| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
(Stroke. 2000;31:3067.)
© 2000 American Heart Association, Inc.
Progress Review |
From the Department of Medicine (Neurology), Duke Center for Cerebrovascular Disease, Center for Clinical Health Policy Research, Duke University, and Durham VA Medical Center, Durham, NC.
Correspondence to Larry B. Goldstein, MD, Duke Center for Cerebrovascular Disease, Department of Medicine (Neurology), PO Box 3651, Durham, NC 27710. E-mail Golds004{at}mc.duke.edu
| Abstract |
|---|
|
|
|---|
Summary of ReviewWe performed a MEDLINE search to identify
controlled studies published during 19661999 that reported the
prevalence of deficiencies of protein C, protein S, antithrombin III,
plasminogen, activated protein C resistance
(APCR)/factor V Leiden mutation (FVL), anticardiolipin antibodies
(ACL), or lupus anticoagulant (LA) in patients with
ischemic stroke. The cumulative prevalence rates (pretest
probabilities) and positive likelihood ratios for all studies and for
those including only patients aged
50 years were used to calculate
posttest probabilities for each coagulopathy, reflecting
diagnostic yield. The cumulative pretest probabilities of
coagulation defects in ischemic stroke patients are as follows:
LA, 3% (8% for those aged
50 years); ACL, 17% (21% for those aged
50 years); APCR/FVL, 7% (11% for those aged
50 years); and
prothrombin mutation, 4.5% (5.7% for those aged
50 years). The
posttest probabilities of ACL, LA, and APCR increased with increasing
pretest probability, the specificity of the tests, and features of the
patients history and clinical presentation.
ConclusionsThe pretest probabilities of coagulation defects in ischemic stroke patients are low. The diagnostic yield of coagulation tests may be increased by using tests with the highest specificities and by targeting patients with clinical or historical features that increase pretest probability. Consideration of these data might lead to more rational ordering of tests and an associated cost savings.
Key Words: cerebral infarction coagulation decision analysis diagnosis
| Introduction |
|---|
|
|
|---|
Before ordering any diagnostic test, the physician should first consider its diagnostic yield (ie, the physician should consider whether the results add to the evaluation, potentially altering therapy and improving patient outcome). The diagnostic yield is a direct reflection of the positive predictive value or posttest probability of the test.2 This is defined as the proportion of patients with positive tests that have the diagnosis (in this case, a coagulopathy). To calculate this estimate, the pretest probability (prevalence of the disease in the target population) and the positive likelihood ratio (LR), which is based on the sensitivity and specificity of the test, must be determined.2 Although topic reviews have been published previously,3 4 5 6 7 this type of diagnostic yield assessment has not been systematically applied to coagulation testing in ischemic stroke patients.
The aim of this review is to compile the available data necessary to estimate posttest probabilities for coagulation tests in general ischemic stroke patients and to indicate where these data are lacking. Separate probabilities are estimated for younger patients who have been traditionally considered at higher risk for coagulopathies.4 5 To accomplish these aims, the currently available diagnostic tests as well as the physiological or pharmacological factors that may interfere with their interpretation are critically reviewed. Next, the reported prevalence rates and odds ratios (ORs) from controlled studies of inherited and acquired coagulopathies in these populations are summarized. Positive LRs are then calculated with the use of assessments of the sensitivities and specificities of the tests. The prevalence rates and LRs are subsequently used for estimation of the posttest probabilities for different subpopulations of patients. On the basis of these data, we suggest a scheme for coagulation testing that should prove useful to physicians evaluating ischemic stroke patients.
| Methods |
|---|
|
|
|---|
50 years and for
all age ranges. These prevalence rates, in addition to the LRs,
provided data for calculation of the posttest probability estimates
described below. Whenever possible, the sensitivity and specificity for an individual diagnostic test were used to calculate the positive likelihood ratio [LR+=Sensitivity/(1-Specificity), expressed as a percentage]. This result was then used to calculate the posttest probability for the corresponding diagnostic test [Pretest Odds=Probability/(1-Probability), where probability is equal to the prevalence of the disease in the target population; Posttest Odds=Pretest OddsxLR for Desired Result; Posttest Probability=Posttest Odds/(Odds+1)].2
| Results |
|---|
|
|
|---|
|
Protein C, Protein S, ATIII, and Plasminogen Deficiencies
Hereditary coagulation defects are evaluated in a similar fashion
by screening functional assays, followed by quantitative
assays.3 However, accurate diagnosis of these defects is
quite complex, since the functional assays are influenced by acute
thrombosis or acute phase reactants. As a result, if the tests are to
be performed, patients should be tested at least 3 months after an
acute thrombotic event and have no ongoing active illness that may
artificially affect the studies. In addition, many other medical
conditions and medications can potentially influence the test results
(Table 1
).9 10 11 12 13 14 15 Abnormal tests must be repeated and
confirmed with the appropriate functional or quantitative assays. If a
patients family history is suspicious for venous thrombosis, then
family members should be considered for testing, because PC, PS, and
ATIII deficiencies have autosomal dominant patterns of
inheritance.3 16
Activated Protein C Resistance
In approximately 95% of patients with APCR, the condition is due
to the factor V Leiden (FVL) mutation.17 This is diagnosed
with a screening assay for APCR followed by polymerase chain reaction
(PCR) for the FVL mutation or by PCR alone. Screening for APCR can be
accomplished with the original or modified activated partial
thromboplastin time ratio (predilution of patient plasma with factor
Vdepleted plasma).18 The most commonly reported APCR
ratio cutoff is 2.1,19 but the ratio is determined by the
individual laboratory performing the test. The original APCR screening
test can be influenced by multiple physiological,
pharmacological, and hematologic factors (Table 1
). If the APCR
is abnormal and a hereditary diagnosis is suspected, ethnic variations
in the prevalence of FVL mutation should be
considered.20
Prothrombin Gene Mutation
The only test available for detection of the prothrombin gene
mutation is the PCR test for the guanine-to-arginine mutation at
position 20210 of the 3'-untranslated region of this
gene.21 The mechanism of hypercoagulability due to the
mutation is still not known but is thought to be related to increased
amounts of thrombin formation once thrombin generation is
triggered.22 Plasma prothrombin levels have been
correlated with the mutation in homozygous and heterozygous
individuals, but to our knowledge, prothrombin levels have not been
evaluated for sensitivity or specificity in comparison to the genetic
test.23 24
Lupus Anticoagulant
The laboratory criteria for establishing a diagnosis of LA have
recently been revised.25 Laboratory diagnosis of LA is
complex because multiple LA tests measuring different types of
phospholipids are available.26 The
heterogeneity of these tests was demonstrated in a
study reporting sensitivities of 62% to 100%, depending on the cutoff
value chosen.27 Therefore, at least 2 different screening
assays should be negative before the diagnosis is excluded (eg,
activated partial thromboplastin time and direct Russell viper
venom time).25 The results are also dependent on the
reagents used,28 with both the tests and the reagents
varying among laboratories.27 Table 1
also lists
factors that may influence LA screening test results. Of note, LA test
results can still be interpreted in warfarin-treated patients, provided
that mixing studies are performed (Thomas Ortel, MD, oral
communication, March 2000).
Anticardiolipin Antibodies
The standard screening assay for measuring ACL is based on an
enzyme-linked immunosorbent assay (ELISA). The ACL ELISA can be
false-positive in the setting of infection, hypergammaglobulinemia,
rheumatoid factor, and heat treatment of serum samples (Table 1
).29 In addition, titers increase with increasing
age and multiparity (Table 1
).29 Unlike LA, which
is based on a functional assay, the ACL ELISA is not influenced by
concurrent anticoagulation with heparin. In the diagnostic
evaluation of antiphospholipid (APL) syndrome, both LA and ACL should
be obtained, because patients may have either one or both tests
positive at the same time.30
Prevalence Rates of Coagulopathies in Ischemic Stroke
Patients
Given this background of limitations in test interpretation,
a total of 107 studies of coagulopathies in ischemic stroke
patients were identified. Fourteen case reports and 38 case series were
excluded on the basis of the listed criteria; therefore, 55 controlled
studies were included in the analysis of prevalence rates, ORs,
and pretest and posttest probabilities. Five studies providing
sensitivities and specificities of ACL, LA, and APCR were used to
calculate positive LRs and posttest probabilities.
Hereditary Coagulation Defects
Only 4 case-control studies of the prevalence of hereditary
deficiencies of PC, PS, and ATIII in ischemic stroke were
identified, ranging from 0% to 21% (Table 2
).31 32 33 34 Case-series
studies reported prevalence rates of 0% to 23%.32 35 36 37 38 39 40
Hereditary fibrinolytic defects such as plasminogen
deficiency have a reported prevalence of 0% to 2.7% in case-control
and case-series studies of ischemic stroke patients aged <45
years.32 35 37 39 These data do not permit calculation of
cumulative prevalence rates.
|
Activated Protein C Resistance/Factor V Leiden
Mutation
Data related to APCR prevalence range from 0% to 38% in studies
of ischemic stroke (Appendix 1).24 31 32 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68
However, of the 31 studies measuring APCR and/or FVL mutation, 5
studies (2 studies of APCR49 68 and 3 studies of
FVL43 51 61 ) found significantly increased odds in
ischemic stroke patients compared with controls (Appendix
1).
The 5 studies that showed a significant association between ischemic stroke and APCR/FVL all had features that limit generalization to unselected ischemic stroke populations.43 51 59 61 68 These limitations include cases identified with obsolete APCR assays that have been modified since elucidation of the genetic mutation,68 highly selected patients with transient ischemic attack only,61 patients aged 0 to 10 years,51 young adults with stroke referred for thrombophilic workup,43 and use of the modified APCR assay without confirmation with PCR for FVL.49
Evidence against an association between FVL and ischemic stroke comes from 2 well-designed prospective studies. The Physicians Health Study followed subjects over many years recording the incidence of venous thrombosis, ischemic stroke, and myocardial infarction. A nested case-control study of this cohort, matched for age, smoking habits, and time since randomization, showed a similar prevalence of FVL for those with strokes (4.3%) and controls without cardiovascular disease (6%).64 The Cardiovascular Health Study also failed to find an association between FVL and ischemic stroke in elderly patients.44 Therefore, although some studies have suggested a trend toward an association between APCR/FVL and ischemic stroke,41 58 62 the majority of the identified studies did not report a significant association and had significant methodological limitations (Appendix 1).
Given these limitations, a cumulative prevalence rate (pretest
probability) of 7% for APCR/FVL was calculated (Table 4
). It
should be noted that this cumulative prevalence rate was derived from a
large, heterogeneous population of patients with a lower
than expected prevalence rate in controls (4.4%). Several studies
excluded patients with traditional stroke risk
factors32 45 49 56 or a history of
thromboembolism.61 Others included patients with only
minor stroke or transient ischemic attack41 or
transient ischemic attack alone.61 Although the OR
appears to be increased in patients aged <50 years (OR, 3.1; Table 3
), patients within this younger group
were also highly selected.32 45 49 61
|
|
Prothrombin Gene Mutation
The prevalence of prothrombin gene mutation in ischemic
stroke patients varied from 1% to 12.5% (Appendix
1).* Of the 13 relevant
studies, 2 found a significant association with ischemic stroke
(Appendix 1)45 69 ; however, the patients were highly
selected. One of the studies included patients with first stroke before
age 50 years, with no history of hypertension, diabetes,
hypercholesterolemia, or
hypertriglyceridemia, and no age-matched
controls.45 The other study included a selected population
of patients with prior myocardial infarction who subsequently had
venous or arterial events. The controls were consecutive
newborns and not age matched.69 In contrast to these
studies with highly selected patients, the majority of the case-control
studies identified (Appendix 1) showed no association between the
prothrombin mutation and ischemic stroke. The cumulative
prevalence of this mutation was 5.7% in studies of those aged
50
years and 4.5% in studies including all age ranges, with a slightly
increased odds of ischemic stroke (1.4; 95% CI, 1.03 to 1.9;
Table 3
).
Anticardiolipin Antibodies/Lupus Anticoagulant
Studies of the prevalence of APL antibodies in ischemic
stroke patients were often reported for either ACL or LA alone.
Appendix 2 shows the wide variability in the reported prevalence
of these antibodies in ischemic stroke patients aged
50
years74 75 76 77 and in studies including patients of all
ages.74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 Eight of 15 controlled studies of ACL or LA
(Appendix 2) found a significant association with ischemic
stroke.77 78 80 81 83 86 87 88 The association between
ACL/LA and ischemic stroke has been well established in
patients with primary APL syndrome and with systemic lupus
erythematosus (SLE)associated secondary APL
syndrome, but a growing body of evidence supports an association in
unselected patients as well.89 A large multiethnic
case-control study recently showed that prevalence rates of ACL in
ischemic stroke were similarly increased in whites (OR, 4.7;
95% CI, 2.3 to 8.2), blacks (OR, 4.0; 95% CI, 2.4 to 6.6), and
Hispanics (OR, 4.0; 95% CI, 1.8 to 8.8).78 In contrast,
the Physicians Health Study reported no significant association
between ACL and ischemic stroke in men.84 However,
the ACL ELISA assay was performed on plasma frozen for
8 years, and
therefore the reliability of the test is uncertain.78
Despite the relationship between detection of ACL and ischemic stroke, the specificity of these antibodies is unclear. DOlhaberriague et al80 reported a significantly increased odds of ACL positivity in ischemic stroke patients compared with patients with other neurological diseases, suggesting that the antibodies are specific to stroke. However, a more recent study by the same group showed that ACL IgG titers may be a general marker of cardiovascular disease.90 Tanne et al91 reported that ACL IgG titers are associated with the presence of multiple stroke risk factors such as age >65 years, atrial fibrillation, valvular heart disease, and congestive heart failure. The increase in ACL prevalence with age is documented in healthy subjects as well.92 93 These results suggest that the ACL IgG titer may represent a risk marker for atherosclerosis but may not be a specific causal factor.
Comparisons among ACL prevalence studies are fraught with difficulty for at least 3 reasons. First, neither the ACL ELISA test94 nor the selection of a cutoff value for a positive test has been standardized among laboratories.95 The significance of IgM titers is also unclear because they may be elevated in diseases other than stroke.80 Second, the diagnosis of ACL or LA positivity must be confirmed with repeated testing3 94 since another important feature of ACL and LA is that the titers can fluctuate over time.96 97 Only 2 studies87 88 mentioned confirmation of abnormal titers. Third, in addition to the widely recognized groups of patients with either SLE-associated or primary APL syndrome, there appears to be a third category of patients with ACL in the setting of multiple atherosclerotic risk factors.78 Prospective studies are needed to ascertain the significance of fluctuations of antibodies over time and the prognosis of further arterial events in this latter group.78
The cumulative prevalence rates of ACL and LA in patients of all ages
and in those aged
50 years are listed in Table 4
. These rates for ACL in
ischemic stroke patients were relatively high (all ages, 17%;
aged
50 years, 21%), with an increased odds of ischemic
stroke with ACL (Tables 3
and 4
).77 78 80 81 83 86
Estimation of Posttest Probabilities
With allowances for the limitations reviewed in the previous
sections, Table 4
lists the calculated posttest probabilities
for specific coagulopathies and available coagulation tests based on
the indicated pretest probabilities, sensitivities, specificities, and
positive LRs.18 19 98 Of note, for the combined LA
screening and confirmatory tests, the sensitivity was 67% and the
specificity was 100%, resulting in a LR+ approaching
infinity.99 Therefore, an arbitrary value of 1000 was used
for this calculation. Table 4
also shows that the pretest
probabilities for ACL and APCR, but not LA, increased by 4% in
patients aged
50 years and that the highest posttest probabilities
resulted from tests with the highest specificities. Because of the
paucity of data on the prevalence of deficiencies of PC, PS, ATIII, and
plasminogen in ischemic stroke, combined with a
lack of sensitivity and specificity data for these coagulation tests,
pretest and posttest probabilities could not be calculated.
Coagulation Tests and Clinical Decision Making in Patients With
Ischemic Stroke
Using data from controlled studies, we calculated cumulative
prevalence rates (pretest probabilities) that ranged from 3% to 21%
(Table 4
). A diagnostic test is typically most
helpful with a pretest probability in the range of 40% to
60%.2 This represents a situation in which the
clinician is undecided about the diagnosis, and the
diagnostic test will aid in decisions concerning further
testing or lead to a change in treatment.2 However, the
posttest probability decreases as the pretest probability decreases.
Because none of the coagulopathies had a cumulative pretest probability
greater than approximately 20%, unselected use of coagulation tests is
unlikely to lead to a change in diagnostic or treatment
strategies.
Given this limitation, a proposed scheme based on the calculated
pretest and posttest probabilities in conjunction with patient-related
factors for each coagulopathy is given in the
Figure
. Because data are lacking
for estimating pretest (and therefore posttest) probabilities for the
hereditary coagulation defects in ischemic stroke patients, the
left column lists the general features of these
disorders.95 3 The testing strategy for these potential
coagulopathies is based on evidence not from controlled trials but from
the practical issues related to coagulation testing as recommended by
coagulation experts.3 95 The middle and right columns
include the cumulative pretest probabilities and associated common
features for ACL/LA (primary or SLE-associated APL
syndrome)89 and APCR/FVL/prothrombin (FII) mutations,
respectively.100 101
|
The strategy of when to test patients after ischemic stroke
applies to the screening tests that are uninterpretable in the setting
of acute phase reactants or infections, such as the functional tests
for hereditary deficiencies, ACL ELISA, LA screening tests, and APCR
screening tests (Table 1
). Therefore, these tests should
generally be performed when the patient is not in an active thrombotic
state and the coagulation and fibrinolytic factors have stabilized,
usually 6 to 8 weeks after the thrombotic event.3 4 15 16
Genetic tests such as the PCR for FVL and prothrombin gene mutation are
not influenced by acute thrombosis.
| Discussion |
|---|
|
|
|---|
Despite the systematic nature of this review, combining heterogeneous studies, ie, those with varying patient populations, leads to limitations in interpretation of the results. Including only studies of unselected ischemic stroke patients would have provided conclusions that are more generalizable. However, combining available prevalence studies of both selected and unselected patients provides larger numbers for calculating the estimates, and many clinicians select patients in a similar manner when justifying further testing.
Another limitation is that the sensitivity and specificity
measurements, crucial for the calculation of posttest probabilities,
were available for only a few diagnostic
studies.18 19 98 99 102 103 Most of these studies did not
meet the methodological standards considered necessary for
diagnostic test research.104 Documentation of
the spectrum of patients tested is an important methodological
consideration,104 and the studies of ACL and LA
diagnostic tests were limited to patients with secondary
APL syndrome in the setting of SLE or to patients with primary APL
syndrome.102 103 To our knowledge, these measurements have
not been performed in unselected populations of seropositive patients
with other atherosclerotic risk factors for ischemic stroke.
Adequate sample size is another methodological standard, and 2 studies
of ACL/LA tests made assessments using only a small number of
patients.99 102 The results of studies with
30 patients
will often produce sensitivity and specificity measurements with wide
CIs and poorer test accuracy.104 Test reproducibility is a
very important methodological consideration, and this is a significant
problem in the diagnostic testing for ACL. Although the
cited studies used an average of multiple test results from the same
laboratory,102 103 the potential for interlaboratory
variability was not addressed.94 On the basis of these
methodological issues, the posttest probabilities calculated in this
study should be viewed as rough estimates.
Information that could potentially increase the pretest probability
includes clinical or historical factors, physical examination findings,
and other diagnostic test results. Because of the low
prevalence of coagulopathies in the general population of stroke
patients, these factors formed the basis for development of the scheme
for coagulation testing shown in the Figure
. For example, the
pretest probability of a positive LA and/or ACL may be 40% to 60% in
a 30-year-old woman admitted with an ischemic stroke who has a
history of miscarriages and marantic endocarditis (3 features of the
APL syndrome; Figure
). In contrast, the pretest probability of
LA and/or ACL may be only 20% in a 65-year-old woman with an
ischemic stroke who has 1 or 2 traditional stroke risk
factors.78 80 The screening ELISA produces a better yield
for the 30-year-old woman with 3 features of APL syndrome because these
historical features increase her pretest probability, thereby resulting
in a in a posttest probability of 60% (Table 4
).
Deciding on a pretest probability estimate for an individual patient is difficult, but studies have shown that this skill can be improved with clinical experience in the target disorder105 or with the use of nomograms or computerized algorithms.106 Another useful method is to choose a range of pretest probabilities and calculate a posttest probability for the middle, upper, and lower ranges.2 This type of sensitivity analysis can help the clinician to appreciate the spectrum in which the coagulation test might be useful. Once the decision is made to pursue the diagnosis, selecting a test with a high specificity will result in a high posttest probability. The diagnostic yield in this situation is optimized by the use of the quantitative approach of pretest and posttest probabilities.
Conclusion
Because of the lack of strong evidence supporting an association
between most coagulopathies and ischemic stroke in general
populations of patients, the difficulty with interpretation of these
tests in the setting of an acute ischemic stroke, and the
methodological problems with the reviewed prevalence studies, our
conclusion is that coagulation tests are often of little value in the
evaluation of general patients with ischemic stroke. On the
basis of the current data, the low diagnostic yield
necessitates careful consideration before these tests are obtained.
Additional prospective, controlled studies of unselected
ischemic stroke patients are needed to better assess the roles
of hereditary coagulation defects, APCR/FVL, prothrombin gene mutation,
and ACL/LA in the etiology of ischemic stroke. The suggested
scheme developed on the basis of the results of this comprehensive
literature review should be further validated in prospective
studies.
| Appendix 1 |
|---|
|
|
|---|
|
| Appendix 2 |
|---|
|
|
|---|
|
| Acknowledgments |
|---|
| Footnotes |
|---|
Received March 9, 2000; revision received August 10, 2000; accepted August 10, 2000.
| References |
|---|
|
|
|---|
A and the MTHFR C677T mutations in childhood
stroke. Thromb Haemost. 1999;2021:81:690694.
Gln) mutation
in ischemic cerebrovascular disease. Haemostasis. 1997;27:105111.[Medline]
[Order article via Infotrieve]
glutamine mutation. Stroke. 1996;27:11631166.
A prothrombin polymorphism
and cerebral ischemia in the young. Stroke. 1997;28:18461847.