceptive use, pregnancy, smoking, and history of
significant pulmonary or cardiac disease?” 9 However, a prospective cross—sectional study
was unable to validate an association between PE and any of these classic risk factors in
an ED patient population suspected of PE.5 Selection bias may account for these findings
since the attending physician’s global clinical impression and suspicion for the diagnosis
of PE was the main criterion used for entry into the study.5 Thus physicians may have
used these classic risk factors as part of their decision to enter patients into the study
cohort. This observational study demonstrated that once a patient is placed in the subset
of patients suspected of having PE, risk factors do not help differentiate between those
with PE and those without PB. The only symptom found to have a statistically significant
association with PE was unexplained dyspnea (RR 1.3), and this association was weak.5
The clinical significance of this finding is uncertain since the majority of patients without
PE also had this symptom. Tachycardia and tachypnea are described as the most common
physical exam signs.6 Unfortunately, the initial vital signs failed to discriminate between

those with PE and those without PE in an at risk ED population.5 Signs and symptoms of

PE are nonspecific and may mimic other disease conditions such as pneumonia or
congestive heart failure, making the determination of the pretest probability for PE a
clinical challenge.”

After completing the history and physical exam, the physician will often obtain a
few initial screening tests. Electrocardiography (EKG) may show non-specific ST-T
wave changes and lefl- or right-axis deviation with PE.10 The classic findings of 81, Q3,
T3 or right-bundle branch block are not commonly found, and require a rather large PE to
be present before they are identifieds’ 6' '0 Chest x-ray is usually obtained to rule-out
other disease processes that could explain the patient’s symptoms. The chest x-ray is
usually normal; however, small pleural effusions, focal atelectasis or ill-defined pleural-
based infiltrates may be observed in patients with PE.‘0 Arterial blood gas analysis may
reveal an abnormal p02, pCOz, and/or A-a gradient in patients with PE.10 Unfortunately,
these tests have low specificity and are not sensitive enough to rule-out pulmonary
embolism.H

Where does this leave the ED physician and the patient suspected of having an
acute PE? Stratification based on classic risk factors, physical exam findings and initial
screening tests, is commonly suggested yet has not been validated in the ED setting.3’ 5
The seasoned emergency medicine physician may be able to use clinical experience and
gestalt to accurately raise or lower the pretest probability from the 15 to 30 % pretest
range. However, it is unlikely that this will place the patient above the treatment
threshold committing them to anti-coagulation or below the diagnostic threshold allowing

ED discharge. Further diagnostic testing is usually required.

The traditional approach to PE has been to utilize ventilation/perfusion (V/Q)
scan as the initial diagnostic modality of choice. The PIOPED study was the largest
prospective investigation evaluating the diagnostic accuracy of V/Q scans using
angiogram as the gold standard.4 This study enrolled over 1,000 patients and generated
the test characteristics for V/Q scan that are still used today. A V/Q scan with high
probability is considered diagnostic for PE with a LR of 18.'2 A normal/near-normal
V/Q scan is considered diagnostic for ruling out PE with a LR of 0.] and a false negative
rate of only 2%.'2 Intermediate and low probability V/Q scan results are considered non-
diagnostic and occur in 50 to 70% of patients, leaving the majority of patients below the
treatment threshold and above the diagnostic threshold.l3 Since the PIOPED study, many
research protocols investigating the accuracy of diagnostic tests for PE use angiogram as
the reference standard only in the non-diagnostic categories of low or intermediate
probability V/Q scan.

Helical computerized tomograhpy (CT) has recently been advocated as an
alternative to V/Q scan. Two recent systematic reviews have determined that helical CT
is not as sensitive as V/Q scan, although helical CT has a much lower rate of non-
diagnostic results. '4’ '5 Helical CT has been demonstrated to have excellent specificity
and has the advantage of being able to visualize other structures in the thorax which may
provide an alternative diagnosis.I 1' '5 The greatest limitation of helical CT is the inability
to visualize sub-segmental emboli and a lack of consistency in reporting.14 The clinical
significance of these small emboli is still uncertain.

Some diagnostic algorithms recommend using bilateral lower extremity venous

dopplers and/or compression ultrasound if the helical CT or V/Q scan is equivocal. '0‘ '2

Ultrasound has been shown to be positive for DVT in 20 to 50% of patients with PE and
would warrant anticoagulation without further diagnostic studies.13 However, this test is
very insensitive for the diagnosis of PE since the original clot may have embolized, be
confined to the calf, or located in the pelvic vessels without lower extremity thrombosis.2
A recent study showed that a single negative ultrasound had a LR of only 0.5, leaving
most patients above the diagnostic threshold.l6 An alternative strategy using serial lower
extremity ultrasounds appears to be effective but is not practical in most ED settings.8
Fortunately, several new diagnostic tests have been introduced to assist in the ED
work-up of suspected cases of PE. D-dimer has been advocated as a diagnostic tool that
may obviate the need for additional diagnostic tests such as V/Q scan, helical CT scan,
and angiogram.l3 D-dimer is a fibrin degradation product that is usually elevated in the
presence of thromboembolic disease. Unfortunately, it is also elevated in such common
diseases as inflammatory arthritidies, cancer, and infection. It may also be elevated
following surgery or trauma. A number of different methods are currently available to
measure D-dimer, including latex agglutination, whole-blood agglutination and enzyme-

linked immunosorbent assay (ELISA).1 " '3

Until very recently, the rapid latex tests and
bedside assays have had inadequate sensitivity to rule-out a life-threatening condition
such as PE.'7’ '8

Since the published studies on the use of D-dimer to rule-out PE are of various
size and quality, the accuracy and utility of the test is still debated. Previous systematic
reviews evaluating the utility of D-dimer in the diagnosis of thromboembolsim were very

11.19.20

broad in scope and included DVT and PE, multiple testing methods, and both

inpatient and outpatient populations. Thus, these findings may have limited applicability

to ED settings. The topic also warrants an updated meta-analysis to include more recent
investigations using a rapid ELISA D-dimer testing method that is more practical in the
ED setting. In order to limit the problems of clinical heterogeneityzl found in previous
systematic overviews and eliminate the problems of inter-observer reliability associated
with qualitative tests, we chose to study only quantitative ELISA D-dimer tests in this
review. The primary objective of this systematic review was to determine the accuracy
of the ELISA D-dimer test in the diagnosis of pulmonary embolism in the ED. A
secondary objective was to determine if the test characteristics change with respect to

covariates such as age, comorbidity, or duration of symptoms.

METHODS

Research Question: What is the accuracy (e.g., sensitivity, specificity, likelihood
ratios) of the ELISA D-dimer test in the diagnosis of pulmonary embolism in the adult
patient presenting to the ED with a suspected PE?

Search Techniques: Computerized searching was performed using MEDLINE
(January 1980 to January 1, 2001) to identify clinical studies assessing the utility of an
ELISA D-dimer test in the diagnosis of PE. The search used the MeSH terms:
(pulmonary-embolism OR PE OR VTE) AND (D-dimer OR fibrin OR fibrinogen-
degradation OR FDP ORfibrinogen-degradation-products) AND (ELISA OR enzyme-
linked-immunosorbent-assay) AND Sensitivity and Specificity. Using a similar approach,
a search of EMBASE was also performed.

Study Selection: Two reviewers (MDB, BHR) independently examined the titles
and abstracts of the references identified in the initial MEDLINE and EMBASE searches
to determine if the study was relevant to the clinical question (relevance search).22
Reviews and editorials were excluded immediately. The reference list of the articles
chosen for inclusion in the meta-analysis and the reference list of prior systematic

19. 20

reviews were also screened to identify further studies for inclusion. In an attempt to

identify other so-called “grey literature”23

, experts in the area of PE and the companies
that market laboratory equipment that utilize ELISA D-dimer methods were contacted.
Non-English language articles passing the initial screen were translated prior to full
review.

Inclusion Criteria: To be included in the meta-analysis, the study must have been

a prospective investigation involving a predominately outpatient population presenting

with symptoms and signs suspicious for PE. If a study included any inpatients, the study
population must have been comprised of at least 80% outpatients or data must have been
available to calculate sensitivity and specificity for the outpatient component of the. study
population.

Final Inclusion: Following the relevance search, the two primary reviewers
(MDB, BHR) compared their exclusion logs to determine if there was any discordance.
Where there was disagreement, consensus was reached by conference. A data collection
form was used to abstract data from each study meeting the inclusion criteria. If a study
met the inclusion criteria, reviewers attempted to contact the author to identify additional
papers, confirm data extraction/estimation for correctness and completeness, and to
obtain missing data. Two reviewers (MDB, BHR) independently confirmed numeric
calculations and graphic extrapolations. The data was evaluated for the presence of
publication bias using statistical methods (see Statistical Analysis)”26

Reference Standards: Although a positive angiogram or autopsy is considered the
gold standard for the diagnosis of PE, we considered any one of the following as
acceptable surrogate reference standards: 1) high probability V/Q scan, 2) CT scan
positive for PE, or 3) positive lower extremity imaging study (ultrasound, impedance
plethysmography or venogram). A negative angiogram was considered the gold standard
for ruling out PE. Acceptable surrogate reference standards for a negative diagnosis were:
1) normal or very low probability V/Q scan, or 2) clinical follow-up documenting the
absence of a thromboembolic event over a minimum of 3 months.27 If a reference
standard is not used in all subjects, the study is susceptible to verification bias (work-up

bias)?”30 To minimize the effect of verification bias and to provide the most

conservative estimate for test sensitivity, any study that did not apply a reference standard
to all subjects had the results analyzed on a “worse case” assumption, i.e. each subject
lost to follow-up was assumed to have the worst outcome.

Quality Assessment: The rigorous inclusion/exclusion criteria functioned as the
primary quality filter in this meta-analysis. The meta-analysis focused the appraisal of
study quality on the potential for differential reference standard bias30 and spectrum
bias.29 Differential reference standard bias is a more subtle form of verification bias and
may occur when a negative test result is verified by a less rigorous standard than those
with a positive test result.” 3 ' For example, a patient with a negative ELISA D-dimer
result is verified by a single lower extremity ultrasound and outpatient follow-up,
whereas a patient with a positive ELISA D—dimer result is verified by serial lower
extremity ultrasound examinations and hospitalization. The reference standard and
patient spectrum for each study was graded in regards to quality parameters (A -
excellent, B - susceptible to some bias, C — indeterminate or poor) as outlined below:

0 Reference standard: The potential for differential reference standard bias was

 

assessed as follows.30 Grade A - Those studies using the same reference standard
regardless of the ELISA D-dimer result. Grade B - Studies using different
reference standards depending on the results of the ELISA D-dimer test.
Grade C - Indeterminate or not meeting the study protocol definition of an
appropriate reference standard.

0 Patient spectrum: The external validity (generalizability) of the meta-analysis
depends upon the Spectrum of disease included in each study and how the patient

population was assembled.3 ' Grade A - Patient spectrum would be expected to

include a consecutive or random sampling of a typical outpatient population

presenting with symptoms and signs suspicious for PE. Grade B - Studies that

selected only a small subgroup of subjects suspected of PE. Grade C -

Indeterminate or not meeting the study protocol definition of an appropriate

patient spectrum.

The blind interpretation of the test under investigation and the reference standard
are typically considered important components in the critical appraisal of diagnostic
tests}:2 However, since the interpretation of a quantitative ELISA D-dimer test is an
objective measurement, it is less critical that the technician performing the D-dimer
quantitative analysis be blinded to the clinical history or the reference standard.28 There
is potential for interpretation bias if the radiologist performing the reference standard was
not blind to the ELISA D-dimer test result.29 This information was obtained from the
manuscript or by author query.

In order to provide the most conservative estimate of test characteristics, after
each study was scored for quality, Grade C studies were excluded from the analysis.

Statistical Analyses: The analysis was based on a summary receiver operating

characteristic (SROC) curves”: 3“

When studies utilize different thresholds for positive
and negative results, the reported sensitivity and specificity will differ among studies. A
graphical display of the variability of the test characteristics between studies can be
assessed with the SROC curve.3 ' To simplify calculations, only single test thresholds and
dichotomous results were used in the analysis. If a study reported results for more than

one test threshold, 3 test threshold of 500 ng/ml was used since this is the most common

cutoff used in clinical practice. The sensitivity and specificity for the single test

10

threshold identified for each study was used to plot an unweighted SROC curve.“ 35 A

correction factor of one-half was added to each cell to avoid calculation problems by
having a value of zero in the 2x2 table.34 This correction has not been found to
significantly alter the results of the SROC curve.34 The SROC curve analysis is based on
a logit transformation of the data which plots the difference (D = logit TPR — logit FPR)
on the y-axis and the sum (S = logit TPR + logit FPR) on the x-axis. The y-axis (D) is
equivalent to the log diagnostic odds ratio and the x-axis (S) is a measure of how the test
characteristics vary with the test threshold. A regression equation (D = or + ,B*S) derived
from the SROC curve analysis can be used to assess the heterogeneity among study
results. If the )6 coefficient is near zero and not statistically significant, then evidence of
significant heterogeneity is not present.23 When there is little variability of test results
(i.e. homogeneity), the SROC curve does not provide additional information over average
sensitivity or specificity values.35 A random-effects model was used to calculate the

:35: 36 The random-effects model

average sensitivity and specificity across studies.3 '
accounts for between-study variability and provides a more conservative estimation as
compared to the fixed-effects model.33 Statistical tests related to the SROC curve were
performed using Meta-Test (version 0.6, Boston, MA). All other statistical tests were
performed using the SAS statistical application program (version 8.0, Cary, NC).
Sensitivity analysis was used to assess the effect of study quality on the overall
results. The SROC curves were compared with and without the specified methodological
flaw. A priori subgroup analyses were performed on studies using traditional ELISA

methods, rapid ELISA methods, those using the 500ng/ml cutoff value, age, comorbidity,

and the duration of symptoms. C omorbidity was defined as having surgery, trauma,

11

myocardial infarction, stroke, acute infection, disseminated intravascular coagulation,
pregnancy, postpartum, or active cancer within the 10 days preceding the ED
evaluation.37

In order to assess for the presence of publication bias, whereby smaller studies
Show effects different from those of larger studies, we used methods previously proposed
by Galbraith24 and Egger25 . The standard normal deviate (SND) of the OR (calculated by
dividing the OR by its SE) was regressed against its precision (as measured by the
inverse of the SE) i.e., SND = a + fl*precision. The intercept or provides a measure of the
degree of asymmetry resulting from publication bias. Data from a homogeneous or
symmetrical set of trials will scatter around a line that runs through the 0 origin, whereas
in the presence of publication bias, the intercept will deviate from 0. The absolute

magnitude of a can therefore be used as one measure of the presence of publication bias.

12

RESULTS
Search: The MEDLINE search yielded 52 references. Eighteen were
immediately deemed ineligible for full review (Tables 1 and 2). The relevance search
had excellent agreement between the two reviewers with the simple agreement 92% and 3
Kappa of 0.83 (95% CI: 0.67, 0.99). An EMBASE search yielded 71 references; 7

additional references were identified as eligible for full review (Figure 1). The search for

grey literature yielded 2 additional published articles.

Table 1: Eighteen studies deemed ineligible after relevance screen (MEDLINE)

38. 39

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Author Year Reason

Lorut C, 1999 Review
Michiels JJ 2000 Review
Caliezi C 2000 Letter

lndik J H 2000 Review

Kline J A 2000 Review

Stein PD 1999 Review
Ndiaye A 1997 DVT after surg
Janssen M 1998 Review
Bounameaux H 1997 Review
Bounameaux H 1997 Review

Lee AY 1997 Review

Perrier A 1997 Review

Perrier A 1995 Review

Sie P 1995 Review
Bouman CS 1995 DVT
Bounameaux H 1994 Review
DVTENOX group 1994 DVT with heparin
Bounameaux H 1991 Review

 

 

 

Table 2. Reasons for ineligibility following initial relevance search of MEDLINE.

Primary Rea_son

Review 14
Letter to editor 1
Subjects suspected for having DVT 3

Total

18

 

Number of Reports

 

 

MEDLINE Search

 

 

Figure 1. Search, inclusion and exclusion flow diagram. The “grey literature” was
defined as those studies that were unpublished or with limited distribution.l4

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

52
l
l 1
34 Ineligible
18
(Table 2)
EMBASE Grey Literature
Search ~— Search
7 2
Full Review
43
l
l l 1
Potential Inclusion Duplicates Excluded
23 3 17
(Table 3) (Table 3)
Author Contact
18/23 —
l 7
Final Inclusion Excluded
11 12
(Table 5) (Table 4)

 

 

 

 

 

Inclusion: A complete manuscript review was performed on the remaining 43
articles. Dutch (1), French (4) and German (1) manuscripts required translation.
Following full review, 3 studies were identified as duplicates and 17 others were
excluded for various reasons (Table 3). In order to clarify important missing information
and confirm data extraction, an attempt was made to contact the authors of the 23
remaining studies. Seventy-eight percent (18/23) of authors responded in some form to

these queries. After obtaining additional information, a further 12 studies were excluded

l4

(Table 4).40'5' Eleven studies therefore met the inclusion criteria and provided a total

Table 3. Reasons for exclusion following full manuscript review.

 

Primary Reason Number of Reports
Review paper 4
Duplicate publication 3
Latex D-dimer method 3
Grade C patient spectrumt 3
Grade C reference standardA 7
Total 20

Notes: I = Patient spectrum: Grade C - Indeterminate or not meeting protocol definition for
adequate patient spectrum; A = Reference standard: Grade C - Indeterminate or not meeting
protocol definition for adequate reference standard.

Table 4. Articles excluded ( 12) after quality rating.

 

Author Year Reason
Sijens PM)“ 2000 Grade C spectruml #
Quinn D“ 1999 Grade C spectrum #
Brimble S42 1998 Duplicate #

Duet M43 1998 Grade C spectrum
Reber G44 1995 Duplicate #

Bonnin F45* 1997 Grade C spectrum #
van Beek E46 1996 Data missing
Rochemaure J47 1995 Grade C spectrum
Flores J48 1995 Grade C reference standardA
Goldhaber S49 1993 Grade C spectrum #
van Beek E50 1993 Data missing
Bounameaux H5 l 1990 Grade C spectrum #

Notes: * = search result exclusively from EMBASE; t = Patient spectrum: Grade C -
Indeterminate or not meeting protocol definition for adequate patient spectrum; A = Reference
standard: Grade C - lndeterrninate or not meeting protocol definition for adequate reference

standard; # = after correspondence with author.

study population of 2,126 subjects. A summary of the major characteristics of each study

is provided in Table

37-39. 52-59
5.

Study Descriptions: The prevalence of disease among outpatients suspected of

PE ranged from 17 to 58%, with most falling in the 20 to 40% range. In almost all

studies, females were represented slightly more than males. Most studies included a

broad range of ages except for one study with a mean age of 72 years59 and another that

.8508 886.0 <m3m 292 m wow: u z 308:8 853 note vim—8% Ho: 8% Rescues u <2 was 23:85 8:80.22 3:5th

no.2 33:83 - m €80 .8865 <m5m 2: mo awe—Eamon Esteem 85:28 08% - < 3220 ”22:83 353.6% n < ”$5 828on 28 3680a
.. m 8ch .mm com msofifimsm 3838.? 53, watcomoa cote—smog 285350 .859 a Mo @2953. 802:2 no 0330828 - < 0220 ”828on
Eocmm n + 63282“ 5:32.285 2: no women Ecumgxoaam 820 Be mofiaflwofiov n .. ”ma Mo 36233 35:3 do 828:: n 2 ”82oz

 

02 < m an .va 5. 8o. .5 2: :22 t. 3 so: 3% mm 2.02..
._w «o cor—pom
3> m < Go as 3 so 28 mm 8m 3. Q So: 3 z. :82 a a as:
3> m < S .t R :3 .03 g 8m cm 2 so: am mm 9.32 .a .0 oeam
a> m < E 5 E 82 2e 2: 8m 3 5 so: 8 O. 2.232 .a a befi
8> < < co .3 on E. .2: 2. 8m 3 mm 88 N: a. .22:
._m 6 5.32
02 < m 34 .5 2. so. .5 3 8m <2 8 so: E
9.32
om ._m we burr—On—
8> < m $4 .03 mm 82 .me 2: 28m 3. B so: we 3. meg. a
“0 802.502 03
02 < m 9% .5 5. so. .me 2: 8m <2 <2 so: of 3 :32 .2.
Ho QmOOZOO—Z cc
02 < m em .08 3 :3 .mg mm o2 am am so: 2: o. 282
.3 Ho humor—NCO..—
8> < m 5. .5 mm 82 .5 2: SM 8 3 so: of t 282
.E 8 90350
8> < < S 53 mm 82 .08 3 Sm m... <2 so: am om :82
.3 8 $5500
2:: 23.525 22253.0. so amen so .33 2%.: 3% am 3 a? 33.3.8 33 3.5
an... 5:25 Sam 8:23an 33 $2283 SE 53.2% 22.8.5 .232 .582 SE 2 m.— a
.315— . 3:25 beam 30,—. 35—96.:—
ego—58.8;

68 9:8:ch ES moafluofinuaao 253. "Sam—onus Eases—.39 no $8531 2: E 35:;— <m5m me 3:56 5.65 .m ~35.

16

focused specifically on an elderly population57 with a mean age of 8] years. Excluding

- 7
these two outliers 5 ’59

, the mean age of the patients in the studies were similar with a
range from 54 to 63 years. The regression analysis showed no statistical evidence of
publication bias based on the p = 0.19, although the absolute value of the intercept was
large (i.e., 98).25‘ 2"

Quality Assessment: Only 2 studies were given a Grade A with respect to both of
the key quality parameters, patient spectrum and reference standard (Table 5).38’ 58 One of
these studies included 20% inpatients which was the maximum percentage of inpatients a
study could have and still meet the inclusion criteria as defined in the research protocol.58
Three of the studies included were given an “excellent” rating with regard to the
reference standard, but were rated Grade B with respect to the patient spectrum. In
contrast, six studies were given a Grade A with respect to the patient spectrum, but Grade
B on the reference standard. In the majority (7/1 1) of studies, the radiologist was blind to
the D-dimer results and any other clinical information.

Analyses: The sensitivity and specificity of each included study was calculated
with the 95% confidence interval (CI) displayed (Figure 2 and Table 5). The pooled
summary estimate using a random-effects model resulted in a sensitivity of 0.95 (95% Cl:
0.90, 0.98) and a specificity of 0.45 (95% CI: 0.38, 0.52). However, this pooled estimate
demonstrated statistically significant heterogeneity with a ,6 of 0.43 (95% CI: 0.004,
0.87). The visual display provided by the logit regression plot (Figure 3) and the SROC
curve shows moderate variability in the results (Figure 4).

Sensitivity Analyses: In order to explore the heterogeneity found in the overall

results, a sensitivity analysis based on the key methodological quality parameters was

17

Figure 2. Sensitivity and specificity plot with the 95% CI displayed as horizontal

lines. The circle at the bottom labeled REM is the pooled sensitivity and specificity

using a random-effects model.

SENSITIUITY
0 20 10 . 60 80

 

 

 

 

I: 2—+——9-—.9—
22:2—o
iiié—u
:2 2222—»
4
2: —<
2 iii—r-
: fi—f—H—
4'

 

REM

SPECIFICITY
109"“. 80 60 40 20

—-'—..—.

(DwVOxUl-wa"

u

0
fl
0

:1
4
l
0

 

 

18

 

Study
Deners
Lenzhofer
Ginsberg
defloerloose
defloerloose
Perrier
Tardy

Heger

Derrier

(OGJVOcL‘I-wa"

Heit

11 Barro

1931
1996
1997
1998
1998
1999

1999

Dis NoDis

17
17
26
51
46
196
40
60
106
16
15

6?
90
124
99
119
475
56
82
338
15
11

Figure 3. Logit regression plot. (D = logit TPR - logit FPR) on the y-axis and the
sum (S = logit TPR + logit FPR) on the x-axis. The y-axis (D) is equivalent to the log
diagnostic odds ratio and the x-axis (S) is a measure of how the test characteristics
vary with the test threshold.

 

 

 

 

 

.16
a:

a.

n.

V

a

'fl

1»

o

d

l

A

m

a.

.—

V
:1 o
m

o
~o

012 3 4 5 6 7

logit (TPR) + logit (FPR)

l9

Figure 4. Summary receiver-operating characteristic (SROC) curve analysis of
ELISA D-dimer in the diagnosis of PE. Plotted in each of the SROC graphs are
individual studies depicted as ellipses. The x- and y—dimensions of the ellipses are
proportional to the square root of the number of patients available to study the
sensitivity and specificity, respectively, within the analysis. Also shown is the
unweighted SROC curve limited to the range where data are available. The cross (x)
represents the independent random-effects pooling of sensitivity and specificity
values of the studies with the shaded box marking the zone of the 95% CI.

 

 

I I I I I I I I I
I I I I I I I I I
I I I I I I I I I
0 5 ...................................................................................................
. I I I I I I I I I
I I I I I I I I I
I I I I I I I I I

SENSITIUITY

I I I I I I I I I
0 4 ......... J ......... J ........ J ......... J ......... J ......... J ......... J ......... J ......... J ......... .4
I I I I I I I I I I

I I I I I I I I I
I I I I I I I I I
I I I I I I I I I

0 3 ooooooiooqdnccouobo‘ooooao..-qnnouoooncqnouvucODUQOIDDOOOCc". ooooooooooooooooooooooooooooooooooooooo
. I I I I I I I I I
I I I I I I I I I
I I I I I I

 

 

 

 

 

 

 

. . . , , firea Under Curve ,
0.2 """"" '3 """"" ‘3 """"" ‘E """""" '3 """"" ‘3 """ Unueighted = 0.5631 "1 """""" *
0.. ----------
0.0 i i I L L

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
1 - SPECIFICITY

20

performed (Table 6). When the SROC curve was derived using only those studies with an

“excellent” or Grade A quality rating (N = 5) in regard to the reference standard, the test

for heterogeneity was not significant with a ,3 of 0.18 (95% CI: - 0.52, 0.88). The pooled

summary estimate for this subgroup using a random-effects model resulted in a

sensitivity of 0.90 (95% CI: 0.83, 0.94) and a specificity of 0.40 (95% CI: 0.26, 0.55). In

contrast, the subgroup of studies with a Grade A patient spectrum (N = 8) was found to

have statistically significant heterogeneity. The proper blinding of the radiologist to the

D-dimer test results had minimal effect on the results (Table 6). There were only 2

studies that used a rapid form of the ELISA D-dimer test, so a SROC curve analysis

could not be performed. The average sensitivity for these 2 studies was higher (100%),

when compared to the pooled results using traditional methods (sensitivity 94%). Using a

Table 6. Sensitivity analysis: random-effects pooled estimates of sensitivity and
specificity for various study subgroups including test for heterogeneity.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Study Group N % Sensitivity, % Specificity, fi,

(95 % CI) (95% CI) (95%CI)
Reference Standard: Grade A" 5 90 (83, 94) 40 (26, 55) 0.18 (-.52, .88)
Patient Spectrum: Grade A1@ 8 97 (92, 99) 48 (41, 56) 0.46 (.10, .81)
Proper B1inding# 7 92 (85, 96) 39 (30, 49) 0.39 (-.l6, .94)
Traditional ELISA 9 94 (88, 97) 45 (36, 55) 0.32 (-.18, .82)
Traditional ELISA 500 ng/ml* 6 95 (87, 98) 40 (32, 50) 0.55 (~31, 1.41)
Rapid ELISA 500 ng/m1*~ 2 100 44(40, 48) N/A
Age 270 1 100 14 (7, 21) N/A
No comorbidity@ 3 89 (75, 96) 55 (46, 64) 0.59 (.04, 1.14)
Duration 24 days 1 73 (59, 86) 33 (19, 48) N/A

 

Notes: N = number of studies included in pooled subgroup/sensitivity analysis; fl = regression
coefficient for the sum (logit TPR + logit FRP) which indicates heterogeneity among studies
when the value is significantly different from zero; 1 = Patient spectrum: Grade A - Consecutive
or random sampling of a typical outpatient population presenting with symptoms suspicious for
PE; " = Reference standard: Grade A - Same reference standard regardless of the ELISA D-
dimer; # = Proper blinding - radiologist interpreting reference standard blind to the ELISA D-
dimer result; ~ = values reported are an average for the 2 studies; * = used a threshold of greater
than 500 ng/ml as criteria for a positive D-dimer test; @ = statistically significant heterogeneity.

N/A = not appropriate/applicable.

21

 

“worse case” assumption did not alter the pooled estimates since, in the majority of
studies, very few cases were lost to follow-up. The only exception was the subgroup
using the rapid ELISA method.” 55 When the “worse case” assumption was applied to
this subgroup, the average sensitivity decreased to 95%.

Although there were only a few studies with sufficient data for subgroup analysis
based on age, comorbidity, and symptom duration, all 3 of these covariates had a
considerable effect on the test characteristics of the ELISA D-dimer test (Table 6). Older
age (270 years) was the only specific age-related subgroup that could be analyzed. The
specificity of the ELISA D-dimer test in this elderly study population57 was much lower
(14%) than all of the other individual studies included in the meta-analysis (Figure 2).
The subgroup analysis of the 3 studies with data on the population without comorbidity
had a higher specificity (55%) but a lower sensitivity (89%). Only one study evaluated
the effect of symptom duration on the ELISA D-dimer results.37 If the duration of
symptoms was greater than 3 days in this study population, both the sensitivity and

specificity decreased (Table 6).

22

DISCUSSION

This systematic review attempted to identify the current published and
unpublished literature regarding the use of ELISA D-dimer in the diagnosis of PE in the
acute care setting. Following an exhaustive search, the application of stringent
inclusion/exclusion criteria and rigorous selection methodology, we included 11 studies
involving 2,126 patients in the study. The results demonstrate that the traditional ELISA
technique is highly sensitive (94%), yet only moderately specific (45%). A highly
sensitive test will assist the clinician in ruling out PE. Using evidenced-based medicine

1. 2
terms,23

the negative LR (-LR) of 0.1 for this test would be clinically helpful,
especially when applied in the face of accurate pretest probability models for PE.8‘ (’0 For
example, a patient with a pretest probability of 20% (low) in the face of a negative
ELISA D-dimer would have a post-test probability of 2%; in such a case, further testing
is probably not warranted. In contrast, the positive likelihood ratio (+LR) for this test is
only 1.7 and is of limited value in most clinical scenarios.

An adequate reference standard (gold standard) and an appropriate spectrum of
patients are key elements of study methodology when considering application of the
results to clinical practice.61 Although angiogram has long been considered the “gold
standard” for the diagnosis of PE, it has been recommended that clinicians now use an
outcome-based standard.27 As expected, the majority of studies included in this overview
used an outcome-based standard such as symptom-free follow-up and survival. Most ED
physicians would consider the absence of a thromboembolic event or death over a 3-

month period a valid outcome measure.27 Studies that use a different reference standard,

depending on the ELISA D-dimer test result, may provide an erroneously high estimate

23

for test sensitivity.3O As expected, the 5 studies using a reference standard not susceptible
to differential reference standard bias30 (Grade A) had a lower pooled sensitivity (90%)
and specificity (40%). However, because they had selected a subgroup of the target
outpatient population suspected of PE, these studies were predominately of lower quality
with respect to the patient spectrum (Grade B) and susceptible to spectrum bias (or
assembly bias). These studies may have been representative of patient populations with
comorbidity or a longer duration of symptoms. For example, Heit et 3137 selected a study
population referred to the Mayo Clinic for angiography. Of all studies included, this was
the only investigation that performed angiography on all subjects and therefore was the
least susceptible to differential reference standard bias. However, the external validity of
the results must be questioned since these subjects were not typical of the ED patient
population suspected of having PE. Another study was rated “excellent” in regard to the
reference standard but was rated Grade B with respect to the patient spectrum.59 This
study included only those patients suspected of PE who were subsequently admitted to
the hospital, which explains the very high prevalence (58%) of PB. The specificity of the
ELISA D—dimer may also be affected by the patient spectrum as demonstrated by the low
specificity (14%) found in the study focused only on the elderly.57 In a subsequent
publication, investigators combined data from two of the studies included in the review”’
56 and performed a subgroup analysis providing similar results with a specificity of 17%
in those over 70 years."2 The reasons for the much lower specificity in the elderly is
unknown but may be related to comorbidity that raises the D-dimer levels in settings not

confined to thrombo-embolism (such as cancer, inflammation, recent surgery).

24

Different forms of interpretation bias are another concern in the evaluation of
diagnostic studies.29 In order to limit diagnostic review bias, the radiologist performing
the interpretation of the reference standard should be blinded to the results of the ELISA
D-dimer test.29 Previous investigators who have assessed the effects of design-related
bias in studies evaluating diagnostic tests have found that the average effect of
inappropriate blinding is small.30 Our meta—analysis had similar findings, afier excluding
studies in which the radiologist was not blinded, there was not a clinically significant
change in the pooled estimates of diagnostic accuracy. Among the studies with proper
blinding that used follow-up as a reference standard (5/7), the personnel performing
clinical follow-up were also blinded to the ELISA D-dimer test results and clinical
course.

The traditional ELISA D-dimer has been shown to have excellent sensitivity but
requires specialized equipment available only at major academic centers and is too time-

consuming to be valuable in most ED settings} '7

Therefore, studies using a rapid ELISA
D-dimer method are important and were consequently reported as a subgroup. The two
studies in this subgroup were among the largest included in the review and reported an
impressive sensitivity of 100%.39' 55 However, both of these investigations were prone to
differential reference standard bias which may yield an erroneously high sensitivity.29 For
the clinician, the pooled results of studies using traditional methods would appear to be a
more conservative estimate for test sensitivity (94%) compared to the perfect sensitivity
(100%) reported by these two studies utilizing a rapid ELISA D-dimer method.

Despite the rigorous methodology used in this systematic review, there are

inherent limitations in this type of analysis. Typically, the studies included in 3 meta-

25

analysis should be evaluated for the presence of publication bias using graphical and
statistical methods.25 Publication bias is expected to be a much greater problem in meta-
analysis of diagnostic tests than in meta-analysis of randomized-controlled trials of
therapeutic effects but there is no agreed upon method to assess publication bias for
diagnostic meta-anlaysis.33 Aware of these limitations, we did perform a regression
analysis, which demonstrated possible evidence of publication bias given the large value
of the intercept. However, the small number of studies included limits the statistical
power of the test, and the intercept was not statistically significant.25 Our search
attempted to examine all literature on this topic but it is possible that some unpublished
or other foreign language studies were missed. It may be that lower sensitivity results
remained unpublished, which would inflate the pooled estimates. However, we searched
extensively for other evidence of unpublished work, so we believe this bias to be limited.

There is also the possibility that some information may have been lost during
translation of those studies published in non-English. For instance, we could not
determine the reason for the unusually high specificity (80%) reported in one study
published in German52 and author contact was unsuccessful. Finally, in the original
protocol, a subgroup analysis was planned to determine if the test characteristics might
differ based on varying levels of pretest probability. While attempts were made to derive
this information from the authors, in most cases pretest probability estimation was not
performed, so additional information was not available. Consequently, there were
insufficient data in the primary studies to comment on this important issue.

Restricting the systematic review to D-dimer tests that use an ELISA method

makes it inappropriate to apply the results to clinical settings where alternative D-dimer

26

testing methods are used. However, a recent investigation has shown that a newer latex-
based method had remarkably similar test characteristics (sensitivity of 96%, specificity
of 45%)"3 to the summary estimates derived in this meta-analysis. A normal whole-blood
agglutination D-dimer assay in combination with a normal alveolar dead-space fraction
has also been shown to have adequate sensitivity to rule-out PE."4 A systematic review by
Kline et al” compared various D-dimer testing methods such as these. However, the
methodology of this meta-analysis differed from ours with respect to the search, the
inclusion criteria, and the quality assessment. In fact, only 4 of the studies were common
to both reviews.53‘ 55'”
Notwithstanding the above concerns, this meta-analysis followed widely accepted
methodology for the selection of studies, the data extraction process, the analysis of study
quality, and the evaluation of subgroups.65 Overall, the summary data reveal that the
ELISA D-dimer is highly sensitive, but only moderately specific. Given these results, the
D-dimer test may be a safe and efficient test to reduce the overall costs associated with
“rule-out PE” assessment in the ED; although the cost-effectiveness of D-dimer testing in
the clinical setting has not been adequately evaluated. Moreover, as rapid D-dimer tests
become more readily available, the number of patients screened for PE will likely
increase, which could result in more V/Q or CT scans ordered by physicians.66 Pooling
the results of studies using traditional ELISA methods provide the most valid pooled
estimates. Although the studies utilizing a rapid ELISA D-dimer were large and would be
useful for many non-academic medical centers, the sensitivity reported in these studies
appear to be suspiciously high. The test characteristics of the ELISA D-dimer appear to

change among certain subgroups of patients such as those having comorbidity or

27

prolonged symptom duration. In addition to assisting clinicians, the summary estimates
may assist researchers performing decision analysis or developing a protocol to test a

diagnostic algorithm for use in the ED.

28

10.

11.

12.

REFERENCES

Pauker SG, Kassirer JP. The Threshold Approach to Clinical Decision Making.
NEJM 1980;302(20):1 109-17.

Goldhaber SZ. Pulmonary Embolism [See Comments]. NEJM 1998;339(2):93-
104.

Hull RD, Raskob GE, Carter CJ, Coates G, Gill GJ, Sackett DL, et a1. Pulmonary
Embolism in Outpatients with Pleuritic Chest Pain. Arch Intern Med
l988;l48(4):838—44.

PIOPED. Value of the Ventilation/Perfusion Scan in Acute Pulmonary Embolism.
Results of the Prospective Investigation of Pulmonary Embolism Diagnosis
(Pioped). The Pioped Investigators. JAMA 1990;263(20):2753-9.

Susec O, Jr., Boudrow D, Kline JA. The Clinical Features of Acute Pulmonary
Embolism in Ambulatory Patients. Acad Emerg Med 1997;4(9):89l -7.

Hyers TM. Venous Thromboembolism. Am J Respir C rit Care Med
1999;159(l):l-14.

Wells PS, Hirsh J, Anderson DR, Lensing AW, Foster G, Kearon C, et al.
Accuracy of Clinical Assessment of Deep-Vein Thrombosis [Published Erratum
Appears in Lancet 1995 Aug l9;346(8973):516]. Lancet l995;345(896l):l326-
30.

Wells PS, Ginsberg J S, Anderson DR, Kearon C, Gent M, Turpie AG, et a1. Use
of a Clinical Model for Safe Management of Patients with Suspected Pulmonary
Embolism. Ann Intern Med 1998;129(12):997-1005.

Heit JA, Silverstein MD, Mohr DN, Petterson TM, O'Fallon WM, Melton L], 3rd.
Risk Factors for Deep Vein Thrombosis and Pulmonary Embolism: A Population-
Based Case-Control Study. Arch Intern Med 2000;160(6):809-15.

Tapson VF, Carroll BA, Davidson BL, Elliott CG, Fedullo PF, Hales CA, et al.
The Diagnostic Approach to Acute Venous Thromboembolism. Clinical Practice
Guideline. American Thoracic Society. Am J Respir C rit Care Med
l999;160(3):1043-66.

Kline JA, Johns KL, Colucciello SA, Israel EG. New Diagnostic Tests for
Pulmonary Embolism [See Comments]. Ann Emerg Med 2000;35(2):l68-80.

Jaeschke R, Guyatt GH, Sackett DL. Users' Guides to the Medical Literature. 111.
How to Use an Article About a Diagnostic Test. B. What Are the Results and Will
They Help Me in Caring for My Patients? The Evidence-Based Medicine
Working Group. JAMA l994;271(9):703-7.

29

l3.
14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

Gallagher EJ. Clots in the Lung. Ann Emerg Med 2000;35(2):181-87.

Mullins MD, Becker DM, Hagspiel KD, Philbrick IT. The Role of Spiral
Volumetric Computed Tomography in the Diagnosis of Pulmonary Embolism.
Arch Intern Med 2000;160(3):293-8.

Rathbun SW, Raskob GE, Whitsett TL. Sensitivity and Specificity of Helical
Computed Tomography in the Diagnosis of Pulmonary Embolism: A Systematic
Review. Ann Intern Med 2000;132(3):227-32.

Daniel KR, Jackson RE, Kline JA. Utility of Lower Extremity Venous Ultrasound
Scanning in the Diagnosis and Exclusion of Pulmonary Embolism in Outpatients.
Ann Emerg Med 2000;35(6):547-54.

Janssen MC, Wollersheim H, Verbruggen B, Novakova 1R. Rapid D-Dimer
Assays to Exclude Deep Venous Thrombosis and Pulmonary Embolism: Current
Status and New Developments. Semin Thromb Hemost 1998;24(4):393-400.

Farrell S, Hayes T, Shaw M. A Negative Simplired D-Dimer Assay Result Does
Not Exclude the Diagnosis of Deep Vein Thrombosis or Pulmonary Embolus in
Emergency Department Patients. Ann Emerg Med 2000;35(2):121-5.

Becker DM, Philbrick JT, Bachhuber TL, Humphries J E. D-Dimer Testing and
Acute Venous Thromboembolism. A Shortcut to Accurate Diagnosis? Arch Intern
Med 1996;156(9):939-46.

Bounameaux H, de Moerloose P, Perrier A, Reber G. Plasma Measurement of D-
Dimer as Diagnostic Aid in Suspected Venous Thromboembolism: An Overview.
Thromb Haemost l994;7](l ):1-6.

Thompson SG. Why Sources of Heterogeneity in Meta-Analysis Should Be
Investigated. BMJ l994;309(6965):]351—5.

Meade MO, Richardson WS. Selecting and Appraising Studies for 3 Systematic
Review. Ann Intern Med 1997;127(7):53l-7.

McAuley L, Pham B, Tugwell P, Moher D. Does the Inclusion of Grey Literature

Influence Estimates of Intervention Effectiveness Reported in Meta-Analyses?
Lancet 2000;356(9237): 1228-31 .

Galbraith RF. A Note on Graphical Presentation of Estimated Odds Ratios from
Several Clinical Trials. Stat Med l988;7(8):889-94.

Egger M, Davey Smith G, Schneider M, Minder C. Bias in Meta-Analysis
Detected by a Simple, Graphical Test. BM] l997;315(7109):629-34.

Steichen TJ. Sbel9:Tests for Publication Bias in Meta-Analysis. Stata Technical
Bulletin Reprints 1999;721 25-33.

30

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

Wolfe TR, Hartsell SC. Pulmonary Embolism: Making Sense of the Diagnostic
Evaluation. Ann Emerg Med 2001 ;37(5):504-14.

Begg CB. Biases in the Assessment of Diagnostic Tests. Stat Med 1987;6(4):41 l-
23.

Mower WR. Evaluating Bias and Variability in Diagnostic Test Reports. Ann
Emerg Med 1999;33(1):85-9l.

Lijmer JG, Mol BW, Heisterkamp S, Bonsel GJ, Prins MH, van der Meulen JH, et
al. Empirical Evidence of Design-Related Bias in Studies of Diagnostic Tests [See
Comments] [Published Erratum Appears in Jama 2000 Apr l9;283(15):l963].
JAMA l999;282(l l):106l-6.

Irwig L, Tosteson AN, Gatsonis C, Lau J, Colditz G, Chalmers TC, et a1.
Guidelines for Meta-Analyses Evaluating Diagnostic Tests. Ann Intern Med
1994;120(8):667-76.

Jaeschke R, Guyatt G, Sackett DL. Users' Guides to the Medical Literature. Iii.
How to Use an Article About a Diagnostic Test. A. Are the Results of the Study
Valid? Evidence-Based Medicine Working Group. JAMA l994;271(5):389-91.

Irwig L, Macaskill P, Glasziou P, Fahey M. Meta-Analytic Methods for
Diagnostic Test Accuracy. J Clin Epidemiol 1995;48(1):l 19-30; discussion 131-2.

Moses LE, Shapiro D, Littenberg B. Combining Independent Studies of a
Diagnostic Test into a Summary Roc Curve: Data-Analytic Approaches and Some
Additional Considerations. Stat Med 1993;12(l4):l 293-316.

Lau J, Ioannidis JP, Balk EM, Milch C, Terrin N, Chew PW, et a1. Diagnosing
Acute Cardiac Ischemia in the Emergency Department: A Systematic Review of

the Accuracy and Clinical Effect of Current Technologies. Ann Emerg Med
2001;37(5):453-60.

Dchimonian R, Laird N. Meta-Analysis in Clinical Trials. Control Clin Trials
1986;7(3):177-88.

Heit JA, Minor TA, Andrews JC, Larson DR, Li H, Nichols WL. Determinants of
Plasma Fibrin D-Dimer Sensitivity for Acute Pulmonary Embolism as Defined by
Pulmonary Angiography. Arch Pathol Lab Med 1999;123(3):235-40.

Demers C, Ginsberg J S, Johnston M, Brill-Edwards P, Panju A. D-Dimer and
Thrombin-Antithrombin Iii Complexes in Patients with Clinically Suspected
Pulmonary Embolism. Thromb Haemost l992;67(4):408-12.

Perrier A, Desmarais S, Miron M], de Moerloose P, Lepage R, Slosman D, et al.
Non-Invasive Diagnosis of Venous Thromboembolism in Outpatients. Lancet
l999;353(9148):l90-5.

31

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

Sijens PE. Rapid Elisa Assay for Plasma D—Dimer in the Diagnosis of Segmental
and Subsegrnental Pulmonary Embolism: A Comparison with Pulmonary
Angiography. Thromb Haemost 2000;84(2)2156-159.

Quinn DA, Fogel RB, Smith CD, Laposata M, Taylor Thompson B, Johnson SM,
et al. D-Dimers in the Diagnosis of Pulmonary Embolism. Am J Respir C rit Care
Med l999;159(5 Pt l):1445-9.

Brimble KS, Ginsberg J S. Evaluation of the Combination of a Bedside D-Dimer
Assay and Enzyme-Linked Immunosorbent Soluble Fibrin Assay in Patients with
Suspected Venous Thromboembolism. Thromb Res l997;88(3):291-7.

Duet M, Benelhadj S, Kedra W, Vilain D, Ajzenberg C, Elkharrat D, et al. A New
Quantitative D-Dimer Assay Appropriate in Emergency: Reliability of the Assay
for Pulmonary Embolism Exclusion Diagnosis. Thromb Res l998;91(1):l-5.

Reber G, Vissac AM, de Moerloose P, Bounameaux H, Amiral J. A New, Semi-
Quantitative and Individual Elisa for Rapid Measurement of Plasma D-Dimer in

Patients Suspected of Pulmonary Embolism. Blood Coagul F ibrinolysis
1995;6(5):460-3.

Bonnin F, Hadjikostova H, Jebrak G, Denninger MH, Vera P, Rufat P, et al.
Complementarity of Lung Scintigraphy and D-Dimer Test in Pulmonary
Embolism. Eur J Nucl Med l997;24(4):444-7.

van Beek EJ, Schenk BE, Michel BC, van den Ende B, Brandjes DP, van der
Heide YT, et al. The Role of Plasma D-Dimers Concentration in the Exclusion of

Pulmonary Embolism [Published Erratum Appears in Br J Haematol 1996
Oct;95(1):218]. Br J Haematol l996;92(3):725-32.

Rochemaure J, Laaban JP, Achkar A, Fretault J, Samama M. [Value of the
Determination of D-Dimers in the Diagnostic Approach of Venous Thrombo-
Embolic Disorders]. Bull Acad Natl Med l995;179(2):299-314; discussion 314-6.

Flores J, Lancha C, Perez Rodriguez E, Garcia Avello A, Bollo E, Garcia Frade
LJ. Efficacy of D-Dimer and Total Fibrin Degradation Products Evaluation in
Suspected Pulmonary Embolism. Respiration l995;62(5):258-62.

Goldhaber SZ, Simons GR, Elliott CG, Haire WD, Toltzis R, Blacklow SC, et al.
Quantitative Plasma D-Dimer Levels among Patients Undergoing Pulmonary
Angiography for Suspected Pulmonary Embolism. JAMA l993;270(23):2819-22.

van Beek EJ, van den Ende B, Berckmans RJ, van der Heide YT, Brandjes DP,
Sturk A, et al. A Comparative Analysis of D-Dimer Assays in Patients with
Clinically Suspected Pulmonary Embolism. Thromb Haemost l993;70(3):408-l 3.

Bounameaux H, Schneider PA, Slosman D, de Moerloose P, Reber G. Plasma D-
Dimer in Suspected Pulmonary Embolism: A Comparison with Pulmonary

32

, 52.

53.

54.

55.

56.

57.

58.

59.

60.

61.

62.

Angiography and Ventilation--Perfusion Scintigraphy. Blood Coagul F ibrinolysis
1990; l (4-5):577-9.

Lenzhofer R, Wimmer F, Haydl H, Kardeis J, Gruber G, Ganzinger U, et a1.
[Prospective Study of Determining the Value of D-Dimer in Diagnosing
Pulmonary Embolism]. Wien Klin Wochenschr 1993;105(17):492-6.

Ginsberg J S, Brill-Edwards PA, Demers C, Donovan D, Panju A. D-Dimer in
Patients with Clinically Suspected Pulmonary Embolism. Chest
1 993;l 04(6): 1679-84.

de Moerloose P, Minazio P, Reber G, Perrier A, Bounameaux H. D-Dimer
Determination to Exclude Pulmonary Embolism: A Two-Step Approach Using
Latex Assay as a Screening Tool. Thromb Haemost l994;72(1):89-91.

de Moerloose P, Desmarais S, Bounameaux H, Reber G, Perrier A, Dupuy G, et
a1. Contribution of a New, Rapid, Individual and Quantitative Automated D-
Dimer Elisa to Exclude Pulmonary Embolism. Thromb Haemost l996;75(l):l 1-3.

Perrier A, Desmarais S, Goehring C, de Moerloose P, Morabia A, Unger PF, et al.

D-Dimer Testing for Suspected Pulmonary Embolism in Outpatients. Am J Respir
Crit Care Med 1997;156(2 Pt l):492-6.

Tardy B, Tardy-Poncet B, Viallon A, Lafond P, Page Y, Venet C, et al.
Evaluation of D-Dimer Elisa Test in Elderly Patients with Suspected Pulmonary
Embolism. Thromb Haemost l998;79(l):38-4l.

Meyer G, Fischer AM, Collignon MA, Benazzouz A, Monge F, Sors H, et al.
Diagnostic Value of Two Rapid and Individual D-Dimer Assays in Patients with
Clinically Suspected Pulmonary Embolism: Comparison with Microplate
Enzyme-Linked Immunosorbent Assay. Blood Coagul F ibrinolysis
l998;9(7):603-8.

Barro C, Bosson JL, Pernod G, Carpentier PH, Polack B. Plasma D-Dimer
Testing Improves the Management of Thromboembolic Disease in Hospitalized
Patients. Thromb Res l999;95(5):263-9.

Wicki J, Pemeger TV, Junod AF, Bounameaux H, Perrier A. Assessing Clinical
Probability of Pulmonary Embolism in the Emergency Ward: A Simple Score.
Arch Intern Med 2001;16l(l):92-7.

Oxman AD, Cook DJ, Guyatt GH. Users' Guides to the Medical Literature. Vi.
How to Use an Overview. Evidence-Based Medicine Working Group. JAMA
1994;272(l7):1367-71 .

Righini M, Goehring C, Bounameaux H, Penier A. Effects of Age on the
Performance of Common Diagnostic Tests for Pulmonary Embolism. Am J Med
2000;109(5):357-61.

33

63.

64.

65.

66.

Bates SM, Grand'Maison A, Johnston M, Naguit I, Kovacs MJ, Ginsberg J S. A
Latex D-Dimer Reliably Excludes Venous Thromboembolism. Arch Intern Med
2001 ;161(3):447-53.

Kline JA, Israel EG, Michelson EA, O'Neil BJ, Plewa MC, Portelli DC.
Diagnostic Accuracy of a Bedside D-Dimer Assay and Alveolar Dead-Space

Measurement for Rapid Exclusion of Pulmonary Embolism: A Multicenter Study.
JAMA 2001;285(6):761-8.

Cook DJ, Sackett DL, Spitzer WO. Methodologic Guidelines for Systematic
Reviews of Randomized Control Trials in Health Care from the Potsdam
Consultation on Meta-Analysis. J Clin Epidemiol 1995;48(l):167-7l.

Goldstein NM, Kollef MH, Ward S, Gage BF. The Impact of the Introduction of a
Rapid D-Dimer Assay on the Diagnostic Evaluation of Suspected Pulmonary
Embolism. Arch Intern Med 2001;16l(4):567-7l .

34

 

l‘l’

1[122122211122111:31]}: