Grubb, Jr RL, Derdeyn CP, Fritsch SM, Carpenter DA, Yundt KD, Videen TO, Spitznagel EL, Powers WJ. Importance of Hemodynamic Factors in the Prognosis of Symptomatic Carotid Occlusion. JAMA. 1998;280(12):1055-1060. doi:10.1001/jama.280.12.1055
From the Department of Neurology and Neurological Surgery (Drs Grubb, Derdeyn, Carpenter, Yundt, Videen, and Powers and Ms Fritsch), the Edward Mallinckrodt Institute of Radiology (Drs Grubb, Derdeyn, and Powers), and the Department of Mathematics (Dr Spitznagel), Washington University, and the Lillian Strauss Institute of Neuroscience of the Jewish Hospital of St Louis (Dr Powers), St Louis, Mo. Dr Carpenter is now at St John's Hospital, St Louis.
Context.— The relative importance of hemodynamic factors in the pathogenesis and
treatment of stroke in patients with carotid artery occlusion remains controversial.
Objective.— To test the hypothesis that stage II cerebral hemodynamic failure (increased
oxygen extraction measured by positron emission tomography [PET]) distal to
symptomatic carotid artery occlusion is an independent risk factor for subsequent
stroke in medically treated patients.
Design and Setting.— Prospective, blinded, longitudinal cohort study of patients referred
from a group of regional hospitals between 1992 and 1996.
Patients.— From 419 subjects referred, 81 with previous stroke or transient ischemic
attack in the territory of an occluded carotid artery were enrolled. All were
followed up to completion of the study, with average follow-up of 31.5 months.
Main Outcome Measures.— Telephone contact every 6 months recorded the subsequent occurrence
of all stroke, ipsilateral ischemic stroke, and death.
Results.— Stroke occurred in 12 of 39 patients with stage II hemodynamic failure
and in 3 of 42 patients without (P=.005); stroke
was ipsilateral in 11 of 39 patients with stage II hemodynamic failure and
in 2 of 42 patients without (P=.004). Six deaths
occurred in each group (P=.94). The age-adjusted
relative risk conferred by stage II hemodynamic failure was 6.0 (95% confidence
interval [CI], 1.7-21.6) for all stroke and 7.3 (95% CI, 1.6-33.4) for ipsilateral
Conclusions.— Stage II hemodynamic failure defines a subgroup of patients with symptomatic
carotid occlusion who are at high risk for subsequent stroke when treated
medically. A randomized trial evaluating surgical revascularization in this
high-risk subgroup is warranted.
THE RELATIVE IMPORTANCE of hemodynamic as opposed to thromboembolic
mechanisms in the pathogenesis of ischemic stroke remains unsettled.1 This distinction is moot for severe symptomatic carotid
stenosis since carotid endarterectomy has been demonstrated to reduce the
risk of subsequent stroke.2 However, there
remains a large number of patients with carotid artery occlusion who comprise
approximately 15% of those with carotid territory transient ischemic attacks
or infarction.3- 6
The overall risk of subsequent stroke is 5% to 7% per year and the risk of
stroke ipsilateral to the occluded carotid artery is 2% to 6% per year.7,8 No surgical treatment has been proven
to be of benefit in preventing subsequent stroke. The efficacy of anticoagulant
treatment or antiplatelet agents in this particular subgroup is not known.
An important role for hemodynamic mechanisms in these patients has been proposed.7 However, studies addressing this issue have either
failed to show that cerebral hemodynamics is important or suffered from potential
bias due to problems in experimental design.1,9
To determine what role hemodynamic factors play in the prognosis and
treatment of patients with carotid artery occlusion, methods for determining
the hemodynamic status of the cerebral circulation, accurately and in awake
subjects under normal conditions, must be available. Measurements of cerebral
blood flow (CBF) alone are inadequate because they cannot distinguish reduced
CBF caused by the hemodynamic effects of arterial occlusion from compensatory
physiological reductions in CBF caused by reduced metabolic demands. It has
been necessary, therefore, to rely on indirect assessments based on the compensatory
responses made by the brain to progressive reductions in cerebral perfusion
pressure (CPP). When CPP is normal (stage 0), CBF is closely matched to the
resting metabolic rate of the tissue. As a consequence of this resting balance
between flow and metabolism, the oxygen extraction fraction (OEF) shows little
regional variation. Moderate reductions in CPP have little effect on CBF.
Vasodilation of arterioles reduces cerebrovascular resistance, thus maintaining
a constant CBF (stage I). As a consequence, the intravascular cerebral blood
volume (CBV) is elevated. This phenomenon is known as cerebrovascular autoregulation.
With more severe reductions in CPP, the capacity for compensatory vasodilation
is exceeded, autoregulation fails, and CBF begins to decline. A progressive
increase in OEF now maintains cerebral oxygen metabolism and brain function
(stage II).10,11 This more severe
form of stage II cerebral hemodynamic failure has also been termed misery perfusion.12
Two basic approaches have been used to assess regional cerebral hemodynamics
in humans. The first approach is based on detecting stage I autoregulatory
vasodilation by either measuring CBF and CBV or determining whether there
is reduced responsiveness of CBF to a vasodilatory stimulus (such as hypercapnia
or acetazolamide). A variety of techniques have been used to make the paired
measurements of cerebral perfusion needed for evaluating the vasodilatory
response.7,10 The second approach
is based on detecting more severe stage II hemodynamic failure by measuring
increases in regional OEF.10 This second approach
is currently possible only with positron emission tomography (PET). In a previous
study using historical controls, we failed to demonstrate a relationship between
stage I autoregulatory vasodilation and the subsequent risk of stroke.13 The current blinded prospective study tests the hypothesis
that stage II hemodynamic failure (increased oxygen extraction) in the cerebral
hemisphere distal to symptomatic carotid artery occlusion is an independent
predictor of the subsequent risk of stroke in medically treated patients.
We enlisted the collaboration of 15 hospitals within the St Louis, Mo,
area to assist with recruitment. Personnel at participating hospitals were
asked to notify the study coordinator about all patients with carotid artery
occlusion, irrespective of the presence or characteristics of cerebrovascular
symptoms. The study coordinator contacted each subject and explained the purpose
of the study. If the subject agreed to participate, clinical, laboratory,
and radiographic information necessary to determine eligibility was obtained.
Original inclusion criteria were (1) occlusion of one or both common or internal
carotid arteries demonstrated by contrast angiography within 120 days prior
to PET and (2) transient ischemic neurological deficits (including transient
monocular blindness) or mild-to-moderate permanent ischemic neurological deficits
(stroke) in appropriate carotid artery territory with last event occurring
within 120 days prior to PET. Following initiation of the study, we made the
following 2 changes in the inclusion criteria to improve recruitment: (1)
carotid occlusion could be demonstrated by either magnetic resonance (MR)
angiography or carotid ultrasound and (2) the 120-day limit for both demonstration
of occlusion and most recent symptom was waived. (At the time of this protocol
change, we also began to enroll asymptomatic subjects with carotid occlusion
into a parallel study.14) Exclusion criteria
were the following:
inability to give informed consent;
not legally an adult;
failure to meet the following functional standards:
self-care for most activities of daily living (may require some assistance),
some useful residual function in the affected arm or leg, intact language
comprehension, mild or absent motor aphasia, and ability to handle own oropharyngeal
nonatherosclerotic conditions causing or likely
to cause cerebral ischemia, including carotid dissection, fibromuscular dysplasia,
arteritis, blood dyscrasia, or heart disease as a source of cerebral emboli.
The latter included significant valvular disease (including mitral valve prolapse),
cardiac arrhythmia (especially atrial fibrillation), cardiomyopathy, and myocardial
infarction within 3 months preceding PET. Mitral annulus calcification, calcific
aortic stenosis, and patent foramen ovale were not considered exclusions.
The cardiac diagnostic assessment was based on information available from
clinical records rather than a standard protocol;
any morbid condition likely to lead to death within
subsequent cerebrovascular surgery planned that
might alter cerebral hemodynamics.
Patients who had undergone endarterectomy for stenosis of the ipsilateral
external carotid artery or contralateral internal carotid artery prior to
PET were eligible whether or not they had had recurrent symptoms. Any subsequent
cerebrovascular surgery after the initial PET caused the patient to be censored
from the study at the time of surgery.
All subjects were studied at Washington University Medical Center, St
Louis, Mo. Just prior to PET, each subject underwent neurological evaluation
including detailed questions regarding any symptoms. Focal ischemic symptoms
in the territory of the occluded carotid artery were categorized as cerebral
transient ischemic attack (<24 hours in duration), cerebral infarction
(>24 hours in duration), or retinal event (any duration) and as single or
recurrent episodes. Time from most recent symptom was recorded. Pertinent
medical records, carotid ultrasound reports, computed tomography (CT) scans,
magnetic resonance images (MRI), and MR and intra-arterial contrast angiograms
were reviewed. The following baseline risk factors were specifically determined:
age, sex, hypertension, previous myocardial infarction, diabetes mellitus,
smoking, alcohol consumption, and parental death from stroke. The degree of
contralateral carotid stenosis and collateral arterial circulation to the
ipsilateral middle cerebral artery (MCA) was determined from intra-arterial
angiograms, if available.15 These arteriograms
were done for clinical purposes at the participating institutions and varied
in the number of vessels injected and views obtained. Blood samples were collected
for determination of hemoglobin, fasting lipid levels (triglyceride, high-density
lipoprotein cholesterol, and low-density lipoprotein cholesterol), and fibrinogen
levels. A noncontrast CT scan of the brain was performed if a CT scan done
as part of usual clinical care did not permit accurate definition of infarction
location. This CT scan was used only to determine the site of tissue infarction
to exclude these regions from subsequent PET analysis.
Eighteen normal control subjects aged 19 to 77 years (mean [SD], 45
 years) were recruited by public advertisement. All were disease free
and taking no medication by their own history. There were 8 women and 10 men.
All underwent neurological evaluation, MRI of the head, and duplex ultrasound
imaging of the extracranial carotid arteries. None had signs or symptoms of
neurological disease other than mild distal sensory loss in the legs consistent
with age, pathological lesions on MR scan (mild atrophy and punctate asymptomatic
white matter abnormalities were not considered pathological), or more than
50% stenosis of the extracranial carotid arteries by duplex ultrasound.
PET studies were performed on 2 different scanners with similar sensitivity
and axial and transverse resolution (Siemens models 953B and 961, Siemens
Medical Systems, Hoffman Estates, Ill).16,17
All normal control subjects were studied with the model 961 scanner. The position
of the head relative to the plane of the PET scan was recorded by a lateral
skull film marked with a radio-opaque line. Each patient underwent a transmission
scan with gallium 68–germanium 68 rod sources to provide individual
attenuation data necessary for the quantitative reconstruction of subsequent
scans. Regional OEF was measured by the method of Mintun et al using H215O, C15O, and O15O.18,19
When technical difficulties precluded collection of arterial time-activity
curves necessary to determine quantitative OEF, the ratio image of the counts
in the raw H215O and O15O images was normalized
to a whole brain mean of 0.40 and substituted for the quantitative OEF image.
The counts in this H215O/O15O image are linearly
proportional to OEF except for small contributions from intravascular oxygen
and recirculating labeled water. The resultant errors are small (<5%) when
regional oxygen metabolism is normal as it was under these circumstances.18
Images were reconstructed using filtered back projection and scatter
correction with a ramp filter at the Nyquist frequency. All images were then
filtered with a 3-dimensional gaussian filter to a uniform resolution of 16
mm full width half maximum. For each subject, 7 spherical regions of interest,
each 19 mm in diameter, were placed in the cortical territory of the MCA in
each hemisphere using stereotactic coordinates.15,20
If any portion of a region overlapped a well-demarcated area of reduced oxygen
metabolism that corresponded to areas of infarction by CT scan or MRI, that
region and the homologous contralateral region were excluded. The mean OEF
for each MCA territory was calculated from the remaining regions and a left-to-right
MCA OEF ratio was calculated. The maximum and minimum ratios from the 18 normal
control subjects were used to define the normal range (0.914-1.082). A separate
range of normal for H215O/O15O images was
determined (0.934-1.062). Patients with left-to-right OEF ratios outside the
normal range were categorized as having stage II hemodynamic failure in the
hemisphere with higher OEF. These categorizations were made without knowledge
of the side of the carotid occlusion or of the clinical course of the patients
since the initial PET study. No information regarding the PET results was
provided to the patients, treating physicians, or investigator responsible
for determining end points.
Patients were followed up by the study coordinator for the duration
of the study through telephone contact every 6 months with the patient or
next of kin. The interval occurrence of any symptoms of cerebrovascular disease,
other medical problems, and functional status was determined. Interval medical
treatment on a monthly basis was recorded as warfarin (with or without other
medication), antiplatelet drugs (without warfarin), or no antithrombotic medication.
The occurrence of any symptoms suggesting a stroke was thoroughly evaluated
by 1 designated blinded investigator based on history from the patient or
eyewitness and review of medical records ordered by the patient's physician.
If necessary, follow-up examination and brain imaging were arranged. This
investigator (R.L.G.) remained blinded to the PET data. All living patients
were followed up for the duration of the study.
The primary end point was subsequent ischemic stroke defined clinically
as a neurological deficit of presumed ischemic cerebrovascular cause lasting
more than 24 hours in any cerebrovascular territory. Secondary end points
were ipsilateral ischemic stroke and death.
Subjects were divided into 2 groups, those with stage II hemodynamic
failure and those with normal (symmetric) OEF. Comparison of 17 baseline risk
factors (Table 1) and subsequent
medical treatment between the 2 groups was performed with unpaired t tests for continuous variables and χ2 analysis for
categorical variables. Uncorrected P values are reported.
No adjustment was made for increased type I error because of the multiplicity
of comparisons. The primary analysis compared the 2 groups with respect to
the length of time before reaching the primary end point by means of the Mantel-Cox
log-rank statistic and Kaplan-Meier survival curves. A value of P<.05 was used as the criterion of statistical significance. Secondary
end points were analyzed in a similar manner. The day of the PET scan was
considered to be the date of enrollment into the study. Survival analysis
of subsequent end points began at that time. No interim analysis was planned
or performed and no subgroup analyses were prespecified by the trial design.
The Cox proportional hazards model was used to test 20 candidate predictor
variables in a univariate analysis. All variables except medical treatment
were treated as time-constant variables whereas medical treatment was treated
as a time-dependent variable. All variables with P<.20
in the univariate analysis were included in a subsequent multivariate analysis.
Both forward and backward stepwise selection based on maximum partial likelihood
estimates were used. Those variables that were significant at P<.05 in the multivariate analysis were included in the final model.
Statistical analyses were performed with SPSS software, Version 7.0 (SPSS
Inc, Chicago, Ill) and SAS software, Version 6.12 (SAS Institute Inc, Cary,
We estimated that we would achieve 80% power to exclude a 4-fold difference
in the primary end point between groups with a sample size of 100 patients,
a 3-fold increase with a sample size of 150, and a 2-fold increase with a
sample size of 350. We projected enrolling a minimum of 250 patients over
5 years. This research was approved by the institutional review boards of
all participating hospitals. Written informed consent was obtained from all
Enrollment began on May 5, 1992. In July 1997, with funding due to expire
in 4 months, we made the decision to stop the study and analyze all subjects
enrolled by November 30, 1996, based on their status as of June 30, 1997.
As of November 30, 1996, 419 subjects had been referred for screening. Eighty-seven
subjects were enrolled in the study. Approximately four fifths of the remaining
subjects refused to participate and the other one fifth were willing to participate
but were ineligible. Of 87 patients who consented to participate, 81 successfully
underwent initial data collection and PET measurements and were enrolled in
the study. In 4 patients, no useful PET data were obtained because of technical
difficulties. In 2 patients, large infarctions made regional analysis of the
PET data impossible (Figure 1).
The diagnosis of carotid artery occlusion was made by intra-arterial
contrast angiography in 75 of the 81 subjects. In the remaining 6, carotid
artery occlusion was demonstrated by MR angiography in 4 and by carotid ultrasound
in 2. In 60 (75%) of 81 patients, carotid occlusion was demonstrated within
the 120-day period specified by the original protocol. Seventy-four (90%)
of 81 patients had demonstration within 1 year prior to PET. There were no
subjects with bilateral carotid occlusion. Prior to PET, 12 patients had undergone
endarterectomy of contralateral internal carotid artery stenosis and one had
undergone endarterectomy of ipsilateral external carotid artery stenosis.
Of 81 patients, 39 had stage II hemodynamic failure (increased OEF)
in one hemisphere and 42 did not. In all 39 patients with stage II hemodynamic
failure, the hemisphere with increased OEF was ipsilateral to the occluded
carotid. The 2 groups were well matched for most baseline risk factors (Table 1). Retinal symptoms were less common
in stage II subjects (3/39 vs 13/26). High-density lipoprotein cholesterol
levels were lower in stage II subjects (1.01 ±0.26 vs 1.16 ±0.39).
Stage II subjects spent a higher fraction of follow-up months on neither warfarin
nor antiplatelet treatment (0.07 vs 0.02). Arteriographic collateral circulation
did not permit distinction between the 2 groups (Table 2). Four subjects who underwent cerebrovascular surgery subsequent
to enrollment were censored at the time of surgery. Three of these 4 subjects
underwent contralateral carotid endarterectomy prior to occurrence of ipsilateral
ischemic stroke and were censored after being followed up for 13 months, 29
months, and 29 months, respectively. Two had not reached any end point and
1 had experienced a vertebrobasilar stroke. The fourth patient experienced
an ipsilateral stroke and underwent subsequent contralateral endarterectomy
at 13 months. All subjects were followed up until the end of the study or
until death (Figure 1).
Mean follow-up duration was 31.5 months. Fifteen total and 13 ipsilateral
ischemic strokes occurred. There were no hemorrhages. In the 39 stage II subjects,
12 total and 11 ipsilateral strokes occurred. In the 42 subjects with normal
OEF, there were 3 total and 2 ipsilateral strokes. The Kaplan-Meier estimates
for the risk of subsequent stroke at 1 and 2 years are given in Table 3. The risks of all stroke and ipsilateral ischemic stroke
in stage II subjects were significantly higher than in those with normal OEF
(P=.005 and .004, respectively; Figure 2). Twelve deaths occurred, 6 in each group (P=.94).
We performed a subgroup analysis of the 57 subjects who met original
entry criteria for symptoms within 120 days prior to PET. All strokes except
1 nonipsilateral stroke in a patient with normal OEF occurred in these patients.
In this subgroup, the risk of all stroke (P=.008)
and ipsilateral stroke (P=.02) was significantly
higher in the 31 stage II patients than in the 26 patients with normal OEF.
The univariate analysis of risk factors for the primary end point of
all stroke is shown in Table 4.
Six variables with P<.20 were entered into the
multivariate model (Table 5).
In the multivariate model, only age and stage II hemodynamic failure remained
significant independent predictors of all stroke. Similar univariate analysis
for ipsilateral ischemic stroke (data not shown) yielded 5 variables for entry
into the multivariate model (Table 5).
Again, only age and stage II hemodynamic failure remained significant independent
predictors of ipsilateral ischemic stroke. The age-adjusted relative risk
conferred by stage II hemodynamic failure was 6.0 (95% confidence interval
[CI], 1.7-21.6) for all stroke and 7.3 (95% CI, 1.6-33.4) for ipsilateral
Due to the previously described lower risk of stroke with retinal ischemia,
we wanted to determine whether the imbalance between the occurrence of retinal
symptoms in the 2 groups could explain our results.21
However, since no strokes occurred in these 16 subjects during follow-up,
it was not possible to use the Cox proportional hazards method for this purpose.
We therefore performed a subgroup analysis excluding the 16 patients with
retinal symptoms. In the remaining 65 subjects, the risk of stroke was significantly
higher in stage II subjects (12/36) than in subjects with normal OEF (3/29, P=.04). Similarly, the risk of ipsilateral ischemic stroke
was also significantly higher in stage II subjects (11/36) than in subjects
with normal OEF (2/29, P=.03).
We have demonstrated that stage II hemodynamic failure (increased oxygen
extraction) distal to a symptomatic occluded carotid artery is an independent
predictor of subsequent ischemic stroke. This study was prospective and blinded
and addressed the possible effect of treatment and other risk factors for
stroke. As with any study that requires informed consent, these patients did
not constitute a consecutive series and thus there remains the possibility
of some bias in the selection because of the high refusal rate, which might
limit the generalizability of the conclusions. However, the rates for stroke
and ipsilateral ischemic stroke in the total group of 81 patients are similar
to those reported by others and the risk factor profile is typical for patients
with carotid artery disease.2,8,22
Following initiation of the study, we waived the 120-day limit for symptoms
and documentation of carotid occlusion in an attempt to improve recruitment.
In retrospect, this action was not necessary and had little effect on our
results. All strokes except 1 nonipsilateral stroke in a patient with normal
OEF occurred in the 57 subjects who met original entry criteria for symptoms
within 120 days prior to PET. In this subgroup, the risks of all stroke (P=.008) and ipsilateral stroke (P=.02)
were significantly higher in stage II patients than in those with normal OEF.
We also included subjects with retinal symptoms who are known to have a lower
risk of subsequent stroke.21 Most (13/16) of
these patients were part of the low-risk group with normal OEF. Due to the
small number of patients and the lack of subsequent strokes, we were unable
to determine if the low risk of subsequent stroke reported in those with retinal
events is attributable to the rarity of stage II hemodynamic failure or is
independent of hemodynamic factors.
The development of modern imaging techniques has made it possible to
indirectly assess the hemodynamic status of the human cerebral circulation
in vivo. Most of these methods rely on identification of preexisting autoregulatory
vasodilation by the measurement of CBV or by the CBF response to vasodilatory
stimuli as a criterion for hemodynamic compromise.7
Physiologically, this approach can be expected to detect less severely affected
subjects than the measurement of OEF.10 We
therefore believe that it would be inappropriate to extrapolate our findings
to other modalities. In fact, Yokota and colleagues9
have recently completed a longitudinal study similar in design to ours in
which the relationship between reduced vasodilatory response to acetazolamide
and the subsequent risk of stroke was evaluated. They prospectively followed
up 105 symptomatic patients with severe stenosis or occlusion in the internal
carotid or the MCA for a median of 2.7 years. There was no difference in subsequent
stroke occurrence between the group with reduced vasodilatory response (7/55)
and the group with normal vasodilatory response (6/50). Yamauchi et al23 have also reported increased risk of stroke in patients
with increased OEF measured by PET in a smaller study with 1-year follow-up.
This study, although consistent with our results, is not entirely comparable
since the absolute value of the OEF, rather than the hemispheric ratio, was
used as the criterion for hemodynamic failure. Furthermore, this study suffered
from possible bias due to lack of blinding and failure to consider the role
of other risk factors.
Although this study establishes that stage II hemodynamic failure is
a strong predictor of subsequent stroke in patients with symptomatic carotid
occlusion, it cannot establish the mechanism for these subsequent strokes.
The demonstration of hemodynamic failure at baseline does not necessarily
prove that all subsequent strokes are hemodynamically mediated. Low-flow states
may predispose to the formation of thromboemboli or, alternatively, thromboemboli
may cause infarction more readily in areas with poor collateral circulation.
The results of medical treatment of stage II patients were poor and
comparable with those reported for medically treated patients with symptomatic
severe carotid stenosis.2 Surgical approaches
to improve cerebral hemodynamics, such as extracranial-intracranial (EC-IC)
arterial bypass surgery, may appear to be logical treatment for these patients.
However, a large, multicenter randomized trial conducted from 1977 to 1985
showed no benefit of EC-IC bypass surgery in preventing subsequent stroke
in patients with symptomatic carotid occlusion.22
At the time that this trial was conducted, there was no reliable and proven
method for identifying a subgroup of patients in whom cerebral hemodynamic
factors were of primary pathophysiologic importance. We have now established
that such a subgroup can be identified and, furthermore, that they are at
high risk for subsequent stroke when treated medically. In stage II patients,
EC-IC bypass surgery will return hemispheric OEF ratios to normal.12,24- 26 However,
in the absence of an empirical trial, it cannot be assumed that the surgery
would be of benefit in this subgroup of patients. The morbidity and mortality
due to surgery and the long-term stroke risk in patients who were operated
on are not known. However, given our documented ability to identify this high-risk
subgroup, it is appropriate at this time to consider performance of a new
trial of EC-IC bypass surgery restricted to patients with stage II symptomatic